Loughborough University Institutional Repository
Slotted and circular pore surface microfiltration. This item was submitted to Loughborough University's Institutional Repository by the/an author.
Additional Information:
•
A Doctoral Thesis.
Submitted in partial fulfilment of the requirements
for the award of Doctor of Philosophy of Loughborough University.
Metadata Record: https://dspace.lboro.ac.uk/2134/13634 Publisher:
c
Alan James Bromley
Please cite the published version.
This item was submitted to Loughborough University as a PhD thesis by the author and is made available in the Institutional Repository (https://dspace.lboro.ac.uk/) under the following Creative Commons Licence conditions.
For the full text of this licence, please go to: http://creativecommons.org/licenses/by-nc-nd/2.5/
University Library
• • Loughborough . , University
AuthorlFtlmg TItle ........ ~!.~.~~ .................................
........................................................................................
,-
Class Mark ...•.......................•.............•............•.........•....
Please note that fines are charged on ALL overdue items. '
,
:'
/
I~nlt~irffi 1111111111111111111111
Slotted and Circular Pore Surface Microfiltration
by
Alan James Bromley
A Doctoral Thesis
Submitted in partial fulfilment of the requirements for the award of Doctor of Philosophy of Loughborough University
27 September 2002 ,
, )
'
.
'
,
, ,
.
,
.
,
L
~
~
© byAlan James Bromley; 2p02
,
, I - .. ~ ~
, .... -
-
~
-
.....
H
................. ,
..
u~~~ ,hl"l'.ry .
~~
Dare
CI_
Ace No.
.......
N." c'l , .... ..... ~
. _.M . . . . . ~'j)_t.,
ABSTRACT Microfiltration, membrane, surface filter, critical flux, model, pore size, geometry, electroless plating The work described by this thesis is a comparison of pore opemng geometry for true surface mlcrofilters. True surface mlcrofilters can be thought of as very fine sieves, with pore sizes less than 10 microns. All other types of so-called microfiltration membranes do not rely on sieving, but obtain their pore retention rating by particle collectIOn mechanisms similar to depth filters. Particle deposition within such microfilters results in permeate flow rate dechne, for a fixed pressure filtration, or pressure drop rise, for a fixed rate filtration. The true surface microfilter pore geometnes considered were circular and slotted, and microfilters with filtering dimension of less than 10 microns were used. The slotted pore microfilters are not commercially available and had to be made in the laboratory as part of this study. The technique used was to plate nickel onto an existmg substrate, thereby reducing the pore dimension until It was within the microfiltration range. The plating was by electroless nickel solution and not by galvanic means. Significant development of the electroless platmg technique led ultimately to the successful manufacture of process scale slotted surface microfilters. Filtration tests investigated the cntical flux that could be obtained under different operating conditions of shear, concentration, etc., where the shear was proVided in a controlled manner using a stirred cell, similar in design to a cone-and-plate viscometer. The critical flux described here is the permeate flow rate that exists in the absence of major particle fouling. It is a stable flux, With respect to time, and can only be present when there is insignificant fouling of the membrane, I.e. the shear at the membrane surface is sufficient to prevent Significant particle deposition. Under these conditions, the microfilter can be used to fractionate, or classify solid particles m suspension. This contrasts significant fouling conditions, where the depOSit and not the filter determine particle retention. The critical flux using the slotted microfilters was significantly greater than ,that of the circular pore filters and, a straightforward shear-dependent mathematical model appeared to represent the data from the slotted filters adequately whereas the cntical flux was not shear dependent when filtering on the circular pore membranes. The circular pore filters very quickly estabhshed a deposit under most operating conditions, thus limiting the possibihty of classificatIOn of particles using such a filter type. Practical applications of the novel slotted geometry membranes include filtration and classification using design information reported in this thesis for the specific material tested.
For Mum and Dad
ACKNOWLEDGEMENTS
The road to completion of this thesis has often seemed long and hard and I would hke to thank the following people for their help and support along the way.
For their continual support, guidance and patience I am indebted to my supervisors Richard Holdich and lain Cumming. I would also like to thank many members of the Department of Chemical Engineering past and present, for their advice and assistance. Particularly I wish to express gratitude to Warren Eagles, Steve Graver, Paul Russell, Martin Kerry, Dave Smith, and Serguei Kostinetsev. Thank you Frank Page for sharing your technical expertise to deliver the electron micrographs of surface microfilters, and PIp Amos for engineering an excellent stirred-cell experimental rig.
.
To the dream team of Paul Izzard and Terry Neale for all your help, support and humour (at my expense) With countless computing conundrums - cheers. We finally made it past Windows 3.1!
To My colleagues in the Department and friends elsewhere thank you for being there: Andrew Darling, Chris Wright, Dave Richardson, Davis Marasco, lsaac Hodgson, John Robinson, Jonathan Pascoe, Kevin Ayadassen, Patricia Morales, Sarah Haydock, and Scott Roane.
I offer my sincere appreciation to the Engineering and Physical SCiences Research
Council (EPSRC) for their financial assistance during this work.
On a personal level I would hke to thank those mdividuals who have supported " myself both emotionally and financially dunng thiS roller-coaster ride of research. To my parents, who have offered unwavering encouragement, support, and friendship throughout, and to Gill, lan, and Ohver, I am eternally grateful. To Les, Claire and Geoff, Sue and Marcus, and Parviz, for your continued friendship, I am indebted.
CONTENTS
1.
INTRODUCTION
1-1
2.
LITERATURE REVIEW: ELECTROLESS NICKEL PLATING
2-1
2.1 2.2 2.3 2.3.1 2.3.1.1 2.3.1.2 2.3.2 2.4 2.4.1 2.42 2.4.3 2.5
2-1 2-2 2-4 2-4 2-5 2-5 2-6 2-7 2-7 2-7 2-8 2-9
2.6.1 2.6.2 2.7
APPLICATIONS OF ELECTROLESS NICKEL PLATING ELECTROLESS NICKEL PLATING SOLUTION ELECTROLESS NICKEL PLATING (HYPOPHOSPHITE BATH) Acid Hypophosphite Bath Deposition Mechanisms Classification of the Metal Substrate Alkaline Hypophosphite Bath PRE-TREATMENT METHODS FOR METAL SUBSTRATES Surface Impunties and Defects Surface Cleaning Surface Activation FACTORS AFFECTING DEPOSITION RATES AND ALLOY COMPOSITION Plating Solution Temperature Plating Solution pH Contaminants Loading of the Plating Bath Solution Agitation Plating Bath Age EFFLUENT TREATMENT AND DISPOSAL OF SPENT PLATING BATHS Pre-treatment Chemicals Plating Bath Components SUMMARY
3. 3.1 3.1.1 3.1.2 3.1.3 3.2 3.2.1 3.2.2
EXPERIMENTAL: ELECTROLESS NICKEL PLATING METAL MICROFILTERS NIckel 'Veconic Plus' Screen with CIrcular Pores (CPO) Nickel 'Veconic Plus' Screen with Slotted Pores (SPO) Stainless Steel Wire Mesh ELECTROLESS NICKEL PLATING (ENP) Ni-Shield M Plating Solution Electroless NIckel Plating Rig
2.5.1 2.5.2 2.5.3 2.5.4 2.5.5 2.5.6 2.6
2-9 2-12 2-14 2-15 2-15 2-17 2-18 2-18 2-18 2-19 3-1
3-1 3-2 3-8 3-8 3-9 3-9 3-10
3.2.2.1 3.2.2.2 3.2.2.3 3.3 3.3.1 3.3.2 3.4 3.4.1 3.4.2 3.4.3 3.4.4 3.4.5 3.4 6 3.5 35.1 3.5.2 3.5.2.1 3.5.2.2 3.5.2.3 / 3.5.2.4 3.5.3 3.5.4 3.6
Pumped System Mechanical and Air Sparge Agitation Disposal of Waste Plating Solution NICKEL ELECTROPLATING (EP) Nickel Electroplating Solution Electroplating Experimental Rig SURFACE PRE-TREATMENT Chemicals Used Alkaline Degreasmg (Stage One) Surface Layer Removal (Stage Two) Surface Activation (Stage Two) Alkaline Coating Treatment (Stage Three) Transfer to Electroless Plating Bath (Stage Three) CHARACTERISATION OF METAL MICROFILTERS Thickness Pore Size Analysis Optical Microscope (OM) Scanmng Electron Microscope (SEM) Image Analysis Coulter Porometer EvaluatIOn of Open Area Quantitative Elemental Analysis SUMMARY
3-11 3-12 3-14 3-14 3-14 3-15 3-17 3-17 3-19 3-19 3-19 3-21 3-21 3-21 3-21 3-21 3-21 3-22 3-24 3-29 3-30 3-32 3-32
4.
4-1
ASSESSMENT OF PRE-TREATMENT STAGES 4.2 4.2.1 Surface Layer Removal by Hydrochloric Acid 4.2.2 Plating Surface Activation usmg Nickel Electroplating 4.2.2.1 No Surface Activation (control experiments) 4.2.2.2 Electroplating of Nickel SPO Screen 4.2.2.3 Electroplating (with Polarity Reversal) of Nickel SPO Screen 4.2.2.4 Electroplating of Stainless Steel Mesh 4.2.3 Alkaline Sodium Carbonate Coating
4-1 4-1 4-4 4-5 4-5 4-9 4-12 4-14 4-14 4-15 4-15 4-16 4-17 4-18 4-19
RESULTS AND DISCUSSION: ELECTROLESS NICKEL PLATING 4.1 ASSESSMENT OF SOLUTION AGITATION METHODS 4.1.1 Pumped Circulation 4.1.2 Mechanical Overhead Stirring 4.1.3 Air InjectIOn 4.1.3.1 Comparative Plating Experiments with and without Agitation 4.1.32 SEM Analysis of Circular and Slotted Pores 4.1.3.3 Elemental Analysis of Substrate Surface and Platmg Deposit
4.3 4.4
ASSESSMENT OF PLATING BATH LIFE CONCLUSIONS
4-22 4-23
5.
LITERATURE REVIEW: SURFACE MICROFILTRATION MEMBRANE PROCESSES GENERAL APPLICATIONS
5·1
5.1 5.2 5.2.1
Removal of Cryptosporidium Parvum Oocysts from Drinking Water 5.3 DEAD-END AND CROSS FLOW FILTRATION 5.3.1 Dead-end Filtration 5.3.2 Crossflow Filtration 5.3.3 Stirred-cell OperatIOn 5.4 MEMBRANE FOULING SURFACE AND DEPTH FILTRATION 5.5 5.6 CRITICAL FLUX 5.6.1 Experimental Determination of the Cntical Flux 5.6.1.1 Direct ObservatIOn Through the Membrane (DOTM) 5.6.1.2 Pressure Profile Technique Critical Flux Experimental Results 5.6.2 5.7 FUNDAMENTAL MODELS 5.7.1 Concentration Polansation Model 5.7.2 Mass Transfer Coefficients 5.7.3 Shear-Enhanced Hydrodynamic Diffusion Inerthil Lift 5.7.4 5.7.5 Interaction Enhanced Migration 5.7.5.1 Electric Double Layer Repulsion 5.7.5.2 Van der Waals Forces 5.7.5.3 Born Repulsion 5.7.5.4 Acid-Base (AB) Interaction 5.7.5.5 Total Interaction Potential Energy OTHER MODELS 5.8 5.8.1 Surface Transport Models 5.8.1.1 Axial Convection
5-1 5-2 5-2 54 54 5-5 5-7 5-8 5-11 5-14 5-15 5-15 5-15 5-16 5-18 5-18 5-20 5-23 5-25 5-28 5-28 5-30 5-31 5-32 5-32 5-34 5-34 5-34
5.8.1.2 Frictional Force Balance Pore Blocking and Cake Filtration Model 5.8.2 5.8.3 Combined Particle Deposition Models Pore-Particle Interaction 5.8.4 5.9 SUMMARY
540
6.
6·1
6.1
THEORY: SURFACE MICRO FILTRATION INTRODUCTION TO PARTICLE DEPOSITION MODEL
5-35 5-36 5-38 5-40
6-1
6.1.1 6.1.2 6.1.3 6.1.4 6.1 5 6.1.6
Gravitational Settling VelocIty (vG) Mass Transfer Coefficient Back Transport Velocity due to Particle DIffusion (Vd) Back Transport Velocity due to Shear-Enhanced Diffusion (vs) Back Transport Velocity due to Inertial Lift (v{) Back Transport Velocity due to Interaction Enhanced Migration (v,)
6-2 6-3 6-5 6-7 6-9 6-10
6.1.6.1 Electrostatic Double Layer Interaction Potential (VDLR) 6.1.6.2 London-van der WaaIs Interaction Energy (VLVA) 6.1.6.3 Total Interaction Energy (Vro'a') 6.1.7 Total Back Transport Velocity (v'o'a') 6.2 SUMMARY
6-10 6-12 6-12 6-17 6-20
7. 7.1 7.2 7.3 7.3.1 7.3.2 7.3.2.1 7.3.22 7.3.2.3 7.3.3 7.3.4 7.3.5 7.4 7.5 7.6 7.7 7.8 78.1 7.8.2 78.3 78.4 7.8.5 7.9
EXPERIMENTAL: SURFACE MICROFILTRATION STIRRED-CELL TEST RIG MEMBRANES
7·1 7-1 7-6 7-8 7-9 7-10 7-10 7-10 7-10 7-11 7-11 7-12 7-12 7-12 7-13 7-13 7-14 7-14 7-15 7-16 7-17 7-17 7-20
8.
RESULTS AND DISCUSSION: SURFACE MICROFILTRATION CONSTANT-RATE FILTRATION WITH NICKEL LARGE PORE (SPO) ANDPOLYCARONATE ISOPORE (CPO) MEMBRANES
8.1
CHALLENGE MATERIAL Particle SIze Determination Alternative Methods of Particle Size Analysis HIAC ROYCO Particle Sensor and Counter Lasentec PAR-TEC lOO Particle Sizer Coulter Multisizer IT Particle Zeta Potential Measurements Latex Suspension Density Latex Stock Suspension Concentration CLEANING NICKEL SPO AND CPO MEMBRANES STIRRED-CELL RIG PREPARATION CLEAN WATER MEMBRANE RESISTANCE TEST (MRT) PREPARATION OF LATEX FEED SUSPENSION CONSTANT-RATE FILTRATION EXPERIMENTS Determination of Critical Flux (lent) Experiments Involving Larger Pore Sizes Effect of Shear Rate on Critical Flux Effect of Concentration on Critical Flux Grade Efficiency c, SUMMARY
8-1 8-1
8.1.1 8.1.2 8.1.3 8.1.4 8.1.5 8.2 8.2.1 8.2.2 8.2.3 8.2.4 8.2.5 8.2.6 8.3 8.3.1 8.3.2 8.3.3 8.3.4 8.3.5 8.3.6 8.3.7 8.3.8 8.3.9 8.4 8.4.1 8.4.2 8.4.3 8.4.4 8.4.5 8.5 8.5.1 8.6 8.6.1 8.7 8.8
Membrane PropertIes Critical Flux Measurement (Penneate Recycled) Particle Carryover from Column and Tubing Walls Critical Flux Measurement (Penneate Collected) Grade EffiCIency CONSTANT-RATE FILTRATION WITH 10 0 J.Lm NICKEL (SPO) MEMBRANE
8-2 8-4 8-8 8-9 8-10 8-12
Membrane Properties Effect of Shear Rate on Critical Flux
8-12 8-14
Type of Critical Flux Grade Efficiency Critical Flux without Surface Shear Comparison of PartIcle DeposItion Model with Flux Measurements
8-18 8-20 8-21 8-22
CONSTANT-RATE FILTRATION WITH 4.9 J.Lm NICKEL (SPa) MEMBRANE Membrane Properties Wetting of Membranes Effect of Shear Rate on Critical Flux Assessment of Backflushing Repeatability of Critical Flux Analysis Order of CntIcal Flux Experiments Type of Critical Flux Grade Efficiency Critical Flux WIthout Surface Shear CONSTANT-RATE FILTRATION WITH 5.0 J.Lm PC ISOPORE (CPO) MEMBRANE Membrane Properties Effect of Shear Rate on Cntical Flux Type of Critical Flux Grade Efficiency Cntlcal Flux without Surface Shear COMPARISON OF CRITICAL FLUX FOR 5 J.Lm SPO AND CPO MEMBRANES Effect of Time on Critical Flux Analysis CONSTANT-RATE FILTRATION WITH 4.1 J.Lm NICKEL (SPO) MEMBRANE Effect of Shear Rate on Critical Flux COMPARISON OF MODEL WITH EXPERIMENTAL MEASUREMENTS EFFECT OF SURFACE SHEAR ON 4.8 J.Lm NICKEL (CPO) MEMBRANE
8-23 8-23 8-25 8-29 8-31 8-31 8-32 8-33 8-34 8-36 8-36 8-36 8-36 8-39 8-40 8-41 8-42 8-44 8-45 8-47 8-48 8-49
8.9 8.10 8.11
SENSITIVITY OF DEPOSITION MODEL TO PARTICLE SIZE EFFECT OF FEED CONCENTRATION ON CRITICAL FLUX CONCLUSIONS
8-51 8-53 8-54
9.
CONCLUSIONS AND RECOMMENDATIONS CONCLUSIONS RECOMMENDATIONS FOR FURTHER WORK
9·1
9.1 9.2
REFERENCES NOMENCLATURE
APPENDICES A B C D
Electroless Nickel Plating Experimental Details and Results Electroless Nickel Plating Procedure for the Production of Metal Microfilters Radial and Tangential Shear Rate Analysis of Stirred-cell Operation Supporting Calculations for Particle Deposition Model
E F G H I J K
Pdesol™ Model For Interaction Enhanced Migration Stirrer Calibration Pressure Transducer Calibration Pore Size Distributions of Nickel SPO and CPO Membranes Constant-rate Filtration Experimental Details and Results Relative Importance of Membrane Open Area to Cntical Flux Author's Publications
9-1 9-4
LIST OF FIGURES
Figure 2.1
Dependence of deposition rates (l1m h· l) on temperature (0C)
Figure 2.2
Relative deposition rate (l1m h· l ) as a function of bath temperature (0C)
Figure 2.3
Dependence of electroless nickel deposition rate (l1m h- I) as a function of pH and solution temperature (OC)
Figure 2.4
Variation In phosphorous content of deposit (%) versus bath pH
Figure 2.5
Deposition rate as a function of bath age for a third-generation 'Novotect' bath
Figure 3.1
OptIcal microscope images of nickel CPO screen With 39 l1m pores, viewed at 20x magnification; (a) shiny 'filtration' Side, and, (b) matt side
Figure 3.2
SEM micrographs of nickel CPO screen With 17 l1m pores; (a) shiny 'filtration' side, and, (b) matt side rotated by 45°
Figure 3.3
SEM micrograph of nickel epo screen with 17 l1m pores. Single matt-side pore viewed at high magnificatIOn to show conical pore entrance
Figure 3.4
SEM micrographs of nickel SPO screen with 13 x 403 l1m slots; (a) many slots (shiny side), (b) single slot (matt side), (c) multiple slots (matt Side), and, (d) single slot (matt side)
Figure 3.5
Optical microscope images of stainless steel mesh exhlbitmg 29 l1m square pore opemngs; (a) 10x magnification, and, (b) 20x magnification
Figure 3.6
Electroless nickel plating ng employing pumped circulation
Figure 3.7
Plating bath agitatIOn by; (a) overhead stirrer, and (b) air injection (air sparge)
Figure 3.8
Schematic representation of nickel electroplating rig
Figure 3.9
Pre-treatment schedule for electroless nickel plating and nickel electroplating.
Figure 3.10
Slotted and circular pore width measurement by SEM; (a) Isolated circular pore opening (untreated), and, (b) single nickel slot (untreated)
Figure 3.11
A typical SEM micrograph of slotted media used for Image Analysis
Figure 3.12
Image processing stages for pore size analysis of slotted pore opening
Figure 3.13
Image Analysis of slotted pore opening; (a) rectangular area selected for profiling, and, (b) density profile of pore area (Plot Profile)
Figure 3.14
Pore width profile along slot length
Figure 3.15
Representation of open area calculation methodology
Figure 4.1
Comparison of average pore size reduction achieved in the presence of air sparge agitation and without solutIOn agitation (17.1 Ilm original pore size)
Figure 4.2
Average electroless nickel plating rate achieved with air sparge agitation
Figure 4.3
SEM micrographs of single circular pore opening (high magnification); (a) original matt side (17.1 Ilm pores) and, (b) matt side after 50 minutes plating time
Figure 4.4
Circular pore width measurement by SEM after 50 minutes plating time. Three-fold pore size reduction achieved from origmal pore size
Figure 4.5
SEM micrographs of nickel SPO screen (high magnification); (a) original shiny side (13.4 Ilm slots), and, (b) shiny side after 60 minutes nickel plating
Figure 4.6
SEM micrograph of nickel SPO screen after 60 minutes nickel platmg (see Figure 45) showing slots on matt side; viewed at high magnification
Figure 4.7
Electron microprobe analysis of nickel SPO membrane surface; (a) untreated surface, and, (b) nickel plated surface followmg slot size reduction to 4.1 Ilm
Figure 5.1
Two Cryptosporidium Parvum oocysts on a Nuclepore membrane filter (note the scale bar)
Figure 5.2
Schematic representation of the cross-flow filtration process
Figure 5.3
Schematic representation of stirred-cell filtration
Figure 5.4
Different filtration resistances experienced dunng microfiltratlOn
Figure 5.5
Penneate flux decay to a pseudo-steady state, and with continuing foulmg
Figure 5.6
A depth effect microfilter and pore size distnbution (note scale bar on SEM)
Figure 5.7
Example surface microfilter rated at 2 !lm
Figure 5.8
A diagrammatIc representation of the Film Theory
Figure 5.9
Forces and torques acting on a charged, spherical particle suspended in a viscous fluid undergoing lanunar flow ID the proximity of a flat porous surface
Figure 6.1
Gravitational settling velocity,
VG
(!lm S-I) as a function of particle
diameter, dp (!lm)
Figure 6.2
Back transport velocity due to particle diffusion,
Vd
(10-6 m
S-I)
as a
function of particle diameter, dp (!lm)
Figure 6.3
Comparison of particle diffusivity (m2
S-I)
by particle diffusion and
shear-enhanced motion over a range of particle size, dp (!lm)- Shearenhanced diffusion coefficient exceeds particle diffusion at 046 !lm
Figure 6.4
Back transport velOCity due to shear-enhanced diffusion mechanism, 6 Vs (10- m S-I) as a function of particle size, dp (!lm)
Figure 6.5
Back transport velocity due to inertial lift,
VI
(10-6 m
S-I)
as a function
of particle size, dp (!lm)
Figure 6.6
Double layer repulsive, VDUI, London-van der Waals attractive, VLVA , and combined, VTotal, particle-particle interaction energies as a function of separation distance, h (nm)
Figure 6.7
Dimensionless total interactIOn energy (VTotalkB 1) as a function of separatIOn distance, h (nm) at vanous NaCl electrolyte concentrations (g mol r\ (a) 5 x IO-s, (b) I X 10-4, (c) 1 X 10-3, (d) 1 x 10-2, and, (e) 0.1
Figure 6.8
Comparison of back transport velocitIes generated from different particle mechanisms over a range of particle sizes. Total back transport is determined from particle, Vd, and shear-enhanced diffusion, Vs, and inertial lift, VI, mechanisms only. The Figure is comparable to that presented in Yoon et al. (1999)
Figure 7.1
Schematic representation of stirred cell rig
Figure 7.2
Schematic representation of stirred-cell membrane holder
Figure 7.3
Diagrammatic representation of stirrer-rod geometries
Figure 7.4
Watson-Marlow 101 VIR peristaltic pump performance charactenstics
Figure 7.5
SEM micrograph of polystyrene latex particles filtered using a
o45 !lm cellulose mtrate membrane
Figure 7.6
Particle size dlstnbutlOn, by mass, of latex stock suspensions measured using a Malvern Mastersizer instrument
Figure 7.7
Schematic representatIOn of gravity-run stirred-cell rig to detennine grade efficiency
Figure 8.1
SEM micrographs of 10.0 Ilm MiIlipore polycarbonate isopore membrane with circular pore openings; viewed at (a) low magnification, (b) high magnification
Figure 8.2
SEM micrograph of nickel slotted membrane with average pore dimensions 13.4 x 4021lm; viewed at standard magnification
Figure 8.3
Comparison of pressure differences (mbar) for 10.0 Ilm Isopore CPO and 13.4 Ilm nickel SPO membranes with permeate recycle (1200 S-I shear; 0.093 kg m-3 latex feed concentration)
Figure 8.4
Pressure difference (mbar) across 13.4 Ilm nickel SPO membrane at elevated permeate fluxes (1200 S·I shear; 0.093 kg m·3 latex feed suspension)
Figure 8.5
Comparison of pressure differences (mbar) for 10.0 Ilm Isopore CPO and 13.4 !lID nickel SPO membranes without permeate recycle (1200 S·I shear; 0 093 kg m·3 latex feed concentration)
Figure 8.6
Grade efficiency of 10.0 Ilm isopore CPO membrane. Penneate sample PI (up to 43 ml filtered). Average of permeate samples P2 - 5 (43 - 184 ml filtered)
Figure 8.7
Grade efficiency of 13.4llm nickel SPO membrane. Penneate sample PI (up to 40 ml filtered). Average of penneate samples P2 - 5 (40 179 ml filtered)
Figure 8.8
SEM micrograph of nickel slotted membrane with average pore dimensions 10.0 x 4021lm; viewed at (a) standard magnification, and, (b) high magnification
Figure 8.9
Pressure profile of 10.0 Ilm nickel SPO membrane up to the critlcal flux (600 S-I shear; 0.020 kg m·3 latex feed concentration)
Figure 8.10
Complete pressure profile of 10 0 Ilm nickel SPO membrane (600 S-I shear; 0.020 kg m-3 latex feed concentration)
Figure 8.11
Critical flux of 10.0 Ilm nickel SPO membrane as a functIOn of surface shear (0 - 1200 S-I shear; 0.020 kg m-3)
Figure 8.12
RelatIonship between applied flux and pressure difference for filtration with 10.0 !tm nIckel SPO membrane compared to pressure difference for same clean water flux. Low shear (0 - 399 S·I), medium shear (400 - 800 S·I) and high shear rates (801 - 1200 S·I) reported
Figure 8.13
Grade efficiency of 10.0 !tm nickel SPO membrane. PI (0 - 40 ml); P2 (40 - 80 ml); P3 (80 - 120 ml); and P4 (120 - 160 ml)
Figure 8.14
Pressure profile of 10.0 pm nIckel SPO membrane without shear (0 S·I shear; 0.020 kg m·3 latex feed concentration)
Figure 8.15
Critical flux due to shear for 10.0 !tm nickel SPO membrane compared to particle deposition model for 5 !tm latex particles
Figure 8.16
SEM micrograph of nickel slotted membrane with average pore dimensions 4.9 x 396 !tm; viewed at (a) standard magnIfication, and, (b) high magnIfication
Figure 8.17
SEM mIcrograph of 4.9 !tm nickel SPO membrane at high magnification showmg corroded regIOn of nickel-phosphorous coating
Figure 8.18
Electron microprobe analysis of 4.9 !tm nickel SPO membrane following Decon 90 treatment (1.0 g rl cleaning solution); (a) nickelphosphorous coating, and, (b) region corroded to origmal membrane surface
Figure 8.19
Pressure profile of 4.9 !tm nickel SPO membrane (600 S·I shear; 0.020 kg m·3 latex feed concentration)
Figure 8.20
Cntical flux of 4.9 !tm nickel SPO membrane as a function of surface shear (0 - 1200 S·I shear; 0.020 kg m·3 feed suspension)
Figure 8.21
Order of cntical flux experiments for 4.9 !tm nickel SPO membrane
Figure 8.22
Relationship between applied flux and pressure difference for filtration with 4.9 !tm nickel SPO membrane compared to pressure difference for same clean water flux. Low shear (0 - 399 S·I), medium shear (400 - 800 5"1) and high shear rates (801 - 1400 S·I) reported
Figure 8.23
Grade efficiency of 4.9 pm nickel SPO membrane. PI (0 - 40 ml); P2 (40 - 80 ml); P3 (80 - 120 ml); and P4 (120 - 184 ml)
Figure 8.24
Pressure profile of 4.9 I1ffi nickel SPO membrane WIthout surface shear (0 S·I shear; 0.020 kg m·3 latex feed suspension)
Figure 8.25
SEM micrograph of 50 I1ffi polycarbonate isopore membrane with circular pore openings; viewed at (a) low magnification, and, (b) high magnification
Figure 8.26
Pressure profile of 5.0 Jlm polycarbonate Isopore CPO membrane (600 S·I shear; 0.020 kg m·3 latex feed concentration)
Figure 8.27
Critical flux of 5.0 Jlm polycarbonate Isopore CPO membrane as a function of surface shear (0 - 1200 S·I shear; 0.020 kg m·3)
Figure 8.28
Relationship between applied flux and pressure difference for filtration with 5.0 Jlm isopore CPO membrane compared to pressure difference for same clean water flux. Low shear (0 - 399 s·\ medium shear (400 - 800 S·I) and high shear rates (801 - 1400 S·I) reported
Figure 8.29
Grade efficiency of 5.0 Jlm polycarbonate isopore CPO membrane. PI (0 - 33 ml); P2 (33 - 63 ml); P3 (63 - 95 ml); P4 (95 - 126 rnI); PS (126 - 175 ml)
Figure 8.30
Pressure profile of 5.0 Jlm polycarbonate isopore CPO membrane without shear (0 S·I shear; 0020 kg m·3 latex feed concentration)
Figure 8.31
Comparison of critical flux due to shear for 4.9 Jlm nickel SPO and 5 0 Jlm Isopore CPO membranes
Figure 8.32
Extended filtration of 4.9 Jlm ~ickel SPO membrane at 84 Jlm S·I for 140 minutes (600 S·I shear; 0.01 kg m·3 latex feed concentration)
Figure 8.33
Extended filtration of 5.0 Jlm isopore CPO membrane at 7 Jl m S·I for 140 minutes (600 S·1 shear; 0.10 kg m·3 latex feed concentration)
Figure 8.34
SEM micrograph of nickel slotted membrane with average pore dimensions 4.1 x 382 Jlm; vIewed at (a) standard magnification, and, (b) high magnIfication
Figure 8.35
Critical flux of 4.1 Jlm nickel SPO membrane as a function of surface shear (0-1200 S·I shear; 0.020 kg m·3 latex feed concentration)
Figure 8.36
Companson of critical flux due to shear for 4.9 Jlm and 4.1 Jlm nickel slotted, and 5.0 Jlm polycarbonate isopore membranes with particle deposition model for 5 Jlm latex particles
Figure 8.37
Optical microscope image of 4.8 Jlm nickel membrane with 'cIrcular' pores; viewed at (a) standard magnification, and, (b) high magnification. Original hexagonal pore shape shown (inset)
Figure 8.38
Companson of critical flux due to shear for 4.9 Jlm and 4.1 Jlm nickel SPO membranes, 5.0 Jlm polycarbonate isopore CPO membrane, and 4.8 Jlm nickel CPO membranes together with particle deposition model for 5 Jlm latex particles
Figure 8.39
Effect of particle size on critical flux for the particle deposition model. Experimental data for all smaller sized SPO and CPO membranes plotted on same axes
Figure 8.40
Comparison of critical flux as a function of feed concentration for 4.9 Ilm nickel SPO and 5.0 Ilm polycarbonate isopore CPO membranes
Figure C.l
Cone-and-plate viscometer theory applied to stirred-cell
Figure C.2
Evaluation of radial shear rate
Figure C.3
Radial shear rate,
Yraw,P
(5"1) calculated as a function of radial
distance from centre of the stirrer cone (mm) using simple radial shear model
JC for NaCl strong electrolyte
Figure D.l
Am vs
Figure F.l
Heidolph RZR 2102 electronic stirrer cahbration
Figure G.l
Calibration graph for Radio Spares pressure transducer (PT)
Figure H.l
Pore size distribution - 10.0 Ilm nickel SPO membrane
FigureH.2
Pore size distribution - 4.9 Ilm nickel SPO membrane
Figure H.3
Pore size distnbution - 4. I Ilm nickel SPO membrane
Figure H.4
Pore size distribution - 4 8 Ilm nickel CPO membrane
Figure J.l
Non-filtration side of 4.9 Ilm nickel SPO membrane following PVA glue addition to block off pores and consistently reduce flow area
FigureJ.2
A study of critical flux behaviour in relation to membrane open area for assorted surface filters
LIST OF TABLES
Table 2.1
Major uses of electroless nickel plating in selected industries (excluding the electronics industry)
Table 2.2
Components of electroless nickel plating baths and their functions
Table 2.3
Reducing agents for electroless nickel plating
Table 2.4
Electroless nickel plating bath compositions used in Industry
Table 2.5
Descnption of Woods Strike surface activation method
Table 2.6
Effect of pH change on electroless nickel plating process
Table 2.7
Contamination problems in electroless nickel deposition
Table 2.8
Typical concentrations of components in a spent Electroless nickel plating bath
Table 3.1
Physical properties of nickel and stainless steel screens
Table 3.2
Optimum electroless nickel plating bath conditions
Table 3.3
Recommended electroplating solution composition
Table 3.4
Nickel electroplating operating conditions
Table 3.5
Cathodic electroplating current values, per electrode
Table 3.6
Pre-treatment schedule matrix
Table 3.7
Pore size determination by SEM analysis
Table 3.8
Pore size measurements of CPO and SPO screens by Image Analysis of SEM and optical microscope images
Table 3.9
Open area measurements of SPO and CPO screens by Image Analysis of SEM and optical microscope images
Table 4.1
Experimental conditions, observations, and pore size analysis for ENP of 17.1 J.UI1 nickel CPO screens with pumped circulation
Table 4.2
Expenmental conditions, observations (including pore size analYSIS) for ENP of different metal screens using mechanical agitation
Table 4.3
Experimental conditions to detennine the effectiveness of ENP
ID
presence of agitation by air injection (17.1 /.lm nickel CPO screens)
\
the
Table 4.4
Pore size analysis (Coulter Porometer
mof nIckel deposition
In
the
presence of air sparge agitation, and without agitation
Table 4.5
Experimental conditions and observations during hydrochloric acid pickling of 17.1 J.Lm nickel CPO screens
Table 4.6
Experimental conditions and observations for ENP of 13.4 J.Lm nickel SPO screens without surface activation or surface later removal
Table 4.7
Experimental conditions, observations, and pore size analysis for ENP of 13.4 J.Lm nickel SPO screens (nickel electroplating pretreatment)
Table 4.8
Experimental conditions, observations, and pore size analysis for ENP of 13.4 J.Lm nickel SPO screens (electroplatmg pre-treatment with polarity reversal)
Table 4.9
Experimental condItions for EP and ENP of stainless steel mesh (29 J.Lm square pores)
Table 4.10
Observations and pore size analysIs for EP and ENP of stamless steel mesh (29 J.Lm square pores)
Table 4.11
Experimental condItIons and observatIOns for ENP of 17.1 J.Lm nickel CPO screens (sodium carbonate treatment evaluation)
Table 4.12
Experimental conditions for EP and ENP of stainless steel mesh with square pores of 29 J.Lm (sodium carbonate pre-treatment evaluation)
Table 4.13
Experimental observations and pore size analysis for EP and ENP of stainless steel mesh (sodium carbonate pre-treatment evaluation)
Table 4.14
Experimental condItions, observations, and pore size analysis of ENP with nickel SPO screens (plating solution life evaluation)
Table 5.1
Filtration processes and theIr characteristics
Table 5.2
Constants in the general mass transfer correlation (Equation 5.6) reported by varIOUS authors.
Table 6.1
Particle size dependence for VariOUS transport mechanisms (sIze range comparable to latex challenge material)
Table 6.2
Shear rate dependence for various transport mechanisms (shear rate range appropriate for constant-rate filtration experiments)
Table 6.3
Summary of transport mechanisms: Performance with increasing parametnc significance (Italics) and overall region of influence
Table 7.1
Selected stirred-cell column dimensions
Table 7.2
Selected stirrer dimensions
Table 7.3
Silicone rubber penstaltic tubmg properties
Table 7.4
Physical properties of polycarbonate track-etched (epO) membranes
Table 7.5
Physical properties of metal surface microfiIters
Table 7.6
Selected experimental conditions for critical flux evaluation as a functIOn of shear rate
Table 7.7
Selected experimental parameters for cntical flux evaluation as a function of latex feed concentratlon
Table S.l
Particles remaining in suspension as a function of initial feed concentratlon
Table S.2
Wettmg capablhties of Bardac-2270-E surfactant and Decon 90 cleaning solutions
TableS.3
Repeatabtlity of 4.9 J.lm nickel SPO membrane cntlcal flux expenments
TableA.1
Experimental conditions for pre-treatment and electroless nickel plating of 17.1 ).lm nickel epo screen
TableA.2
Observations made during pre-treatment and electroless nickel plating of 17.1 J.lm nickel epo screen
TableA.3
Experimental conditions for pre-treatment and electroless nickel plating of 13.4 J.lm nickel SPO screen
TableA.4
Observations made during pre-treatment and electroless nickel plating of 13.4 J.lm nickel SPO screen
TableA.5
Experimental conditions for pre-treatment and electroless nickel plating of 29.2 ).lm stainless steel mesh
TableA.6
Observations made during pre-treatment and electroless nickel plating of 29.2 J.lm stainless steel mesh
Table C.1
Equipment specifications and operating conditions
TableD.1
Sedimentation calculation data
TableD.2
Stirrer and membrane geometries for use with correction factors
TableD.3
Particle diffusion mechanism calculation data
Table D.4
Shear-enhanced hydrodynamic diffusion mechanism calculation data
TableD.S
Inertial lift calculation data
Table D.6
ElectrostatiC double layer repulsion calculation data
TableD.7
Particle transport velocities for 5 llm diameter particles
TableD.S
Experimental data and calculated parameters
TableE.1
Pdesol™ output table for interaction enhanced migration model
Table 1.1
Clean water backflush (BF) and membrane resistance tests (MRT) prior to filtration experiments with 10.0 llm nickel SPO membrane
Table 1.2
Clean water backflush and membrane resistance tests (MRT) prior to filtration experiments with 4.9 llm nickel SPO membrane
Table 1.3
Clean water backflush and membrane resistance tests (MRT) prior to filtration experiments with 4.1 !lID nickel SPO membrane
Table 1.4
Clean water backflush and membrane resistance tests (MRT) prior to filtration experiments wIth 5.0 llm PC isopore CPO membrane
Table I.S
Clean water backflush and membrane resistance tests (MRT) prior to filtration experiments with 4.8 llm nickel CPO membrane
Table 1.6
Critical flux evaluated for 10.0 llm nickel SPO membrane across a range of shear rates (experimental parameters: 0 - 1200 S·I surface shear; 0 020 kg m-3 latex feed suspension)
Table 1.7
Critical flux evaluated for 4.9 llm nickel SPO membrane across a range of shear rates (experimental parameters: 0 - 1400 S·I surface shear; 0.020 kg m-3 latex feed suspension)
Tablel.S
Repeatabihty of critical flux analysis for 4.9 lLm nickel SPO membrane across a range of shear rates (expenmental parameters: 0 1200 S-1 surface shear; 0.020 kg m-3 latex feed suspension)
Table 1.9
Critical flux evaluated for 4.1 lLm nickel SPO membrane across a range of shear rates (experimental parameters: 0 - 1200
S-1
surface
3
shear; 0.020 kg m- latex feed suspension) Table 1.10
Critical flux evaluated for 5.0 lLm PC isopore CPO membrane across a range of shear rates (experimental parameters: 0 - 1400
S-1
surface
3
shear; 0.020 kg m- latex feed suspension) Table 1.11
Cntical flux evaluated for 5.0 lLffi PC isopore CPO membrane across a range of shear rates (experimental parameters: 0 - 1200
S-1
surface
3
shear; 0.093 kg m- latex feed suspension) Table 1.12
Critical flux evaluated for 4.8 lLm nickel CPO membrane across a range of shear rates (experimental parameters: 0 - 1400
S-1
surface
3
shear; 0.020 kg m- latex feed suspension) Table 1.13
Critical flux evaluated for 10.0 lLm nickel SPO membrane across a range of latex feed concentrations (experimental parameters: 600 S-1 surface shear; om - 0040 kg m-3 latex feed concentration)
Table 1.14
Critical flux evaluated for 4.9 lLm nickel SPO membrane across a range of latex feed concentrations (experimental parameters: 600
S-1
3
surface shear; 00052 - 0040 kg m- latex feed concentration) Table 1.15
Critical flux evaluated for 5.0 lLm PC isopore CPO membrane across a range of latex feed concentrations (experimental parameters: 600 S-1 surface shear; 0.0035 - 0040 kg m-3 latex feed suspensIOn)
Table 1.16
Extended filtration of 4.9 lLm nickel SPO membrane at various concentrations of feed suspension (expenmental parameters: 600
S-1
3
surface shear; 0.01 - 0.10 kg m- latex feed concentration) Table 1.17
Extended filtration of 5.0 lLffi PC isopore CPO membrane at various concentrations of feed suspension (experimental parameters: 600 3
surface shear; 0.01 - 0.093 kg m- latex feed concentration)
S-1
CHAPTER ONE. Introduction
CHAPTER ONE
INTRODUCTION In microfiltration the problem of irreversIble membrane fouling by internal deposition of material within the filter medium is well documented. It arises with most conventional micro filtration membranes because the pore ratmg of the filter is determined by tortuous flow channels through the membrane, rather than by the true surface pore opening dimension. There is a common misconception that the membrane pore rating is much like a sieve size. Hence, a microfilter rated at I J.lIIl may have pore openings many microns wider allowing fine particulate material to enter, and Irreversibly foul, the membrane. There are several commercially available microfilters designed to overcome this problem, where the pore rating is equal to the pore opening size, i.e. similar to sieves, but these are restricted to laboratory scale use only. Most of them are manufactured by a track-etching process and the Nuclepore filter is one such example. Other non-commercial examples have recently been reported (Kuiper et al., 1998; Holdich et al., 1998; BromIey et al., 2002) and they all filter by retaining the particles on the surface of the filter, rather than allowing internal filtration. Experimental evidence suggests that these true surface filters provide high, long-term fluxes compared to the conventional macroporous type of microfiltration membrane. However, fouling may still take place on a true surface filter due to cake, or gel layer formation and because of pore plugging. If the deposit is retained on the surface of the filter, the operating conditions may be altered to minimise the foulmg resistance to permeate flow; e.g. a high crossflow rate may reduce the deposit layer thickness, or even prevent deposition altogether. This effect has become known as critical flux; which represents a sensitive balance between the permeate rate, operating conditions and the material to be filtered (Field
et al., 1995; Kwon and Vigneswaran, 1998; Li et al., 2000). For fluxes below a critical value there is little fouling because the deposit can be removed by the shear, or other hydrodynamic conditions. Hence, the permeate flow rate during filtration may be similar to that found when permeating clean liquid. These conditions represent the
ideal operating regime for a mIcrofilter. However, critical flux is very sensitive to operating parameters and it is often found that a slight change in the conditions can lead to a considerable reduction in permeate rate. It is likely that critical flux is easier I-I
CHAPTER ONE. Introduction to maintain with a true surface filter because, in the absence of surface pore plugging, the membrane medium resistance remains constant because particles finer than the pore size will pass through the filter. Whereas, a conventional macroporous membrane will experience internal deposition of the finer particles giving internal pore plugging, leading to a greater liquid flow through the remaining pore channels, which will cause further fouling. It is possible, therefore, to argue that the optimum microfilter design is for a true surface filter with a pore geometry that does not easily surface plug. It is well known that when the pore and particle size distributions overlap the tendency for membrane fouling is greatest (Tarleton and Wakeman, 1993). Thus, a surface filter with circular pores filtering a suspension containing particles of similar diameter to the pore opemng size can easily plug, or block. Under these circumstances any advantage of filtering on a surface filter is lost. Suitable pore geometry to overcome this type of plugging is that of a slot and it is possible to commercially obtain filters with slot sizes slightly greater than the microfiltration range, i.e. 13 x 420 J.lID. Thus, a technique that can modify these filters to bring them into the micro filtration range will resllit in a slotted surface microfilter that may be less susceptible to pore blockage than a circular pore filter. The main project aim, reported by this thesis, was to investigate the filtration performance of surface filters and to compare the geometry of the pores of such filters. One advantage of a true surface filter is that the finer particles should not deposit within the membrane, unlike a conventional macroporous filter. So, it should be possible to select a surface filter to retain a species of specified particle size, e.g. bacteria, Cryptosporidium Parvum oocyts, precipitates etc., and to avoid the retention of finer material provided a secondary membrane, or cake, does not form. This investigation considers the formation of a secondary membrane on the surface filters with different pore geometry and its influence on the critical flux. Clearly, if It can be shown that the slotted geometry has advantages over a circular pore, with the size of particles used in this study, then there is an Incentive for further research and development to produce finer pore size slotted microfilters that are capable of removing the materials mentioned earlier.
1-2
CHAPTER ONE. Introduction The following chapters were written with the intention of providing the background and results to fulfilling the above main aims. During the research, investigation into different aspects designed to contribute towards this overall objective were required Primanly, there was a need to produce microfilters consistently within the desired pore size range and likewise to understand, and predict, the different types of critical flux behaviour. The following chapters have this aim and will be introduced here, very briefly, by way of a thesis overview. In CHAPTER TWO there is a review of literature on the subject of electroless plating.
This is the experimental technique that was used to reduce the slot width from 13 J.UIl into the microfiltration range of less than 10 J.UIl. Electroless nickel plating was chosen in preference to galvanic depOSition because of the requirement to close the slot width slightly rather than deposit a layer of nickel on the top surface of the filter. In galvanic nickel deposition the metal will deposit on the cathode. To make the metal deposit on the side of a membrane pore, rather than on the top surface of the filter, would reqUire an anode running through each pore. Clearly, this is not feasible. Conversely, deposition from electroless nickel solution will occur on all appropriately pre-treated surfaces. And there is the possibility furthermore that deliberately passing the fluid through the pores will encourage higher mass transfer rates at the pore surface, compared to the membrane surface, since the flow Will be greatest within the pores. CHAPTERS THREE and FOUR describe the experimental methodology and corresponding results for the development of high-quality surface filter production using the electroless nickel deposition method. A similar technique had been used before to successfully make circular pore microfilters (Srruth, 1998). In the earlier work, the membrane substrate was formed mto tubes before plating with the plating fluid pumped through the holes. In contrast, the membrane material handled during this project was predominantly slotted, with a much higher open area to filtration. The intended membrane geometry was flat discs for the filtration tests, rather than the tubes used preVIOusly. Such variations in the substrate material and process operation from previously successful plating conditions led to considerable problems in the adequate removal of the hydrogen gas evolved during the electroless depOSition process. The successful development of an innovative manufacturing process for
1-3
CHAPTER ONE. Introduction high-quality (uniform) process scale slotted surface filters represents a major achievement for this project. The vanous process challenges encountered during process development are reviewed in Chapters Three and Four. Some of the literature on microfiltration is presented in CHAPTER FIVE. This is a subject covered by a vast amount of publIshed literature over a period of thirty to forty years. The review presented here concentrates on a basic understanding of the most important mechanisms that may take place dunng microfiltration and how they might influence the occurrence of critical flux. Other literature reviews are referenced and the interested reader may wish to consult those in addition to reading Chapter Five. One important aspect of the work is that given the multitude of different models, operating conditions and results it is highly unlikely that there will ever be a single unifying model for the process of crossflow microfiltration under all conditions.
Many of the models reviewed in Chapter Five were tested against constant-rate filtration results presented in Chapter Eight. The experimental equipment and procedures for these investigations are reported in CHAPTER SEVEN. It became apparent that a simple mass transfer model could be applied to explain the results of the slotted microfilters, when there was significant retention of particles at the membrane surface. T1us mass transfer model is developed in CHAPTER SIX giving an insight in to how the operatmg conditions influence resulting permeate flux performance. The model does not appear to accurately descnbe neither the performance of the very large 13 llll1 slotted filter nor the commercial membranes with circular pore geometry. In the former case, the amount of latex challenge material retained by the membrane was limited and it is likely that a concentration gradient, needed for a diffusion-based model, was not realistically established. It is notable that the predicted fluxes were considerably less than the measured ones for critical flux operation. In the case of the circular pore membranes, the predicted fluxes were, m general, in excess of the measured ones. This suggests that the mass transfer model is not appropriate for tlus type of geometry and that an additional membrane resistance must have occurred; a resistance that is likely to be associated with the pore geometry because it affected only the circular pore membranes. This may be explained by surface plugging. Thus, on the basis of these results the slotted pore
1-4
CHAPTER ONE. Introduction geometry is shown to provide less of a foulmg environment than the circular one and further research into this type of filter is recommended. Practical applications of the novel slotted geometry membranes include filtration and the ability to classify particles into different Size ranges, or grades. This type of performance will only be possible so long as a deposit, or secondary membrane, does not form on the surface of the membrane filter. In CHAPTER EIGHT there is strong evidence to show that such a deposit rapidly forms on a circular pore membrane at even low feed concentration. Under identical operating conditions, no such deposit forms on the slotted filters. At some point, such as a higher feed concentration or permeate flux, it is possible to induce a deposit, which is likely to lunit the application of the slotted filters to classification duties. This point is equal to the cntlcal flux value for the filter. When operating below critical flux, the particle removal mechanisms from the surface of the filter are greater than the convective flow of particles towards the membrane surface. Hence, the filtration flux, or required pressure drop to effect a filtration, is stable. Classification, or fractionation, of particles is possible under these conditions. However, above the critical value deposition of particles on the membrane surface results and the abihty to classify particles is lost. Thus, for the purpose of predicting the ability of the slotted microfilter to classify particles, and for throughput calculations, the two most important pieces of information are (a) the filter grade efficiency curve (at sub-critical operation) and (b) the critical flux as a function of challenge concentration and shear rate. The mathematical model predicts the latter shear rate relation and the grade efficiency curves are reported in the Results Section of this thesis. This work has enabled the detailed design of slotted microfllters as sub-IO nucron classifiers for the challenge material investigated. The principal design data is presented in graphical form for grade effiCiency (Figure 8.23) and critical flux as a function of solid concentration (figure 8.40) and surface shear (Figure 8.38) in Chapter Eight. Finally, conclusions and recommendations from this work are summarised in CHAPTER NINE.
1-5
CHAPTER 2. Literature Review (Electroless Nickel Plating)
CHAPTER TWO LITERATURE REVIEW: ELECTROLESS NICKEL PLATING Electroless nickel plating (ENP) is an autocatalytic immersion process whereby a uniform coating of nickel and phosphorous alloy is deposited onto a catalytic surface. In contrast to nickel electroplating it does not use an external electric current to produce a deposit on the substrate material. Deposition occurs in an aqueous solution containing
primarily
metal
ions,
a
reducing
agent,
complexing
agent,
stabilisers/inhibitors and a catalyst (platmg part). Non-catalytic surfaces can also receive e1ectroless plating once they have been activated by deposition of a sUItable catalytic material. Chemical reactions on the surface of the part being plated cause deposition of the metal or alloy. Once a primary layer of metal has formed, that layer, as well as each subsequent layer, becomes the catalyst that allows the reaction to contmue. Electroless nickel plating produces very uniform, hard and lubricious coatings, and by varymg the percentage of phosphorus in the coating (up to 15 per cent), deposits can be produced to exhibit non-magnetic and highly corrosion resistant characteristics
or
hard
deposits with excellent
wear resistance
(Website
http://www.imagineenng-mc.comlsolutionslelectroless.html). As a platmg techmque, ' it is particularly suited to covering complex geometry's and irregular surfaces to which the solution has access. Introduced by Brenner and Riddell in 1946, electroless plating has since found widespread use as a functional coating of metals, alloys and plastics alike in many different industries.
2.1 APPLICATIONS OF ELECTROLESS NICKEL PLATING Electroless nickel plating is the most important catalytic plating process in use today. The principal reasons for its widespread commercial and industrial uses are to be found in the unique properties of the electroless nickel deposits. Excellent physical characteristics such as hardness, wear resistance, unifonnity of coating, corrosion resistance as well as the ability to plate non-conductors such as glass, ceramics, plastics and graphite make the electroless nickel coating highly desirable (Colaruotolo and Tramontana, 1990, p.207). To varying degrees, these properties are utilised by all segments of industry, either separately or in combination. Examples of electroless nickel plating can be found in many industries including aerospace, automotive, 2-1
CHAPTER 2. Literature Review (Electroless Nickel Plating) chemical, electronIcs and computer, materials handlmg, medical and pharmaceutical, military (armaments), mining, paper and texUles. Table 2.1 summarises some major uses for electroless nickel plating in selected industries. Comprehensive reviews of electroless nickel plating applications are presented by Colaruotolo and Tramontana (1990, pp.207-228), and Riedel (1991, pp.261-285) for general engineering and Duffek et al. (1990, pp.229-259) specific to the electronics industry. Table 2.1 Major uses of electroless nickel plating in selected industries (excluding the electronics industry), in Colaruotolo and Tramontana (1990, pp.207-227). Sector
Application
Basis metal
Phosphorous content
Aerospace
Landing gear components Compressor blades Fuel mJectors Shock absorbers Heat exchangers Pumps and Imnellers Tubes Ball valves Gears and clutches Jettmgpump heads
AI
M,H
Coating thickness (mm) 1-2
WR, bUildUp
Alloy steels Steel Steel Steel
M,H
I
CR,WR
M,H M,H H
1.0 0.4 3.0
CR,WR CR,LU CR,ER
Cast Iron/steel Steel Steel Steel
L,M,H
3.0
CR
H H M,H
24 1-3 1.0
CR, WR, U CR,WR WR, bUildUp
Steel
L,M,H
24
CR,ER
(%)
Automotive ChemIcal
Oil and Gas Matenal handlin" Mmmg
Property of interest
Phosphorous content (%): H = 9 to 12; M = 5 to 8; L = 1 to 2 CR = corrosion resistance; LU = lubricity; ER = erosion resistance; AR = abrasion resistance 2.2 ELECTROLESS NICKEL PLATING SOLUTION Electroless nickel plating baths generally consist of a source of nickel ions, a reducing agent, suitable complexing agents, and pH regulators, in addition to water. They may also contain amongst other components accelerators, stabilisers, buffers and wetting agents. The purpose (with examples) of the various bath components is shown in Table 2.2, on the following page. Nickel sulphate is generally the preferred source of nickel ions, although alternative nickel salts, such as nickel chloride and nickel acetate
2-2
CHAPTER 2. Literature Review (Electroless Nickel Plating) Table 2.2 Components of electroless nickel plating baths and their functions, in Riedel (1991, p.13). Comnonent Nickel ions
Hypophosphite ions Complexing agents (complexants)
Accelerators
StabIlisers
Buffers
pH regulators
Wetllng agents
Function Source of metal
Reducing agent
These form nickel complexes, prevent excess free Ni ion concentration so stabilising solution and preventing Ni phosphite precipitallon. Also act as I oH buffers Activate hypophosphite ions and accelerate deposition. Mode of action ooooses stabilisers and comolexants Prevent solution breakdown by shielding catalytically acllve nuclei
Examnle nickel chloride nickel sulphate nickel acetate sodIUm hypophosphite sodium borohydnde hydrazme monocarboxylic acids dicarboxylic acids hydrocarboxylic acids ammonia, alkanolarnines
anions of some monoand di- carboxylic acids, fluorides, borates lead, tin arsenic, molybdenum, cadmium or thallium ions, thiourea, etc. For longer term pH control sodIUm salt of certain complexants. Choice depends on pH range used For subsequent pH adjustment sulphuric and hydrochloric acids, soda, caustic soda, ammonia Increase wettability of surfaces to be ionic and non-ionic coated surfactants
are available for more specialised applications (Mallory, 1990a, p.3). The reducing agent is considered the most important plating bath component and is often used to classify the electroless plating process (Riedel, 1991, p.14). There are four different reducing
agents
commonly
used
in
electroless
nickel
plating
solullons.
Table 2.3, on the next page, summarises these reducing compounds with certain useful properties. They are all structurally similar WIth each compound containing two or more reactive hydrogen atoms. Of these four however sodium hypophosphite is by far the most widely used reducing agent, WIth over 90 per cent of autocatalytic deposition based on reduction by sodium hypophosplute (Denms and Such, 1993, p.311). It is one of the strongest reducing agents available followed closely by boronhydrogen compounds. Hydrazine is another powerful reducing agent and produces 2-3
CHAPTER 2. Literature Review (Electroless Nickel Plating) nickel that is over 99 per cent pure (Dennis and Such, 1993, p.3I1). Plating baths based on hydrazine however are not very stable and are rarely used in practice. Table 2.3 Reducing agents for electroless nickel plating, in Mallory (199030 p.5), and Riedel (1991, p.16). Compound
Formula
Mol. wt. Sodium hvoophosphite NaH2P02.H20 107 Sodium borohvdride NaBH! 38 Dimethvlarrune borane (CH3hNHBH3 59 Hvdrazme 32 H2N.NH2
pH Free Redox ranl!e electrons potential 4-6 2 -1.4 -1.2 12 -14 8 -1.2 6-10 6 4 -1.2 8 - 11
Mol. wt. refers to the molecular weight of the compound.
2.3 ELECTROLESS NICKEL PLATING (HYPOPHOSPIDTE BATH) 2.3.1 Acid Hypophosphite Bath Acidic hypophosphite baths offer a number of advantages in comparison with alkaline ammonia solutions, which were preferred in the early days of development. These include higher deposition rates, increased stability, greater simplicity of bath control and nickel-phosphorous depOSIts with Improved properties (Gawrilov, 1979, p.24; Riedel, 1991, p.20). The success of the acid hypophosphite bath has unsurprisingly led to development of over 100 commercial plating solutions for general and specialist applications. A summary of typical plating bath compositions and operating parameters is reported in Table 2.4. Table 2.4 Plating bath compositions used in Industry, in Riedel (1991, p.20). Concentration (g I~) Nickel as Ni nickel chloride nickel sulphate Sodium hypophosphite Temperature range (0C) IpH Deposition rate (J.U11 h· 1) Phosphorous content of deposit (wt %)
6.7 5.5 20 85-95 4.6-5.0 10-30 5-14
2-4
CHAPTER 2. Literature Review (Electroless Nickel Plating) 2.3.1.1 Deposition Mechanisms
Nickel deposition by hypophosplute can be represented by the following equations (Mallory, 1990a, p.6):
,,' ) H 2 PO-3 + H 2
(2.2)
Overall equation:
AIl the reactions take place on the active surfaces of the catalyst with external energy provided by higher solution temperatures, i.e. 60 - 95°C. Nickel and Phosphorous are reduced simultaneously and the coatings produced are metastable solutions of phosphorous in nickel. The formation of W ions results in the baths becoming more acidic over time (Gorbunova, 1963, p.l). The continual reduction of mckel and hypophosphite requires fresh additions of nickel sulphate and sodIUm hypophosphite. Electroless nickel plating is always accompanied by the evolution of hydrogen gas. The deposition mechanism is incompletely represented by the stoichiometric reactions of Equations 2.1 to 2.3. Crucially there is no account for the phosphorous component of the alloy coating. Although the precise nature of the partial reactions is even today not fully understood, four principal reaction mechanisms have been proposed to explain electroless nickel depositIOn (Mallory, 1990a, p.6). These reaction schemes are 'atomic hydrogen', 'hydride transfer', 'electrochemical' and the 'co-ordination of hydroxyl ions'. Mallory (1990a, pp.7-11) presents a comprehensive review of each
mechanism. Of particular mention is the inclusion of a secondary reaction of hypophosphite to elemental phosphorous.
2.3.1.2 Classification of the Metal Substrate
The surface to be coated must be capable of initiating the heterogeneous mechanisms outlined previously. Electroless nickel plating will only occur on specific surfaces 2-5
CHAPTER 2. Literature Review (Electroless Nickel Plating) without catalytic treatment. The plating matenals (substrates) can be classified according to their catalytic activity as follows (Riedel, 1991, p.34):
Class 1. Intrinsically catalytically active materials These are metals that are capable of sustaining the electroless deposition process unaided. Nickel with its ability to initiate and sustain catalytic deposition, is a Class 1 material.
Class 2. Extrinsically catalysed materials The surfaces of these materials are not themselves catalytically active. It is necessary to deposit an intrinsically active metal (Class 1) on the surface to provide catalytic activity.
The intrinsically catalytic metals (Class 1) for hypophosphite-based baths are the dehydrogenation catalyst metals of Group VID of the periodic table including nickel, cobalt (in alkali only), palladium, and rhodium. The extrinsically catalysed materials (Class 2) represent metals, which are more electropositive than nickel, such as iron and aluminium, beryllium and titanium. These will initially displace nickel from a solution of its ions (Mallory, 1990a, pp 6-7). If the substrate metal is more electronegative than nickel it can be made catalytic by adopting one of the surface treatments: electrolytic deposition of a thin nickel deposit on the substrate surface; or fonning a galvanic cell in the plating solution between the metal substrate and a more electropositive metal (Mallory, 1990a, p.7). After appropriate catalytic activation, non-metallic materials can equally well be coated with eIectroless nickel, preferably in neutral or alkaline hypophosphite solutions at temperatures of 70°C (Riedel, 1991, p.36). Zinc, cadmium, tlD, lead and sulphur are all considered as 'blocking elements' , or poisons, towards the hypophosphite bath (Riedel, 1991, p.35). 2.3.2 Alkaline Hypophosphite Bath The industrial uses of alkaline hypophosphite solutions are restricted to a few applications where plating at low temperatures or low-phosphorous deposits is a
2-6
CHAPTER 2. Literature Review (Electroless Nickel Plating) requirement. Deposition rates tend to be somewhat slower in alkaline solutIOns compared to acid baths and the depOSIts formed can be more porous and less corrosion resistant, though often brighter (Riedel, 1991, p.24). Alkaline plating baths can operate at temperatures between 30 and 90°C and are frequently used below 40 °C for applying thin initial conductive coatmgs onto etched and activated plastics, and deposition onto printed circuit boards (Dennis and Such, 1993, p.316). At temperatures higher than 80°C, which are necessary for commercially acceptable deposition rates, ammonia is lost rapidly from hot alkaline solutions, causing unpleasant fumes and resulting in the bath becoming unbalanced.
2.4 PRE-TREATMENT METHODS FOR METAL SUBSTRATES 2.4.1 Surface Impurities and Defects The pre-treatment of surfaces in preparation of electroless nickel deposition is extremely important because the presence of dirt, dust, oils or other contaminants on the surface will prevent satisfactory adhesion from bemg achieved. All materials to be plated are considered as being contaminated with oil, grease, dust and impurities. Typical impurities found on metal surfaces include carbonised hydrocarbons, oils, lubricants such as soaps, oxides, passive layers, foreign metals, and other adsorbed species (Riedel, 1991, p.185). A successful pre-treatment removes all contaminants leaving a clean and nominally OXIde-free surface. Poor surface preparatIOn conversely can cause a lack of adhesion, porous deposits, surface roughness and coatings and/or dark deposits (Aleksmas, 1990, p 103). 2.4.2 Surface Cleaning The selection of an appropriate pre-treatlnent procedure is based on the nature of the contarnmants and type of plating matenal. Foreign contarrunants and oxide layers can \ be removed usmg commercIal alkaline cleaners and acid pickling solutions respectively. Good rinsing is also Important to eliminate carryover of solution ions and pOSSIble poisoning of the plating bath components. It has been suggested that initiation of the deposition reaction and coatmg adhesion can be improved if the material enters the plating solution with an alkaline rather than acid film (PMD (UK) Ltd., 1993).
2-7
CHAPTER 2. Literature Review (Electroless Nickel Plating) 2.4.3 Surface Activation Materials that are catalytically inactive require activation steps in order to initiate the nickel deposition process. Pre-treatment schedules involving surface activation are widely reported in literature (Gawrilov, 1979, pp.l11-132; Hajdu, 1990, pp.193-206; Riedel, 1991, pp.l84-203). One such example is the activation of stainless steels (18/8 chromium-nickel steel, 12% chromium steel, etc.) using an electrolytic 'Woods Strike' in a chloride nickel-hydrochloric acid bath. During a Woods Strike, a thin layer of active nickel is deposited on the surface to overcome passivity and provide a catalytIc surface in readiness for electroless nickel plating. Operating conditions specific to the Woods Strike are reported in Table 2.5. Table 2.5 Description of Woods Strike surface activation method, in Metal Finishing Guide Book and Directory (1994).
Solution concentration: Nickel chloride Hydrochloric acid, (32%) Current denSIty: anodic cathodic
240 g rl 320 ml rl 1 A dm·2 (0.5 to 1 minute)
2 A dm·2 (2 to 6 minutes)
Riedel describes a typical pre-treatment sequence for stainless steels, inclusive of surface activation, as follows (Riedel, 1991, p.188):
1. Degrease using organic solvent; 2. Alkaline clean (60 - 80°C) for 5 to 10 minutes; 3. Water rinse (room temperature); 4. Cathodic electrocleaning (room temperature) for 2 to 3 minutes; current reversal; 5. Water nnse (room temperature); 6. Activation with 'Woods' nickel strike (room temperature) for 2 to 6 mInutes; 7. Water nnse; and 8. Electroless nickel plating.
2-8
CHAPTER 2. Literature Review (Electroless Nickel Plating) Non-metals such as alununa, graphite, plastics, and silicon can also be coated after appropriate pre-treatment.
2.5 FACTORS AFFECTING DEPOSITION RATES AND ALLOY COMPOSITION The kinetics of the chemical reactions in electroless nickel deposition govern not only the rate of metal build-up on the surface but also the composition of the alloy formed. Most of the individual operating parameters and solution components have been studied and reported (Riedel, 1991, p.38). An expression for the complex dependence of deposition rate on the various parameters involved in electroless plating can be written as (Riedel, 1991, p.38):
Where T is the depositlon temperature, pH is the solution pH, and CoRed
CN;'"
CRe
and
represent the concentrations of nickel, reducing agent and the spent reducing
agent (orthophosphite) respectively. The term O/V describes the tank loadmg (area to volume ratio). The type and concentration of the complexing agent, accelerator and ' stabiliser are identifiable by respective terms K, B, and S. Factors nI' n 2 etc., refer to additional factors such as agitation and the extent of bath contamination. Of all these parameters, it is in general the temperature and pH value as well as the type and concentration of the stabiliser, which most affects the deposition rate (Riedel, 1991, p.38). 2.5.1 Plating Solution Temperature The temperature of the plating bath is the most important parameter to affect the deposition rate (Gawrilov, 1979, pp.27-28). The rate of deposition increases with increasing temperature as illustrated in Figure 2.1, on the following page. There is very little plating below 60°C, but beyond this the plating rate increases exponentially with rismg temperature (Mallory, 1990b, p.72). The higher solutlon temperature leads 2-9
CHAPTER 2. Literature Review (Electroless Nickel Plating)
25~------------------------------------~
-
, .c
20
E
:::1.
ai
1ii ...
15
c 0
:;:::
·iii
10
0
Q. Q)
C
5
O+-----,,----~----~------r_----._----~
40
50
60
70
80
90
100
Plating solution temperature, QC. Figure 2.1 Dependence of deposition rates (J.lII1 hol) on temperature (QC), Fields et al. (1984) cited in Riedel (1991, p.39). Plating bath composition at pH 5: nickel chloride (30 g rl), sodium hypophosphite (10 g rl), sodIUm hydroxyacetate (10 g rl).
to a higher rate of diffusion to the surface. Practically useful deposition rates of 8 J.lII1 h· 1 and greater are achieved above 80°C at this particular pH. In Figure 2.2, on page 2-11, the relative deposition of nickel IS considered for the temperature range 90 to 100 QC. The experimental conditions are the same as Figure 2.1. At these elevated temperatures a rise in temperature of 10 °c produces a twofold increase in deposition rate. While the higher deposition rates make this regIOn attractive, there is a possible danger of bath instabilIty at these temperatures. Prolonged operation of the plating bath above 90°C increases the risk of solution plate-out or even solution decomposition (Mallory, 1990b, p.72). With this in mind acid hypophosphite plating solutions are usually operated between 80 and 90°C (PMD (UK) Ltd., 1993). In addition to the deposItion rate, the operating temperature can also affect the phosphorous content of the deposit, and hence ItS properties. Baldwin and Such (1963) cited in Riedel (1991) reported that the phosphorous content of deposits
2-10
CHAPTER 2. Literature Review (Electroless Nickel Plating) decreases when the temperature of the plating bath is increased beyond its normal operating range. Accurate temperature control of plating baths is therefore essential if operation within the optimum temperature range is to be maintained. As discussed in a later Section 2.5.5, a uniform bath temperature is difficult to achieve without extremely vigorous agitation. Poor circulation of solution adjacent to the bath heater can lead to localised overheating which can cause plate-out, roughness or even bath decomposition (Aleksinas, 1990, p.104; Riedel, 1991, p.39).
100~------------------------------------~
rfl. ~
90
~
.!:! c
..
80
"tJ
G)
·in 0
70
Q. G)
Q
60
50~------r-----~------~------~----~
90
92
94
96
98
100
Plating solution temperature, °c. Figure 2.2 Relative deposItion rate (Ilm hol) as a function of bath temperature eC), Gutzeit (1959) (1960) cited in Riedel (1991, pAO). Plating bath composition at pH 5: nickel chloride (30 g r1), sodium hypophosphite (10 g r1), sodium hydroxyacetate (10 g r1).
Electroless nickel plating baths can be heated internally or externally using process heating equipment such as PTFE, stainless steel immersion heaters and PTFE coated steam coils or heating jackets for large volumes (PMD (UK) Ltd., 1993).
2-11
CHAPTER 2. Literature Review (Electroless Nickel Plating) 2.5.2 Plating Solution pH The pH of the plating solution will reduce over time due to accumulation of hydrogen ions. Laboratory tests have shown that three moles of W are generated for every mole of Ni2+ deposited, which is in agreement with Equation 2.1 (Mallory, 1990b, p.58). Figure 2.3 shows the dependence of electroless nickel deposition rate (J.1m h- I) as a function of pH. It is seen that the deposition rate is sensitive to solution pH. As the acidity of the plating bath increases, deposItion rate decreases. At pH 5 the deposition rate is 20 J.1m h- I, falling to 16 J.1m h- I at pH 4. A very low plating rate is observed if the pH is allowed to drop below pH 4.0.
25~----------------------------------~
:c
~
20
E
::1.
of
e
15
:;:::
10
C 0
'iij 0
Q,
CD
C
5 o+---~--~----~--~--~----~--~--~
7
6
5
4
3
pH of plating solution Figure 2.3 Dependence of electroless nickel deposition rate (J.1m h- I) as a function of pH and solution temperature eC), Gawnlov (1979, p.28). The bath contained NiCh-6H20 (30 g rl), NaH2P02 H20 (l0 g rl) and sodium glycoIIate (10 g rl).
Electroless plating baths incorporate a buffer to maintain pH. The concentration of hydroxide ions can also be increased during process operation by continuous dosing with dilute alkali
(N~OH
or NaOH), (Riedel, 1991, p.4I). The most dramatic effect
of lowering the pH is on the composition, and the properties of the nickel-
2-12
CHAPTER 2. LIterature Review (Electroless Nickel Platmg) phosphorous deposit. Figure 2.4, below, IS a plot of variation in phosphorous content against bath pH. A decrease in the phosphorous content of the coating is observed with increasing pH. Mallory (1990b, p.60) suggests an operating pH range of 5.0 -7.0 for an acid nickel plating bath. The deposit quality and solution stability will be affected if the solution is operated outside this range.
12
ffl i
10
cQ) c
8
.:
0
()
1/1 ::J
...00
~
Cl. 1/1 0
~
c..
6 4 2 0 3.5
4.0
4.5
5.0
55
6.0
6.5
pH of plating solution Figure 2.4 VarIation in phosphorous content of deposit (%) versus bath pH, Grunwald (1983) cited in Riedel (1991, p.42).
Table 2.6 Effect of pH change on electroless nickel process, in Mallory (1990b, p.62). Chanl!e Effect on solution Effect on deposit Raise Increase deposition rate; lower Decreased phosphorous content; shift in pH phosphite solubility stress to tensile direction.
Lower pH
Decreased stability with Poorer adhesion on steel resultant olate-out Decreased deposition rate; Increased P content; sluft compressive direction. Imoroved ohosohite solubilitv
In
stress to
Imoroved adhesion on steel
2-13
CHAPTER 2. Literature Review (Electroless Nickel Plating) The influence of pH on process parameters, originally described by Gawrilov (1979, p.30), is summarised in Table 2.6.
2.5.3 Contaminants All electroless nickel processes are sensitive to contamination from metals, sulphur compounds and particulate matter such as dust (PMD (UK) Ltd., 1993). It is important therefore that impurities do not enter the plating bath with the substrate after pre-treatment or dunng replenishment of components where they can severely inhibit nickel deposition. The water supply to the plating bath is another possible source of contamination if It is not managed properly. Tap water quality is adequate provIding it is soft and has low metals content (PMD (UK) Ltd., 1993). The levels of tolerable impurities are summarised in Table 2.7 (Riedel, 1991, p.51). Also mentioned are the effects of contaminants on deposit quality and methods of treating contaminated baths.
Table 2.7 ContaminatIOn problems in electroless nickel deposition, Innes and Kunces
(1979), cited in Riedel (1991, p.52). Acid hypophosphite bath with 6 - 7 g rl nickel, 16 - 37 g rl sodium hypophosphite at pH 4.2 to 5.0 and temperature 85 DC to 91 DC. Contaminant
Limiting cone. (DDm. m£lrI)
Pb,Cd
~5
Zn Cu
>300 ~ 15
Fe AI
> 150
Pb Cr(llI)
>3 ~ 15
Cr(VI) S2.
>3
~3oo
~1O
N03'
~50
H3P03
80-150 gl~
Symptom
no depositIOn or stepwise deposition low deposItion rate dark deposit low deposItion rate dark deposit, low deposition rate decomposition stepwise deposition, low deposition rates stepwise depositIon no depositIOn or dark deposits no deposition low deposition rate
Action
dummy deposItion to deplete bath discard and replace bath dummy deposItion to deplete bath discard and replace bath discard and replace bath discard and replace bath discard and replace bath dIscard and replace bath dummy deposItion hold at pH 4 and 85 DC for 2 hours discard and replace bath (raise pH and temperature)
2-14
CHAPTER 2. Literature Review (Electroless Nickel Plating) In many cases the only option is to discard the entire bath and start again with fresh
solution. Lead, cadmIUm, chromium (VI) and sulphur can affect the deposlllon process at very low concentrations. Lead and cadmium for example are deposited on the substrate disproportional to their concentration in solution. Metals such as zinc, iron, and aluminium are co-deposited in the pH range 4.6 - 4.8, albeit at higher solution concentrations (Riedel, 1991, p.52). Generally speaking small quantities of impurities in the solution will plate out and have lIttle effect on the process. 2.5.4 Loading of the Plating Bath The bath loading describes the ratio of the area of substrate immersed to the volume of solution in the tank. Gninwald (1983), cited in Riedel (1991, p.53) found a very strong dependence of deposition rate on bath loading for a particular plating solution. A nickel coating of 15 I1m thickness was measured after immersion in the plating bath for two hours at 4 dm2 r1. Reducing the plating loading to 0.5 dm2 rl resulted in a thicker deposit of 40 I1m (Riedel, 1991, pp.53-54). Bath loading can also affect the phosphorous content of the deposit. Experiments by Wiegand et al. (1968), cited in Riede1 (1991, p.54) found that the lower the bath loading, the lower the phosphorous content in the deposit. A bath loading of between 0.5 and 1.5 dm2 r 1 is recommended by PMD (UK) Ltd., (1993). Operating at too high a ratio of sample area to solution volume can cause a drop in plating rate of 15 - 20 per cent. 2.5.5 Solution Agitation The selection and operation of an efficient agItation system represents an important consideration for electroless plating. Agitation of the plating solution can help maintain constant temperature conditions and avoid temperature layering within the bath. By establishing a flow pattern in the bath fresh solution is continuously delIvered to the reaction surface and depleted solution quickly removed. In addition hydrogen bubble retention on the surface, which can inhibit plating reaction sites, is mmnTIlsed. Effecllve circulation throughout the tank is difficult to achieve however and requires a good understanding of process conditions. Energetic flow movement at the tank walls and past immersion heaters can prevent localised hot spots and solution plate-out. Gawrilov (1979, p.31) suggests that solution movement can raise the rate of nickel deposition, possibly by raising the pH within the diffusion layer. Certainly no 2-15
CHAPTER 2. Literature Review (Electroless Nickel Plating) agitation or poorly implemented agItation can lead to solution stratification, resulting in gas pitting, patterns, and/or streaking of the deposit. Solution agItation can be achieved in one of three ways or by a mixture: (a) Air injection (b) Pumped circulation
(c) Mechanical stimng
Although the use of air agitation can affect some plating baths, it is the easiest and most widely used method. Low-pressure oil-free positive displacement blowers wIth filter units are typically used. The filtered air IS directed downwards to generate a rolling motion of the solution. According to technical information air agitation can reduce overall chemical efficiency by approximately 5 per cent (PMD (UK) Ltd., 1993). SolutIOn agitation using high flow pumps is considered to be the best method for setting up optimum flow patterns in the plating bath and there is no air source to consider (Kuczma Jr., 1990, pp.159-160). Intensive agitation close to the substrate surface has been found to be more effective than high flow rates. Stallman (1986), cited In Riedel (1991, p.55) showed that increasIng the flow rate led to a decrease In
depositIOn rate, although a smoother deposit resulted (acid hypophosphite bath). The outflow from the pump, following a filter stage, is directed through positional sparge tubes rather than a single orifice. Mechanical stirrer mixing as a source of agitation is not a first choice method. The flow patterns derived from mixers are relatively weak and easily broken by the immersed substrate. Multiple lower-powered stirrers produce better results than a single high-powered unit. The use of ultrasonics can provide a special type of enhanced solution agitation. Ultrasonic radiation of frequency from a few kHz to around 2 MHz is known to accelerate both electrodeposition and electroless processes (Kuzub and Mukhlya, 1963; Rich, 1955; Shenoi, B & M, 1970, cited in Riedel, 1991). In the case of alkaline 2-16
CHAPTER 2. Literature Review (Electroless NIckel Plating) hypophosphite baths, up to fifteen-fold accelerations have been reported. For acid hypophosphite baths the effect is less marked, somewhere between two- and four-fold (Riedel, 1991, p.56). Gawrilov (1979, p.150) reported that ultrasonics can reduce bath stability and their action tends to produce deposits with higher porosity. The relatively large capital cost of ultrasonic equipment for an already well-developed process coupled with detrimental plating effects however has restricted the widespread use of this technology. 2.5.6 Plating Bath Age Electroless nIckel baths have a finite operating life defined by the number of times the entire nickel ion content (g rl) is consumed and replenished. Each replenishment is known as a 'turnover'. The optimum life of the Ni-Shield M solution is eight complete replenishments of metal content (PMD (UK) Ltd., 1993). After this volume of replenishment the corrosion resistance and other important functional properties of
14
-.c
12
E
10
....I'llai
8
C 0
6
,
:::1.
...
~
'iij 0
Q. Q)
4
C
2 0 0
10
20
30
40
Nickel consumed and replaced, 9
50
r1.
Figure 2.5 Deposition rate as a function of bath age for a third-generation 'Novotect' bath operating at constant deposition conditions (6.7 g rl Ni, 35 g rl hypophosphite, temperature 88 ± 1°C; pH
=4.8 ± 0.1). Linka and Riedel (1986) cited in Riedel (1991, pp.56-57). 2-17
CHAPTER 2. Literature Review (Electroless Nickel Plating) the deposit deteriorate severely. Figure 2.5, which shows deposition rate as a function of bath age for an acid hypophosphite bath illustrates this on the preceding page. One bath turnover is approximately equivalent to 7 g
rl
nickel consumed and replaced
(Riedel, 1991, pp.57 -58). The rate of nickel deposition on the substrate is at its highest initially in fresh plating solution. A similar behaviour is observed for other bath types.
2.6 EFFLUENT TREATMENT AND DISPOSAL OF SPENT PLATING BATHS 2.6.1 Pre-treatment Chemicals Pre-treatment chemicals used for normal degreasing, cleaning and pickling are considered as simple acids and alkalis with no significant metal content, and as such can be neutralised. Mixing the various waste pre-treatment solutions reduces the volumes of fresh acid or alkalis required for the duty. Tills cocktail of waste products is neutralised to a point where the pH drops Within the range pH 7.0 - 8.5 (Riedel, 1991, p.225).
2.6.2 Plating Bath Components The electroless plating bath is discarded after a number of turnovers, usually between six and eight turnovers, when phospillte, sulphate, sodium and other harmful components have accumulated in the bath (PMD (UK) Ltd., 1993). ,Operating costs can be greatly reduced by optinusing the solution to extend the number of turnovers before disposal is required. Table 2.8, on the following page, shows typical concentrations for a spent electroless nickel hypophosphite bath. Electroless plating wastes are generally treated with either conventional techniques, such as chemical precipitation, and reduction, or more advanced recovery techniques such as electrolytic reclamation, ion exchange, reverse osmosis, or electrodialysis (Li et aI., 1999). Recovery technology separates harmful ions such as phosphite, sulphate, sodium etc., from the useful components of the solution such as nickel, hypophosphite Ions, and organic acids. Capaccio (1990, pp.526-527) has reviewed the regeneration of spent baths using ion exchange and electrolytIc recovery. Conventional chemical treatment of electroless plating baths involves neutralisation and subsequent separation of the precipitated metal hydroxides, and oxidatIOn of ,
2-18
-----------------
-
-
CHAPTER 2. Literature Review (Electroless Nickel Platmg) organic acids. According to Ying et al., cited in Riedel (1993, pp.227-228) nickel and phosphorous can be removed from spent baths m a three stage process. Initially slaked lime solution is added to precIpitate most of the nickel. After this potassium permanganate is used to oxidise hypophosphite and orthophosphlte to orthophosphate. At least some of the organIc acids are simultaneously OXIdised by this treatment and destroyed. The final stage involves a further addition of lime to precipitate the residual phosphorous and nickel. Separation of the nickel can be accomplished by further precipItation WIth sulphides or organo-sulphides. This will successfully reduce the concentration of dissolved nickel to less than 1 mg
rl
in readiness for a safe
disposal. Table 2.8 Typical concentrations of components in a spent
electroless nIckel plating bath, in Riedel (1991, p.227). Component Concentration nIckel 2 - 6 g r' hypophosphlte IO-30gr' orthophosphlte < 200 g r' organic aCIds 5 - 40 g r' sodIUm sulphate unknown ammonia unknown decompOSItion products unknown heavy metals < IOOppm
2.7 SUMMARY The successful applIcation of electroless nickel plating (ENP) technology enabled pore size reduction of nickel screen with circular (CPO) and slotted (SPO) pore openings. During thIS autocatalytic process a uniform coating of nickel and phosphorous alloy is deposited onto a catalytic surface without the need of an external power source, i.e. nickel electroplating. The plating of non-catalytic surfaces is also possIble once the surface has been activated by deposition of a suitable catalytic material. The uniform, hard and lubricious coatings formed by electroless nIckel plating have led to its widespread use in industries such as aerospace, chemical and pharmaceutical, electronics and computer, mining, paper and textiles. Commercially available plating baths consist of a source of nickel ions, a reducing agent, suitable
2-19
CHAPTER 2. Literature Review (Electroless Nickel PlatIng) complexing agents, and pH regulators, in addition to water. Accelerators, stabilisers, buffers and wettIng agents may also be present. Acid hypophosphite baths offer a number of advantages in companson with alkahne ammonia solutions including higher deposition rates, increased stability, greater simplicity of bath control and nickel-phosphorous deposits with improved properties. The pre-treatment of surfaces is extremely important because the presence of dirt, dust, oils and other contamInants on the surface will prevent satisfactory adhesion from being achieved. Deposition rate is affected by factors such as plating solution temperature, solution pH, component concentrations, contaminants, bath loadIng, solutIOn agitation, and plating bath age, although temperature of the plating bath is the most important parameter; deposition rate increases with increasing temperature.
2-20
CHAPTER 3. Experimental (Electroless Nickel Plating)
CHAPTER THREE EXPE~NTAL:ELECTROLESSNICKELPLATING
In this Chapter experimental details concerning the pre-treatment (PT) and electroless nickel plating (ENP) of metal screens for use as rnicrofilters, and methods of characterisation are presented. Compnsing of five sections, the Chapter opens by introducing the nickel and stainless steel screens whose pore openings of various geometries were reduced by nickel deposition. The electroless nickel plating process is outlined in the second section. Operational considerations associated primarily with solution agitation are reported. In the following two sections the optimisation of chernical pre-treatment and application of nickel electroplating to activate the substrate surface are discussed. Finally, methods appropriate for characterising pore size; open area and surface elemental composition are presented.
3.1 METAL MICROFILTERS Four different metal screen products sUItable for use as microfilters were selected against criteria such as pore size, pore geometry, ease of plating, open area, and availability. Table 3.1 highlights important characteristics of the nickel and stainless steel screens used throughout the electroless nickel plating work. Table 3.1 Physical properties of nickel and stainless steel screens. Reference
Material
Pore geometry
Average pore size (urn)
NiCPO-40 nickel circular 39.8 NiCPO-17 nickel circular 17.1 NISPO-13 nickel rectangular 13.4 x 402.8 SSPO-29 stainless steel square 29.2
Open area Thickness (%)
(urn)
3.5 0.7 5.7 31.0
29 53 283 41
Stork Veco B.V., located in the Netherlands, supphed 'Vecomc Plus' screens made from 100 per cent nickel. The screens are manufactured from a metal substrate with a specially textured surface using a patented electrofonning process. Once cleaned and degreased, the surface receives a photosensitive coating. The product image is then transferred onto the surface by UV exposure through a photomask and once
3-1
CHAPTER 3. Experimental (Electroless Nickel Plating) developed the solved photo-resist is rinsed away. Electrodepositlon technology is used to deposit nickel molecules to matrix foundation on areas, which are not masked with the photo-resist. Finally the desired electroformed product is separated from the metal substrate. The reader is directed to the Website (http://www.storkveco.com) for further information about the 'Veconic Plus' screens. The applications of Veconic metal screens are many and varied including filtration of chemicals, polymers, wastewater, coolant, paint, hydraulics and oil, as well as the separation of other particles. Perforations in the screens are available as either circular (CPO) or slotted (SPO) pore openings with a mmimum pore size of 10 J.U11 currently achievable. Nickel screens with circular pore geometry were supplied with nominal pore sizes of 39 J.U11 and 17 J.U11, as measured. According to the manufacturer's product information the slotted screen offers rectangular slots 10 J.U11 wide and 420 !lm long. Accurate pore size measurement using techniques described in Section 3.5.2 (on page 3-21) confirmed that the actual slot size (average) was closer to 13.4 x 403 J.U11. The nickel slotted material was around ten times thicker than the corresponding nickel screens with circular pores. Each screen was supplied as a 0.9 x 0.9 m flat sheet as standard. The nickel surface was ideal for receiving electroless plating because it can sustain deposition without prior activation. Pre-treatment of the metal surface however remained of great importance. A thin stainless steel mesh was supplied by United Wire of Edmburgh as a general purpose wire mesh. 3.1.1 Nickel 'Veconic Plus' Screen with Circular Pores (CPO) Figures 3.1 and 3.2 (on pages 3-3 and 3-4) are images of the nickel CPO screen With 39 J.U11 pores (optical microscope Image) and 17 !lm pores (Scanning Electron Microscope image) respectlvely. Both metal screens were untreated. Each circnlar pore formed part of a regular pattern. As seen with clarity in Figure 3.2, both metal screens had two distinctively different sides: one shiny and the other matt. Pore channels were conical, as illustrated in Figure 3.3 (on page 3-5), with each pore
3-2
CHAPTER 3. Experimental (Electroless Nickel Plating)
Figure 3.1 Optical microscope images of nickel CPO screen with 39
~m
pores,
viewed at 20x magnification; (a) shiny ' filtration ' side, and , (b) matt side. 3-3
CHAPTER 3. Experimental (Electroless Nickel Pl ating)
Figure 3.2 SEM micrographs of nickel CPO screen with 17 Ilm pores;
Ca) shiny 'filtrati on' side, and, (b) matt side rotated by 45 °.
3-4
CHAPTER 3. Experimental (Electroless Nickel Plating) tapering from the base of the cone and larger pore opening (matt side) to the top of the cone and smaller pore opening (shiny side). The cone base is visible as a raised circle surrounding each pore on the shiny side. Direct passage through the screen by specific conical channels is in contrast to the tortuous path offered by many commercially available microfilters. It was desirable to use the shiny side as the filtration surface, utilising slightly smaller pores whilst avoiding the wider 'funnel' entrance on the reverse side. All pore size measurements refer to pore openings on the shiny side.
Figure 3.3 SEM micrograph of nickel
epo screen with
17 )lm pores. Single
matt-side pore viewed at high magnification to show conical pore entrance.
A disadvantage of usi ng either screen as a microfilter, was the low percentage open area available to flow (porosity). The measured open area for the
epo
screen with
39 )lm pores was equivalent to 3.5 per cent of the total surface. This compares to only 0.7 per cent open area for the smal ler pore size. Such a relati onship between pore size and open area is to be expected (See Section 3.5.3 on page 3-30). Both
epo screens
were essentially identical; the on ly difference being pore size. 3-5
CHAPTER 3. Experimental (Electroless Nickel Plating)
Figure 3.4 SEM micrographs of nickel SPO screen with 13 x 403 Ilm slots; (a) many slots (shiny side), and , (b) single slot (matt side).
3-6
CHAPTER 3. Experi mental (Electroless Nickel Pl ating)
Figure 3.4 SEM micrographs of nickel SPO screen with 13 x 403 !lm slots; (c) multiple slots (matt side), and, (d) sin gle slot (matt side).
3- 7
CHAPTER 3. Experimental (Electroless Nickel Plating) 3.1.2 Nickel 'Veconic Plus' Screen with Slotted Pores (SPO) Scanning Electron Microscope (SEM) micrographs of the nickel SPO screen , Figure 3.4 (a) to (d), on pages 3-6 and 3-7, show the consistent dimensions of the slot and regular placement across the surface. The fl ow channel of rectangul ar cross-section narrowed slightly from the slotted opening on the matt side through to the shiny side. The manufacturer's stated slot dimensions of 10
x 420 Jlm did not agree with the
measured slot size of 13.4 x 403 Jlm . The slot width was not entirely uniform with a minimum value at each end where the slot tapers slightl y. The added thickness of the SPO screen contributed to a rigid microfilter with deep flow channels. 3.1.3 Stainless Steel Wire Mesh The stainless steel mesh pictured
In
Figure 3.5 consisted of a single layer of
interwoven steel fibres providing pores of 29.2 Jlm square geometry. Whil st the measured open area was greater than that of other screens, the steel mesh required acti vation in advance of electro1ess nickel plating. The fabric-like nature meant that the thin steel mesh was less robust as a microfilter. Having identical sides, the filtration surface debate is irrelevant for the metal wire mesh.
3-8
CHAPTER 3. Experimental (Electroless Nicke l Plating)
Figure 3.5 Optical microscope images of stainless steel mesh exhib iting 29 ~m square pore openings; (a) 10x magnification, and, (b) 20x magnification.
3.2 ELECTROLESS NICKEL PLATING (ENP) 3.2.1 Ni-Shield M Plating Solution Ni -Shi eld M electroless nickel solution was supplied by PMD (UK) Ltd, Coventry. The Ni-Shield M has been speciall y developed to produce hard, corrosion and wear resistant deposits of nickel phosphorous alloy (6 - 8 % phosphorous). Table 3.2 on the next page lists the recommended operating conditions for Ni-Shield M solution (PMD (UK) Ltd. , 1993). Each fresh electroless nickel plating solution was prepared using a 3:7 ratio of concentrated Ni -Shield MMU to deioni sed water. Ni-Shield MRA (nickel) and Ni-Shield MRB (hypophosph ite) replenishing solutions were available to maintain
opt imum
nickel
and
hypophosphite
concentrations.
Six
comp lete
replenishments of the Ni-Shield solution were possible before the rate of deposition progressively decreased (PMD (UK) Ltd. , 1993). The solution was agitated for 10 minutes and fi ltered using Whatman 0.8
~m
cellul ose . nitrate filters before use. 3- 9
CHAPTER 3. Experimental (Electroless Nickel Plating) Although electroless plating bath volumes varied with each experiment, they were typically in the range 400 ml - 1900 ml according to the loading requirements. The electroless
plating solution
was
replaced
regularly
to prevent component
concentrations falling below optimum levels. Table 3.2 Optimum electroless nickel plating bath conditions, in (PMD (UK) Ltd., 1993). Concentration (g r') Nickel as Ni Sodium hypophosphite Temperature range (OC) pH Agitation Filtration (!lm) Loading (dm" r') Plating rate at 85 °C (Ilm h'l) Plating bath life
5.5 - 6.5 28.0- 32.0 80 - 90 (optimum 85 ± 1 0c) 4.5 -4.9 mechanical or air movement 5 -10 0.5 - 1.5 15 -20 6 - 8 turnovers
3.2.2 Electroless Nickel Plating Rig The main features of the electroless plating rig were a medium-sized water bath, two I kW heaters , electroless Ni plating bath, an appropriate method of agitation and
temperature measurement. Figure 3.6 on the following page, is a schematic representation of the electroless plating rig with a pumped circulation system. Prefiltered Ni plating solution was transferred to a 2000 ml glass Pyrex beaker immersed in the water bath. The rate of evaporation from the plating bath was high because of the operating temperature. A barrier layer of plastic anti-evaporation balls was used in respective Ni plating and water baths to minimise fluid losses (PMD (UK) Ltd., 1993). Separate sets of anti-evaporation balls associated with each bath prevented contamination of the plating solution. The water bath was heated to the desired setpoint temperature of 85°C by two I kW heaters before introduction of the plating solution bath. Operating temperature of the plating bath was monitored closely using a glass mercury thermometer. Temperature control, although basic, was suitably effective. The bath temperature was lowered below 60 °C between plating experiments to prolong the solution life. Fume extraction was used to remove steam and fine droplets of solution. Safety goggles, gloves and protective clothing were worn at all times. The Ni-Shield M solution is acidic and classified as a skin irritant. 3-10
CHAPTER 3. Experimental (Electroless Nickel Plating) Inhalation of the hot vapour was avoided; ammonia can cause asthmatic reactions and nickel is a skin and mucous membrane sensitiser (PMD (UK) Ltd. Hazard Data Sheet,
1998).
Ni plating solution return
heater
I I
~
Ni plating solution
re
peristaltic pump
Membra e holder
Stirrer bar Water bath
Personal computer
Figure 3.6 Electroless nickel plating rig employing pumped circulation.
3.2.2.1 Pumped System In the early experiments hot plating solution was continuously drawn through the
plating filter sample using a peristaitic pump and recycled back to the plating bath. This pumped operation provided a controlled flow of solution through each pore and, according to on-line pressure measurement, gave some indication of the rate of nickel deposition. The plating sample was securely located plating side up in a modified 47 mm diameter transparent Sartorious membrane holder; the top nozzle and twothirds of the top flat surface had been removed leaving an 8 mm rim around the holder top. Holes drilled in the filter holder side promoted gas removal from the system. The Sartorious membrane holder with its thick rubber o-ring supplied a particularly good
3-11
CHAPTER 3. Experimental (Electroless Nickel Plating) seal. A 6 mm bore silicon rubber recirculation line was attached to the bottom nozzle of the membrane holder. The return line from the peristaltic pump discharged at the bottom of the plating solution. Normal operation of the peristaltic pump was in the direction shown in Figure 3.6 on the previous page, i.e. drawing liquid though the membrane holder. This method of pump operation was preferable to pushing liquid up through the filter where overpressurisation could lead to a sudden release of hot plating solution. Two peristaltic pumps of different capacity fulfilled this recirculation duty. A radio spares pressure transducer (l bar g, 0 - 100 m V) connected to the recirculation line (before the peristaltic pump) provided real-time pressure data useful for monitoring and ultimately controlling the plating process. An increase in the pressure difference across the filter indicated a reduction in the size of pores caused by plating. Pressure readings were displayed locally. The pressure transducer was also connected to a Pi co Technology 16 bit analogue-digital converter and Pc. Using PicoLog for Windows software (release 5.02.5) pressure measurements were logged at 5-second intervals during a plating experiment. A Druck DPI 603 portable pressure calibrator was used to calibrate the pressure transducer.
3.2.2.2 Mechanical and Air Sparge Agitation Alternative methods of agitation by mechanical and air movement applications were investigated following a catalogue of problems with the pumped system. A critical problem caused by hydrogen gas evolved from the plating reaction, on the substrate surface, gravely impeded the pumping operation. The pumped system including membrane holder was removed from service. The reader is referred to Chapter Four (Electroless Nickel Plating Results and Discussion) for further discussion. Mechanical agitation of the plating solution bath by variable speed overhead stirrer incorporating a plastic stirrer shaft and paddle blades was considered . Figure 3.7 (a) on the next page is a schematic representation of the overhead stirred system, which employed a plastic stirrer shaft with paddle blades. The pre-treated filter was introduced to the Ni plating bath and agitated at a low speed sufficient to keep the filter mobile in the bulk solution. Experimental trials with overhead agitation proved unsuccessful. Limited
3-12
CHAPTER 3. Experimental (Electroless Nickel Plating) control of process parameters restricted the plating output to low quality, inconsistent deposits.
air compressor
(b)
(a)
Figure 3.7 Plating bath agitation by (a) overhead stirrer,
and, (b) air injection (air sparge).
An experimental plating rig, which supported agitation by air movement, is illustrated
in Figure 3.7 Cb). A Pyrex sintered glass fimnel connected to a small air compressor generated the air sparge. Tubing from the compressor entered the glass fimnel through a rubber bung. Tests performed in hot water prior to use in the plating bath confIrmed the high temperature stability of the components. The air sparge apparatus rested on the bottom of the plating bath supported in the upright position. Filter samples of 47 mm diameter were cut from the virgin metal sheet, each one with a rectangular 60 x 15 mm tab. The pre-treated filter was secured firmly in a clamp, positioned just above the sintered glass fimnel in the plating bath. A delay of 60 seconds was introduced from the time of immersion in the plating bath to activation of the airflow. The column of bubbles rising continuously up the sides of the metal filter could clear gaseous plating products from the surface. 3-13
CHAPTER 3. Experimental (Electroless Nickel Plating)
3.2.2.3 Disposal of Waste Plating Solution Strict waste disposal regulations dictate the safe disposal of spent nickel plating solution, which contains amongst other components residual nickel. The disposal procedure was followed as outlined in the PMD process data sheet (PMD (UK) Ltd., 1993). Addition of sodium hydroxide to the spent solution (under agitation) raised the solution pH, to pH 12 and initiated a change in solution colour from dark green to dark blue. When heated to 95°C the remaining nickel precipitated out of solution. The resultant clear solution was filtered off using Whatman grade filter paper, cooled, and treated with dilute sulphuric acid to realise neutral pH. A specialist contractor was employed to handle the dark brown nickel residue according to the waste disposal procedure of the Chemistry Department, Loughborough University.
3.3 NICKEL ELECTROPLATING (EP) 3.3.1 Nickel Electroplating Solution Nickel slotted media and stainless steel mesh required nickel electroplating as part of the pre-treatment schedule. Although nickel is autocatalytic when immersed in electroless plating solution, a surface finish on the nickel slotted material proved resistant to standard pre-treatment methods. Deposition of nickel onto the surface assisted the electroless nickel reaction by providing sites for plating. Electroplating of stainless steel mesh was a necessity. Nickel electroplating solution was prepared according to the simultaneous activation/nickel strike procedure cited in literature for stainless steel (Metal Finishing Guidebook and Directory, 1994). This recommended a I: I ratio of nickel chloride and hydrochloric acid (Table 3.3) Table 3.3 Recommended electroplating solution composition. Nickel chloride (g r ) Hydrochloric acid (ml r l)
100 -400 50-200
600 ml nickel electroplating solution was prepared in a 1000 ml glass Pyrex beaker. 30 ml hydrochloric acid of 10.2 molarity (0.3 moles) was added to the beaker and diluted up to the 600 ml mark with deionised water. 60 g nickel chloride (NiCI2·6 H20; 0.3 moles) was added to the acid solution under agitation. The electroplating 3- 14
CHAPTER 3. Experimental (Electroless Nickel Plating) solution was strongly acidic having a pH of 0.9. The bright green nickel chloride powder was stored in an airtight container to prevent ingress of moisture. 3.3.2 Electroplating Experimental Rig The electroplating rig comprised of a 1000 ml glass Pyrex beaker, one or two nickel electrodes, Thurlby model PL320 QMD (30V, 2A) power supply, colour coded wires with crocodile clip attachments, and clamp stands to support each electrode. Pertinent features of the electroplating rig are presented in Figure 3.8.
+ Cl)
"a
o C
er:
•
+
Cl)
Cl)
"a
"a
0
0
rI~~ Figure 3.8 Schematic representation of nickel electroplating rig.
A surface activation procedure for stainless steel applying cathodic and anodic settings detailed in Table 3.4 (on the following page) was adopted and developed further for Ni electroplating. All electroplating experiments were carried out at room temperature. Electroplating commenced anodically with current density I - 2 A dm·2 for 2 minutes. During this initial anodic treatment stage the filter sample and nickel electrode(s) represented the positive anode and negative cathode electrodes respectively. Positive metal ions stripped from the filter sample anode moved across to the nickel cathode(s) leaving a 'fresh' surface on which to electroplate. After a
3- 15
CHAPTER 3. Experimental (Electroless Nickel Plating) short time the electrode polarity was reversed to enable cathodic treatment. With this arrangement, positive nickel ions, Ni 2+ were attracted to the negatively charged plating material resulting in a thin layer of nickel on the metal surface. The thickness of this electroplated layer was determined by the plating time and current achievable in the electrolysis cell. The electroplating treatment activated the filter surface ready for electroless nickel plating. Table 3.4 Nickel electroplating operating conditions. Treatment anodic cathodic
Anode
Cathode(s)
Current density Plating time (A dm-2) (mins) 1-2 filter sample Ni electrode(s) I-4 Ni electrode(s) filter sample 2 - 20 2
Nickel electrodes measuring lOO x 60 mm were prepared from 40
~m
nickel CPO
screen material. Each electrode was treated with 5 M sodium hydroxide 'degrease' and deionised water ' rinse' steps before entry into the electroplating bath. 47 mm diameter circular filter samples were cut from the virgin screen material with a 60 x 15 mm tab for connection to the electrical circuit. Pre-treatment of the filter sample and Ni electrode(s), and agitation of solution preceded any electroplating treatment. One or two pre-treated Ni electrodes were utilised in the electroplating stage. Where one electrode alone was required the current density was adjusted accordingly. Colour-coded electrical wires coupled to the power supply terminals were connected to each electrode by crocodi le clips. Electrodes were set up in the strongly acidic electroplating bath moments before introduction of the filter sample. Sample transfer from the pre-treatment stage to Ni electroplating bath was rapid yet smooth. A clamp-stand held each electrode and the filter sample fixed in position. Separation distances between electrode(s) and filter sample of 35 - 40 mm were common. Table 3.5 Cathodic electroplating current values, per electrode. Description
Current, per electrode (mA)
Nickel SPO screen Stainless steel mesh
325 230
3-16
CHAPTER 3. Experimental (Electroless Nickel Plating) Current reversal during electroplating was unusual, favouring continuous cathodic treatment instead. Table 3.S, on the previous page, highlights typical cathodic current values (per electrode) for nickel slotted and stainless steel mesh samples. These current values were derived from a recommended current density of 2 A dm·2 A maximum voltage of 4 volts was set. The filter surface was electroplated for between 10 and IS minutes. Rapid and direct transition between plating stages ensured the
survival of the active surface into the electroless Ni plating bath.
3.4 SURFACE PRE-TREATMENT Electroless nickel processes are sensitive to contamination by metals, sulphur compounds and particulate matter such as dust (PMD (UK) Ltd.,
1993).
Comprehensive pre-treatment of the sample was vital to its success in the plating bath. The metal surface was processed through three pre-treatment stages, carefully developed to optimise the starting surface for nickel deposition . In stage one strong alkaline solutions of sodium hydroxide and Ultrasill 11 removed oils, grease, organic contaminants and dust from the surface. Stage two offered two parallel treatments of concentrated hydrochloric acid capable of 'stripping' away the top oxide layer or nickel electroplating for removal and subsequent deposition of a thin nickel layer. The final stage before electroless plating was particularly important. A thin film of sodium carbonate delivered on the raw surface provided the best possible start to the deposition process and improved adhesion (PMD (UK) Ltd., 1993). Good rinsing between pre-treatments protected the Ni-Shield M solution from contamination by ' foreign' solution ions. The pre-treatment schedule adopted for electroless Ni plating and electroplating experiments is outlined in Figure 3.9, on page 3-18.
3.4.1 Chemicals Used Sodium hydroxide pellets and hydrochloric acid S.O. 1.18 (37 %) for general laboratory applications were supplied by Fisher Scientific UK. Ultrasill II was provided by the Department of Chemical Engineering, Loughborough University. Fisher Scientific UK also supplied anhydrous sodium carbonate, a specified laboratory
reagent.
All
solutions
were
prepared
using
deionised
water.
3-17
()
::r: Ultrasill 11 2%w/w
STAGE ONE
Deionised water rinse
SURFACE CLEANING (ALKALINE DEGREASING)
...
5 M sodium hydroxide
...
Deionised water rinse
~
trl ;:Q
w
trl x
'"0 Cl>
:J.
3
Cl> 0
I
~
OPTfONAL
!:!. trl
~
[....
Ni electroplating
STAGE TWO
Hydrochloric acid 8 M
...
Deionised water rinse
SURFACE LAYER REMOVALffiEPLACEMENT (ACID PICKLINGI ELECTROPLATING)
@~
0
Deionised water rinse
(;-
'" '"
Z
o·
"" ~
::g ~. 0
(JQ
OPTfONAL
OPTIONAL
I
Electroless Ni plating Sodi urn carbonate 2%w/w
STAGE THREE
w I
PREPARATION FOR ELECTROLESSNICKEL PLATING BATH
...
'dummy' Ni plating solution
...
...
00
Figure 3.9 Pre-treatment schedule for electroless nickel plating and nickel electroplating.
Deionised water rinse
~
CHAPTER 3. Experimental (Electroless Nickel Plating)
3.4.2 Alkaline Degreasing (Stage One) All filter samples were contacted with 5 M sodium hydroxide in a tall-sided 250 ml glass beaker, and sonicated for 5 minutes at room temperature. A tab attached to the 47 mm diameter filter was carefully reshaped to allow submergence of the filter in the degreasing solution. Waste sodium hydroxide was transferred safely to a waste storage container. The surface of nickel slotted and stainless steel mesh filters was particularly oily and greasy. Tests showed that cold sodium hydroxide alone was insufficient to attack this greasy layer. An aggressive degreasing stage was developed to treat such screens in advance of the sodium hydroxide stage. 50 ml Ultrasill 11 solution 2 % w/w was contacted with the filter sample and heated to approximately 60
0c. The solution was
sonicated for 15 minutes, reheating every 3 minutes. Following safe di sposal of the strong alkali solution the filter was rinsed in deionised water and sonicated for two further minutes. 3.4.3 Surface Layer Removal (Stage Two) 50 ml hydrochloric acid 8 M was added to the glass beaker containing the filter sample. The beaker was placed in a vacuum desiccator and degassed for 90 seconds to remove air from the pores thus maximising the contact area for acid pickling solution. Acid solution was heated to approximately 60 °C before sonicating for 15 minutes, whilst reheating at 5-minute intervals. Operations involving the warm acid solution were conducted in a fume cupboard at all times. 3.4.4 Surface Activation (Stage Two) Nickel electroplating to activate the metal surface replaced the hydrochloric acid pickling stage in the pre-treatment schedule. The filter sample was degreased as usual in either Ultrasill II or sodium hydroxide solutions followed by a deionised water rinse. Electroplating of the filter material followed the procedure described in Section 3.3 on page 3-14.
3-19
CHAPTER 3, Experimental (Electroless Nickel Plating)
Table 3.6 Pre-treatment schedule matrix,
STAGE THREE
STAGE TWO
STAGE ONE
c:
U 0
0
Cl)
~
0
.. . --.. ........ ........ - u c:
CD
0
Cl) ~
Cl)
I'G
Cl)
Cl)
c: Cl)
E
';: Cl)
c. )( w
0
VI
.!!!.
I'G
Cl)
:S :S
.!!!.
..
'iij
'iij
I'G
I'G
<
::l
0
Cl)
u
~
:!::
~
:!::
::l
()
0
0
.!!!.
I'G
== "C
0
U) :;:;
"0
I'G
I'G
,!!!
0 'Qj
Z ::!;
..,
Cl)
c: 0
Cl)
0
VI I'G
en
c:
J:
0
0
I'G
.!!!. 3:
,!!! c:
Cl)
..
Q. 0~
u
Cl)
w
0
en
Cl)
0
~
CD
0
.!!!.
0
+
Cl)
I'G
Cl)
VI
I'G
Cl)
()
-
.!!!.
.c:
()
()
()
()
J:
J:
J:
J:
J:
::!;
::!; ex>
::!; ex>
::!; ex>
::!; ex>
.!!!.
CD
"C
0
..- - -..
0
~
Cl)
Cl)
c:
0
+
Cl)
0 :;:;
~
.c:
I'G
==
~
0
.!!!.
...
0
() N
I'G
"C
Z
.!!!
~
Cl)
c:
0 'Qj
0
~ 0
'"
en c:
:;:;
"
.!!!
c:
u
c. en Gi ..I<
0
VI
:;:; I'G
Q.
'c VI VI
.
-'" Cl)
== "C
Cl) >. Cl) ,!!! E 0 c: E u 0 Total
" ~
0
~
w
'Qj
0
Time
'so' and ' degass' refer to sonication and vacuum degassing respectively
3-20
CHAPTER 3. Experimental (Electroless Nickel Plating) 3.4.5 Alkaline Coating Treatment (Stage Three) A final rinse in 1 - 2 % w/w sodium carbonate solution was important before immersion in the plating bath. An optional deionised water rinse was sometimes incorporated after acid pickling. Otherwise the filter sample was contacted with 50 ml sodium carbonate 2 % w/w and sonicated for 3 minutes. 3.4.6 Transfer to Electroless Plating Bath (Stage Three) The speed with which the filter sample was conveyed between the sodium carbonate and Ni plating solutions was crucial to the success of the plating process. Gripped by tweezers, the filter was dipped briefly in a 1000 ml 'buffer' electroless Ni plating solution at room temperature to remove excess chloride ions present on the surface. Table 3.6, on the previous page, is a pre-treatment matrix, which identifies all the pretreatment combinations adopted during the electroless Ni plating experiments. Each pre-treatment schedule is referenced according to the codes PT I - 16.
3.5 CHARACTERISATION OF METAL MICROFILTERS 3.5.1 Thickness The thickness of each metal microfilter was measured at ten points across the surface using a pair of Vernier callipers. An average thickness value was calculated from all ten measurements (in microns). Analysis of CPO screen samples after electroless nickel plating suggested a deposit of 9 ~m thickness on the surface was typical after a 60-minute experiment (Experiment Reference 70).
3.5.2 Pore Size Analysis
3.5.2.1 Optical Microscope (OM) A simple, non-destructive pore size analysis technique employing an optical microscope, video camera and PC with image capture software allowed for preliminary assessment of pore dimensions. This method, which offered mediumlevel precision and a reasonably fast analysis time, was particularly suited to development of the nickel deposition process. Pore openings were studied using a
3-21
CHAPTER 3. Experimental (Electroless Nickel Plating) Leitz optical microscope (Leitz Wetzlar, manufactured in Germany) at either 20x or 50x magnification of the incident lenses. Microscope lighting units and condenser lenses were carefully set-up. The microscope stage was usually illuminated by transmitted and a little incident light. A video camera mounted to the microscope direct! y overhead relayed the images to a Pc. Selected microscope views were captured using image capture software (Videologic TV Snap, version 1.0.03) and saved as compressed image files. Care was taken to ensure that each image was focused as sharply as possible and the aspect ratio (640 x 480) was constant for all images. A graticule (I 0
~
divisions, Graticules Ltd., Tonbridge, Kent) provided the
measurement scale for pore size analysis. New graticule images were acquired at the beginning of each session. Around 20 pores located across the surface were measured in the horizontal and vertical directions using graticule images captured at the same magnification. In the case of slotted pore openings the slot width was evaluated at six positions along the slot length. Stainless steel mesh proved difficult to analyse because only one layer of the interwoven fibres could be focu sed at anyone time. Polarised incident light reduced glare from the steel surface. Optical microscopy suffers from a relatively small depth of focus, i.e. approximately 10 I-lm at 100 times and I I-lm at 1000 times magnification respectively (S haw, 1996, pp.46-47). The measurement of distances less than 2 I-lm is likely to be in serious error. The visibility of an object can also be limited due to a lack of contrast between object and surrounding background.
3.5.2.2 Scanning Electron Microscope (SEM) The Scanning Electron Microscope (SEM) provides a representative high-resolution image of a sample's surface morphology. The depth of focus is far superior to that of optical microscopy (some 300-500 times bigger than for light microscope) so that many different levels of object can remain in focus (Hunter, 2001, p.209). Equally the resolution possible is of the order of 150 - 250 angstroms (Dennis and Such, 1993, p.16). Depending on the composition and nature of the sample it is sometimes necessary to prepare the surface with a thin layer of gold. The surfaces of untreated nickel CPO and SPO screens were examined directly using the SEM instrument.
3-22
CHAPTER 3. Experimental (Electroless Nickel Plati ng)
Figure 3.10 Slotted and circular pore width measurement by SEM; (a) isolated circular pore opening (un treated) , and, (b) si ngle nickel slot (un treated).
3- 23
CHAPTER 3. Experimental (Electroless Nickel Plating) Each screen was viewed at simil ar magnifications. Pertinent SEM pictures were captured either digitally or using a standard SLR camera. The width of discrete circular and slotted pores, as illustrated in Figure 3. 10 (a) and (b), on page 3-23, could be detemlined using on-screen cursors. SEM analysis of circular and slotted pore dimensions is reported in Table 3.7. Table 3.7 Pore size determination by SEM analysis.
Side
Pore size
shiny side shiny side matt side
(Ilm) 14.0 11.6 x 362 11.5 x 402
Screen NiCPO-17 NiSPO- 13
3.5.2.3 Image Analysis Circular and slotted pores examined by electron optical microscopy were analysed using 'Scion Image for Windows' (Scion Corporation, Beta version 4.0.2), an image processi ng and analysis program for the Pc. Scion Image for Windows can acquire, displ ay, edit, enhance, analyse and animate images. It supports many standard imageprocessing functions, including contrast enhancement, density profiling, smoothing, edge detection , and median filtering (Website http://www.scioncorp.com).Scion Image can also measure the area, mean, centroid, and perimeter of selected regions and performs automated particle analysis. Slotted and circular pore openings were viewed at the max Imum resolution (magni fication) possible of electron and optical microscopy to afford optimum precision for Image Analysis. Images saved as Windows Bitmap files (.BMP file ex tension) were resized to just fit in the ' image window'. Consider the Image Analysis of pore openings with rectangular slot dimensions. As seen in Figure 3.11 on the following page, there were typically two complete slots per SEM image. An image consists of a two-dimensional array of pixels (picture elements). Each pixel is individually described by integers ranging from 0 to 255. Scion Image follows the IBM PC convention di splaying zero pixels as white and those with a value of 255 as black (reference). A Look-up Table (LUT) is used within Scion Image to map pixels (in the range 0 - 255) to screen colours.
3-24
CHAPTER 3. Experimental (Electroless Nickel Plating)
Figure 3.11 A typical SEM micrograph of slotted media used for Image Analysis.
The spatial scale used within the program was calibrated from measurement of a known distance such as the 'scale bar' displayed at the top of the SEM image. The maximum spatial scale achievable was 0.46
~m
per pixel. Pictures derived from the
optical microscope were scaled according to graticule dimensions viewed at the same magnification. The spatial scale, typically in pixels per micron, was reset for each image. Figure 3.12, on the next page, illustrates the image processing stages for a slotted pore shown in Figure 3. 11 above. Individual pore openings were selected from the image, as indicated, usi ng the rectangular selection tool and copied as a duplicate image to a new window (Figure 3.12 (a)). Spatial calibration was preserved in the new window. A thresholding technique called 'density slici ng' (Figure 3.12 (b)) was applied to discriminate the pore opening (darker area) from the surrounding metal surface (lighter area) based on their grey values. All pixels shaded between upper and lower thresholds were highlighted in red. Background pi xels were left unchanged. Threshold 3-25
CHAPTER 3. Experimental (Electroless Nickel Plating)
Figure 3.12 Image processing stages for pore size analysis of slotted pore opening.
(a) Slotted pore selected for analysis and duplicated in new image window.
(b) Density slicing of selection. LUT window (below) adjusted to highlight darker slot region. Pixels within density slice thresholds are shaded red .
(c) Selection converted to a binary image (black and white only).
(d) Binary image processed with 'erosion' filter removing pixels adjacent to and distant from slot region.
3-26
CHAPTER 3. Experimental (Electroless Nickel Plating) levels were adjusted by manipulating the density slice (in the LUT window) until the pore opening (darker area) was highlighted. During this process some unrelated areas, which shared grey shades with the pore opening, were also highlighted. The resultant greyscale image was duly converted to a binary image, i.e. black and white only, to enable further processing (Figure 3.12 (c». Pixels highlighted by density slicing were set to black (255) and all other pixels to white (0). An 'erosion' process filter removed pixels touching the edges of the pore and isolated pixels in the binary image (Figure 3.12 (d» . A pixel was removed (set to white) if four or more of its eight neighbours were white. Remaining pixels distant from the pore area were removed carefully with the 'eraser' tool. A density profile of the pore area was generated for a rectangular selection slightly
larger than the pore outline (Figure 3.13 (a». The width of this ' Plot Profile' was equal to the length of the rectangular selection, in pixels. The selection was divided into pixel-wide columns. Each point in the plot represented the average grey value of the pixels in the corresponding column in the selection. Given that the image was binary the grey value of each pixel was either 255 (black) or 0 (white). A Plot Profile of average grey along the length of the selection (pixels) is presented in Figure 3.13 (b). EJ
E!l lil 42 IIDd'IC!.o omp
3.13 (a)
3.13 (b) O~o~------------------------~~ =---------------------------m~
Figure 3.13 Image Analysis of slotted pore opening; (a) rectangular area selected for profiling, and, (b) density profile of pore area (Plot Profile). 3-27
CHAPTER 3. Experimental (Electroless Nickel Plating) The Plot Profile data was exported as a text file to an Excel spreadsheet. The average grey value was converted to a meaningful pore width, in microns, by dividing the average grey value by 255 (average value if all pixels in a column were black) and mUltiplying by the selection height, in microns. The selection length was simply converted to a distance along the slot, in microns, using the spatial scale factor and discarding white background regions at either edge of the selection. Figure 3.14 is a profile of the slot width measured at 0.4 ].l.ffi increments along its length. The pore length was derived from Plot Profile data. The maximum spatial scale achieved was equal to 2.2 pixels per micron. Pore size distribution for the slotted media was quantified from twenty pores.
14 12
E
10
::l.
-...
..s:::
8
"C
';: Q)
0 D.
6
4 2 0 0
100
200
300
400
Distance along pore length, ~m.
Figure 3.14 Pore width profile along slot length.
Greyscale images of circular pores from optical microscope studies were also analysed using Scion Image. The spatial scale was calibrated according to graticule dimensions viewed at the same magnification. The simple pore geometry removed the need for image processing prior to analysis. Horizontal and vertical dimensions of
3- 28
CHAPTER 3. Experimental (Electroless Nickel Plating) each circular pore were measured for comparison with a 'measuring' tool. 88 and 33 circular pores were examined for large and small pore openings respectively. The optical microscope resolution restricted the spatial scale to 1.9 pixels per micron. Average pore sizes, including standard deviation , confidence intervals, and measurement details are presented in Table 3.8. The average pore size was concluded from all pore measurements. Table 3.8 Pore size measurements of CPa and Spa screens
by Image Analysis of SEM and optical microscope images. Reference
Image
Pores viewed
Measurements per pore
NiCPO-40 NiCPO-17 SSPO-20 NiSPO-13
OM OM OM SEM
88 33 30 20
2 2 2 - 1000 (width)
Pore width 'horizontal'
Std dev
(urn)
39.8 + 0.3 17.1+0.2 29.2 + 0.04 402.8 + 0.6
Pore width 'vertical'
Std dev
(urn)
lA 0.7 3.2 2.3
38.9 +0.3 17.3 + 0.3 28.1 + 0.03 13A + 1.0
lA 0.9 2.9 lA
'std dev ' refers to the standard deviation of the population. The range of pore sizes on either side of the mean, known as the 'confidence interval' was calculated from the standard deviation, number of samples, and confidence level. A confidence level of 9S per cent was assumed. The confidence interval is given by:
(3.1)
where x is the mean value,
(J
is the standard deviation, and n represents the number
of samples.
3.5.2.4 Coulter Poromeler
The Coulter Porometer IT is a fu lly automated, microprocessor controlled version of the classical 'bubble-point' method for analysing the pore size distribution of media such as membrane filters, paper filters , cloths, etc. It uses a liquid displacement technique based on the ASTM F-316 and BS 6410 specifications. The Porometer monitors the expUlsion of a wetting fluid (Porofil) from the pores of the test material
3-29
CHAPTER 3. Experimental (Electroless Nickel Plating) exposed to increasing air pressure. The wetting fluid is displaced from progressively smaller pores and the total airflow rate through the membrane can be related to the pore size via a version of the Washburn equation. Statistical functions are also applied to the data. The Coulter Porometer can determine pore size distributions for test samples containing pores in the overall size range of (approximately) 300 - 0.05 !lm diameter. 3.5.3 Evaluation of Open Area The open area (per cent) of a porous medium refers to the ratio of area available to flow compared to the total 'active' surface area of the membrane. In filtration applications a large open area is usually desirable to allow a high throughput of permeate (filtrate). Open area was evaluated using optical microscope and SEM images of individual pores viewed within the Scion Image Analysis software. Figure 3.15 is a simple diagrammatic representation of the methodology used to evaluate open area.
o
o o
o Figure 3.15 Representation of open area calculation methodology.
Having already determined the average pore dimensions for circular and slotted pores, additional linear dimensions were measured using the appropriate measurement tool. The average horizontal distance between pores on the same row (blue arrow) and average vertical distance between pores directly in line (green arrow) were measured from several images. In the case of the CPO membrane for example, the average horizontal and vertical dimensions were typically made up from 70 and 30 individual 3-30
CHAPTER 3. Experimental (Electroless Nickel Platmg) measurements respectively. Each linear dimension was measured from the edge of one pore to the edge of the next. A rectangular area was considered (red dashed box) encompassing the area of two pores, i.e. one complete pore and four quarter pore areas. The ratio of total area of ' pores within this selection to the total area of the selection, and converted to a percentage, provided the open area of the screen. The open areas of CPO and SPO screens, calculated according to the same methodology, are summarised in Table 3.9. A simple relationship exists between pore size and open area for CPO media whereby reduction in pore diameter Jp.", to
Jpo,,2
results in a new open area Aopen2 according
to the relationship:
(3.2)
Equation 3.2 clearly shows the effect that pore size reduction by nickel plating had on the open area. If the diameter of pores was made smaller then the new open area would be significantly smaller than before as a result of the squared term. For example nickel plating of the 17 !lm circular pores to 2 !lm would result in a new open area of a mere 9.7 X 10.3 per cent!
Table 3.9 Open area measurements of SPO and CPO screens by Image Analysis of SEM and optical microscope images. Reference
Data points
Distance between pores, X-direction
Data points
0,1111)
NICPO-40 NiCPO-17 NiSPO-13 SSPO-20
73 71 30 20
162.5 181.5 141.0 22.9
34 31 60 20
Distance between pores, Y-direction
Open area
(!IJ11)
(%)
3105 327.1 143.6 22.7
3.5 0.7 5.7 310
3-31
CHAPTER 3. Experimental (Electroless Nickel Plating) 3.5.4 Quantitative Elemental Analysis Electron probe microanalysis (EPMA) provides detailed information of the surface composition of a sample. The scanning electron microscope (SEM) is coupled to a solid-state X-ray detector capable of determining the intensity and characteristic Xrays emitted by surface atoms during electron bombardment. It can qualitatively or quantitatively determine elements in the sample with an atomic number greater than 5 (Boron) according to limits of detection in the order of 0.01 per cent (Website http://www.nhm.ac.uklmineralogy/facilities/probe.htm). EPMA provided elemental composition information about the virgin nickel SPO screen and electroless nickel coatings.
3.6 SUMMARY
Nickel 'Veconic Plus' screens available WIth circular and slot pore openings, and stamless steel mesh with square pore openings, provided the substrate for electroless nickel plating. The electroless nickel plating rig was developed with the aim of producing a consistent and uniform deposit. To achieve this, various modes of solution agitation were considered including pumped circulation, mechanical stirring and rur injection. Pre-treatment schedules specific to the requirements of individual metal substrates were optimised to ensure good adhesion of the nickel-phosphorous deposit. The physical nature of the stainless steel mesh caused difficulties for surface activation by nickel electroplating. Although the nickel slotted media is by definition autocatalytic the presence of an organic surface fimsh dictated a more aggressive pretreatment. Techniques of varying precision were used to characterise the untreated and nickel coated screens. Pore size analysis using Image Analysis of optical microscopy images and electron micrograph was the favoured method. A similar method for determining open area available to flow was used.
3-32
CHAPTER 4. Results and Discussion (Electroless Nickel Plating)
CHAPTER FOUR RESULTS AND DISCUSSION: ELECTROLESS NI PLATING Experimental results describing the development of an electroless nickel plating (ENP) process and associated pre-treatment methods are presented in this chapter. Three distinct sectIOns form the discussion. In the first section the effect of solution agitation on the consistency and uniformity of nickel-phosphorous deposits is presented. The modes of agitation considered include circulation by pumping, mechanical overhead stirring and air injection. As explained in Chapter Three (Experimental: Electroless NIckel Plating) the pre-treatment stage IS extremely important if a successful deposition is to be achieved. The following pre-treatments specific to each metal substrate are evaluated in the second section: surface layer removal by hydrochloric acid; surface activation using nickel electroplating (EP); and alkaline treatment prior to bath entry. Fmally, nickel deposition behaviour as a function of plating solution life receives comment. A summary of the experimental findings concludes the Chapter. Further work by colleagues in the Chemical Engineering Department, Loughborough University, to optimise the ENP process for production of microfilters is also outlined. A complete record of all electroless nickel plating and related experiments is presented in Appendix A.
4.1 ASSESSMENT OF SOLUTION AGITATION METHODS Three different agitation methods were independently compared according to the umformity of pore size achieved by electroless mckel plating. Solution agitation was provided by pumped circulation, mechanical agitation and air injection beneath the sample. 4.1.1 Solution Circulation by Pumping Pumped circulation of plating solution through the pores was evaluated following previous plating success with tubular membranes. The plating sample was located in a mollified 47 mm transparent Sartorious membrane holder. Its design incorporated a thick o-ring, which on compression generated a critical seal around the sample. The metal screen to be plated was swiftly transferred from the pre-treatment stage into the plating solution; a cold 'buffer' plating solullon restricting carryover of excess sodium
4-1
CHAPTER 4. Results and Discussion (Electroless Nickel Plating) carbonate into the plating bath. Tests performed to assess the impact of a buffer solution on electroless plating had previously shown no adverse effects. Submerged in the buffer solution, the membrane holder was quickly assembled once the sample was in position. Speed and care were both essential for a successful transfer. Air was excluded from the peristaltic tubing and membrane holder pnor to use. In early experiments, the membrane holder was inverted in the plating bath with the
exposed sample surface facing downwards. Problems with holder stability were addressed using a specially designed support, which fitted neatly around the edge of the holder. A magnetic stirrer (base-section submerged in the water bath) operated directly beneath the membrane holder to provide additional fluid mixing. The pumping action from high capacity (HC) or low capacity (LC) peristaltic pumps circulated the plating solution around the bath and importantly, from a plating perspective, regulated flow through each pore. In the normal pumping direction the plating solution was drawn through the sample and around the recirculation line before discharging at the bottom of the plating bath. Hydrogen bubbles generated from the plating reaction at the surface were carried into the peristaltic tubing where they collected. If the flow of bubbles Was allowed to continue without interruption the suction-side pressure (monitored locally) steadily increased until the peristaltic tubing eventually collapsed. The pressure difference across the sample in tlus situation exceeded 400 mbar. This steady pressure rise was temporarily relieved by the action of a short backflush. In its chosen orientatlon the membrane holder was seen to act as a barrier to the natural escape of hydrogen bubbles from the surface. Bubbles trapped in the pores or touching the sample caused blinding of the plating surface. Wherever this occurred, nickel deposition was prevented or at best, severely restricted. The pumped system was modified with this problem in mind. The membrane holder was repositioned so that the holder top faced upwards in the bath. Plating solution was now pushed up through the sample pores (Experiment reference 44 onwards) and in this mode of operation the pressure data provided a forewarning of overpressurisation. An occasional backflush removed gaseous reaction products from the pump tubing. Electroless plating using pumped operation was often unsuccessful and suffered from non-uniform deposition across the surface; many pores were untouched, others plated completely. The pumped system represents a paradox whereby the potential benefits
4-2
Table
4.1 Experimental conditions, observations, and pore size analysis for ENP of 17.1 J11l1 nickel epo screens with pumped circulation.
Description of pumping duty Contmuous Contmuous Contmuous Contmuous Occasional
Pump
HC20R (pull) HCR (pull) HC13 (pull) HC 14 (pull) HCR (push)
Plating time (mins) 60
Temp
Pore width
(DC) 79.4
15.0± 1.3
Comments/Observations
Expt. ref.
Bubbles seen nsmg up tubIng. Pressure dropped after flow reversed. SuctIOn Side pressure
33
(urn) > 400 mbar. Some holes completely blocked
85
785
no analysIs
90
780
19.0±07
60
81.0
no analysIs
110
845
40±16
120
84.5
14.0±1.1
120
848
7.0± 19
120
84.9
no analysIs
107
84.5
40±1.7
Interrmttent (60s on - 60 s off) Intermittent (40 s on - 300 s off) Delayedlcontmuous (5 rmnute delay) Delayedlconllnuous (5 minute delay)
HCR (push) HC 14R (push) He 10R (push) LC70R (push)
Delayed/contInuous (5 rmnute delay)
LC70R (push)
110
852
90±18
Delayed/contmuous (10 rmnute delay)
LC70 (push)
80
846
100±14
Small bubbles observed nsmg up tubmg. Pressure drop higher with sllrrer bar at low serung Surface coated m a black depOSit which could be removed by somcatlon. Pores appear untouched when viewed with an optical rmcroscope. Tubmg filled With arr mltlally. Consider submergmg filter holder pnor to use. Pores margmally smaller. approximately 15 I1m (optical microscope) Bubbles seen nSIng from surface towards end of Immersion lime. Pressure IOcreased slightly when solutIOn pumped every 10 minutes. O-nng became stuck to sample. Random platIng. There was a shmy Hmsh to the surface Bubbles noticed on sample. A Few bubbles rose up through solutIOn. Some pores were completely blocked. Pressure Increased to 20 mbar after 10 - 20 rmnutes (no pumpmg). Small bubbles seen nsing from surface. Bubbles remamed at the end. A reactIOn was observed after 60 seconds Pump started m reverse after 5 mInutes Gas pulled through tubing m reverse mode. No sign of bubbles when pump stopped. There were a few bubbles on the surface Imllally. Pump started m reverse after 5 rmnutes. Bubbles passed through tube from sample. Pump direction changed after tubIng filled with gas Many small bubbles passed through the sample all at once. When the pump stopped more bubbles were observed on the surface than usual. Pressure Increased to 20 mbar at end of expenment Manv pores plated (OptiCal microscope). Pump activated after 5 mInutes (run for 2 rmnutes) Flow reversed tmtlally to remove air from tubIng. Bubbles nSIng from surface. Most pores received plating (optical microscope) Pressure remaIned high after backflush. Flow reversed after 32 rmnutes. Bubbles passed through sample on return to normal directIOn. Pressure 21 - 23 mbar at end ofplatmg
35 38 41 50
51 52 53 59
56
55
CHAPTER 4. Results and Discussion (Electroless Nickel Plating)
to plating of improved flow are restricted by the presence of gaseous deposition products trapped in the tubing. Many electroless nickel plating experiments were performed using pumped circulation, a selection of which are presented in Table 4.1, on the previous page. The plating substrate was nickel CPO screen, which unplated has a pore size of 17.1 ± 0.2 ~m. Nickel deposition behaviour was evaluated against parameters such as solution flow rate, plating time, plating bath temperature, flow direction, stop-start pump operation, use of an initial pumping delay, and backflush operation. Where specified the pore size analysis was performed using optical microscopy. As seen in Table 4.1, the reduction of pore size varied greatly between samples experiencing similar conditions. The comments and observations report a catalogue of plating inconsistencies and process problems. Of particular concern was the apparently random coverage of deposited nickel. Regions of variable plating intenSity were clearly visible on the surface and optical microscopy revealed a wide range of pore sizes within each region. Intermittent pump operation failed to improve the situation (Experiment references 51 - 53). Average pore openings between different samples ranged from 12 ~m down to 4
~m
in size. Pump activation was also delayed to
encourage initiation of the deposition reaction in the first 5 to 10 minutes (Experiment references 53, 55, 56, and 59). The average pore size range for this operational mode however was similar to that reported for experiments without a pumping delay 4.1.2 Mechanical Overhead Stirring Solution agitation by overhead stirring was investigated following problems with the pumped system. A pre-treated metal screen was introduced to the nickel plating bath and agitated at a speed suffiCient to maintain mobility in the bulk solution. The sample was subjected to the flow pattern generated by the stirrer but occasionally rested against the plating bath wall. Experimental studies performed using nickel CPO and SPO screens, and stainless steel mesh are presented on the following page in Table 4.2. As can be seen the electroless plating results were mixed; the plating rate and, quality and consistency of nickel deposits were essentially random. The operating performance of the stirred system was restricted by its simplicity. The random movement of the sample through a comparatively large solution volume
4-4
CHAPI'ER 4. Results and Discussion (Electroless Nickel Plating) afforded only limited control of the plating process. Plating bath temperature was the only parameter that could be manipulated, although temperature effects such as 'dead' spots were a distinct possibility in the relatively large plating bath. On some occasions the stirrer also damaged the delicate stainless steel mesh. Table 4.2 Experimental conditions, observations (including pore size analysis) for ENP of different metal screens using mechanical agitation. Screen
Temp
NiCPO-17
Plating time (mins) 20
NiCPO-17
23
83.0
NiCPO-17
65
82.5
NiCPO-17 SSPO-20
60 20
86.3 80.0
NiSPO-13 NiSPO-13
45 60
79.0 80.3
Observations
Expt. ref.
Small bubbles seen on surface immediately after immersion. Pore size reduced to 14.0 ± 1.5!1Jll Copious small bubbles on surface and rismg through plating bath. Pores completely plated. Small bubbles observed leaving the sample after 60 minutes. Complete plating of most pores. Ni sample was completelv plated No bubbles observed. Steel mesh did not seem to react in ENP bath. Microscope studv indicated unsuccessful platine:. No bubbles observed Shadmg on sample surface was not uniform. Some dark and lie:ht
10
(0C)
80.0
12
7
60 10
34 36
4.1.3 Air Injection
4.1.3.1 Comparative Plating Experiments with and without Agitation Solution agitation by air injection offered the advantageous removal of hydrogen bubbles from the plating surface. A small compressor (external to the bath) generated a continuous column of tiny bubbles from a Pyrex sintered funnel. The air sparge (AS) apparatus was pOSitioned centrally at the bottom of the plating bath. A 47-mm diameter pre-treated sample with rectangular tab was secured in position just above the sintered funnel in the plating solution. Activation of the air sparge was delayed to encourage initialisation of the deposition reaction on entry. The plating surface was studied closely for signs of reaction during the first minute. When the air sparge was
4-5
CHAPTER 4. Results and Discussion (Electroless Nickel Plating) operating the sample experienced a wall of fine bubbles sweeping past both faces, strong enough to cause slight swaying. It has been reported
In
literature that solution agitation by air injection can, in some
circumstances, disrupt the electroless plating reaction; air bubbles close to the surface may interfere with mass transfer between the surface and bulk solution (Riedel, 1991). In a series of experiments the performance of electroless plating was evaluated in the presence 17.1
of
air
flow
movement.
Pre-treated
nickel
CPO
screens
with
± 0.2 jlI11 diameter pores were exposed to the electroless plating solution (1400
ml volume) for varying durations, with either air sparge based agitation or no agitation. The experimental conditions are highlighted In Table 4.3. Contact time with the plating solution varied from 10 to 60 minutes. The air sparge apparatus remained located In position during all experiments. The pore sizes were analysed post-plating using a Coulter Porometer IT instrument (see Section 3.5.2.4). All samples were evaluated in a 47 mm filter holder with water as the wetting fluid. Pore size measurements including pore range, mean flow pore size and bubble point are presented in Table 4.4 (on page 4-7) for nickel CPO screens plated with air flow agitation and equally for those Without any solution agitation. The virgin nickel screen had a mean flow pore size of 14.6 jlI11 with pores ranging from Table 4.3 Experimental conditions to determine the effectiveness of ENP in the presence of agitation by air injection (17.1 jlI11 nickel CPO screens). Agitation
Plating time (mins)
air snarge air sparge air sparge air snarge air sparge air soarge air snarge
20 30 40 45 50 60
no acitation no a!!ltation no agItation no agItation
20 40 50 60
10
Solution Experiment temperature reference (OC) 84.8 65 80.0 67 79.6 69 66 80.0 793 71 800 68 80.5 70 80.0 79.3 80.0 79.9
73 72
75 74
4-6
CHAPTER 4. Results and Discussion (Electroless Nickel Plating) Table 4.4 Pore size analysis (Coulter Porometer II) of nickel deposition in the presence of air sparge agitation, and without agitation!. P lating time (mins)
Minimum pore size (fl-m)
Maximum pore size (fl-m)
Mean flow pore size (fl-m)
Bubble point (fl-m)
0 10 20 30 40 45 SO 60 OT 20 T 40 T SOT 60 T
12.3 9.4 7.8 6.3 4.4 6.2 3.1 4.1 12.3 8.4 6.5 5.0 3.9
21.1 12.6 12.0 8.6 6.9 9.0 5.0 6.0 21.1 11.0 8.7 7.1 6.1
14.6 10.8 9.2 7.6 5.5 7.9 4.1 5.2 14.6 9.8 7.9 6.2 5.2
20.4 12.6 9.1 8.6 7.0 12.1 5.0 6.2 20.4 11.0 8.7 7.2 6.1
Plating rate (fim br-I) nla
Expt. ref.
11.4 8.1 7.0 6.8 4.5 6.3 4.7
65 67 69 66 71 68 70
nla
nla
7.3 5.0 5.0 4.7
73
nla
72
75 74
16 ___ solution ag itation by air sparge - 6 - no solution agitation
E
:::!.
Cl
12
.!:! UI
~
0
Q.
8
3:
0
I;
c:
Q)
4
::E O+----.----~----._--_r----._--_.----~
o
10
20
30
40
50
60
70
Electroless nickel plat ing time, minutes.
Figure 4.1 Comparison of average pore size reduction achieved in the presence of air sparge agitation and without solution agitation (17.1 f.lm original pore size).
4-7
CHAPTER 4. Results and Discussion (Electroless Nickel Plating) 12.3 J.!m to 21.1 J.!m in size, when analysed using the Coulter Porometer. It is notable that these pore sizes are different to those determined by Image Analysis; a technique described in Section 3.5.2.3. This is not surprising however, because the pore sizes were determined using two inherently different approaches; one based on flow rate through the pores and the other by optical measurement. As with any analytical method the Coulter Porometer provides valuable comparative analysis between samples, but it is not necessarily appropriate to compare these results with those of other analytical techniques. As expected, pore sizes were consistently smaller following longer plating times, irrespective of solution agitation. Pore size reduction however, as shown clearly in Figure 4.1 (on the previous page), was generally greater when using air sparge agitation. An average plating rate of 7.0 J.!m hr· 1 was achieved with solution movement compared to 5.5 Ilm hr' I without. Identical pore size reduction reported after the longest plating time (60 minutes), regardless of agitation, may be coincidental. The trend exhibited by experiments carried out with agitation suggests a spurious result. In the absence of more experimental data however one can
12 , - - - - -- - - -- - - - - -- - - -- - - -- - - - - - - - - - - - - - - - - - . - - - Linear regression
8 Pore size (Ilm) ; 12.60 - 0.17 t t; plating time (mins)
4
O+-----,-----.-----.-----.-----.-----~
o
10
20
30
40
50
60
Electroless nickel plating time, minutes.
Figure 4.2 Average electroless nickel plating rate achieved with air sparge agitation.
4-8
CHAPTER 4. Results and Discussion (Electro less Nickel Plating) only su rmise a trend . The pore size reduction of samples subjected to ai r sparge agitation was linear with plating time. Figure 4.2, on the preceding page shows the pore sizes achieved following progressively lon ger exposure times to the plating solution. The equati on displayed on the plot predicting pore size reduction as a function of plating time was derived from a best fit of the data.
4.1.3.2 SEM Analysis of Circular and Slotted Pores Nickel screens with circular and slotted pores were exam ined following electroless nickel plating with air injection. Figure 4.3 (a) and (b), on pages 4-9 and 4-10, are SEM micrographs of pores viewed from the matt side before and after nickel deposition respectively. In Figure 4.3 (b) the nickel membrane was immersed in the plating solution for 50 minutes (Experimental reference 68). A smooth nickel deposit of considerable thickness is evident coating the conical pore entrance, which funnels into a visibly smaller pore opening. The new pore size of 4.7
~m
was confirmed by
SEM analysis (Figure 4.4) and compares favourably to a mean flow pore size of 4. 1 ~m estimated by the Coulter Porometer. The electroless nickel plated 'filtration' surface of the SPO membrane was also compared to the original surface. SEM micrographs are presented for both situations in Figure 4.5 Ca) and (b), on page 4-1 1. The slots were clearly much smaller after nickel deposition . Importantly for practical
4-9
CHAPTER 4. Results and Discussion (Electroless Nickel Plating)
Figure 4.3 SEM micrographs of single circul ar pore opening (high magni fication); (a) ori ginal matt side (1 7. 1 iJ.m pores) and, (b) matt side after 50 minutes pl ating time.
Figure 4.4 C ircular pore width measurement by SEM after 50 minutes plating time. Three-fo ld pore size reducti on achieved from original pore size.
4-1 0
CHAPTER 4. Results and Discuss ion (Electroless Nickel Plating)
Figure 4.5 SEM micrographs of nickel spa screen (high magnification); (a) original shiny side (13.4]lffi slots), and, (b) shiny side after 60 minutes nickel plating.
4-11
CHAPTER 4. Results and DisclIssion CElectroless Nickel Plating)
....1..-......., .. _'.
-
~
••• , . , : "
••
- .
--- .........---....... --
•
7
-
-
~
... . . --
~
~
Figure 4.6 SEM micrograph of nickel SPO screen after 60 minutes nickel plating (see Figure 4.5) showing slots on matt side; viewed at high magnification.
purposes the slots remained open along the length despite the extended 60 minutes plating duration (Experiment reference 83). The slot width was reduced to approximately 2.4 Ilm. Figure 4.6 is an SEM micrograph of the reverse (matt) side of the membrane. Despite the individual examples of successful plating described in this Section, the overall picture describes wide variation in pore sizes throughout each filter, making the final product unsuitable for use as a micro filter.
4.1.3.3 Elemental Analysis of Substrate Surface and Plating Deposit Elemental analysis of the nickel SPO screen surface was carried out using Electron Probe Microanalysis (EPMA). A brief description of EPMA is avai lable in Section 3.5.4. Figure 4.7 Ca) and Cb), on the following page, show the respecti ve EPMA traces of a vi rgin 13.4 Ilm nickel SPO membrane and plated nickel SPO membrane ex hibitin g pore sizes of 4 f..lm . The results support the manufacturer's
4-- 12
X-RAY: Li ve:
0 - 20 ke;l.) 50 s Preset:
50s Remai ni n9:
Os
X-RA'!:
Li't..le:
4.7 (a)
0 - 20 kelJ 50 s Preset:
50s Remaining:
4.7 (b)
I
I
f
f t
I
II
, ,
t
i
,
I
I
P
I
.,.
< .5 FS= 2K MEt'11 :
t
w
Os
...
N 5.660
keV
ch
293~
H
J .Jl
N
:~
I
I
.,~
10.8 > " 1 .0 1f8 cts FS= 2K t'lEt11 : SAt1PLE 1 ~
i
6.100
f(elJ
ch
315=
Figure 4.7 Electron microprobe analysis of mckel SPO membrane surface; (a) untreated surface, and, (b) nickel plated surface following slot size reduction to 4.1 !lm.
11.2 > 28 cts
CHAPTER 4. Results and Discussion (Electroless Nickel Plating)
claims that the screens are produced from 100 per cent nickel. As expected, the coating deposited during ENP is a mixture of nickel and phosphorous.
4.2 ASSESSMENT OF PRE·TREATMENT STAGES Pre-treatment processes were meticulously investigated to see if changes to the substrate preparation could improve the uniformity of the plated pores.
4.2.1 Surface Layer Removal by Hydrochloric Acid Surface layer removal by hydrochloric acid is an important preparation stage for electroless plating. Vigorous chemical attack at the surface strips away the oxide layer to expose a 'fresh' plating surface. Development of the hydrochloric acid pretreatment step benefited from the qualitative assessment of reaction and mass transfer parameters. IS-mm square samples of mckel screen with 17.1 I1m circular pore openings were degreased using a 20 ml solution of SM sodium hydroxide and rinsed in deionised water. Each screen was sonicated for five minutes in IM, 2M, or 8M hydrochloric acid solutions (20 ml) at temperatures up to 60 °C. The metal surface Table 4.5 Experimental conditions and observations during hydrochloric acid pickling of 17.1 I1m nickel CPO screens. Acid strength
HC) solution temperature
Observations
(OC) I
IM
20
2
IM
40
3 4 5
2M 2M 2M
20 40 50
6
2M
60
7
8M 8M 8M 8M
20 40 60 60
8 9 10
No effervescence after 5 minutes Sample sonicated for a further 15 minutes but still no sign of reaction A few bubbles were observed after 30 seconds, however this activity stopped after 90 seconds No Significant sign of reaction No effervescence A few bubbles were seen originatmg from the sample, but this ceased after a short time A few bubbles were observed instantly on immersion, but no sign of activity after 60 seconds No effervescence No effervescence No effervescence No effervescence (shinv side roughened) 4-14
CHAPTER 4. Results and Discussion (Electroless Nickel Plating) was studied for signs of reaction With the acid; observations are summarised in Table 4.5. There were few visible signs of chemical oxidation on the sample surface for the range of hydrochloric acid solutions considered. Even the combination of a rough nickel surface in the most severe conditions did not appear sufficiently corrosive for surface layer removal. Stainless steel mesh on the other hand, seemed to corrode under identical conditions; the acid solution colour changed to pale green. The aggressive conditions of 8M hydrochloric acid, heated to 50°C were duly adopted from the assessment although the contact time was raised to 15 minutes. Vacuum degassing was later introduced to maximise the contact area for the acid reaction by de-aerating the pores. 4.2.2 Plating Surface Activation using Nickel Electroplating
4.2.2.1 No Surface Activation (control experiments) Nickel SPO screens were briefly prepared before immersion in (fresh) 650 ml plating solution for 60 minutes, with air sparge agitation. Neither surface layer removal by hydrochloric acid nor surface activation by nickel electroplating were adopted in the pre-treatment. Plating conditions and initial observations around the sample surface are described in Table 4.6. The absence of any immediate reaction on entry to the plating bath was not completely unexpected. Pore size analysis of the few experiments performed suggested that there was random plating in the absence of surface treatment. The pore dimensions between consecutive deposition experiments 109 and 110 were reduced to 1.9 ± 0.3 J.1m and 8.7 ± 0.4 J.1ffi, respectively. Table 4.6 Experimental conditions and observations for ENP of 13.4 J.1ffi nickel SPO screens without surface activation or surface later removal. Plating time (mins) 60 60 60
Temp
60 60
80.3 80.0
Observations
Expt. ref.
No sign of bubbles mltiallv. Difficult to observe at end. No significant reaction observed before AS No reaction seen in first 60s. No obvIous Sign of effervescence at end. No effervescence in 60s. Few bubbles witnessed at end No sign of bubbles m 60s. Bubbles seen at end.
106 107 108
(0C)
79.3 79.8 80.2
109 110
4-15
CHAPTER 4. Results and Discussion (Electroless Nickel Plating)
4.2.2.2 Electroplating of Nickel SPO Screen Surface activation represents an important stage in the preparation of nickel SPO screens. The oxide layer is usually corroded by a vigorous reaction in warm concentrated acid solution. This process coupled with the deposit of an alkaline film (sodium carbonate treatment) offers the substrate an excellent start in the plating bath. Although hydrochloric acid pickling is favoured for its relative simplIcity, nickel electroplating is an equally valid process. Here catalytic nickel ions are electroplated on to the top surface layer of the nickel screen material. The hydrochloric acid stage in the pre-treatment schedule was substituted for nickel electroplating. Fresh 600 ml nickel chloride electroplating solution and two new nickel electrodes were prepared. o2
2 A dm
current density was used. Pre-treated Nickel SPO screen samples were
contacted for 60 minutes with a fresh 650 ml electroless plating solution, in the presence of air sparge agitation (Experiment references III - lIS). Operating parameters,
observations
and
pore
size
measurements
are
reported
in
Table 4.7. Table 4.7 Experimental conditions, observations, and pore size analysis for ENP of 13.4 Itrn nickel SPO screens (nickel electroplating pre-treatment). Plating time (mins) 60
Temp (0C) 80.0
Slot width (urn) 1.0 ± 0.2
60
80.0
30±0.2
60
80.0
6.9 ± 0.3
60 60
80.0 80.0
96±0.4 7.7±O.2
Slot length (urn) 59.5 ± 14.1
Observations
Surface effervesced vigorously, instantly after addition and at end 241.0 ± 141.1 Surface effervesced vigorously on addition 4020±224 Surface reactIOn nnmedtate. More vigorous than other experiments 374.3 ±7.2 Sample effervesced immediately 3906+100 Slgmficant reactIOn on surface
Expt. ref. III
112 113 114 115
Pore size analysis was supported by optical microscope measurements. As seen, the average slot widths varied randomly after plating, i.e. pore sizes observed between 1.0 !Lm and 9.6 !Lm. In contrast to hydrochloric acid treatment, the plating surface effervesced immediately on addition to the plating bath after electroplating. Although it does not necessarily follow that instant mltiatIOn in the platmg bath is indicative of a favourable plating outcome. Indeed, comparison of plating performance between
4-16
CHAPTER 4. Results and Discussion (Electroless Nickel Plating) hydrochloric acid and electroplating pre-treatments favours the fonner where pore size reductions of 12.S I1m - 14.6 Ilffi, were more consistent during the first five comparable runs. It is important to note however that the bath loading in these experiments was three times greater than for experiments descnbed in Table 4.7. As was the case for most nickel deposition of nickel SPO screens the final slot length varied widely between sinular experiments.
4.2.2.3 Electroplating (with Polarity Reversal) of Nickel SPO Screen
Electroplating with polarity reversal specified for stainless steel activation (see Section 3.3.2) was evaluated for nickel SPO media. Electroplating commenced anodically (current denSity of 2 A dm·2) in an existing nickel chloride solution utilising two electrodes. During this stage positive metal ions were stripped from the sample anode leaving a clean surface on which to deposit nickel. After one minute, the electrode polarity was reversed and electroplating continued for 10 minutes using the same current settings. Nickel samples were immersed in the 'new' electroless plating solution for 40 minutes, with agitation by air injection. Samples were studied closely on entry to the electroless plating bath. Experimental conditions and corresponding results are set out in Table 4.S. Table 4.8 Expenmental conditions, observations, and pore size analysis for ENP of 13.4l1m nickel SPO screens (electroplating pre-treatment with polarity reversal). Plating time
Temp
Slot width
Slot length
(mins) 40
(OC)
(urn)
(urn)
SO.S
3.3 ±0.3
257.9±56.2
40
SO.O
6.3 ±0.3
343.1 ±24.0
40
79.5
7.0+0.3
381.1 + 3.5
Observations
Expt. ref.
Sample effervesced vigorously (most to date). Still reacting at end. Sample effervesced vigorously from start Effervesced immediately
116
117 118
The nickel SPO screen reacted vigorously immediately after entry to the plating solution. Pore sizes were reduced to between 3.3 I1m and 7.0 Ilffi in diameter following 40 minutes plating time. Slot length measurements also varied greatly across the plated surface and lacked consistency between repeated experiments. 4-17
CHAPTER 4. Results and Discussion (Electroless Nickel Plating) Although the average slot width appeared to increase with plating throughput the small sample population made inference of a trend uncertain.
4.2.2.4 Electroplating of Stainless Steel Mesh
Prior to electroless nickel plating the stainless steel mesh experienced hydrochloric acid pickling and catalytic activation by electroplating. An overhead stirrer operated at low speed to limit damage to the sample provided solution movement in the electroless plating bath. Table 4.9 summarises the pre-treatment and electroless plating operating parameters. Table 4.9 Experimental conditions for EP and ENP of staInless steel mesh (29 IJlI1 square pores). Electroplating Current Voltage Plating time Temp Expt. time ref. (0C) (mins) (mA) (V) (mins) none nla nla 20 80.0 10 15 231 0.75 40 85.9 13 15 231 0.73 60 86.2 15 15 230 0.69 60 85.8 20
A distinct change in solution colour was indicative of stainless steel mesh corrosion in warm hydrochloric acid. The lightweight sample effervesced during electroplating, sometimes curling towards the anode. Standard nickel electroplating pre-treatment is detailed in Table 4.10, on the following page. A distinct line observed on the sample after electroless plating highlighted the electroplating coverage. A white precipitate formed immediately afterwards during contact with 2 % w/w sodium carbonate solution. There was no visible sign of reaction in the electroless plating bath during the 20 minute plating experiment (Table 4.10); a microscope study confirmed there had been minimal nickel deposition. After 40 and 60 minutes plating time the square pore openings were measured as 11 IJlI1 and 13
~m
respectively. The extrinsically
catalytic nature and mechanical properties of the stainless steel mesh made successful surface preparation and consistently uniform nickel deposition difficult to achieve. The interwoven structure formed from narrow stainless steel strands offered an uneven surface for nickel deposition. The thin mesh-like material also lacked rigidity causing operational problems for both electroplating and electroless deposition stages. 4-18
CHAPTER 4. Results and Discussion (Electroless Nickel Plating) Table 4.10 Observations and pore size analysis for EP and ENP of stainless steel mesh (29 !llTI square pores). Pre-treatment Observations
Pore width
Electroless plating Observations
Expt ref.
No bubbles observed. Steel mesh did not react in platmg bath. Microscope study indicated unsuccessful plating Small bubbles observed on surface. Stirrer speed set low to prevent damage to sample. Distinct mark from electroplatmg solution level.
10
(urn)
HCl turned green
no analysis
Surface covered in bubbles during electroplating. Cathode pushed away from anode. White precipitate after Na2C03. Small bubbles on mesh during electroplating. White nrecIDitate after Na2C03. Mesh curled towards anode in electroplating solution.
11.0 ± 1.0
13
15
13.0± 1.2 "
no analysis
Mesh did not plate well. Thin stnp across mesh where no plating has occurred.
20
4.2.3 Alkaline Sodium Carbonate Coating Sodium carbonate treatment immediately before entry to the electroless plating bath establishes a desirable alkaline film on the surface, which can improve the adhesion of the deposit (PMD (UK) Ltd., 1993). The role of sodium carbonate within the pretreatment schedule was evaluated for nickel CPO screens having 17.1 !llTI pores. In the absence of sodium carbonate treatment the sample was briefly contacted with a cold 'buffer' plating solution. An existing electroless nickel plating solution was made up to 1800 ml with 400 ml fresh solution. Plating solution was pumped up through the nickel CPO screen, supported in a membrane holder, using high capacity (HC) and low capacity (LC) peristaltic pumps. Experimental conditions and relevant observations are outlined in Table 4.11, on the following page. Similar results were achieved regardless of sodium carbonate inclusion (Experiment references 44 and 45). The plated surface was noticeably darker and pore size reduction seemed to be slightly less successful following alkaline pre-treatment. Any useful study of parameters such as deposit coverage, plating uniformity and quality was not possible from such a small population of experiments.
4-19
CHAPTER 4. Results and Discussion (Electroless Nickel Plating) Table 4.11 Experimental conditions and observations for ENP of 17.1 lUll nickel CPO screens (sodIUm carbonate treatment evaluation).
Description Plating Temp time (mins) ("C) no Na2C03 60 83.2
Pump
Flow rate (m1 min·I )
Observations
Expt. ref.
HC
43
Pressure remained low throughout experiment, i e. -0.06 mbar (suggesting suctIOn). Flow reversed every 5 minutes Pressure did not rise above 0.06 mbar. Platmg result to similar experiment 44, although darker surface. Pore width reduced slightly to - 15 !lm Area behind o-ring plated Sample less tarnished than usual. Pore Size reduced to 18 !lm. Area behmd o-nng plated
44
Na2C03
60
85.0
HC
43
no Na2C03
120
850
LC
18
Na2C03
110
84.5
LC
18
45
48
49
The appropriateness of a sodium carbonate film on the substrate surface in assisting nickel deposition onto stainless steel mesh was also investigated. Each sample was subjected to a typical pre-treatment schedule including electroplating before passage to the electroless nickel bath. The experimental environment is outlined in Table 4.12. Table 4.12 Experimental conditions for EP and ENP of stainless steel mesh with square pores of 29 lUll (sodium carbonate pre-treatment evaluation). Description
EP time (mins)
no Na2C03 no Na2C03 dID in Na2C03 standard Na2C03 standard Na2C03 10 mins m Na2C03 15 mInutes Na2C03
15 15
20 230
0.77
15 15 15 15
231 231 230 230
075 0.73 0.72 0.7
Current Voltage ENPtime Temp Expt. ref. (mA) (V) (mins) COC) 60 60 60 40 60 60 60
85.9 850 852 85.9 862 85.2 85.1
23 28 25 13 15 26 30
4-20
CHAPTER 4. Results and DiscussIOn (Electroless Nickel Plating) In the absence of sodium carbonate treatment the sample was briefly contacted with a cold buffer plating solution. Agitation provided by an overhead stirrer was restricted to protect sample integrity. Observations during pre-treatment and electroless plating. together with pore size evaluation are presented in Table 4.13. A white precipitate formed immediately in the 2 % w/w sodium carbonate solution. Omission of sodium carbonate pre-treatment did not appear to affect the plating result. Pore size determination by optical microscope analysis suggested average pore sizes of 8 to 9 llm without sodium carbonate treatment. The surface was comparatively firm and dark in colour. Results from a single experiment incorporating a quick dlp in alkalIne solution reported a final pore size of 14 Jlm. Sample preparation using sodium carbonate led to a reduction in pore size of between 11 Jlm and I3 llm after electroless plating. The adoption of longer immersion times in the electroless solution coincided with a drop in depOSition performance. The sample did not appear to be damaged despite deposition of a coating of silver colour on the vessel sides and stirrer.
Table 4.13 Experimental observations and pore size analysis for EP and ENP of staInless steel mesh (sodium carbonate pre-treatment evaluation). Pre-treatment Observations
Pore width
Electroless plating Observations
Expt. ref.
Many tiny bubbles m platmg bath surrounding sample. Mesh reacting at end Surface was hard and dark.
23
90+0.9 , , " ,,;&, ' N,C :',' 14.0 ± 1.2 '''!;' "~ Surface covered m bubbles 11.0± 1.0 Small bubbles observed on surface. during electroplatmg. Cathode SUrrer speed set low to prevent pushed away from anode White damage to sample. Distinct mark precipitate after Na2C03. from electroplatmg solution level. , Small bubbles on mesh dunng l3.0± 1.2 • electroplating. White precipitate , after Na2C03. White precipitate formed after no Sample did not plate well. Bottom Na2C03 analysis and sides of plating bath coated in metal. White precipitate formed no Platmg solution full of bubbles. immediately. analysis Stirrer covered in silver metal. Bottom of plating bath and Sides coated too. Sample undamaged.
28 25
(um)
Small bubbles sample.
nsing
from
"
I
8.0±0.9
'
"
"
",'
.
,
,
,
13
15
.
26
30
4-21
----------------------------
--
CHAPTER 4. Results and Discussion (Electroless Nickel Plating)
4.3 ASSESSMENT OF PLATING BATH LIFE Electroless nickel plating baths have a finite active life, which is influenced by metal loading, bath temperature and the presence of impurities. A typical Ni-Shield M solution, If operated at specified conditions, is usually exhausted after between six and eight turnovers (PMD (UK) Ltd., 1993). The life of a standard plating solution was evaluated from the plating performance of nine sequential plating experiments. A fresh 1900 ml electroless plating solution prepared from the Ni-Shield M solution and distilled water was evaluated using pre-treated nickel SPO screen samples. The plating solution was filtered using Whatman 0.8 llm cellulose nitrate membranes whilst the temperature was raised to the desired value (80°C). Consecutive samples were treated in the plating bath for 60 minutes in the presence of air sparge agitation. The plating bath was allowed to cool below the activation temperature whenever the plating bath was not required. The sample surface could be studied in the first 60 seconds before air-based agitation commenced. Care was taken to ensure the reproducibility of all process activities. Slot dimensions were analysed using optical microscopy and, as seen in Table 4.14 on page 4-23, there was little noticeable reduction of the 13.4 llm slot width by nickel deposition during the initial five plating experiments (Experiment references 85 - 89). The average slot length once again varied greatly between experiments. On completion of the ninth and final plating experiment, and after 1.7 dm2 plating surface had passed through the bath, the average slot width was restricted to 6 llffi. This increase in plating rate was further supported by visual inspection of an mstantaneous surface reaction occurring in the plating bath following the fifth experiment. Such observations are in contrast to those reported prior to experiment six where no initiation was witnessed in the first 60 seconds. From the results available it appears that the solution became more active with throughput, first showing noticeable signs of plating after approximately five plating runs of 0.2 dm2
r1
bath loadmg. This
finding contradicts the manufacturer's information, which suggests that the plating rate of nickel solution is at its optimum initially before settling to down to a consistent rate.
4-22
~--------------------CHAPTER 4. Results and Discussion (Electroless Nickel Plating) Table 4.14 Experimental conditions, observations, and pore size analysis of ENP with Dlckel SPO screens (plating solutIOn life evaluation). Temp
Slot width
Slot length
("C)
1
Plating time (mins) 60
79.5
(Ilm) 13.4 ± 0.3
392.5 ±5.1
2
60
79.9
13.3 ±0.2
3860± 1.5
3
60
800
12.8±0.2
396.3±2.1
146±0.3
406.8±5.9
No.
4
Observations
Expt. ref.
1900 m1 fresh solution. Water vacuum used during pre-lreatment. No sign of bubbles on sample after 30 seconds. Sample did not effervesce No bubbles seen before AS (30 s) IDtroduced Electroless bath held at 80 C for 60 minutes before ext. Air sparge knocked slightly at angle. No effervescence in first 40s before AS No effervescence in first 40s before AS No effervescence ID first 40s Effervescence after lOs ID bath Effervescence after lOs ID bath Surface reaction observed Immediately Sample effervesced Immediately
85
(J.lm)
5 6
60 60
79.9 79.9
136+10 no analysIs
376.8±7.9 no analysis
7
60
80.0
no analysIs
no analysis
8
60
80.3
no analysis
no analysis
9
60
80.0
6.1 ±0.8
393.4±5.6
86
87 88 89 90 91 92 93
4.4 CONCLUSIONS Three different solution agitation methods were evaluated for their effectiveness in producing consistently uniform pores. Circulation of plating solution through the substrate pores delivers fresh plating solution to each pore and has the potential for monitoring the pore size by measuring pressure drop across the membrane. It was found that for flat sheet membranes the hydrogen generated by the plating process could not be satisfactorily vented. The plating results for the mechanical agitation based system were equally poor. Despite air injection providing the most effective method of solution agitation, the electroless nickel plating process produced nonuniform membrane pores. Microfilters produced in this way would clearly be unsuitable for filtering particles of a consistent size.
4-23
--
CHAPTER 4. Results and Discussion (Electroless Nickel Plating) The expenmental investigation of deposition rate with plating bath life proved inconclusive. Development work carried out by colleagues in the Department of Chemical Engineering, Loughborough University, has recently found that a much more vigorous pre-treatment is reqUIred in conjunction with sonication during electroless nickel plating. The pre-treatment and plating schedules for this enhanced plating procedure are presented in Appendix B. Nickel slotted and circular pore microfilters produced USIng this optimised plating process were tested in constant-rate filtration experiments described in Chapters Seven and Eight of this thesis.
4-24
CHAPTER 5. Literature Review (Surface Microfiltration)
CHAPTER FIVE LITERATURE REVIEW: SURFACE MICROFlLTRATION 5.1 MEMBRANE PROCESSES - Membrane processes include reverse osmosis (RO), nanofiltration (NF), ultrafiltration (UF) and microfiltration (MF). These are pressure driven using a semi-permeable
membrane for filtration of particles and solutes and they are normally characterised by the same parameters: water flux and rejection. Separation may be based on sieving mechanisms: whereby particles bigger than the pore diameter of the membrane will be retained by the membrane, except for NF and RO where exclusion and diffusion mechanisms are more appropnate. The main differences between these processes are the pore size of the filter, the species size rejected and driving force applied for each process (Wiesner et aI., 1992), which can be seen in Table 5.1. Table 5.1 FiltratIOn processes and their characteristics, in Droste (1997) and separately from Slrkar et al. (1999). Membrane process Microfiltration Ultrafiltration Nanofiltration Reverse osmosis
Species separated
Size range (microns) 0.1 - 10
Driving force
Suspended particles, Pressure difference emulsions (0.03 - 0.3 MPa) Macromolecules, 0.001 - 0.1 Pressure difference emulsions (0.2 - 0.7 MPa) 0.0004 - 0.006 Pressure difference Dissolved salts (0.4 - 2 MPa) Dissolved salts 00001-0001 Pressure difference (2.7 - 14 MPa)
In terms of filtration rate (flux) which is normally reported
ID
volume of filtrate per
effective filter area per hour Cl m·2 h-'), among these three processes, MF produces the highest flux, typically higher than 100 1 m-2 h'" followed by UF with less than 100 I m·2 h- I and RO with less than 10 I m·2 h· l . According to a commercial report compiled in 1998 (Nunes et al., 200 1) the microfiltration sector contnbuted 50 per cent of the overall membrane separation market in Europe. This is followed by UF: 23 per cent; RO: 15.2 per cent; and
5-1
CHAPTER 5. Literature Review (Surface Microfiltration) electrodialysis: 11.4 per cent. It was estimated that the membrane separation market would expand WIth strong growth as quality standards become more stringent, particularly in the pharmaceutical industries.
5.2 GENERAL APPLICATIONS Microfiltration processing is widely used in the Pharmaceutical industry where stringent regulations require consistently safe and stenle product manufacture. A 0.22 !llll rated membrane is typically selected to remove particles and importantly all viable bacteria. Applications in this field include cell and bacterial harvesting, treatment of antIbiotics and viral containing solutions, and stenlIsation of valuable biological molecules, DNA and tissue culture medium (Belfort et al., 1994). Microfilters also provide an essential polishing duty in the electronics industry where a 0.1 !llll pore size membrane is installed as a final point-of-use filter for ultra pure water. The food industry uses microfiltration technology in clarification and sterilisatIOn of wine, beer and juice. A 1.0 !lm rated depth filter can remove yeast cells and common bacteria during sterilisation operations. Other notable applIcations of microfiltration include latex product recovery and concentration, oil-water separations, recycling of pigment dispersions, wastewater treatment, textile finishing effluents, and thickening of radioactive contamInated solids. Hydrophobic microporous membranes are used in many waterproof breathable fabrics. 5.2.1 Removal of Cryptosporidium Parvum Oocysts from Drinking Water Future applications will largely depend on the development of membrane and module technology to offer greater operating characteristics. A potential market of great significance, which may benefit from this, is the production of drinking water (Jacangelo et aI., 1997; Hillis, 1997; and Owen et al., 1995). The evolution of stringent water quality regulations for removal of particulates and more recently Cryptosporidium Parvum oocysts and Giardia Lamblia has increased the drive to
develop suitable treatment processes. The filtration of parasites from potable water could become a major application for InlcrofiltratIon. Such objects are particulate in
5-2
CHAPTER 5. Literature Review (Surface Microfiltration) nature and have sizes greater than 2 Ilm. An example of two oocysts is shown in Figure 5.1 , where they have been filtered on a Nuclepore track-etched membrane.
Figure 5.1 Two Cryptosporidium Parvum oocysts on a Nuclepore
membrane filter (note the scale bar), in Rushton et al. (2000).
Cryptosporidium Parvum has received considerable attention in recent years because
of widespread cases reported, high-profile outbreaks, and the life-threatening risk to the growing population of people with compromised immune systems. There are numerous publications relating to the parasite (Adin et al., 2000) and an increasing amount of relevant information available on the World Wide Web. Cryptosporidium Parvum, a protozoan parasite, is usually transmitted by drinking water and can cause
gastrointestinal illness of varying severity. The Cryptosporidiosis infection is selflimiting in healthy adults but can be serious or even life threatening to infants, the elderly, pregnant women and people with compromised immune systems (e.g. persons with HIV/AIDS, cancer patients, and transplant patients). The situation is exacerbated by the fact that there is no anti-microbial treatment for the parasite at present (Fricker and Clancy, 1998). Cryptosporidium Parvum can survive for extended periods in the natural environment as an oocyst. Oocysts are nearly spherical in shape with a diameter of 4 to 8 Ilm (Roessler, 1998). Giardia Lamblia cysts in comparison are larger egg-shaped particles about 9 Ilm wide and 12 Ilm long.
5-3
CHAPTER 5. Literature Review (Surface Microfiltration) The world's worst outbreak of Cryptosporidiosis occurred in 1993 when at least 400,000 people in MIIwaukee became ill after drinking municipal water contaminated with Cryptosporidium Parvum (Friedman-Huffman and Rose, 1998). A total of 54 deaths were attributed to this outbreak, primarily involving people with weakened immune systems. In 1998 an outbreak in Sydney, Australia affected 3.7 million residents who were required to boil drinking water following high levels in drinking water supply (MacCormick,
1999). Oocysts are considered significant at
concentrations of 1 per litre or less (Gregory, 2000). It has been suggested that the infectious dose 50 (ID50) in humans is around 132 oocysts (Fricker and Clancy, 1998). Cryptosporidium Parvum oocysts are extraordinarily resistant to normal means of
disinfection and chlorination in drinking water, with 8000 - 16000 mg rl chlorine treatment (Roessler, 1998) required to inactivate Cryptosporidium Parvum oocysts (levels some ten times greater than for Giardia). This level of chlorination well exceeds the level that is practical in potable water (0.5 mg rl). The principal means for control of CryptosporidlUm Parvum is therefore through physical-chemical treatment processes in conjunction with inactivation processes such as ozone and chlorine dioxide (Ozekin and Westerhoff, 1998). Some more recently developed techniques include ozonation, activated carbon, ion exchange, reverse osmosis and ultraviolet radiation. Crossflow microfiltration could be partIcularly suited for removal of oocysts as they are particulate in nature, although any commercial system would have to treat large quantities of water.
5.3 DEAD-END AND CROSSFLOW FILTRATION 5.3.1 Dead-end Filtration Conventional dead-end filtration involves the flow of suspension perpendicular to a se1TI1-permeable medium under applied pressure. The particles in the feed suspension, which are sufficiently large not to pass through, accumulate on the membrane surface as a cake layer or in its mterior leading to an increased resistance to filtration. The cake itself acts as a filter for the remaining suspension, and grows continuously with time. This cake growth leads to a decreasing filtrate flux and may result in unacceptably low fluxes. Dead-end filtration is by nature a batch process and must be stopped periodically to remove particles or replace the filter medIUm. In a few
5-4
CHAPTER 5. Literature Review (Surface Microfiltration) filtration processes, e.g. a rotary drum filter, continuous solids removal is used which can facilitate a continuous operation. Dead-end mode of operatIOn is not generally appropriate for separation of finer particles and colloids from more dilute suspensions. Plugging of the filter medium can occur rapidly after the start of filtration with a subsequent increase in the medium resistance and low fluxes. Generally, excellent rejection of particles can be achieved by dead-end filtration, because the cake accumwated on the surface of the membrane acts as a secondary filter medium, which can capture smaller particles, but it requires frequent cleaning of the membrane. However, as the filtration progresses, a concentrated layer at the membrane surface may continuously bUild up and this will slow down the movement of fluid through the membrane, resulting in flux decline and finally this will stop the fluid flow entirely (Davis, 1992a). Crossflow filtration was developed in an attempt to overcome difficulties expenenced in the dead-end mode of operation. 5.3.2 Crossflow Filtration
In crossflow filtration the feed stream flows parallel to the membrane surface to
impose a high shear on the surface. The tangential feed stream separates into two streams: the 'penneate' stream and the 'retentate' stream. Figure 5.2 on the following page, is a diagrammatic representation of the cross flow filtration process. The penneate stream is equivalent to the filtrate in conventional filtration and the retentate descnbes the residual suspension from the filtration. The velocity of the penneate stream through the membrane is generally low compared to the crossflow velocity of the suspension. The scounng action of bulk tangential flow and low penneation flow through the membrane reduces the tendency for particles to accumulate at the membrane surface, and so maintain penneate flux at a much higher level than for dead·end filtration. The penneate filtration rate is referred to as the 'flux' which in microfiltration is conveniently reported in units of litres per square metre of membrane per hour (l m·2 hot). This rate is eqUivalent to the superficial liquid velocity through the membrane. The common mode of operation utilised by industry and large-scale laboratory rigs is to pump the suspension to be filtered through a narrow tube or channel havmg microporous membrane walls. The relative flow rates of penneate and retentate are establtshed by controlling the backpressure on the retentate
5-5
CHAPTER 5. Literature Review (Surface Microfiltration) to establish the pressure drop across the filtration surface. Typical velocities and 02
pressures employed are 1 - 8 m sol and 0.1 - 0.5 MN m respectively.
Penneate (lower pressure)
11111111111111
___ Membrane
Penneate (lower pressure)
Figure 5.2 Schematic representation of the cross-flow filtration process, in Rushton et al. (2000).
Since the growth of cake is limited by the shear stress acting on the surface of the filter medium, crossflow filtration has the advantages of thinner cake deposition, higher filtration flux, and a continuous mode of operation (ConnelI et al., 1999). The major advantages of crossflow filtration are summarised as follows:
•
Particle accumulation at the surface filter is minimised;
•
Filtration performance does not depend on particle size since it is assumed that a filtration barrier can retain the particles;
•
Feed additives or filter aid such as flocculating agents are not required; and
•
Retentate or products are not contaminated with filter aid.
5-6
CHAPTER 5. Literature Review (Surface Microfiltration) A drawback of crossflow microfiltration is the incurrence of additional pumping costs for the circulation of feed suspension over the membrane surface. 5.3.3 Stirred-cell Operation Crossflow microfiltration can be mimicked on a small-scale using a batch stirred-cell. Indeed, the first recognised crossflow experiments performed by Bechold in 1907 were carried out using a stirred-cell. In this filtration mode a stirrer operates slightly above the membrane, providing shear across the membrane surface, see Figure 5.3. Stirred-cell operation, also referred to as 'pseudo-crossflow', is often used as a starting point to investigate the behaviour of a particular membrane-suspension system. Operation of the stirred-cell is simple enabling precise control of experimental parameters in a compact system with relatively small quantities of feed suspension, and membrane area. Further experimental work can then be carried out using a traditional crossflow experimental rig where process control is considerably more demanding.
Membrane ~
~ Penneale
~
Figure 5.3 Schematic representation of stirred-cell filtration,
in Rushton et al. (2000).
5-7
CHAPTER 5. Literature Review (Surface Microfiltration) 5.4 MEMBRANE FOULING For solid-liquid separation using crossflow microfiltration, results show similar general characteristics in terms of flux profiles. At the initial stage of the filtratIOn process, there is a sharp decline of flux compared to the initial clean water flux. After this initial sharp fall, the rate of decline reduces until an equilibrium flux is aclueved (Mir et al., 1992). This situation reflects that there is some degree of fouling of the filter with a high fouling rate at the beginning of the filtration and an equilibrium fouling layer thickness may be reached where no further growth of the thickness of the fouling layer occurs later. Particles are transported to the membrane surface by the permeate flow and may deposit on the surface and within the porous structure of the membrane. There are a number of different fouling mechanisms involved and these are illustrated in Figure 5.4.
Crossflow
I
Concentration "'o'o,,u° polarisation ·,,'r'1n""6.-": resistance 00
~~~~~~
Membran~
resistance Pore adsorption resistance
or cake resistance Pore blocking resistance
Figure 5.4 Different filtration resistances experienced during microfiltration, in Rushton et al. (2000).
In the literature, two mechanisms of flux decline have been explained, which are due
to internal and external fouling of the membrane (Tarleton and Wakeman, 1993). For internal fouling, the rapId depOSItIon and capture of finer particles from the feed stream and their intrusion into the pores occurs. The particles may adsorb onto the pore walls or block the pore or precipitate within the filter. Particles entering the pores
5-8
CHAPTER 5. literature Review (Surface Microfiltration) can deposit in the pore channel and block the pore from further flow. Increased flow through remruning pore channels increases the likelihood of further pore blockmg. Small particles may adsorb on the walls reducing the pore channel dimensions. Altering the composition of the feed strerun, pre-treating the medium, or selecting a non-adsorbing filter medium usually controls adsorption. Internal fouling is generally a problem for microporous membranes with significant depth and open pore structures. If the membrane acts as a true surface filter the internal foulmg should be negligible and the membrane resistance should equate to the clean resistance. The external fouling IS caused by cake formation resulting in further deposition of solids on the membrane surface and subsequently further solid layers usually called a 'dynruruc' or 'secondary' membrane. During filtration an accumulation of particles generally occurs near the surface of the membrane due to convection of fluid through the membrane pores. The particle concentration then increases significantly above that of the feed. This effect known as 'concentration polarisation', typically results in the deposition of particles at the membrane, either as a fouling layer, or if the concentration at the surface is sufficiently high, as a filter cake. The cake layer resistance is often compared to the 'gel layer' formed m ultrafiltration systems by deposition of polymer and fine species at low concentrations. For each fouling mechanism described a resistance to permeate flow is produced. In general these resistances are irreversible with the exception of concentration polarisation. The clean membrane has a natural resistance to flow, which depends on the membrane thickness, its nominal pore size, and various morphological features such as tOrtUOSity, porosity and pore-size distributIOn. Clean water membrane resistance is measured using a clean water membrane flow test. Darcy's Law can be used to calculate the permeate flux for applied transmembrane pressure (Field et al., 1995):
(5.1)
where, R~
= R" + R,
5-9
CHAPTER 5. LIterature Review (Surface MicrofiltratIOn) J is the permeate flux, V is the total volume of permeate, A is the external membrane
surface area, t is the filtration time, !lP is the transmembrane pressure drop (TMP) and
Jl is the permeate viscosity. Rm refers to the clean membrane hydraulic resistance and
R~ is the sum of an irreversible term, R", corresponding to fouling which is not removed when pressure is released, and of a reversible term, Rr. including reversible fouling and concentration polarisation. The foulmg behaviour of the membrane depends on various factors such as process parameters (i e. pressure), pore shape, particle shape and membrane morphology which have been studied extensively by various researchers such as Persson & Nilsson (1991), Davis (1992b), Belfort et al. (1994) and Tarleton & Wakeman (1993, 1994a, 1994b). The net effect of fouling during membrane filtration is that the permeate flux will diminish and in some cases it may become too low to be economically viable. Flux decay wIth time is illustrated in FIgure 5.5.
Pseudo-eqUilibrium filtratIOn flux rate deposit depth stabilised
___
J
e
Flux decay with increasing resistance
Filtration time, S Figure 5.5 Permeate flux decay to a pseudo-steady state, and
WIth continuing fouling, in Rushton et al. (2000).
Membrane fouling can be reduced by several techniques such as back flushing of the membrane, VIbrating the membrane, and operating at a sub-critical condition to avoid
5-10
CHAPTER 5. Literature Review (Surface MicrofiltratIon) cake formation (Fane, 1996). Cleaning of the membrane by chemicals has also been reported in the literature and is in common commercial use. Chemicals are used to dissolve or remove a species adhering to the membrane, which can reduce the fouling. Proper selection of the membrane can also reduce the problem of fouling. According to Holdich et al. (1998), membranes that contain uniform pores, which are not connected to each other and pass directly from one side to another side of membrane can be used to reduce internal clogging/fouling. A 'surface filter', such as this, does not require a tortuous flow channel to achieve its pore size rating. Considerable research effort has been directed towards techniques which can avoid or reduce the rate of permeate flux decline (severity of membrane fouling) other than by increasing crossflow velocIty. Three major approaches can be identified: hydrodynamic (changmg flow regime across the membrane surface), surface modifications (changing surface/foulant affinity), and regular cleaning (Wenten, 1995; Ma et al., 2001; Wakeman and Williams, 2002). These include: Electric fields (Wakeman and Tarleton, 1986; Bowen and Sabuni, 1992; Wakeman and Sabri, 1996; Chuang and Hsu, 1998; Weigart et al., 1999); Ultrasonic fields (Matsumoto et al.,
1996); Combined ultrasonic and electric fields (Wakeman and Tarleton, 1991); Oscillatory or pulsatile flow of the feed stream (Howell et al., 1993; Li et al., 1998); Abrasive particulate material addition; periodic reverse filtration of permeate back through membrane backflushing (Mugnier et al., 2000; Kuberkar and Davis, 2001), and backpulsing (Redkar et al., 1996; Mores et al., 2000; Sondhi et aI., 2000; Heran and Elmaleh, 2001; Sondhi and Bhave, 2001); Turbulence promoters (Millward et
al., 1995), helical inserts (Holdich and Zhang, 1992; Holdich et al., 1998), baffles (Wang et al., 1994); Rotating membrane or surface near to membrane (Rushton and Zhang, 1988); and Modification of surface chemistry (Al-Malack and Anderson, 1997; Ma et aI., 2001). Fouled membranes can also be treated wIth cleaning solutions replacmg feed or backflushing streams and sponge-ball cleaning for tubular systems (Bartlett et al., 1995; SiHassen et aI., 1996; Gan et aI., 1999).
5.5 SURFACE AND DEPTH FILTRATION Surface filtration functions mainly by direct interception of particles on the surface filter medium (membrane), where particles larger than the pore size of the membrane
5-11
CHAPTER 5. Literature Review (Surface Microfiltration) are screened at the membrane surface. This screening process prevents the larger particles from entering or passing through the pore opening (Dickenson, 1997) and provides the result that no internal fouling can occur (Davis, 1992a). Thus, using surface filters, until they suffer from mechanical failure, can eliminate the problem of internal fouling. In comparison, depth filtration occurs by both direct interception of particles on the surface of the membrane and by particle retention in the membrane. For this type of filtration, larger particles may be trapped on the surface layer and finer particles trapped in inner layers of the membrane as was reported by Gupta et al. (1995), and Lopez et al. (1995). An example of a depth (membrane) filter is shown in Figure 5 6, together with its pore size distribution obtained from the Coulter Porometer™.
'$.
100
30
80
25
g' -5
" "co ~
~
0
20
60 40
10
">
20
5
3 E
0
0 55
:l
U
~
...."'" 0
15
Co
ia
ri' 0
~
1a
'$.
25
35
4
45
5
~
"E
.0
:l C
";;j
c0
'J f!
~
Pore Size, J.ll11
Figure 5.6 A depth effect microfilter and pore size distribution
(note scale bar on SEM), in Rushton et al. (2000).
Accordmg to the pore size distribution, there are no pores bigger than 5.5 Ilm, yet the SEM clearly shows pores 50 to 60 Ilm in diameter. Thus, the pore rating is only achieved by retaining particles less than 50 to 60 Ilffi within the filter. A true surface filter has pore sizes that are the dimension of its rating; it does not have to rely on depth filtration mechanisms (see Figure 5.7 on the following page). Surface filters can be manufactured using various materials including polymeric materials or metal sheets. For polymeric sheets (e.g. polycarbonate), surface filters 5-12
CHAPTER 5. Literature Review (Surface Microfiltration) may be produced by nuclear bombardment of a polymer sheet followed by chemical etching. This technique was developed by Nuclepore (Rushton et al., 2000) and has since been extended (Yamazaki et al., 1996). Typical thickness of this kind of surface filter is 10 IJ.II1 and an example of this type of filter was used in Figure 5.1, on page 5-3. A method for producing surface filters from metal plate based on electroforming and photo-etching has been developed recently. Stork Veco B.V. of the Netherlands used this technique to produce a surface filter from 100 per cent nickel plate. The perforations of either round or slotted holes are aVaIlable down to 13 IJ.II1 (Website http://www.storkveco.com). Having a substantial thickness of up to 600 Jlm the material can be formed into a tubular filter suitable for crossflow nucrofiltration.
Figure 5.7 Example surface microfilter rated at 2 Jlm, from Kuiper et al. (1998).
Currently, surface microfilters have been used to study the separation of suspensions containing various matter including yeast, blood cells and oil droplet dispersions. It has been reported that stable fluxes can be obtained in the range of 100 1000 I m·2 h· l ; a flux range which is higher than that associated with depth filters (Holdich et al., 1998; Kuiper et al., 1998).
5-13
CHAPrER 5. Uterature Review (Surface Microfiltration) Clearly, a surface filter resembles a very fine sieve and could be used for fractionation, or classification, of liquid based materials provided that the secondary membrane or surface fouling is prevented. Figure 5.7 on the preceding page shows a surface filter with a high open filtering area, which is likely to provide high fluxes and low fouling. This is the goal for any commercial surface filter. However, the method of production of the surface filter illustrated is similar to that of silicon clups used in computers, With 10 by 10 mm dimensions, and it is unlikely that this method would be economically viable for production of industrial scale use filters (Kuiper et al., 1998). An alternative, and cheaper, method of production of surface microfilters is therefore required.
5.6 CRITICAL FLUX There remains some ambiguity as to the exact definition of critical flux (lent). Field et
al. (1995) and Howell (1995) defined critical flux as the flux below which a decline of flux with time does not occur; above it fouling is observed. In the strong form of the hypothesis the pressure difference across the membrane is equivalent to the clean water value at the same permeate flux. Alternatively the weak form describes a situation where there is no progressive fouling but the pressure difference is greater than the clean water value. Metslimuuronen et al., (2002) present a schematic representation of weak and strong forms of critical flux (Figure 1). Neither of these forms of critical flux should be confused with the 'equilibrium flux', as illustrated in Figure 5.5 (on page 5-10), found in membrane filtration which occurs after fouling or by concentration polarisation and is a value of only a few percent of the clean water flux. The critical flux is dependent on the factors that mfluence concentration polarisation such as nature and concentration of the feed, shear rate, and membrane characteristics. For larger particles (ca. > 1 micron), shear-enhanced diffusion and inertial lift effects dominate and fouling is dominated by pore blocking and cake formation (Bacchin et
al., 1995). For the smaller particles (10 - 100 nm), short range interparticle forces including electrostatic and van der Waals forces dominate and coagulation and adsorption dominate fouling (Harmant and Almar, 1998).
5-14
CHAPTER 5. Literature Review (Surface Microfiltration) 5.6.1 Experimental Determination of the Critical Flux The identification of critical flux is difficult to determine for an experimental system because cntical flux is very sensitive to operating parameters and it is often found that a slight change in these conditions can lead to a considerable reduction in permeate rate. Authors have generally studied particle deposition on the membrane surface by mass balance using the direct observation through the membrane (DOTM) method or a pressure profile techmque (Li et aI., 1998 and 2000).
5.6.1.1 Direct Observation Through the Membrane (DOTM) There have been a number of non-invasive techniques used to study concentration polarisation and particle polarisation at the membrane surface: VIlker et al. (1981); McDonogh et al. (1990, 1995); Romero and Davis (1991); MackIey and Sherman (1992); Hodgson et al. (1993); Wakeman (1994); AItmann and Ripperger (1997);
Li et al. (1998, 2000); Mores and Davis (2001). Hodgson et al. (1993) developed a new techmque, termed direct observation through the membrane (DOTM), to investigate membrane-fouling interactions. The DOTM technique provides means of critically examining the nature of deposit-membrane interactions as permeate flux is increased. Their studies involved mounting an optical rmcroscope on the permeate side of a transparent membrane and observing particle deposition on the surface of a transparent membrane from the permeate side in a crossflow microfiltration module during filtration of yeast and resin. Li et al., (1998) conducted filtration tests of yeast, algae and latex beads in the imposed flux mode, so that the flux could be controlled, at, below, or above the critical flux. The fouling was observed through a microscope placed on the permeate side of the module. Mores and Davis (2001) explored direct visual observation (DVO) of yeast deposition and subsequent removal via backwashing
and
single
backpulses
using
micro video
photography
with
microfiltration membranes.
5.6.1.2 Pressure Profile Technique Many authors have used a pressure profile method to evaluate particle deposition on the membrane surface (Chen et al., 1997; Madaeni, 1997; Kwon and Vigneswaran, 1998; Chan and Chen, 2001). Constant flux operation is employed in this experimental approach rather than constant pressure to provide better control of 5-15
CHAPTER 5. LIterature Review (Surface Microfiltration) partIcle transport whilst allowing foulmg to be monitored more accurately through transmembrane pressure (TMP). According to a 'flux-stepping' technique the feed suspension is filtered initially at a low flux followed by slow and controlled flux increase in increments of constant flux until the transmembrane starts to increase wIth time (Kwon and VIgneswaran, 1998). Typical step size increase of 60 I m·2 h- I permeate flux has been reported. At each step, the permeate flux value is maintained constant for a fixed time; 20 - 30 minutes is typIcally considered sufficient to momtor cnttcal flux. When the permeate flux is below the critical value the assocIated transmembrane pressure is low, independent of time (does not change even with extended filtration time) and increases Imearly with the imposed flux. Above the cntical flux, the transmembrane pressure rises rapidly with the imposed flux and is dependent of time. The highest flux for which the transmembrane pressure IS stable is taken as the cntical flux. However, given the considerable gap between each step of permeate flux, the value between highest flux with no increase of transmembrane pressure and lowest flux with pressure increase is taken to represent the critical flux (Kwon and Vigneswaran, 1998). In order to determme whether a strong or weak form of the critical flux has been observed plots of critical flux against transmembrane pressure are made. Below the critical flux, where there is little or no accumulation of deposited material on the membrane, the pressure is directly proportional to permeate flux. If the experimental data points exhibIt the same flux as pure water up to the critical flux this represents the strong fonn. Conversely the weak fonn of critical flux is descnbed by expenmental data falling below this pure water line. 5.6.2 Critical Flux Experimental Results There are many instances in the literature where mvestigattons of cntical flux have been reported, including: Raasch (1987); Turker and Hubble (1987); Atmar and Howell (1989); Chang and Hwang (1994); Bacchin et al. (1995); Chang et al. (1995); FIeld et al. (1995); Howell (1995); Chen et al. (1997); Madaeni (1997); Kwon and Vigneswaran (1998); Li et al. (1998); Defrance and Jaffrin (1999); Fradm and Field
(1999);
G6san-Guiziou
et
al.
(1999);
Huisman
et
al.
(1999);
Madaeni et al. (1999); Wu et al. (1999); Kwon et al. (2000); Li et al. (2000); Madec
5-16
CHAPTER 5. Literature Review (Surface Microfiltration)
et al. (2000); Metsfunuuronen and Nystrom (2000); Silva et al. (2000); Carrere et al. (2001); Chan and Chen (2001); Metsltmuuronen et al. (2002). Field et al. (1995) introduced the critical flux concept based on microfiltration with yeast cell suspensions. Howell (1995) has also presented a review of pertinent data. Both authors compared constant pressure and constant permeate flux experiments, and observed that when experiments were carned out at fixed flux, the transmembrane pressure remained generally constant or moderately increasing, and recommended this procedure over the classical constant pressure method because it should reduce the seventy of the fouling. Li et al. (2000) used a microscope to watch partIcles being deposited on the membrane and observed a critical flux for mass deposition. Li and co-workers found that a modified shear-induced diffusivity model predicted critical fluxes for particles of 5 - 12 microns acceptably but not for smaller particles. Madaeni and co-workers (Madaeni et al., 1999) observed with activated sludge that transmembrane pressure (TMP) was higher for clean water flux even though cake formation was not evident: They observed also that the discrepancy between the colloid and water transmembrane pressure was greatest for larger pore membranes. Chen et al. (1997) using colloidal silica particles obtained little (for microfiltratlOn) or negligtble (for ultrafiltration) hysteresis in flux-pressure profiles below critical flux but significant hysteresis above it due to cake formation. Kwon et al. (2000) filtered a synthetic suspension of spherical polystyrene latex particles with retentive membranes by a constant flux method. They observed that the critical flux increased with increasing particle size and cross-flow velocity and decreased with increasing concentration, while a different pore size of the membrane had no effect on critical flux. Similarly, Madaeni (1997) studied the effects of operating conditions on critical flux for latex feed suspension and hydrophilic membranes. The feed suspension described the weak form of critical flux. The critical flux increased with particle size and was lower for a mixture of particle sizes compared to the corresponding mono-size values. Critical flux was dependent on the nature and concentration of the feed and shear rate, being higher for greater crossflow velocity, lower concentration, and larger particles.
5-17
CHAPTER 5. Literature Review (Surface Microfiltration) 5.7 FUNDAMENTAL MODELS 5.7.1 Concentration Polarisation Model
Membrane
dP
Flux
J
......1 - - -
-----c
b
Figure 5.S A diagrammatic representation of the Film Theory.
Early models for the steady state permeate flux in crossflow microfiltration were based upon the concentratIOn polarisation model, or 'film theory', which had been successfully applied to ultrafiltration of macromolecules (Zydney et al., 1986), see Figure 5.8. When particles are rejected during filtration a thin fouling layer will quickly form (within a few minutes or less) on the membrane surface. A concentration polarisation boundary layer develops adjacent to the fouling layer across which there is a concentration gradient. The particle concentration in this boundary layer decreases rapidly from the edge of the membrane surface to the bulk suspension. At steady state the convection of particles towards the membrane by permeate flow is balanced by Brownian particle diffusion away from the membrane caused by the concentration gradient, and by convection of particles toward the filter exit due to tangential flow of suspension. The Steady state flux is given by Equation 5.2 (Fick's law):
J = _D
dC dy
(5.2)
5-18
CHAPTER 5. Literature Review (Surface Microfiltration) where J is the permeate flux or velocity, C the particle concentration, D the particle diffusion coefficient and dC/dy is the concentration gradient over a differentlal element in the boundary layer. If the diffusion coefficient is assumed constant, then this expression may be integrated over the boundary layer thickness, {j (film thickness) subject to the boundary conditions C=Cw at y=O, and C=Cb at y={j, giving:
J
=Dln(C )= kln(C {j w
w
Cb
Cb
)
(5.3)
k = Dj {j is the mass transfer coefficient between the bulk suspension and membrane
surface, and Cw and Cb are the particle concentration at the wall and in the bulk suspension respectively. If the permeate flux is in excess of the 'critical' value then particles will accumulate on the membrane surface forming a constant-concentration cake or 'gel' layer with a substantial resistance to filtration. The gel concentration, Cg, (constant) replaces membrane the concentration at the wall in Equation 5.3. Use of a suitable mass transfer correlation enables the solution of Equation 5.3 to provide an expression of the form:
J =
kln(~: )
(5.4)
Brownian particle diffusivity, DB, of a particle of radius, rp, can be calculated for macromolecules and other smaller colloids according to the Stokes-Einstein relationship:
(5.5)
5-19
CHAPTER 5. Literature Review (Surface Microfiltration) where kB is the Boltzmann constant (1.381 x 10023 J K"I) and T is the absolute temperature (K). Concentration polarisation analysis appears to work well for macromolecular solutions in ultrafiltration predictmg flux values within 15 to 30 per cent for macromolecular solutions where the diffusion coefficient is fairly large (Porter, 1972;' Bhattachrujee and Datta, 1996, 1997). When applIed to filtration of particulate suspensions, however, where the diffusion coefficient is evaluated by the StokesEinstem formula, the predicted flux from concentration polarisation theory is one or two orders of magnitude too low. This shows that particles go back into bulk solutIOn: with a velocity faster than that of diffusIOn. ThiS has been termed the 'flux paradox' for colloidal suspensions (Green and Belfort, 1980). Two principle explanations, first presented by Blatt et al. (1970) have been offered for this flux discrepancy. The first explanation, favoured by Blatt et al. is that closely packed layers of micron or submicron sized particles at the membrane offer a smaller than expected hydraulic resistance. As described by Kozeny-Carman equation larger colloidal species form much less resistive cakes than the considerably smaller macromolecules. Porter pointed out that this hypothesis is inconsistent with experimental data, which shows, pressure-independent flux in the crossflow ultrafiltration of colloids. Porter and others supported a second mechanism where transport away from the membrane by Brownian particle diffusion is augmented by shear-enhanced phenomena, such as shear-enhanced hydrodynamic
diffusion, inertial
lift and surface transport
mechanisms. 5.7.2 Mass Transfer Coefficients EquatIOn 5 6 is the general mass transfer correlation for transport to a stationary membrane surface in a stirred-cell:
Sh = ARe m Sc'
(56)
Where A, and exponents m and n are constant parameters depending on the cell geometry.
5-20
CHAPTER 5. Literature Review (Surface Microfiltration)
Sh=kb D
rob 2
Il v Sc=-=-
Re=--
pD
V
D
(5.7) to (5.9)
In these equations, k is the mass transfer coefficient, b is an arbitrary characteristic
dimension of cell radius (stirrer radius has been used by some authors), D is the diffusivity, ro the stirring velocity (stirrer speed), and v is the kinematic viscosity (v =
III p). The Sherwood number (Sh) is a measure of the ratio of convective mass
transfer to molecular mass transfer. It can also be considered as the ratio of the channel dimensions to the boundary layer thickness
o.
D
k =-
o
(5.10)
The Reynolds number (Re) is a measure of the ratio of inertia effects to viscous effects and the state of turbulence in the system, and the Schmidt number (Sc), a dlmensionless measure of the ratio of momentum transfer to mass transfer. Theory indicates that the Schmidt number exponent, n is equal to -213 for sufficiently high Schmidt numbers in both laminar and turbulent boundary layers. ThIS has been confmned by experiments (Smith et al., 1968). Smith et al. (1968), and Colton and Smith (1972) reported experimental measurements for the mass transfer coefficient controllIng dIssolution of a stationary, homogeneous disc of benzoic acid (m place of membrane) m a stirred batch dialyser and found the following correlations (Equations 5.11 and 5.12) to be applicable: Laminar boundary layer over the membrane surface:
k b = 0.285 (Re)oss (SC)033 D
8,000 < (Re) < 32000
(5.11)
Turbulent boundary layer over the membrane surface: 5-21
CHAPTER 5. Literature Review (Surface Microfiltration)
kb
D
= 0.0443 (Re)o7s (SC)033
32,000 < (Re) < 82000
(5.12)
The transition from a laminar to turbulent boundary layer occurred at a Reynolds number of approximately 30 000. A turbulent boundary layer above Re'" 32 000 was verified by changing the impeller radius at a constant stirring rate. Mathematical analysis of the convective-diffusion equation by Colton and Smith (1972) for the specified experimental system confirmed the exponent values m and n for laminar and turbulent flow conditions. Smith et al. (1968) proposed correction factors, which can be used to adjust thelT experimentally determined parameter A for use in cells of similar geometry (Iaminar boundary layer):
Sh =
The factor
(9 'IF 8)
(5.13)
A Re"' Sc'
9 accounts for relative membrane permeability, 'IF corrects for the ratio of
(active) membrane radlUs/impeller radius, and 8 corrects for the ratio of axial gap distance/membrane radius. The corrected mass transfer equation is appropriate for the laminar boundary layer only. Table 5.2 Constants in the general mass transfer correlation (Equation 5.6) reported by various authors. Authors Malone and Anderson (1977) Mitchell and Dean (1986) Opong and Zydney (1991) Nicolas et al. (2000) Kaufmann and Leonard (1968)
m
Membrane heterogeneous nnca membranes track-etched polycarbonate
Feed potassium chloride BSA
A 3.75'
0.57'
-
0.537
asymmetric UF polyethersulphone UF membranes cellophane membrane
BSA
0.23 T
-
sugars sugars
0.104 0.105
0.66 0.68
t values corrected for Smith et al. (1968) stirred system.
5-22
CHAPTER 5. Literature Review (Surface Microfiltration) In the application of Equations (5.11) to (5.13) the diffusion constant is a critical
parameter, see Table 5.2. Malone and Anderson (1977) measured the diffusion rates of potassium chloride in aqueous solution, in a diaphragm cell uSing heterogeneous mica membranes. The system geometry and membrane permeability were corrected for Smith et al. (1968). Although the stirring rate dependence was in close agreement with Smith et al., the magnitude of the correlation was considerably greater. Malone and Anderson attnbuted this discrepancy to the contribution of Iow membrane porosity and impeller design to transport phenomena. Mitchell and Dean (1986) in their analysis of BSA rejected by track-etched polycarbonate membranes reported a system-specific Reynolds number exponent similar to values reported for homogeneous and heterogeneous membranes. Nicolas et al. (2000) determined the constants A and m experimentally for their system. Their results found good agreement with the diffusion coefficient specified by Kaufmann and Leonard (1968) for a sugar solute/membrane system. Opong and Zydney (1991) evaluated the general correlation of Smith et al. (1968) for filtration of BSA solutions through a highly retentive membrane. The system coefficient A = 0.23 was comparable to Smith et al. The convective mass transfer correlations for turbulent and laminar flow of fluids in channels and tubes are outlIned by Gutman (1987) and Rushton et al. (2000). 5.7.3 Shear-Enhanced Hydrodynamic Diffusion As a possible resolution to the flux paradox, Zydney and Colton (1986) proposed that the concentration-polarisation model could be apphed to mlcrofiltration provided that the Brownian particle diffusivity, DB, was replaced by the shear-enhanced hydrodynamic diffusivity, D s, first measured by Eckstein et al. (1977). Shearenhanced hydrodynamic diffusivity of particles occurs because individual rotating particles undergo random dlsplacements from streamlines in a shear flow as they Interact and tumble over other particles. The resultant lateral migration of the particle is towards a region where particle concentration is lower. It should not be confused with Brownian particle diffusion, which results from the Interactions of particles with the surrounding fluid molecules and is present even in the absence of shear. Eckstein et al. (1977) reported the earliest work on shear-enhanced diffusion. These researchers measured the lateral displacement of individual radioactively tagged, 5-23
CHAPTER 5. Literature Review (Surface Microfiltration) spherical particles in a sheared suspension using a Couette flow device, and proposed an empirical correlation for the diffusion coefficient as a function of shear rate, particle size and suspension concentration. The smallest particle used in the experiments had a diameter of 1000 microns and the maximum shear rate appeared to be I
S·l.
The effective particle diffusivity was found to increase with the square of
particle radius, rp, and linearly with shear rate,
r
w.
It increased with particle
concentration up to a volume fraction of 0.20 and was approximately constant thereafter. Although their data had considerable scatter, Eckstein et al. estimated the shear-enhanced diffusion coefficient, Ds. to be:
D, '" 0.025
r: r
w
o < t/J < 0.2
(5,\4)
0.2 < t/J < 0.5
(5.\5)
where t/J is the volume fraction of particles. Ten years later Lelghton and Activos (1986, 1987) employed a much simpler experimental technique, which avoided wall effects in the Couette viscometer, to estimate the shear-enhanced diffusion coefficient. They measured the viscosity of concentrated suspensions as a function of time and attributed its progressive decline to a diffusive migration of particles from the gap in a Couette viscometer. The particles used were of 46 and 87 microns diameter, and the maximum shear rate was 76
S·l.,
They reported shear-enhanced coefficients of self-diffusion for both concentrated and dilute suspensions (005 ::;; t/J ::;; 0.4) correlated by:
(5.16)
These two independent sets of results are in reasonable agreement for particle volume fractions less than 0.2. Above this volume fraction, however, the data of Eckstein et
al. (1977) gives an approximately constant value for the diffusion coefficient whereas 5-24
CHAPTER 5. Literature Review (Surface Microfiltralion) Leighton and Acrivos (1986, 1987) found that it increased rapidly with particle volume fraction and was approximately five times greater. This discrepancy has been attnbuted to the greater experimental dIfficulties experienced by Eckstein et al. (1977). Based on the above work, it has been suggested that shear-enhanced hydrodynamic diffusion is the dominant mechanism of back-transport in filtration of particulate systems. Zydney and Colton (1986) used the approxImate relationship for the shearenhanced diffusion coefficient measured by Eckstein et al. in a study of membrane plasmapharesis. Their result suggested that flux is proportional to shear rate to the first power. Although this correlation gave good agreement with experiment for large particles (> 2 microns), blood cells etc., the fit deteriorated as the particle size decreased. Davis and Leighton (1987) incorporated the empIrical correlations of Leighton and Acnvos into a model, which described the shear-induced expansions of particle layer away from a membrane surface, and its simultaneous tangential flow along the membrane surface. In this local model, the structure of the concentrated particle layer was predIcted at any axial position along the filter. Romero and Davis (1988) extended this method to predIct the axial variation of the particle layer thickness and structure at steady state. 5.7.4 Inertial Lift
Attempts to explain the 'flux paradox' associated with the concentration-polarisation model have led some workers to suggest that the back-diffusion of particles away from the membrane is supplemented by a lateral migration of particles due to Inertial lift. If the condItions are such that the inerual lift veloclly is sufficient to offset the opposing permeate velocity, then the particles are not expected to be deposited on the membrane. The inertial lift theory is the pnmary alternalive to the shear-enhanced model of crossflow. Indeed, inertial lift theory was the first mechanism to be widely accepted as an explanalion for the flux paradox of crossflow microfiltration. The lift force is generated by the pressure difference, which eXIsts on either side of a particle moving through the parabolic velOCIty field of a tube or slit flow. An isolated,
5-25
CHAPTER 5. Literature Review (Surface MicrofiItration) neutrally buoyant particle in a duct under laminar flow conditions (and a non-porous surface) will migrate across the fluid streamlines due to the inertia resulting from interactions with the flow field and the duct wall; the so-called 'pinch effect'. The strongest inertial hft occurs when the sphere is much closer to one wall than the other. Thus it has been suggested that two major causes of lateral migration exist in crossflow membrane filtration: a 'permeation drag force' exerted on the particle due to convection flow of fluid through the membrane and an 'inertial lift force' which carries particles near the membrane away from it. Segre and Silberberg (1962) were the first researchers to quantitatively observe the pinch effect. Cox and Brenner (1968) developed the basic theory for the inertial lift force acting on a spherical particle in a laminar flow duct bounded by non-porous walls. Theoretical expressions for the lateral force and velocity of the particle, for channel Reynolds numbers much less than one, were derived by Ho and Leal (1974), and Vasseur and Cox (1976) for a two dimensional channel, and by Ishii and Hasimoto (1980) for a tube. Under these conditions, the lateral migrations appear to be caused by the weak inertial effects in the fluid flow. Belfort and co-workers have extended the analysis of Cox and Brenner (1968) to include the effect of wall porosity for particle motion in a slit (Altena and Belfort, 1984) and in a tube (Altena et al., 1985). They have also measured (Otis et al., 1986) particle trajectories in a slit with one porous wall under slow laminar flow (order one channel Reynolds numbers) and their results agreed WIth the Altena and Belfort (1984) model. The inertial lift velocity,
VI,
of a small sphere under very slow laminar flow conditions
in dilute suspensions, where particle-particle interactions are negligible is:
VI
=b
(5.17)
Where Pt is the fluid density, Yw is the shear rate at the Wall, and b is the dimensionless distance from the wall, made dimensionless by the tube diameter or channel height. The equation is valid for a 2 Re«I, where a=rp /2H o and Re=2Ho/v. In the
5-26
CHAPTER 5. Literature Review (Surface Microfiltration) region near the wall, b is positive, indicating that the inertial lift velocity carries the particles away from the wall (Belfort et al., 1994). The maximum value for slow laminar flow occurs near the wall and is b "" 1.6 for a two-dimensional channel and
b "" 1.3 for a tube. Schonberg and Hinch (1989) have extended the theory for Reynolds numbers up to 150. They also showed that the parameter b decreases as the channel Reynolds number is increased. Most crossflow filtration operations are carried out under fast laminar flow conditions with Reynolds numbers typically of the order 102 or 103• Drew et al. (1991) conducted a study for relatively fast flow rates (Re
» 1) in a two-dimensional
channel with porous walls and found that the maximum value is given when
b = 0.577 . It was also demonstrated that the eqUilibrium position of particles moves toward the wall as the Reynolds number increases.
V,
= 0.577
(5.18)
When applied to crossflow micro filtration, the basic premise of the inertial lift theory is that particles are carried to the membrane walls only if the permeate flux exceeds the maximum inertial lift velocity. If this condition is met for a clean membrane, then a stagnant layer will form due to particle deposition. If this foulmg layer has a high resistance then It Will reduce the permeate flux until it just balances the inertial lift velocity. The maximum inertial lift for fast laminar flow (Re »1) in a clean channel (with thin fouling layers) is:
J
= V, = 0 036
Pf
,3
r2
p
w
(5.19)
Jl
A limitation of inertial lift models is that they only strictly apply for dilute suspensions. While they may be valid for the bulk, the enhanced concentration and added effects of particle-particle interactions in the polarised layer would need to be
5-27
CHAPTER 5. Literature Review (Surface Microfiltration) considered. Although significant lift forces are encountered in thin-channel membranes, the inertial lift velocity becomes negligible in tubular systems.
5.7.5 Interaction Enhanced Migration Electrostatic interactions can affect the rate of particle deposition on the membrane surface. There are a variety of interactions of wluch the most significant are:
•
Electrostatic double-layer;
•
London-van der Waals;
•
Born repulsion; and
•
Acid-base interaction
Electrostatic double layer (usually repulsive) and London-van der Waals (usually attractive) interactions form the basis of the Dejagun-Landau-Verwey-Overbeek' (DLVO) theory of colloidal stability. The interactlon potentials depend on the distance between surfaces and electrostatic double layer mteractions and are strongly influenced by surface charge and ionic strength. van der Waals mteractions predominate at small and large interparticle distances, whereas double layer repulsion dominates at intermediate distances. The summation of the individual interaction potentials between particles gives the total interaction energy, which if repulsive, can result in release of the fine particle from the surface.
5.7.5.1 Electric Double Layer Repulsion Double layer mteraction theory is extensively investigated independently by Deryagin and Landau, and Verwey and Overbeek, DLVO theory. Shaw (1989) and Klular et al. (1998) review the application ofDLVO theory to colloids. The constant charge,
VXLR' and constant potential, V;LR' double layer potential
between two spherical particles of radii rpJ and rp2 for Stern potentials
lI'OJ
and
lI'02,
are
expressed by Equations 5.20 and 5.21 respectively.
5-28
CHAPTER 5. Literature Review (Surface Microfiltration) Constant potential (Hogg et al., 1966):
... (5.20)
Constant charge (Wiese and Healy, 1970):
... (5.21)
The reciprocal Debye length,
K,
is descnbed by Equation 5.22,
(5.22)
Where Ew is the permittivity of dispersing medium (dielectric constant), h the separation distance between particles, e is the charge of the electron, no the particle concentratIon, and z is the charge number of the electrolyte. In practical applications of the DLVO theory, the electrostatic interactions are usually quantIfied in terms of zeta potential, the potential of the shear plane close to the particle surface as calculated from electropheretic mobility measurements. The zeta potential depends on the ionic strength, with increasing ionic strength leadmg to decreased zeta potential due to compactness of the diffuse double layer at the particle surface (decrease of Debye-length) (McDonogh et al., 1992). For equal spheres and Stern potentials (zeta potential) and small electric double layer overlap, such that [- 7dl] «I, these expressions both reduce to (Shaw, 1996):
5-29
CHAPTER 5. Literature Review (Surface Microfiltration)
(5.23)
Another approximate expression for
VDLR
is that given by Reerink and Overbeek. For
equal spheres, repulsive interaction potential energy between particles for constantpotential is given by (Yoon et al., 1999):
=64tr Sw rp (k. T)2 (tanh z e !l'o )2 exp{- K" h) ze
4k. T
This equation is valid for low to moderate potentials
«
of double layer is smaller than the fine particle size,
lCrp
(5.24)
60 mV), when the thickness
»
1 (Klular et al., 1998).
Expressions for electrostatic repulsive forces instead of energies have also been considered in literature for crossflow filtration (Buffham and Cumming, 1996).
5.7.5.2 Van der Waals Forces The London-van der Waals energy of interaction,
VLVA,
between two similar particles
is usually attractive. It is electrostatic in nature and originates due to interactions between permanent and/or OSCIllating dipoles of atoms. It decays slowly and acts over a distance as small as 10 nm (Khilar et al., 1998). The London attractive energy (also known as dispersion force) accounts for nearly all of the van der Waals attraction with few exceptions. Hamaker denved the following expression for the 'unretarded' London-van der Waals attractive potential between two equal spheres:
v = _Am [ LVA
6
2{I+H) +In(~)~ H{2+H) 2+H ~
(5.25)
Where, H= hlrp The separation distance between particles and particle radius are denoted by h and rp respectively.
5-30
CHAPI'ER 5. Literature Review (Surface Microfiltration)
Am is the effective Hamaker constant between two particles (1) and (2) in a dispersing medium (3). The negative sign accounts for the fact that it is an attractive potential. The van der Waals potential is overestimated when the separation is large (H > 10 nm) and retardation occurs.
The Hamaker constant can be evaluated by using either the London-Hamaker microscopic or Lifshiftz macroscopic approaches (Visser, 1972). The values of A obtained by the microscopic and the macroscopic approach tend to be similar in the non-retarded range. The Hamaker constant has been obtained experimentally for many materials and these values are generally in agreement with theoretical calculations. Its value varies with matenal, but is generally between about 10.20 and 10.19 J. The presence of water will alter the Hamaker constant and it is usual to apply an effective Hamaker constant based on the geometric mean of the individual constants. Thus, to calculate the attraction between particle (1) and particle (2) in the presence of phase (3) the effective Hamaker constant, A\32, is:
(5.26)
If the two particles are of the same material, this expression becomes:
(5.27)
Where A131 is the effective Hamaker constant for two identical particles (1) in a dispersing medium (3).
5.7.5.3 Born Repulsion
The Born repulsive interaction, VBR , is a short range potential resulting from the overlap of electron clouds as the particles approach the point of contact. Feke et al. (1984) and Ruckenstein and Prieve (1976) developed expressions for the Born repulSIve potential of sphere-sphere and sphere-plate systems respectively. Equation 5.28 is the Ruckenstein and Prieve expression for a sphere-plate system.
5-31
CHAPTER 5. LIterature Review (Surface Microfiltration)
(5.28)
Here Am is the Hamaker constant and (J is the atomic collision diameter in LennardJones potential (Khilar et al., 1998).
5.7.5.4 Acid-Base (AB) Interaction
ACId-base (AB) interaction, VAB , is among a number of important non-DLVO shortrange interactions (Christenson, 1988). This interaction can contribute substantially to the adhesion energy between two phases as well as govern, m some cases, the stabIlity of colloidal systems. An equation for the energy due to AB mteraction is given as (Chedda et al., 1992; Khllar et al., 1998):
(5.29)
Where V:O is the hydrophobic interaction, ho the minimum eqUlltbriurn dIstance, and
A. is the decay length of the liquid molecules. The negative sign implies hydrophobic attraction, and the positive sign implies hydrophilic repulsion. For further dIscussion on AB interaction the reader is referred elsewhere to van Oss (1994) and Khilar et al. (1998).
5.7.5.5 Total Interaction Potential Energy
The total interaction potential,
VTotal,
describes the variation of potential with the
distance between the particles and is calculated by summation of the individual effects. It is implicitly assumed that the various dIfferent types of interactions are independent of one another.
(5.30) 5-32
CHAPTER 5. Literature Review (Surface Microfiltration) The magnitudes of these potentials are reported in "Table 3.1.1 Magnitudes of various contnbutlOns to VT at a partIcular set condition", in Khilar and Fogler (1998, pAO) for a clay-water system. Of these four interaction potentials, VDI.R and VLVA contribute significantly towards the total interaction energy. Aimar et al. (1989) and Bacchin et
al. (1995) proposed a theoretical model, which describes colloidal deposition on a membrane surface accounting for surface interactions (Yoon et al, 1999). A mass transfer equation links the deposition rate to hydrodynamic conditions (permeation and tangential flow through a boundary layer thickness 8) and to physicochemical properties of the suspension (diffusivity, DB, and potential barrier between particles, VB). Equation 5.31 predicts the existence of a critical flux, Jent. for membrane
filtration of large-size colloids as (Yoon et al., 1999):
v,
=
i In(; )
(5.31)
and, (5.32)
Where gis the boundary layer thickness. The boundary layer thickness, 0, is obtained by Uveque's equation:
d
; = 1.62 [Re'Sc, (d h
/
L)]1f3
(5.33)
Where, dh is the hydraulic diameter of channel entrance.
Bacchin et al. (1995, 1996) claimed that the surface interacttons between colloids play significant roles on the particle deposition and the hydraulic resistance of the deposit in ultrafiltration. By taking electrokinetic effects into consideratton for the depOSIt properties, a reduction in porosity owing to a decrease in repulsive interaction 5-33
CHAPTER 5. Literature Review (Surface Microfiltration) between colloids was found. McDonogh et al. (1989) considered the particle interactions in terms of double layer theory to discuss the charge effect on the permeating flux in crossflow ultrafiltration. The predicted results using their theory agreed with experimental data over two orders of magnitude of the particle size.
5.8 OTHER MODELS 5.8.1 Surface Transport Models As an alternative to back-transport of particles away from the membrane by mechanisms such as diffusion and inertial lift, it has been proposed that particles are carried to the membrane surface by permeate flow and then roll or slide along the surface due to the tangential flow. Axial convection (continuum) models and singleparticle models have both been developed to describe this situation.
5.8.1.1 Axial Convection Axial convective-flow mathematical models descnbe the simultaneous deposition of particles into the cake layer and the flow of this layer towards the filter exit. The fully developed laminar flow equations are solved for the velocity profiles in the bulk suspension and the cake layer, and steady state cake thickness and permeate flux are also determined. 1n general, the cake thickness increases With increasing distance from the filter entrance and with decreasing axial flow rate. Quantitative predictions to compare with experiments are difficult because the particle concentration, effective viscosity, and specific resistance of the flowing cake layer are not known a priori. Leonard and Vassilieff (1984) presented a model in which the rate of convection perpendicular to the membrane was balanced by axial convection of a unifonuly concentrated layer. Particle diffusion and lift force effects were specifically neglected, and a linear velocity profile in the concentrated layer and cake viscosity equal to that of the bulk suspension were assumed. Super-micrometer particles of red blood cells (2 IJlll
X
5
~m)
were studied. Davis and Birdsell (1987) extended the convection
analysis of Leonard and Vassilieff by proposing parabolic velOCity profiles for both the bulk suspension and the flowing cake. The assumption of equal viscosity of the cake and suspension was also relaxed. Davis and Birdsell found some agreement
5-34
CHAPTER 5. LIterature Review (Surface Microfiltration) between theory and experiments for 'large' particle systems of acrylic beads (150 - 212
~)
where the cake layers were observed directly. Romero and Davis
(1991) further extended this model to distinguish flowing and non-flowing cake layers on membrane surfaces, assuming shear-enhanced diffusion between a bulk solution and the flowing cake layer. The main weakness of convection models is their inability to make predictions of static filter cake formation without applying very restrictive condItions on the nature of the cake. In most crossflow processes the filtration performance is not affected by loosely packed flowing layers, but by static deposits of much higher hydraulic resistance. Thus, convection models predicting the formation of flowing cake layers, are of limited use in understanding the significant mechanisms of filtration flux decline.
5.8.1.2 Frictional Force Balance In the single-particle model, the basic concept is to consider a spherical particle on the
surface of the membrane, or on the surface of a stagnant cake layer, and perform force and torque balances on the partIcle to determine if It Will adhere to the surface or be transported along the surface. This type of model is based on the balance between the resultant axial force, which tends to remove particles, and the normal force, which tends to retain the particle. The axial force is thought to be constant since the bulk velocity is constant, and the normal force will decrease due to the decay of the permeate flow rate. In the initial stage, the normal force is stronger than the axial force and hence causes deposition. The steady state occurs when the two forces are balanced and results in constant cake thickness and permeate flux. The advantage of this form of model is that no assumptions about the fluid flow regime are needed; however, predictions of filtration flux are not possible a priori, as the friction coefficient and boundary layer velocity profile are seldom known accurately. Flscher and Raasch (1986) using limestone suspensions demonstrated that different crossflow velocities permit degrees of particle classification from the bulk suspension to the cake. Rautenbach and Schock (1988) proposed that a particle reaching the surface of a membrane (or filter cake) either sticks at the point of contact of moves tangentially along the surface. Their work was supported by experiments using
5-35
CHAPTER 5. Literature Review (Surface Microfiltration) Lu and Iu (1989) obtained similar conclusions to Rautenbach by considering the moments of hydrodynamic forces acting on a particle at a rough surface. Expressions for the limiting conditions for deposition of each particle size in a broad distrIbution were presented and supported by experiments with calcium carbonate particles. Cohen and Probstein (1986) and McDonogh et al. (1989) showed that interparticle forces, in addition to hydrodynamic forces are likely to influence the behaviour of particles as they approach the membrane or cake surface in crossflow filtration. Blake
et al. (1990) developed a frictional force balance model involving both hydrodynamic and interparticle forces, to predict the steady state flux of a crossflow filter. The' Model successfully predicted both the linear nature of permeate velocity dependence on shear stress, and the associated gradient as being a function of particle diameter for the filtration of spherical monodisperse latex particles (0.55 and 1.8 Ilm diameters). Kuiper et al. (2000) developed a simple single-particle model for cross-flow with microsieves. The model described the crossflow conditions required to release a trapped spherical particle from a circular pore. Release of trapped particles (polystyrene spheres and yeast cells) was determined by flux measurements as well as by in-line observation through a microscope. Output from the model gave a fairly good Indication of what cross-flow should be applied to keep pores free for conditions specified. 5.8.2 Pore Blocking and Cake Filtration Model Membrane fouling is often analysed using the cake filtration theory or one of the classical 'blocking laws'. Hermans and Bredee first introduced pore blocking filtration mechanisms, which were developed further by Gonsalves, Grace, Shirato and co-workers, and Hermia. The blocking laws are based on the well-known filtration models applied to dead-end filtration:
2
d t _ k (dt )"
dV 2
-
"
dV
(5.34)
5-36
CHAPTER 5. Literature Review (Surface Microfiltration) Where t is the filtration time and V the total filtered volume. The exponent n characterises the specific filtration model: representing mechanisms due to cake filtration (n = 0). standard filtration (n = I). intermediate blocking (n = 1.5). and complete blocking (n =2). Usually one of these terms is used to analyse the fouling data with the standard blocking or cake filtration models often favoured. In practice however it is probable that these mechanisms occur in consecutive stages or even successively. The blocking models implicitly assume that the membrane has straightthrough non-interconnected pores such as track-etched membranes. Bowen et al. (1995) considered pore-blocking mechanisms to descnbe the stages of permeate flux decline during dead-end rnicrofiltration ofBSA solutions through tracketched polycarbonate membranes. with pore sizes 0.1 - 1.0
~m.
Blockage of the
smallest pores first through to formation of a cake was observed; exponent n decreased from 2 to O. Filippov et al. (1994) studied membrane fouling by a sieve mechanism in which particles greater than the pore blocked the pore. Kostinetsev et
al. (2002) used a sieving-cake filtration model to predict the behaVIOur of track-etched membranes challenged by mono-size latex particles. Three separate periods of flux decline were identified. and a definite and consistent transition from blocking filtration to cake filtration eXisted. Most commercial rnicrofiltration membranes have highly interconnected pore structures and internal deposition within the tortuous membrane structure is significant. Holdich and Zhang (1995) showed that the standard filtration model for pore channel constriction could accurately describe internal foulmg of commercially available macroporous membranes. with pore sizes 0.45
~
and 1.2
~.
rather than
pore plugging. In the crossflow filtration study deposition of latex particles reduced the equivalent pore diameter. Ho and Zydney (1999) developed a theoretical model accounting for effects of permeate flow through membranes with different interconnected pores for protein rnicrofiltration. The different blocking laws have been used for over 50 years to analyse and interpret filtrate flux data. There has been considerable recent interest in the application of these models to protein ffi1crofiltration. with the fouling attributed to the physical deposition of large protein aggregates on or within the membrane pore structure.
5-37
CHAPTER 5. Literature Review (Surface Microfiltration) However, these models have often been applied to the fouling of polymeric microfiltration membranes having highly interconnected pore structures, e.g. polyethersulfone, polyvinylidene fluoride and mixed cellulose ester membranes. Ho and Zydney (1999) showed that membranes having interconnected pore structure do foul much more slowly than membranes with straight through (non-connected) pores, with data suggesting that aggregate deposition only disturbs the filtrate flow over a small penetration distance into the interconnected pore structure of these membranes. 5.8.3 Combined Particle Deposition Models Pore blocking and cake layer formation mechanisms are commonly used to explain the flux decline during microfiltration of colloidal suspensions. A cake is formed by the layer-wise random packing of particles depositing on the membrane surface. Numerous computer simulations have modelled the random packing structure of these partIcle beds. Suzuki and Oshima (1985) proposed a simple particle-packing model using the random packing assumptions to calculate the co-ordination number, i.e. the number of touching spheres. Houi and Lenormand (1986) used a sticking model to describe the depositing process of spherical particles on filter fibres, and Tassopoulos et al. (1989) developed a discrete stochastic model to simulate the phenomena of particle deposition on a substrate surface. Several authors have also studIed particle-packing theory with the aim of understanding particle deposition phenomena and resulting fluxes on porous membranes. Lu and Hwang (1995) defined the critical incident angle to the cake layer in order to conclude whether particles deposit of not in their two-dimensional study, while McDonogh et al. (1998) considered the interparticle distance produced by electrostatic repulsions. Kawakatsu et al. (1995) analysed cake layer porosity and cake layer formations by assuming a rhombohedral tetragonal structure as the basic packing method of cake layers. Most of these particle-packing studies suffer from limitations: in the case of the 2-D model there is no pore blocking phenomena, and mono-dispersed particles in a 3-D model are unrepresentative of a typical particle system. Yoon et al. (1999) simulated the flux decline for microfiltration of multi-dispersed particles, which produces a three-dimensional randomly deposited cake. The rate of 5-38
CHAPTER 5. Literature Review (Surface Mlcrofiltration) particle deposition on the membrane was derived from the difference between permeation velocity and the overall back transport velocity, as shown in Figure 5.9. Diffusion and shear-enhanced mechanisms, i.e. shear-enhanced diffusion and inertial lift terms, together with charge interactions were considered in the calculation of the back-transport velocity. The resultant simulation was compared with experimental results obtained with multi-dispersed iron oxide particles and polycarbonate tracketched (PTFE) membrane filters with pore diameters of 0.2 l1II1. For moderately charged particles between 0.1 and 10 microns the particle back-transport was controlled mostly by surface interactions between the particle and surface. Simulated flux decline compared favourably with the experimental results.
Double layer repulsion
Shear-enhanced diffusion
• Inertial hit
Drag torque
Diffusion
l" (I
Permeation drag
Axial drag
Sedimentation
Ir ,
" van der Waals attraction
1"···········1··········· s;;;;i~p~~;;;~~·bi;t·;:;;~~·b~;~~···························l·········1 .............. ............................................................................................................ J
J
J
Figure 5.9 Forces and torques acting on a charged, spherical particle suspended in a viscous fluid undergoing larninar flow in the proximity of a flat porous surface, in Wiesner and Chellam, 1992.
5-39
CHAPTER 5. Literature Review (Surface Microfiltration) 5.8.4 Pore-Particle Interaction In most of the forgoing models the resistance of the membrane is assumed to remain
constant, at the value provided by clean water, or to have some value determined by fitting the experimental data to a model where Rm remains constant, but other membrane resistances change with flow conditions. The former approach is far too simplistic and has been heavily criticised (Rushton et al., 2000). It is not valid in conventional cake filtration, where Rm must be determined in-situ, and it is not likely to be a valid approach in membrane filtration. The latter approach requires another empirically determined variable, limitmg the applicability of the models. However, recent progress has been made on fully characterising the membrane-pore-particle interaction using Atomic Force Microscopy (AFM), where a particle of interest is attached to the cantilever probe and scanned over the membrane surface. This results in a direct measurement of the force between the membrane and the particles in suspension (Bowen et al., 1996; Hilal et aI., 2002). In time, this research could well lead to a fundamental prediction of the membrane resistance, by relating the membrane pore-particle interaction to the number of particles deposited, as determined by a material balance using the permeate volume. The deposited particles and their arrangement on the membrane would be coupled with flow equations around the particles in order to provide a complete description of the membrane resistance.
5.9 SUMMARY There are numerous models for crossflow microfiltration in the literature; the subject is a very fruitful one for academic investigation. The interested reader is directed to reviews by Davis (1992b), Lojkine et al. (1992), Belfort et al. (1994), Bowen & Jenner (1995), and Sethi et al. (1997). However, as has been mentioned in many literature surveys before, the nature of finely suspended particulate material is very complex and unlikely to yield to a single unifying model. Most of the models descnbed here have been validated by their authors using specific experimental conditions. It is highly likely that the model was appropriate, under those conditions, but has limited validity when applied under different conditions. This rather apparent negative statement shouldn't be taken as condemnation of modelling of this process, merely that when investigating an application that the model selected from the literature has to be done so with care and with an understanding of the process to
5-40
CHAPTER 5. Literature Review (Surface Microfiltration) which it will be applied. Perhaps the least contentious subject in crossflow filtration modelling is that of critical flux. Since its first reporting (Raasch, 1987) many researchers have shown that it can occur and have provided useful data and hypotheses of how it can be used for best effect. Clearly, for fractionation of fine particles the fonnation of a secondary membrane must be avoided; otherwise the grade efficiency of the system will be defined by the secondary membrane and not the rnicrofilter. Hence, It is important to operate below the critical flux for the system under consideration. For practical applications of fractionation, the highest flux (whilst still being below the critical) at the highest solid concentration in suspension would be required. This is technically demanding and fractionation would be impossible with any filter type other than the surface filters described in Section 5.5. For the purpose of filtration of finely divided particulate material, such as Cryptosporidium Parvum, both surface filters and depth membrane filters could be
used. However, as should be evident from the multitude of models described in this chapter, membrane fouling is complex and strongly dependent upon the nature of the material being filtered: more so than the nature of the filter. Hence, a very tight
microfilter could be used to exclude all particulate material, say at 0.1 Il11l pore size, but most fluids of commercial interest will also posses material finer than this. It is this finer material that will penetrate the membrane filter, relying on depth filtration mechanisms, leading to long-tenn and irreversible flux decline. The alternative approach is to not use a depth filtration membrane at all. Hence, surface rnicrofilters SUitable for industrial use, rather than the Nuclepore laboratory filters that are commercially avrulable, are being investigated for both rnicrofiltration and fractionation applications.
5-41
CHAPTER 6. Theory (Surface Microfiltration)
CHAPTER SIX THEORY: SURFACE MICROFILTRATION In the absence of significant fouling of the membrane a model of microfiltration can
be formulated around what is taking place between the particles above, or on, the membrane surface. Where there is significant membrane fouling described by both internal and surface pore plugging, the membrane resistance will increase. There have been many models based on empirical fits of the data to classic blocking filtration models; e.g. standard filtration law, intermediate filtration law, etc. This project was formulated to investigate conditions when membrane fouling should be minimal; hence, an appropriate model would be one in which it is possible to predict the flux and pressure behaviour under different operating conditions based on the particle properties and the prevailing hydrodynamic conditions only. The experimental equipment used to provide the hydrodynamic conditions will be descnbed in Chapter Seven (Surface Microfiltration Experimental). A design based on a cone-and plate-viscometer was employed in order to provide a well-defined shear field above the membrane. The suitability of the constant shear model for stirred-cell operation was evaluated using a simple mathematical model reviewed in Appendix C.
6.1 INTRODUCTION TO PARTICLE DEPOSITION MODEL In the model development forces acting on particles above, and on, the membrane
surface including those due to particle diffusion, shear-enhanced diffusion, inertial lift and collOidal interaction are conSIdered, to provide a total back transport velocity, or effective mass transfer coefficient. The variation in mass transfer coefficient with operating conditions is then correlated with standard published correlations for that variation with respect to hydrodynamic effects; e.g. Reynolds number. It is this model that is then evaluated in Chapter Eight (Surface Microfiltration Results & Discussion), by comparison with the experimental data. In addition to the forces and phenomena already mentioned, due consideration is also given to particle sedimentation and, in common with all previously published work, the assumption of spherical particles will be made. Supporting calculations for each of the various transport mechanisms presented here are presented in Appendix D. 6-1
CHAPTER 6. Theory (Surface Microfiltration)
6.1.1 Gravitational Settling Velocity (vG) The gravitational settling velocity,
VG,
of spherical particles moving towards the
membrane, was calculated using Stokes' law:
(6.1)
Where rp is the particle radius, Pp and Pi are particle and fluid densities respectively, g is the gravitational acceleration term, and fJ is the fluid viscosity. Gravitational
settling velocity is dependent on suspension properties of particle size, particle-fluid density difference and fluid viscosity. Figure 6.1 illustrates the relationship between setthng velocity and particle size. Settling velocity increases with the square of
~
'I/)
E ::1. ....::
300
~t:J .....
250
'u0
200
>~ Cl
.5:
EQ) I/)
150 100
iij
c 0
-.
i
50
'5O
III
CJ
0 0
20
40
60
80
100
Particle diameter (d,). Ilm.
Figure 6.1 Gravitational settling velocity, VG (Ilm S·l) as a function of particle diameter, dp (Ilm).
6-2
CHAPTER 6. Theory (Surface Microfiltration) particle radius and is more important for larger particles. As a forward acting particle transport mechanism to the membrane surface, gravitational settling is negligible compared to the permeation velocity. The gravitational settling velocity of a 5 llm diameter particle for example is 7.1 x 10.7 m s·1, some three orders of magnitude smaller than a typical permeation (superficial) velocity of 1.9 x 104 m S·I.
6.1.2 Mass Transfer Coefficient A general mass transfer correlation for transport to a stationary membrane surface in a stirred-cell is given by (Smith et aI., 1968):
Sh
= A (Re)
M
(SC)033
(6.2)
There are many mass transfer coefficients available in literature for stirred-cell geometries (Kaufmann and Leonard, 1968; Smith et al., 1968; Colton and Smith, 1972; Malone and Anderson, 1977; Mitchell and Dean, 1984; Opong and Zydney, 1991; Nicolas et al , 2000). Smith et al. (1968) reported experimental measurements for the mass transfer coefficient in the laminar boundary layer. The constant parameters A and exponent m were evaluated as A
= 0.285
and m
= 0.55
in this
region.
Sh = 0.285 (Re)oss (SC)033
(6.3)
The mass transfer coefficient describes a speCific stirred-cell geometry and membrane permeability. Correction factors proposed by Smith et al. are available to accommodate different stirred-cell systems. The corrected form of EquatIOn 6.3 for the stirred-cell system described in Chapter Seven, is
Sh = 0.299 (Re)oss (SC)033
(6.4)
6-3
CHAPTER 6. Theory (Surface Microfiltration) Dimensionless Sherwood (Sh), Reynolds (Re) and Schmidt (Sc) numbers are represented by,
Sh=kb D
wb 2 Re=--
V
Sc =D
V
(6.5) to (6.7)
Rearrangement of Equation 6.4 in terms of k and simplification results in,
k
_ (D067 Yw -0.069 v
OSS
022
b
OI
)
(6.8)
The validity of the coefficient k is restricted to the laminar regime, or more accurately Re < 32000 for rotational systems (Smith et al., 1968). This equates to shear rates
Yw < 1000 S·I
for the stirred-ceIl system.
Back transport away from the membrane surface by diffusional mechanisms was evaluated using Equation 6.9. The dIffusivity parameter was replaced by DB or Ds for particle, or shear-enhanced diffusion respectively.
J == v = k In (
v = 0.069
(
~: )
(6.9)
D067 t, oss bO I) In (C) w
vo n
-..!!:.
Cb
(6.10)
6-4
CHAPTER 6. Theory (Surface Microfiltration) 6.1.3 Back Transport Velocity due to Particle Diffusion (Vd) Particle transport away from the membrane surface and across the boundary layer by a particle diffusion mechanism, Vd, was estimated using Equation 6.11,
(6.11)
where kd is the mass transfer coefficient due to particle diffusion, and Cw and Cb, the particle concentrations at the membrane surface (wall) and bulk fluid respectively. The mass transfer coefficient, kd, denved for the specific experimental conditions was evaluated by
kd
DB is the particle diffusivity,
D
067
r
055
= 0.069 ( B ;22
b0 1
J
v
rw
(6.12)
is the shear rate at the membrane surface, b
represents the active membrane radius and v is the fluid viscosity. Particle diffusivity was calculated from the Stokes-Einstem relationship:
(6.13)
Here kB is the.Boltzmann constant, T is the average feed suspension temperature, J1 is the fluid viscosity and
rp
the particle radius. Particle diffuSIVity IS inversely
proportional to particle size, and is relatively insignificant for particle diameters larger than 1 J.1Ill. The particle diffusivity for a 3 J.1Ill diameter latex particle is 1.5 x 10. 13 m 2 S-I. The average feed suspension temperature was determined from column measurements recorded either side of filtration experiments.
6-5
CHAPTER 6. Theory (Surface Microfiltration)
Vd
= 0.069
D ro"v.oss JIn (C) ~ ( v Cb 067
bOt
(6.14)
B
The shear rate was related to the stirrer rotational speed by the simple relationship
Y...
= 1.5 n. The reader is directed to Appendix C for further discussion about shear
rate determination using a constant-shear model. Shear rates at the membrane surface of up to approximately 1400
S·I
were evaluated empirically. Particle concentration at
the membrane surface, Cw, was estimated from packing considerations. For monodisperse particles, which are rigid and spherical in shape, the packing density cannot exceed a maximum value of 60 per cent by volume at high fluxes nor fall below Cb at low fluxes). The respective random packing density for compressible or poly-disperse
Cl)
::I
..en
E
'C::!.
~....::
'u0
100~----------------------------------~
10
,:>-'" .......
a; c
> 0 t: 'iij
o
::I
1
0.:1:: en .-
c'C
... .--
Cl) _III u ~
ut:
0.1
III III
ID
a.
.s
o
20
40
60
100
80
Particle diameter (dJ. J1m.
Figure 6.2 Back transport velocity due to particle diffusion, Vd (10.6 m
S·I)
as a
function of particle diameter, dp (lIID). Additional information: Xv = 600 S·I,
Cb = 0.02 kg m·3 , Cw = 525 kg m·3 , T = 297.3 K.
6-6
CHAPTER 6. Theory (Surface Mlcrofiltration) particles is 80 to 90 per cent by volume. AIl particles were assumed to be nonadhesive. A considered packing density of 50 percent by volume (c.f. 62 per cent by volume obtained from particle packing simulation by Yoon et al. (1999) was used for mono-size particles in the model. The bulk concentration Cb was carefully measured at the start of each experiment. The predicted back transport velocity for particle diffusion increases with shear rate to the power 0.55 and decreases with particle radius to the two-thirds power. Transport away from the membrane by particle diffusion mechanism, as seen in Figure 6.2 on the preceding page, has little effect on large particles and, it is widely accepted that the particle diffusion mechanism is significant for sub-micron particles. The back transport velocity,
Vd,
for a 5 I11I1
diameter particle is 6.0 x 10.7 m S·I. 6.1.4 Back Transport Velocity due to Shear-Enhanced Diffusion (vs) The effect of shear-enhanced hydrodynamic diffusion,
Vs,
on particles at the
membrane surface was modelled using,
= 0.069
V s
D ( s
067" OSS bO I
w
von
I
)
In
(C) ~
Cb
(6.15)
The back transport Equation 6.15 is primarily the same as that descnbing particle dIffusion with particle diffusivity DB, replaced by shear-enhanced diffusivity, Ds. Eckstein et al. (1977) reported an experimentally detennined approximation for the effective shear-enhanced hydrodynamic diffusivity,
(6.16)
Shear-enhanced diffusivity is proportional to the square of particle size multiplied by the shear rate. Particle diffusivity however, is independent of shear rate and inversely proportional to particle size. The comparative significance of these diffusion-based
6-7
CHAPTER 6. Theory (Surface Microfiltration) mechanisms is demonstrated in Figure 6.3 across a range of particle sizes and for a shear rate of approximately 600
S·I .
Shear-enhanced diffusivity is clearly the dominant
mechanism for micron-sized and larger particles whereas Particle diffusivity is in ascendancy below 0.5 J.!m. The shear-enhanced diffusivity of a particle with a diameter of3 J.!m at a modest shear rate (600
S· I)
is 4.1 X 10. 11 m2
S·I,
which is more
than 2 orders of magnitude greater than its particle diffusivity. Shear-enhanced diffusivity becomes the larger coefficient above a particle size of 0.46 J.!m (vertical line displayed on graph).
10.7 ';"
fII
'"E
-
~
Cl
Diffusion dominated by Brownian motion
10.8 10.9
Diffusion dominated by shear-enhanced motion
~
:~ fII
10.10
;E "C
10.11
Q)
U
:e (\J
10.12
a..
10.13 0.01
0.1
1
10
100
Particle diameter (dp ) , Ilm.
Figure 6.3 Comparison of particle diffusivity (m 2 S·I) by particle diffusion and shear-
enhanced motion over a range of particle size, dp (J.!m). Shear-enhanced diffusion coefficient exceeds particle diffusion at 0.46 J.!m. Additional information:
y"
= 600 S·I.
Based on a plot presented by Wiesner et al. (1992).
A curve of shear-enhanced hydrodynamic diffusion as a function of particle size is plotted in Figure 6.4, on the following page. As expected the back transport term is important for spherical particles of I J.!m diameter and larger, a size range describing
6--8
CHAPTER 6. Theory (Surface Microfiltration) the latex challenge material used in filtration experiments. The back transport velocity,
Vs,
for these 1 to 221UTI latex particles is considerably greater than that
exhibited by particle diffusion. The back transport velocity,
Vs,
for a 5 I1m diameter
particle is 7.3 x lO·s m S·I.
-
10·
~
'1/1
0
G)
:I
-
E
:::!. ~
"C-;: >--'uo .2c Gi 1/1 > :I 1::.== o"C a."C 1/1 G) c u f! :I _"C ~ U
ItS
ID
.= ~ ItS
103 1()2 101 10° 10.1
G)
s:.
1/1
10.2 0
20
60
40
80
100
Particle diameter (dJ. l1m.
Figure 6.4 Back transport velocity due to shear-enhanced diffusion mechanism, Vs (10-6 m S·I) as a function of particle SIze, dp (IUTI). Additional information: ]{v = 600
s·t. Cb = 0.02 kg m·3, Cw = 525 kg m·3, T = 297.3 K
6.1.5 Back Transport Velocity due to Inertial Lift (VI) Transport of particles away from the membrane caused by inertial lift mechanism is descnbed by (Drew et al., 1991): 3 • 2
VI
= 0.036 P, rp Yw
(6.17)
Jl
6-9
CHAPTER 6. Theory (Surface Microfiltration) Where Pt IS particle densIty,
Tp
is particle radius,
tw
represents the shear rate at the
membrane surface and J1IS the viscosity of the fluid. The Inertial lift velocIty for fast' laminar flow increases WIth the cube of partIcle size and the square of tangential shear rate. Inertial lift is independent of partIcle concentration gradient. Figure 6 5 shows the performance of inertial lift velocity as a function of particle size. As observed, the back transport velocity
VI
is significant only for large particles. The back transport
velocity VI of a 5 Jlm diameter particle is 2.0 x 10.7 m s·l.
2000 Cl)
.
:;, "Co;111 >0
:= E u 0
1500
:::1.
Gi..-i
>.:.
t::= 0::
1000
Q.-
ca C .cat:: ... Cl)
III
.... .-c
~
u .... 0 ca m
500
o
20
40
60
80
100
Particle diameter (dJ, Ilm.
Figure 6.5 Back transport velocity due to inertial lift, VI (10-6 m S·I) as a function of particle size, dp (Jlm). Additional information: Xv
=600 S·I.
6.1.6 Back Transport Velocity due to Interaction Enhanced Migration (v,)
6.1.6.1 Electrostatic Double Layer Interaction Potential (VDu) The electrostatic double layer repulsive energy, Stem potential
~,at
VDIR,
between identical spheres with
a separation distance, h, is 6-10
CHAPTER 6. Theory (Surface Microfiltration)
(6.18)
Where &v is the permittivity of water,
rp
is the particle radius, k8 the Boltzmann
constant, T is the average feed temperature, Z represents the charge number of the electrolyte, e is the elementary charge of the electron, and
KD
describes the inverse
Debye length (Shaw, 1992, p.214). The measurable zeta potential of the particle ,"was used in place of the Stem surface potential. A surface charge for latex particles of - 55 mV was measured using a Malvern Zetasizer. The
VDLR
model is valid for low to
moderate potentials, i.e. '" < 60 mV (Kmlar and Fogler, 1998, p.32). The inverse Debye length,
KD,
was calculated using the relationship (Shaw, 1992,
p.179):
(6.19)
Here, no is the particle number concentration. The thickness of the double layer can be estimated by
K D -1.
The double layer model is based around an ionic dispersing
medium and could not be applied directly to the latex feed suspension prepared using ultra pure water. An equivalent molar concentration, c, was calculated as if the ultra pure water medIUm were NaCI electrolyte (Atkins, 1998; Maron and Prutton, 1959, pp 443-451). Although this was not a true representation of the stirred-cell particle system, it would nonetheless indicate the significance of
V DLR
at siIDllar molar
concentrations of electrolyte. Not surprismgly the molar concentration of equivalent electrolyte, c, was low, i.e. 5.2 x 10.5 g mol
r1.
Likewise the equivalent particle
number concentration, no, was 3.1 x 10.22 m·3• The corresponding calculation is presented in Appendix D, where it is described in some detail.
6-11
CHAPTER 6. Theory (Surface Microfiltration) The interaction energy due to electric double layer repulsion,
VDLR,
is strongly
influenced by the surface charge and potential and the ionic strength of the feed suspension. The double layer repulsion energy is described approximately by the function VDIR =f(e
h
with a range of the order of double layer thickness
),
I (K - ). D
Double layer repulsive energy for 1 flID latex particles at a separation distance of 1 nm is 1.1 x 10-7 J.
6.1.6.2 London-van der Waals Interaction Energy (VLV.v Equation 6.20 is the classic expression for London-van der WaaIs attraction between particles (Khilar and Fogler, 1998, p.33):
v
= _ Am [2(I+H)
6
LVA
H(2+H)
+In(...!!....-)~ 2+H ~
(6.20)
Where Am is the effective Hamaker constant, and the ratio of particle separation distance h to particle radius
rp
is represented by H.
VLVA
acts over separation distances
up to 10 nm (Khilar and Fogler, 1998, p.32). Beyond this distance the retardation effect should be considered. Individual polystyrene latex and water Hamaker constants have been reported by Visser (1972). The effective Hamaker constant was calculated from these experimental values as 2.27 x 10-21 J. The magnitude of London-van der WaaIs attraction is primarily influenced by the effective Hamaker constant value. London-van der WaaIs interaction potential for 1 Ilm particles separated by 1 nm is - 1.9 X 10-19 J.
6.1.6.3 Total Interaction Energy (VToto!) The total interaction energy between charged particles, VTotal , was obtained by summation of VDLR and
VLVA
(Khilar and Fogler, 1998, pp.38-44) It is assumed that
the interaction potentials are independent of each other.
(6.21) 6-12
CHAPTER 6. Theory (Surface Microfiltration) Born repulsive, VBR and Acid Base VAB short-range interaction potentials were not considered to contribute significantly towards the total interaction energy. The magnitudes of these two potentials in addition to
VDLR
and
VLVA
are reported in "Table
3.1.1 Magnitudes of various contributions to VT at a particular set condition", in Khilar and Fogler (1998, pAO) for a clay-water system.
.-; p
1.5
,
0 ....
c
0 +I
,,-
,
10
{
u
I!
.sc
05
0
>0-
, I
lOo
E2 CD
I I
.... . . .................... ,-
0.0
C
Electnc double layer repulSIon (VDLR)
CD
iii
-0.5
London - Van der Waals attractoon (VLVA)
+I C
.s0
D..
---
Totallnteractoon energy (VT018')
-1.0 0
4
2
6
10
8
Separation distance between particles (h). nm.
Figure 6.6 Double layer repulsive, and combined,
VTotal,
V DLR,
London-van der Waals attractive,
VLVA,
particle-particle interaction energies as a function of separation
distance, h (nm). Additional information:
"'0 =- 55 mY, no = 3.13 x 10
22
number m·3,
Aw = 2.27 X 10.21 J.
The particle interaction energy curves of electric double layer repulsion, London-van der Waals attraction and total interaction are presented in Figure 6.6 for particles of the same material. Particle interaction energy, reported in units of Joules (scaled to x 10-18) is plotted against separation distance, in nm. London-van der Waals shortrange interaction dominates when particles are in very close proximity to each other. There is a relatively large increase in potential energy over a short range of influence
6-13
CHAPTER 6. Theory (Surface Microfiltration) when h is small. As particles move slightly apart and thereafter up to 10 nm separation distance, the total interaction energy is described by the double layer repulsion curve. The potential curves, with the exception of close contact, show little variation with separation distance. This is partly a result of low equivalent electrolyte concentration. The model describing double layer repulsion (Equation 6. 18) is a simplified expression for identical spherical particles sharing equal Stem potentials. A system containing different particle types and a higher electrolyte concentration would exhibit the well·known "primary minimum" part of the energy curve, which is noticeably absent here (cf Figure 8.2 in Shaw, 1992, p.220).
3,-------- ------ ------ -----------------,
f " .. · .... ·· ..............
If)
o -0
-
So 0"If)
>--
2
\
Q) Q)
\
t: t:
If)
\ \
Q)
t:
t: 0 Q)+i
E ~
(b)
\ \
If)~
.2
(a)
\ \
1
\ \
-
\
o lii
\
,
\
t:
\
"-
"
- .- ._ ._ ' -..
(e)
------
(d) ' - ' - ._ ._ ._ -
OL--.------.-----~==----._----_.----~
o
2
4
6
8
10
Separation distance between particles (h), nrn.
Figure 6.7 Dimensionless total interaction energy (VT% lleB 1) as a function of
separation distance, h (nm) at various NaCI electrolyte concentrations (g mol rl); (a) 5 x 10'5, (b) I x 10-4, (c) I x 10,3, (d) I x 10'2, and, (e) 0.1.
6-14
CHAPTER 6. Theory (Surface Microfiltration) The double layer repulsion energy between like particles is strongly influenced by the concentration of electrolyte. Figure 6.7, on the previous page shows the general variations of VTotal with h for increasing N aCI electrolyte concentrations (represented by (a) to (e». Dimensionless total interaction energy is defined by electrolyte concentrations range from (a) 5 x 10-5 g mol
r1
VTo,aJ
I kBT . Dilute
(equivalent electrolyte
concentration used in model) to (e) 0.1 g mol r1. At very low electrolyte concentration (curve (a» the double layer thickness (1ITCD) is greater leading to strong double layer repulsion. At relatively high electrolyte concentrations, (curves (d), (e» the doublelayer repulsion is weak. The relative influence of surface interactions on the rate of colloid deposition to a membrane surface was investigated using a theoretical model proposed by Bacchin et
al. (1995), (1996). A mass transfer equation relating colloid deposition to hydrodynamic conditions (permeation and tangential flow) and suspension properties (particle diffusivity and stability) has been developed for laminar cross-flow conditions_ The velocity of colloid particles away from the membrane due to interaction-enhanced migration, v" is described by
v =DBln(VB) , 0 0
(6.22)
Where DB is the Particle diffusivity, 0 is the boundary layer thickness and VB symbolises the potential barrier between particles enhanced by surface interactIOns. The Particle diffusivity is calculated from the Stokes-Einstein relationship, and the boundary layer thickness is estimated from Uveque's expression for laminar flow in a crossflow channel, Equation (6.23).
(6.23)
6-15
CHAPTER 6. Theory (Surface Microfiltration) Where dh is the hydraulic mean diameter of the channel entrance. Reynolds number for a channel is given by Equation (6.24)
.pJ....f_v_d.::.. h Re =-
(6.24)
Jl
Vs, the potential barrier between particles, is obtained by integrating the exponent of dimensionless total interaction energy with respect to separation distance, h, between the limits 0 < h <
00,
(6.25)
Yoon et al. (1999) applied the surface interaction model to cross-flow microfiltration of multi-dispersed iron oxide particles. Experimental condiUons are reported in open literature. To summarise, a 0.01 wt % iron oxide (dso
=0.24 J1ID) was dispersed in
10-3 molar NaCI electrolyte. Particle zeta potential of + 50 mV was specified, and a modest cross flow velocity of 0.24 ms'! was maintained in the cross-flow channel. A simulation model was established within the context of this project according to Yoon et al.' s expenmental data, and evaluated using the partial (PDE) and ordinary differential equation (ODE) solver PdesoP'M (v2.0) supplied by Numerica.
V,
was
calculated for different particle sizes, (rp) across a range of interparticle distances, h, from 0 to 10 nm. Equation 6.25 was solved as an ordinary dtfferential equation using an integrator based on the Runge Kutta (RK) Fehlberg fonnulas. The reader is directed to Appendix E for a full text version of the Pdesol™ model for interaction enhanced migration. For a 0.1 J.Ull diameter iron oxide particle, equation 6.25 reached a plateau at a separation distance of h
=H = 3 nm, as h -7
00
(6.26)
6-16
CHAPTER 6. Theory (Surface Microfiltration) This resulted in values of
V; and
V,
of 5.11
X
1058 and 4.44 x 10.5 m S·I, respectively.
Calculation of V, could not be resolved beyond 0.47 J.lffi diameter iron oxide particles separated by a distance of 1.46 nm. On closer inspection it became apparent that the exponential term within the ODE describing VB, became extremely large, very quickly. At the point where the calculation reached an upper limit, VTol./kBT was equal to 709.8, which on taking the exponential became 10308 , i.e. the largest number which the computer could process. This finding, that the model describing interaction enhanced migration is limited to sub-micron particles (for a set of experimental data), contradicts the work of Yoon et al. (1999). Figure 10 suggests that the 'interaction enhanced' back transport model can be evaluated up to particle sizes of 100 !lm. Using the same experimental data however it has been shown that the constituent term for determining VB, and hence v" can be calculated only up to 0.5 J.lffi. A co-author, Anthony Fane was contacted about the severe limitations of v, calculation, but he was unable to help directly. An enquiry was also made to P. Aimar with regards to Equation 6.22, but without reply to date.
6.1.7 Total Back Transport Velocity (Vtotal) The total transport velocity opposite to the membrane surface, V'o,al, was calculated from summation of individual particle diffusion, shear-enhanced diffusion and inertial lift mechanistic terms,
V,olal
= Vd
+
Vs
+
V,
(6.27)
The effect of particle surface interactions on the total velocity opposing permeate flux were considered negligible for the typical particle sizes encountered in the experimental work, i.e. particles larger than 2 !lm in diameter. The mass transport model described by Bacchin et al., was applicable to cross-flow operation only, and limited to small iron oxide colloidal particles. The magnitude of such interactions could not be considered for larger micron-sized latex particles subjected to microfiltration in a stirred-cell. Questions posed by the shortcomings of the model remain essentially unanswered. The importance of double layer repulsion and 6-17
CHAPTER 6. Theory (Surface Microfiltration) London-van der Waals interactions were evaluated as far as possible, and ultimately omitted from the total back transport velocity term. The difference between back transport and permeation velocity is:
(6.28)
The relative magnitudes of the particle transport mechanisms of particle diffusion, shear-enhanced diffusion and inertial lift depend strongly on the particle size and shear rate, and to a lesser extent on the bulk concentration of particles in the feed suspension. Figure 6.8 compares the back transport velocity of the three transport mechanisms across a wide range of particle sizes. The model parameters reflect the
104 Particle diffuSion (v.) Shear-Induced diffuSion (v.>
";"
1/1
E
103 ___ w
:1.
~
'u0
"i > t::
102
---
Inertial 11ft (v,)
/
,,-
-
101
c
10°
..lI: U
10-1
...-
,,-/
."-
././
0
-
,,-
Total (v. + v, + vJ
c..
1/1
/
/
.E'''-
I
I
I
I
I
I
I
I
I
I
I! I'll
rn
10-2 001
0.1
1
10
100
Particle diameter (d,). Ilm.
Figure 6.S Comparison of back transport velocities generated from different particle
mechanisms over a range of particle sizes. Total back transport is determined from particle, Vd, and shear-enhanced diffusion, Vs. and inertial lift, VI, mechanisms only. The Figure is comparable to that presented in Yoon et al. (l999).
6-18
CHAPTER 6. Theory (Surface Microfiltration) total back transport velocity for 30 J.lIll diameter particles and above. Inertial lift velocity is more sensitive to particle size than shear-enhanced diffusion because it varies with particle radius cubed as opposed to a squared term in the latter mechanism. The particle size range of latex challenge material is considered for the three back transport mechanisms in Table 6.1. Shear-enhanced diffusion is the most dominant term in the size range 1 - 22 J.lIll (average dso
= 5 J.lIll) whilst contribution from
particle diffusion and inertial lift is strong for sub-micron particles and larger micronsized particles respectively. Table 6.1 Particle size dependence for various transport mechanisms (size range comparable to latex challenge material). Particle size Particle diffusion Shear-enhanced diffusion Inertial lift Vd (m S·l) V, (m S·l) V, (m S·l) urn 2.13 x Wc, 2 1.11 X 10-6 1.29 X 10.8 7 5 5 5.98 X 10. 7.27 X 10. 2.02 X 10.7 7 10 3.76 X 10. 1.84 X 10-4 1.62 X 10"
All terms are dependent on the shear rate at the membrane surface. The stirred·cell column was operated at shear rates
(S·I)
in the range 0 <
i'w < 1400.
Table 6.2
presents velocity outputs from the three mechanisms for selected shear rates. As seen, shear-enhanced diffusion is the largest term over the shear rate range. Particle diffusion represents the second most significant term up to shear rates of 1200
S·I.
Inertial lift velocity has little influence at low stirrer speeds; achieves the same order of magnitude as particle diffusion at Inld-shear rates; and becomes significant at a relatively high system shear rate, i.e. 1400 S·I. Table 6.2 Shear rate dependence for various transport mechanisms (shear rate range appropriate for constant-rate filtration experiments). Shear rate Particle diffusion Shear-enhanced diffusion Inertial lift S·l Vd (m S·l) V, (m S·l) V, (m S·l) 7 100 2.23 X 108.16 x 10-:0 5.61 X 10-9 7 5 600 5.98 X 107.27 X 102.02 X 10-7 7 1200 8.76 x 101.69 X 10-4 8.08 X 10-7 1400 9.53 x 10"7 2.04 x 10-4 1.10 x 10-6 6-19
CHAPTER 6. Theory (Surface Microfiltration) The dependence of back transport mechanisms to particle size, shear rate, and concentration parameters is summarised in Table 6.3 below. The range of comparison reflects the experimental operating conditions. Table 6.3 Summary of transport mechanisms: Performance with increasing parametric significance (italics) and overall region of influence. Particle diffusion Vd
decrease sub-micron particles Shear rate increase low shear Concentration decrease low Particle size
Shear-enhanced diffusion
Inertial lift
V.
V,
increase small - large particles increase low - hil(h shear decrease low&medmm
increase large particles increase hil(h shear no effect none
6.2 SUMMARY Consideration of the most significant forces and phenomena acting on particles above, and on, a membrane surface suggests that for the particle size range considered in this work the shear-enhanced diffusion is the dominant effect. Hence, equation (6.15) is the most appropriate to apply when calculating the back transport velocity, or equation (6.8) provides the variation in the mass transfer coefficient. Equations (6.4) to (6.9), together with (6.16), provide the full model that will be used in Chapter Eight to understand, or model, the experimental observations. During the development of this model it became apparent that the approach adopted by Aimar appears to have two uncertainties. Firstly, it is difficult to understand how a short-range force can have physical meaning over a long range. Integrating from zero to infinite distance. Secondly, short-range interactions are not normally applicable to particles of 100' s microns diameter, where forces and effects such as weight and sedimentation usually dominate.
6-20
CHAPTER 7. Experimental (Surface Microfiltration)
CHAPTER SEVEN EXPERIMENTAL: SURFACE MICROFILTRATION This chapter describes the experimental methods and materials used to measure the critical flux (Jcnt). A comprehensive account of all the constant-rate filtration experiments is presented in Appendix I.
7.1 STIRRED-CELL TEST RIG The stirred cell is illustrated in Figure 7.1. The cell consisted of a clear acrylic column, with a feed volume of 195 m1 up to the overflow. Towards the top of the colunm were two ports which could be used to recycle the penneate stream back to
~
_ _ Adjusting screw
.....-.I.....;;.~--,
Bearing Operating level
. .- - / - - Stirrer with shallow cone
_~~~~+-_ Membrane
Permeate
Recycled to column, or collected in receiver
Peristaltic pump Figure 7.1 Schematic representation of stirred cell rig.
7-1
CHAPTER 7. Experimental (Surface Microfiltration) the column, or to introduce fresh ultra pure water. Selected column dimensional measurements are presented in Table 7.1. Table 7.1 Selected stirred-cell column dimensions. External diameter (do) 50.2 Internal diameter (d,) 40.0 Height to overflow (Y) 219.0 Capacity to level (Q) 195
mm mm mm ml
A removable membrane holder, shown in Figure 7.2 below, was also manufactured from acrylic and polished to a transparent finish. The membrane to be tested was located on a perforated brass support plate. A thick, soft, rubber o-ring was fitted between the filter and the column. The flanges of the column and base section were bolted together tightly and the compressed o-ring provided a tight seal around the filter. The original screw-thread seal, which caused the thin track-etched membrane to twist on tightening, was replaced with the compression sealing design depicted in the Figure. The acrylic stirred cell was supported firmly by a specially designed clamp stand.
d,
I
1l1li
rt-i I
!lJ
;
---l.~1
Rubber O-ring Membrane
-!~_\-
I
I I-f-,. LJ--+-l
PTFE spacer
Perforated brass support plate
Figure 7.2 Schematic representation of stirred-eell membrane holder.
7-2
CHAPTER 7. Experimental (Surface Microfiltration) A lightweight plastic stirrer rod incorporating a shallow (4° angle) cone (36 mm diameter) was operated close to the membrane surface, to reduce fouling and keep the suspension well mixed. The design of the stirrer was based on a cone-andplate constant shear viscometer. In a constant shear viscometer the tangential shear rate throughout the fluid is independent of radial position (Holland, 1995, pp.96-98). Hence, it is possible to calculate the uniform shear field used in these experimental studies. The reader is referred to Appendix C for an explanation of the adopted constant-shear model. The stirrer rod was driven by a Heidolph RZR 2102 electronic stirrer motor. A two gear system provided rotational speeds, Q, from 12 - 400 rpm (gear I) and 60 - 2000 rpm (gear 11). The desired stirrer speed was selected and monitored using a local LED display. A Radio Spares TM-30ll hand tachometer measured the rotational speed of the stirrer rod. Presented in Appendix F, the stirrer rod calibration data was compared to the corresponding rotational speeds displayed locally, on the stirrer motor.
a
m
~.. _ .. _ .. _ .. _ .. _
.. _ .. __ ....
s .. _ ..
------------------~--Membrane surface
Figure 7.3 Diagrammatic representation of stirrer-rod geometries.
A phosphor-bronze bush provided support to the top section of the stirrer and reduced whip generated along the shaft. The bush was seated in a white nylon, bearing cap,
7-3
CHAPTER 7. Experimental (Surface Microfiltration) which fitted tightly over the column top. Vent holes were added to the bearing cap to ensure that the column was at atmospheric pressure. A brass adjusting screw, located above the bearing, was used to accurately position the stirrer cone tip above the membrane surface. The stirrer cone tip was typically set at a distance s = 1 mm above the membrane prior to each experiment. Figure 7.3, on the previous page, is a schematic representation of the stirrer rod arrangement (not to scale) showing specific dimensions reported in Table 7.2. Table 7.2 Selected stirrer dimensions. Cone diameter (2a) 35.7 1.2 Cone height (m) Cone angle «(J) 0.07 Shaft diameter (w) 21.0
mm mm radians mm
A Radio Spares pressure transducer (1 barg, 0 - 100 mY) was connected to the permeate line between the membrane and peristaltic pump. Tape-sealed screw thread fittings were used to minimise the ingress of air. The transducer was connected to a PICO Technology 16 bit analogue-digital converter and PC. Using PicoLog for Windows software (release 5.02.5) pressure measurements were logged every two seconds. Pressure data could be easily downloaded to a spreadsheet The pressure transducer was calibrated using a Druck DP! 603 portable pressure calibrator. A typical calibration graph is presented in Appendix G. A Watson-Marlow low-flow model 101UIR peristaltic pump with a maximum flow rate of 50 rnl mm-! (5 mm bore tubing) was situated on the permeate line. Silicone rubber peristaltic tubing with different bore dimensions was employed providing flow rate ranges appropriate to each experiment. The tubing dimensions and corresponding flow rate ranges are shown in Table 7.3. Table 7.3 Silicone rubber peristaltIc tubing properties. Tube bore (wall thickness), mm 1.6 (1.6) 3 1.0) 5 (1.5) 0.2 - 7.3 08 - 26.0 1.6 - 53.0 Flow rate ranee, rnl min-' The peristaltic pump is a positive displacement device, but generates pulsatile flow, which has been reported in literature to reduce membrane fouling. The pump flow 7-4
CHAPTER 7. Experimental (Surface Microfiltration) rates measured are average values and do not reflect the higher peak flow rate. Increasing the number of rollers on the pump-head can dampen such flow oscillations. Rotational speed of the pump-head was measured for different bore peristaltic tubing. Figure 7.4 illustrates the performance of the peristaltic pump across its operating range 1 - 32 rpm.
1000
-+-- 1.6 mm bore tubing -&- 3 mm bore tubing
~
'/11
800
-e-
E
6 mm bore tubing
:1.
><
. :J
;;:::
CII I'I:J
600
400
C ==
I'I:J CII
(3
200
o
5
10
15
20
25
30
35
Peristaltic pump rotational speed, rpm.
Figure 7.4 Watson-Marlow lOlUlR peristaltic pump performance characteristics.
A Watson-Marlow 502S peristaltic pump utilising 6 mm bore tubing was used for high flow applications. Depending on the experiment, permeate was either recycled back to the column, as in most cases, or collected in a receiver on an electromc balance (Ohaus precision plus). Once connected to a PC using a purpose built twoway data cable, balance readings could be logged in real-time using a simple QBasic software program. The permeate line was 6 mm bore silicone rubber tubing.
7-5
CHAPTER 7. Experimental (Surface Microfiltration) 7.2 MEMBRANES Polycarbonate (PC) isopore track-etched membranes supplied by Millipore were used with nominal pore sizes of 10 ~ and 5 Ilm. The isopore membranes were manufactured from bisphenol polycarbonate using a unique non-radioactive manufacturing process consIsting of two steps: a tracking step followed by an etching step. Hence the tenn 'track-etched' membrane. During the tracking step fast moving Argon ions (Ar~ break the polymer chains leaving "tracks" in the polymer material. The activated film is submitted to chemical etching phase in which the degraded regions are converted preferentially into straight, cylindrical pores of a defined, unifonn diameter (Millipore Corporation, 1992). The pore density and pore diameter of the membrane can be controlled by adjustment of the tracking and etching steps respectively (Gutman, 1987). The membrane properties as described are typically used for small-scale laboratory applications such as sieving of particles or cells.
CH3
/~
~-o-o-g-o I CH3 n
bisphenol polycarbonate
The physical properties quoted for these track-etched membranes having circular pore openings (CPO) are given in Table 7.4, on the following page. Each 47 mm diameter membrane was used as supplied. The manufacturer states a pore size distribution of +0 % to -20 %, which equates to a minimum pore size of 4
~
for a 5 Ilm membrane.
The pore size distribution reported is detennined either by counting pores of different diameter sizes on SEM micrographs or by performing an air flow test using a Coulter Porometer instrument, according to ASTM method F 316. The membranes have two distinct sides: a shiny side and a matt side. It is recommended that the membranes are used shiny side up (Millipore Corporation, 1992). The data sheet states that the nonnally hydrophobic polycarbonate surface has been treated with a wetting agent,
7-6
CHAPTER 7. Experimental (Surface Microfiltration) polyvinylpyrrolidone (PVP) to make the surface hydrophilic. The membrane was handled using tweezers at all times. Table 7.4 Physical properties of polycarbonate track-etched (CPO) membranes. Pore size t Material
Pore Open geometry area t
(%)
(~)
10.0 5.0
PC film PC film
circular circular
8.0 80
Pore densityt (pores cm·2) 10' 4x 105
Pore size Thickness t range t (~)
(~)
8 -10 4-5
100 15.0
fManufacturer's physical properties (Millipore Corporation, 1992)
Slotted surface filters formed from "Veconic Plus" metal screen were supplied by Stork Veco, with rectangular slot dimensions of 10 x 420 I1IIl (Website http://www.veco.storkgroup.com). The area open to flow was reported to be 5 per cent, similar to the manufacturer's stated value. From pore size analysis it transpired that the actual slot dimensIOns were closer to 13.4 I1IIl wide with a slot length of 402 11IIl. Surface filters with smaller width slots were produced from the 'VeCOnIC Plus' screen using a patented electroless plating technique, the development of which
IS
discussed in Chapter Three (Electroless Nickel Plating Experimental).
Physical property data for untreated and electroless nickel plated slotted filters is presented in Table 7.5. Pore size (slot) dimensions, open area, and filter thickness were measured according to techniques descnbed in Section 3.5. Filters of 47 mm diameter surface filters were used in the stirred-cell rig.
Table 7.5 Physical properties of metal surface microfilters. Pore width
Pore length
(~)
(~)
13.4+ 1.0 10.0+0.7 4.9±0.9 4.1 ±0.4 4.8+02
402.8+0.6 402.1 ± 1.6 396.2± 1.5 382.0±2.5 n1a
Material
nickel nickel nickel nickel nickel
Pore Plating Open Thickness, geometry treatment area slotted slotted slotted slotted circular
virgin plated plated plated plated
(%)
(!lIll)
5.7 4.4 2.1 1.8 0.3
283 -330 -330 -330 88
7-7
CHAPTER 7. Experimental (Surface Microfiltration) 7.3 CHALLENGE MATERIAL Polystyrene latex particles produced by the Balance Swell Method (BSM) were selected as the challenge material (Goodwin et al., 1974) for constant-rate filtration experiments. The uniform, spherical nature of the polystyrene latex particles is evident in Figure 7.5. Several other particle systems including porous polystyrene particles, silica wide pore (VLS) from phase separation and Coulter latex standard suspensions were assessed but too few particles in the deSired size range and wetting problems in the former case, made them unsuitable.
Figure 7.5 SEM micrograph of polystyrene latex particles filtered using a 0.45 ILm cellulose nitrate membrane.
Latex stock suspensions of 1000 ml volume were prepared using filtered (0.1 ILm) ultra pure water (18.2 Ma cm,l conductivity), and stored at room temperature. The addition of Coulter dispersant (Coulter Electronics Ltd) was found to have no significant effect on measured particle size distribution by Coulter Multisizer IT and it was, therefore, concluded that the stock suspension was fully dispersed.
7-8
CHAPTER 7. Experimental (Surface Microfiltration) 7.3.1 Particle Size Determination The particle size and size distributions of the test suspensions were evaluated using a Mastersizer supplied by Malvern Instruments. The Mastersizer employs a laser diffraction technique to generate a particle distribution by volume, which is equivalent to the weight distribution. The concentrated sample was added to a well-mixed feed tank and re-circulated through the cell. The operating envelope of the instrument limited
its use to the relatively concentrated stock suspensions (Website
http://www.malvern.co.uk). The number of microfiltration experiments performed necessitated the use of several latex 'working' suspensions from the original batch which are denoted A-D. Figure 7.6 provides a particle size distribution, by mass, of the stock suspensions. As seen the mean particle sizes (d50 ) varied from 4.3 (working suspension D) to 7.3
~m
(working suspensions A and B). There were significant
particle numbers measured with particle diameters between I
~m
and 22
~m.
Working latex suspensions C and D were used in the filtration experiments
10.-----------------------------------------, A B
suspension suspension suspension suspension
8
C
: .. .. .. / /" .: ."-:" r \ i/ v. \
0
6
; /
.: I
.... "
/
"'. •••.
; I
4
"".
.... I. . . I. .... I
;i f/
j
\ 1.
~
2
\1
.-i
'. \
:1.
".\ ...."\
..
o+---~~~~~~--~~~~~~--~~~·~~ "·~ "~
0.1
10
1
Particle size,
100
~m.
Figure 7.6 Particle size distribution, by mass, oflatex stock suspensions
measured using a Malvern Mastersizer instrument.
7-9
CHAPTER 7. Experimental (Surface Microfiltration) of 10 !lm, 5 !lID, and 4 !lm membranes descnbed later in this chapter. Considering the particle sizes of interest approximately 50 per cent of all particles had diameters greater than 5 !lm, and 32 percent of particles were in excess of 10 !lID size. 7.3.2 Alternative Methods of Particle Size Analysis 7.3.2.1 HIAC ROYCO Particle Sensor and Counter
Located in the permeate line downstream from the filter, the HlAC particle sensor 346-BCL and particle counter (model 4100) measured the number of particles across a given size range. The particle sensor uses a laser light source as a stable reference with constant intensity. A sapphire sensing chamber allows accurate measurement of particle size and concentration (Website http://www.partIcle.com). High number concentration of fine material (particles less than 4 !lID in size) in the latex suspension meant that the HlAC frequently operated beyond the maximum concentration limit. Use of the HlAC therefore was restricted to lower-concentration systems, for example the stirred-cell rig with 2 - 4 !lm pore size membranes.
7.3.2.2 Lasentec PAR-TEC 100 Particle Sizer
A Lasentec PAR-TEC 100 particle sizer was evaluated as an in-Iine/stand-alone device for monitoring latex particle concentratIOns. The Lasentec is a scanning microscope, which measures the size of particles (reported as a chord length) by moving a focused laser beam at a constant velocity across particles in the suspension (Website http://www.lasentec.com). As the focused beam moves across the surface of a particle, back-scattered light is collected and converted to an electrOnic pulse. The mode of operation means it is particularly suitable for non-spherical particles. Unfortunately the low concentrations of latex suspensIOn used in filtration experiments were outside the lower sensitivity limit of the Lasentec unit.
7.3.2.3 Coulter Multisizer 11
Latex suspension samples from the stirred-cell column and permeate stream were often analysed using a Coulter Multisizer IT stand-alone analytical instrument. The Coulter Multisizer IT determines the number and size of particles suspended in a conductive liquid by monitoring the electrical current between two electrodes, on
7-10
CHAPTER 7. Experimental (Surface Microfiltration) eIther side of an aperture through which the suspension is forced to flow (Webslte http://www.particle.com). As each particle passes through the aperture, it changes the impedance between the electrodes and produces a short electrical pulse. The magnitude of this pulse is essentially proportional to the particle volume. Samples of 2 - 10 m1 were carefully collected from the well-mixed stirred-cell column using a syringe and wide-bore sample tube. Extremely dilute permeate samples were also submitted for analysis. A Coulter sample tube WIth a 70 J.Un orifice was used to analyse the latex suspension in the presence of I % w/w salme solution as electrolyte. Analytical raw data was exported into Excel spreadsheets to create a number distribution of the particles. Sample analysis was straightforward but time-consuming as calibration and frequent washing of the Coulter beaker and sample tube was a requIrement. Unfortunately problems associated with satisfactory operation of the instrument restricted its use to occasional analysis. 7.3.3 Particle Zeta Potential Measurements Latex particle surface charge was determined using a Malvern Zetasizer 3000 HSA. The surface charge or zeta potential (~) is a measure of the magnitude of the repulsion or attraction between particles. The Zetasizer equipment uses laser light scattering and electrophoresis to measure the electrophoretic mobility, and hence zeta potential of a dilute suspension (Website http://www.malvem.co.uk). 7.3.4 Latex Suspension Density The density of polystyrene latex suspension was measured as 999.6 kg m·3 using the density bottle technique. Two 25 cm3 glass density bottles with associated stoppers were dried in an oven, removed and cooled in a desiccator. The empty weight of each complete density bottle was recorded. Ultra pure water was added to each bottle, and the stopper carefully replaced ensuring that water filled the capillary space. Any excess water was removed before the density bottles were weighed. Glassware was dried thoroughly in the oven before determining the mass of latex suspension completely filling each bottle. The average density of the latex suspension could then be calculated from water and latex suspension measurements.
7-11
CHAPTER 7. Experimental (Surface Microfiltration) 7.3.5 Latex Stock Suspension Concentration Latex suspension concentration was evaluated by passing a known volume of the suspension through a membrane and determining the mass of solids deposited on the membrane. A 0.4 Ilm track-etched membrane was weighed prior to use. A known' volume of latex suspension was pushed gently through the membrane located in a Whatman filter holder. After drying for 24 hours in an oven, the track-etched membrane was weighed and the solids content calculated. The average concentration of solids in the suspensions was estimated to be 1.8 kg m'3.
7.4 CLEANING NICKEL SPO AND CPO MEMBRANES Nickel SPO and CPO membranes were cleaned in 1 g rl Decon 90 solution prior to an experiment so as to remove particles and enhance the surface wetting properties. Decon 90 is an alkaline emulsion of anionic and non-ionic surface-active agents. The pH of a 1.0 g
rl
solution of Decon 90 was measured as pH 10.6. 200 rnl cleaning
solution was prepared in a glass beaker with ultra pure water. The fouled membrane was contacted With the cleaning solution and sonicated for ten minutes, turning occasionally. Optimisation of the cleaning process saw the Decon 90 solution heated during sonication to approximately 60°C. When SEM images and elemental analysis indicated possible corrosion of the plated membrane surface however, an ambienttemperature cleaning step was adopted. Before use in the stirred cell rig the membrane was sonicated and rinsed well in ultra pure water.
7.5 STIRRED-CELL RIG PREPARATION A 47 mm test membrane was carefully placed on the perforated support plate. In the case of nickel SPO and CPO membranes where each side of the membrane was slightly different, use of an optical microscope ensured that the same side was consistently facing upwards. The thin polycarbonate isopore membrane was wetted slightly to help locate it in position before the column was then lowered onto a rubber o-ring and bolted tightly to the base-section. The compressed o-ring made a seal around the filter, which was clearly viSible as a thin black line.
7-12
CHAPTER 7. Experimental (Surface Microfiltratlon) A combmation of Millipore Elix and Milli-Q 185 PLUS water punfication systems operating in series provided polished ultra pure water with very low levels of ions, organics and particulates (conductivity,
7(=
18.2 MQ cm· l ) for all experiments. Fresh
ultra pure water was collected daily. The stirred cell column and permeate line were flooded WIth ultra pure water. Peristaltic tubing was located in the pump housing and secured. Ultra pure water was backflushed through the membrane for two minutes to remove air from beneath the membrane and from the mdividual membrane pores. Pressure measurements downstream of the membrane were logged at two-second intervals.
7.6 CLEAN WATER MEMBRANE RESISTANCE TEST (MRT) A clean water membrane resistance test (MRn was carried out before each experiment involving a nickel SPO or CPO membrane. The clean water MRT prOVIded a baseline resistance for the membrane to clean water flow, a result that was indicative of the state of fouling. The stirred cell column was filled with ultra pure water to the normal operating level with the stirrer in position, but inactive. Permeate flow was recycled to the column. The peristaltic pump was operated for three mmutes at each flow rate. Pressure measurements were logged every two seconds. The pressure difference across the membrane was monitored at six flow rates spread across the pump output range. Penneate flux was measured usmg an electronic balance with 3 decimal place precision (Ohaus precision plus) and a stopwatch. Temperature of ultra pure water in the column was also monitored using a glass mercury thermometer. Infrequent evaluation of clean water MRT was acceptable for the isopore track-etched membranes since a new membrane was used each time.
7.7 PREPARATION OF LATEX FEED SUSPENSION Polystyrene latex stock suspension was sonicated for 2 minutes in an ultrasonic bath and a further minute using a sonic probe (Branson Sonifier 250) delivering 70-Watts power. A senSItive balance with 4 decimal place precision (Mettler AJlOO) was used to weigh the mass of latex 'working' suspension required for a particular experiment. This weighed amount was carefully transferred to the column half-filled with ultrapure water. The stirrer was duly relocated in position and the column level made up to
7-13
-
--------
CHAPTER 7. Experimental (Surface Microfiltration) the 195 m1 mark (operating level depicted in Figure 7.1, on page 7-1) with ultra pure water. The feed suspension was mixed well before commencmg an experiment. All experiments were performed under condItions of constant permeate rate, rather than constant pressure; hence pressure measurements were used to assess the occurrence of fouling and cntical flux. In most experiments permeate was recycled back to the column top to maintain the bulk feed concentration, in the absence of significant cake formation. It also served to maintain the liquid level. Where a 'once-through' arrangement was specified permeate was collected m a receiver located on the balance. A constant hydraulIc head was maintained by regular addition of ultra pure water to the column.
7.8 CONSTANT-RATE FILTRATION EXPERIMENTS 7.8.1 Determination of Critical Flux (lent) Critical flux was evaluated for all membranes described in this chapter according to a 'flux-stepping' approach. Both constant-pressure and constant-rate filtration has been used to evaluate critical flux for particular systems. Constant-rate operation is generally preferred because it allows for filtration to take place initially at a low flux followed by slow and controlled flux increase in increments of constant flux. The solute transport is better controlled allowing fouling to be accurately monitored using pressure dIfference across the membrane (Metsamuuronen et al., 2002). In constantpressure mode filtration, flux, the dependent variable tends to produce constantly changing conditions in the boundary layer due to the uncontrolled convection of solute toward the membrane surface (FIeld et al., 1995). The detennination of critical flux by constant-rate filtration has found widespread application in literature: Chen et
al. (1997); Madaeni (1997); Kwon and Vigneswaran (1998); Li et al. (1998); Defrance and Jaffrin (1999); Gesan-Guizou et al. (1999); Huisman et al. (1999); Madaeni et al. (1999); Chan and Chen (2001); and Metslimuuronen et al. (2002). The stirred-cell system was operated at constant flux with pressure readings logged at two-second intervals. Below critical flux the pressure dIfference across the membrane is directly proportional to flux. The permeate flux was increased stepwise from an initially low value relative to pore size until the pressure drop across the membrane increased 'signlficantly' with time. The peristaltic pump operated at constant 7-14
CHAPTER 7. Experimental (Surface Microfiltration) permeate flow for 10-15 minutes in the normal direction during each 'flux step'. Permeate was recycled back to the column top to maintain bulk suspension
In
the
absence of cake formation and to preserve a constant static head. The pump flow rate was measured between flux steps using an electronic balance (3 decImal place precision) and stopwatch. Critical flux was reported when membrane fouling caused the pressure difference to increase rapidly during constant flux operation. IrrevefSlble fouling was confirmed by further filtration at a sub-critical flux where fouling had not previously been observed. The membrane was backflushed for two minutes before a final low-flux filtration run. A critical flux was reported for the membrane and experimental conditions between the highest flux under stable pressure conditions (lower value) and the successive flux step observed in a fouling regime (upper value). As the flux step size was small it was considered acceptable to report critical flux as the mid-point value. In this thesis critical flux is primanly reported in units of JlIIl S-I allowing quick comparison with the particle deposition model. Wherever possIble, however, the corresponding flux value in tradItional units of I m-2 h· 1 is shown in parentheses. The simple conversIOn between flux measured in micrometres per second and traditional flux units (I m-2 h· l ) is:
J (I m·2 h·')
= 3.6 x
J (JIms·')
(7.1)
Thorough cleaning of the stirred-cell rig between experiments using detergent and a hard-bristled brush removed any residual latex particles from walls, stirrer, membrane support. etc. SIlicone rubber tubing was flushed through with ultra pure water. 7.8.2 Experiments Involving Larger Pore Sizes Early constant-rate filtration experiments provided comparative performance data between a virgin nickel membrane with slot dimensions of 13.4 x 402 JlIIl and a commercially aVaIlable
10 JlIIl polycarbonate isopore membrane. Based on
manufacturer's quoted pore size and optical microscopy average slot dImensions were imtially understood to be 10 x 420 JlIIl. a size comparable to the isopore membrane. Later pore size analysis of greater accuracy by SEM confirmed that the rectangular pores were actually larger with an average width of 13.4 J.lm. The open area available 7-15
CHAPTER 7. Experimental (Surface Microfiltration) to flow for the slotted membrane was measured as 5.7 per cent, which compared favourably to the manufacturer's quoted open area for the circular pores of 8 per cent. A constant stirrer speed of 800 rpm was employed in these experiments. This corresponds to a shear rate at the filter surface of approximately 1200 s·\. It is roughly equivalent to a crossflow velocity of 1 m s·\ in a 5 mm flow channel filter. A feed concentration of 0.093 kg m·3 (,working' latex suspension A) was used in all experiments described in this section. The experimental method was similar to that introduced in the preceding section with the exception of a backflush between constant-rate flux steps. After a short break, and with the pump direction reversed, the filter was gently back-flushed for two minutes at 269 llm s·\ (968 I m·2 h·\). In these experiments permeate was recycled back to the column top. New nickel slotted and track-etched circular pore membranes were used for each experiment. Samples were collected from the column for particle size analysis using a 10 rnl syringe with a 200 mm metal sample tube attachment (1.5 mm bore). Typically 4 rnI10 rn1 sample volumes were collected from the column middle. Ultra pure water was added to reinstate the column level. Constant-rate filtration experiments were also performed without permeate recycled to the column. In these 'once-through' filtration experiments, permeate was pumped and collected in a receiver, situated on an electronic balance. The permeate flow rate was measured at one minute intervals. The liquid level in the column was maintained by manually adding ultra pure water. 7.8.3 Effect of Shear rate on Critical Flux The influence of shear rate across the membrane surface on membrane fouling and particularly critical flux was studied in a series of expenments for the latex particle challenge system and a selection of membranes reported in Table 7.6, on page 7-19. Nickel slotted membranes with reduced slot dimensions of 10.0 llID, 4.9 llm, and 4.lllID were evaluated. Experiments were also performed with 5.0 llm polycarbonate isopore membranes and a nickel membrane with circular slots of 4.8 llID diameter. The open area of these membranes was reported earlier in Table 7.5, on page 7-7. Stirrer speeds in the range 0 - 933 rpm were investigated in a random order. T1us corresponds to an operating envelope for shear at the filter surface of 0 to 1400 s·\. A 7-16
CHAPTER 7. Experimental (Surface Microfiltration) two-gear Heidolph motor provided stirrer speeds up to 267 rpm and 933 rpm through gears I and IT respectively. Latex feed concentrations of 0.020 kg m·3 were prepared from 'working' latex suspensions C and D. It is worth noting that each set of experiments was completed using the same suspension. In the case of zero shear at the filter surface the stirrer rod remained located in position, stationary above the membrane. In Table 7.6, on page 7-19, experimental conditions including flux step size and duration, temperature of feed and the number of experiments completed are presented. Whilst a new isopore membrane was supplied for each experiment such convenience was not afforded for manufactured metal surface filters. Nickel SPO and CPO membranes, with the exception of the virgin filter were subjected to the vigorous cleaning step described in Section 7.4. 7.8.4 Effect of Concentration on Critical Flux The influence of feed concentration on critical flux was studied in a series of experiments at latex feed concentrations in the range 0.0035 - 0.40 kg m·3 (latex 'working' suspensions C and D). The resulting data together with grade efficiency work (described in the next Section) provides critical design information for use as a classifier. 4.9 J.1lIl nickel slots and 5.0 Ilm polycarbonate circular pores were compared. A constant stirrer speed of 400 rpm was employed in all experiments. This is equivalent to a shear rate at the membrane surface of approximately 600
S·I.
Selected experimental conditions for each membrane are presented in Table 7.7. A new MiIIipore isopore membrane was tested each time. 7.8.5 Grade Efficiency The grade efficiency of each membrane was investigated. In this case, gravity drainage was used instead of the peristaltic pump so that permeate samples, for size analysis, were obtained as close to the filter as possible. Figure 7.7, on page 7-18, shows the simple experimental arrangement used. The short permeate line discharged into a collection beaker, and two valves in the permeate line were used: the lower was set to control the flow, whilst the upper valve opened/closed the permeate line. There was no pressure measurement but the control valve provided an initial permeate flux well below that of the critical flux for corresponding conditions. The column was allowed to drain and the experiment was stopped when the liquid level reached the
7-17
! o H A P T E R 7. Experimental (Surface
MicrofiItra~On)
stirrer cone. Penneate samples were collected for particle concentration and size analysis. Samples from the column were also gathered at the beginning and end of the experiment for concentration and size analysis. The size distribution of latex particles in the samples was measured using a Coulter Multisizer H. A sample tube with a 70 I1IIl orifice was used for all analyses. Aliquots of each well-mixed sample were transferred to the electrolyte solution, using a Volac high precision micropipette, enabling calculation of the particle concentration present
in the stirred cell for each size grade.
_ _ _ Adjusting screw
+_+-_
Stirrer with shallow cone
Membrane
Penneate collection beaker
Electronic balance
Figure 7.7 Schematic representation of gravity-run stirred-cell rig to determine grade efficiency.
7-18
Table 7.6 Selected experimental conditions for cntical flux evaluation as a function of shear rate. Membrane
Ni Ni Ni Ni PC isopore
Pore geometry slotted slotted slotted circular circular
Pore size (urn) 10_0 x 402.1 4.9 x 396.2 4.1 x 382.0 4.8 5.0
Shear rate range (S-l)
Latex feed concentration (kg mol)
0-1200 0-1400 0-1200 0-1400 0-1400
0.02 0.02 0.02 0.02 0.02
Average flux step size (urn S·l) 83 (299) 7 (25) 9 (32) 4 (14) 6 (22)
Average fluxstep duration (mins) 10.2 11.2 9.8 10.0 10.7
Feed suspension temperature
Sample size
(OC)
22.5 - 26.0 22.5 - 28.0 21.0 - 24.0 22.0 - 26.0 22.5 - 24.5
12 22 14
13 11
Values reported In parentheses refer to flux measured in traditional units of I m-2 h- 1
Table 7.7 Selected experimental parameters for critical flux evaluation as a function of latex feed concentration. Membrane
Pore geometry
Ni PC isopore
slotted slotted
Pore size (urn) 4_9 x 396.2 5.0
Shear rate range (S-I) 600 600
Feed concentration range (kg mol) 0.0052-0.4 0.0035-04
Values reported in parentheses refer to flux measured in traditional units of! m-2 h- 1
Average flux step size (urns-I) 14 (50) 4 (14)
Average flux-step duration (mins) 10.5 10.4
Sample size 19 21
CHAPTER 7. Experimental (Surface MicrofiltratlOn)
7.9 SUMMARY Application of a small-scale stirred-cell experimental rig enabled the incremental investigation of critical flux (lent) for a range of circular and slotted pore surface microfilters. The design of the stirrer used in the acrylic cell was based on a cone-andplate constant shear viscometer in which the tangential shear rate throughout the fluid is independent of radial position. The effect of shear rate and concentration on critical flux were evaluated independently for polycarbonate Isopore circular pore (5 and 10 microns), nickel slotted (4, 10 and 13 microns) and nickel circular pore (5 microns) membranes. Complete record of these experiments are presented in Appendix I. Spherical polystyrene latex particles produced by the Balance Swell Method (BSM) represented the challenge material. Critical flux was evaluated according to a 'flux stepping' approach whereby the permeate flux was mcreased stepwise until the pressure drop across the membrane increased 'significantly' with time. Below critical flux the pressure difference across the membrane is dlfectly proportional to flux. A mid-point value was reported for the membrane and experimental conditions between the highest flux under stable pressure conditions and the successive flux step observed in a fouling regime. Grade efficiency of each membrane was evaluated using gravity drainage operation so that permeate samples were obtamed as close to the filter as possible.
7-20
CHAPTER 8. Results and Discussion (Surface Microfiltration)
CHAPTER EIGHT RESULTS AND DISCUSSION: SURFACE MICROFILTRATION This chapter comprises essentially three main sections. Stirred-cell experiments to compare the filtration performance of two true surface microfilters With circular and slotted pores of approximately 10 ~ are considered in the first instance. The focus of the study then shifts to membranes having smaller pore openings. Constant-rate filtration experiments involving circular and slotted pore openings approximately 5!lm in diameter are presented together with results for smaller 4
~
slots. The
relatIOnship between the hydrodynamic parameter of shear rate at the membrane surface and cntical flux (lent) forms the basis of the discussion. The particle deposition model introduced in Chapter Six (Surface Microfiltration Theory) is evaluated against the experimental results. In addition the significance of the specific membrane material is briefly outlined. Finally, the effect of particle feed concentration on membrane fouling for the smaller pore-size membranes is discussed. A concise summary concludes this chapter. All the results presented were obtained using the stirred-cell rig and latex feed suspension, and by following the experimental methodology described in Chapter Seven (Surface Microfiltration Experimental). Results and selected operating conditions for all filtration experiments are tabulated in Appendix I.
8.1 CONSTANT·RATE FILTRATION WITH NICKEL LARGE PORE (SPO) AND POLYCARBONATE ISOPORE (CPO) MEMBRANES The comparative filtration performance of membranes having circular and slotted pore geometries was evaluated in the stirred cell under identical experimental conditions. Constant-rate filtration experiments were performed using microfiltration membranes with relatively large pore sizes. A commercially available 10.0 !lm polycarbonate isopore (epO) membrane and virgin nickel (SPO) membrane with slot dimensions of 13.4 !lm x 402 ~ were studied. When the work was originally carried out tlie rectangular pores were analysed using an optical microscope and tlie average slot dimensions were tliought to be approximately 10 x 420 !lm, a slot size comparable witli tlie track-etched membrane and in agreement witli the
8-1
----------- CHAPTER 8. Results and Discussion (Surface Microfiltration)
manufacturer's dimensions. However, it has tran spired from subsequent pore size analysis by a scanning electron microscope that the rectangular pores were larger with an average width of 13.4 llm . Whilst this removes the possibility of direct comparison between equal sized s lotted and circul ar pores, the results nonetheless provide useful information about the fi ltration capabi lities of the membranes in question .
8.1.1 Membrane Properties The isopore track-etched (CPO) membrane was exam ined using electron microscopy. As seen in Figure 8.1 (a) and (b) the manufacturer's quoted nominal pore sizes are similar to the s izes observed. The manufacturer states that the pore size distribution is +0 % to -20 %, giv ing a minimum pore size of 4 Ilm . The random nature of the irradiation process used to manufacture these membranes can be seen in the indiscriminate scatter of pores. The ex istence of 'doublets' or ' double shots' is also evident in these Figures. Doublets arise when two or three holes are so cl ose that they become one single pore, larger than it should be. The membranes are not symmetric
8-2
CHAPTER 8. Results and Discussion (Surface Microfiltration)
Figure 8.1 SEM micrographs of 10.0 JJ.m Millipore po]ycarbonate isopore membrane with circular pore openings; viewed at (a) low magnification , (b) high magnification.
Figure 8.2 SEM micrograph of nickel slotted membrane with average pore dimensions 13.4 x 402 JJ.m ; viewed at standard magnification.
8-3
CHAPTER 8. Results and Discussion (Surface Microfiltration) and the appropriate filtration side is shown for each membrane. A micrograph of the nickel slotted membrane, Figure 8.2 on the previous page, shows the consistent width of the slot with the exception of the extremities where the slot tapers slightly. The open area available to flow for the SPO membrane was measured as 5.7 per cent, which compares favourably to the manufacturer's quoted open area of 8 per cent for the circular pores.
8.1.2 Critical Flux Measurement (Permeate Recycled) A technique of 'flux stepping', described in Section 7.8.1, was employed to determine the critical flux for the circular and slotted membranes. The permeate flux was increased step by step from an initially low flux of 35 ~m
S·I
(126 I m'2 h' l) until the
pressure drop across the membrane increased 'significantly' with time. At each step the permeate flux was maintained for 10 minutes with the permeate recycled to the column top, followed by backflushing (reversal of flow direction) for two minutes at a measured flux of 269 ~m
S' I
(968 I m'2 h'\ Incremental flux steps were not uniform.
A constant stirrer speed of 800 rpm was employed in all the experiments described. This corresponds to a shear rate at the membrane surface of approximately 1200
S'I.
The initial latex feed concentration (working latex suspension A) was 0.093 kg m'3 Following each lO,minute flux run, the recycle stream was temporarily sampled to allow flux measurement by stopwatch and electronic balance. Pressure readings were recorded at 2-second intervals. As the experiments were performed at a constant rate, any increase in the pressure drop across the membrane was indicative of membrane fouling. Figure 8.3 on the following page shows the pressure difference across the membrane (in mbar) for the CPO and SPO membranes, as a function of filtration time. Permeate flux values (10,6 m S'I) are displayed next to the relevant data points. The oscillatory nature of the peristaltic pump cycle is evident in the pressure measurements, which vary between maximum, and minimum values. The pressure drop profile was analysed according to 'peak' pressure values only. Initially, the CPO membrane showed no sign of fouling at an imposed flux of 35 ~m S· I (126 I m'2 h'\ Doubling the permeate flux resulted in a rise in the pressure drop during the ten-minute step, which suggests that the membrane fouled under these conditions. Moving from a non-
8-4
CHAPTER 8. Results and Discussion (Surface Microfiltration) fouling regime to a fouling reglffie, after increasing the flux may indicate the existence of a critical flux (Jcrit) between 39 - 75 flm
S-I
(140 - 270 I m-2 h- I) . In a
repeat experiment (not shown), there was further evidence of a critical flux between
36 - 73 flm
S-I
(130 - 263 I m-2 h- I) for the CPO membrane. Following a backflush,
and for the experiments at permeate rates less than 223 flm
S-I
(803 I m-2 h- I), the
pressure drop returned to the clean water value of 5 mbar, which suggests that the fouling was reversible. This behaviour is consistent with fouling by cake layer formation rather than pore plugging, where backflushing is usually less successful in returning the membrane to its original state.
250 -r-----------------------------------------------,
..:
x 13.4 micron nickel SPO membrane + 10.0 micron isopore
200
0
E
epo membrane
epo membrane
Flux, Ilm So, .
epo membra ne
150
o
20
40
60
80
100
140
120
Time, mins.
Figure 8.3 Comparison of pressure differences (mbar) for 10.0 flm isopore CPO
and 13.4 flm nickel SPO membranes with permeate recycle (1200
S-I
shear;
0.093 kg m-3 latex feed concentration).
Studying the feed suspension in the column provides a further insight into the fouling of the membrane. Because any particles passing through the membrane are returned back to the column, it is to be expected that the challenge size distribution would
8- 5
CHAPTER 8. Results and Discussion (Surface Microfiltration) change significantly during the experiments when a surface cake formed. Five samples were taken from the middle of the colunm at various times and analysed using the Coulter Multisizer II. Particles in the recycle tube at the time of sampling were not considered significant because the recycle volume was small and the residence time low.
Table 8.1 Particles remaining in suspension as a function of initial feed concentration. Sample collection Flux Flux Flux Flux
step 2 (end) step 4 (end) step 6 (end) step 8 (end)
Percentage of initial number of par ticles 2um Sum 8 um 67.6 70.6 86.7 40.5 58.5 76.4 16.9 35.3 72.2 12.2 25.7 53.2
Table 8. 1 shows the change in the number of 2 Ilm, 5 Ilm and 8 Ilm particles in suspension at specific times throughout the experiment. The rate at which particles were deposited on the CPO membrane does appear to be related to particle size. On completion of the experiment following a permeate flux of 578 Ilm
S-I
(2081 I m-2 h-\ there were only 12 per cent of 2 J.!m particles still in suspension compared to 53 per cent of 8 Ilm particles. Superficially, it is surprising that there should be less reduction in larger particles than smaller ones. However, segregation of particles within a deposited cake is well known (Blake et al., 1992) and the size distribution data in Table 8.1 show that stirring was more effective at removing the larger particles than the smaller ones from the thin cake layer deposited on the membrane. It is noticeable that the percentage of all particles reduced as the flux was increased, indicating that the cake thickness increased with increasing permeate flux. Using this membrane, the pressure drop increased whenever the stirred cell was operated at a flux of 75 Ilm
S-I
(270 I m- 2 h-I), or higher. The pressure drop across a
surface filter will increase whenever pores become obstructed. It is clear from the particle size analyses that large numbers of particles were deposited on the membrane surface during the filtration runs. As the challenge material was substantially smaller than the pore size it is likely that pore bridging occurred, leading to the formation of a cake layer.
8-6
CHAPTER 8. Results and Discussion (Surface Microfiltration) The SPO membrane showed no sign of fouling at a flux of 479 ~m
S-I
(1724 I m-2 h-I).
The pressure difference however was consistently in excess of the clean water value. At a sub-critical flux of 122 ~m
S-I
(439 I m-2 h- I) for example the pressure difference
was double that for clean water suggesting there was some deposition on the membrane. In a further experiment, the permeate flux was gradually increased to 12,000 ~m
S-I
(43,000 I m-2 h-\ over fourteen constant flux steps, to investigate the
existence of a critical flux for this membrane. All other experimental conditions were as stated previously. The performance of the membrane across the flux range 40 12000 ~m
S-I
(144 - 43,000 I m-2 h- I) is shown in Figure 8.4.
Figure 8.4 Pressure difference (mbar) across 13.4
~
nickel SPO membrane at
elevated permeate f1uxes (1200 S-I shear; 0.093 kg m-3 latex feed suspension).
Permeate flux values (10-6 m
S- I)
are displayed next to the relevant data points with
underlined values representing identical peristaltic pump settings. The pressure plot has been split into four zones A, 8, C and D. Zone A was essentially a repeat of the experiment featured in Figure 8.3 (on page 8-5). There was no suggestion of fouling during these e ight flux steps. When the flux was increased from 437 ~m 2405 ~m
S-I
S-I
to
(1573 to 8658 I m-2 h- I) over three steps (zone 8), there was a 2 mbar 8- 7
CHAPTER 8. Results and Discussion (S urface Microfiltration) increase in the pressure drop. In an attempt to elucidate the critical flux, the flu x was increased consecutively to 5083 flm
S-I,
7779 flm
S·I ,
and 12 173 flm
S- I
(182991 m-2 h-I , 280041 m-2 h· l , and 43823 I m-2 h-I) . The results from these high flux runs can be seen in zone C. There was evidence of fouling as the pressure drop increased to 55 mbar at the highest flux , and rose during each step. The pressure drop consistently returned to the clean water value after backflushing, which shows that the fouling was not permanent. On reducing the permeate flux to 267 flm
S- I
(96 1 I m-2 h-I) (zone D) the clean water
pressure drop of 5 - 7 mbar, was once again maintained, suggesting that the previous fouling was completely removed. 8.1.3 Particle Carryover from Column and Tubing Walls The affinity of pol ystyrene latex for surfaces meant that particles were always deposited on the acrylic column walls, plastic stirrer rod and silicone rubber peristaltic tubing during an experiment. Kwon and Vigneswaran (1998) experienced a similar problem during crossflow microfiltration of latex particles where particle adhesion to the pipe, tank and other solid surfaces was reported. Although the column walls and stirrer rod were scrubbed with detergent solution between experiments, the tubing was maintained in service for considerable time. The potential therefore for carryover of particles to subsequent experiments was an unknown quantity. The collective significance of these residual latex particles in relation to the feed suspension was tested by means of a simple experiment. Particle re-entrainment from the tubing wall was also evaluated. Latex suspension was carefully drained from the stirred-cell column and related tubing on completion of a filtration experiment involving permeate recycle. The test membrane was removed and the column re-assembled without the usual thorough cleaning step (see Chapter Seven 'S urface Microfiltration Experimental'). The stirred cell was operated under normal conditions with no membrane (al though the filter support was present), at a relatively fast rotational stirrer speed of 800 rpm. Ultra pure water was pumped around the column in a continuous loop for 20 minutes at a modest flux . During this time, the silicone rubber tubing was tapped firmly at regular intervals to make as many particles as possible leave the tube walls. A column sample analysed using the Coulter Multisizer II showed that residual latex from the walls and tubing represented just 3.7 per cent of 8-8
CHAPTER 8. Results and Discussion (Surface Microfiltration) the final column concentration from the filtration experiment. The particle size distribution was simi lar to the feed suspension. The impact therefore of particle reentrainment from the tubing walls was negligible. 8.1.4 Critical Flux Measurement (Permeate Collected)
The filtration performance of these CPO and SPO membranes was also assessed using the 'once through' experimental set-up described in Section 7.8.2. In this case, the column volume was maintained by the addition of ultra pure water. Removal of the recycle meant that as the experiment progressed there was less challenge material available to foul the membrane. The other operating conditions were as stated previously. The performance of both membrane types can be seen in Figure 8.5. The CPO membrane first showed signs of fouling at a flux of 148 Iilll S'I (533 I m'2 h'\ when the pressure drop increased steadily as the membrane fouled. This result differs from the experiments described in Figure 8.3 (on page 8-5), where there was significant fouling at 75 Iilll S'I (270 I m'2 h'I). However, critical flux is not unique to
250
x 13.4 micron nickel SPO membrane
.; ctI
.Q
E
200 '
+ 10.0 micron isopore GPO membrane
D
a)
Flux.).lm s" .
()
c: ~
150 .
"C
...::::I
~ \riID~r
100 .
Cl)
VI VI
... Q. Cl)
I 353 I
epo membrane
Cl)
;;:
@J
50 '
0 0
20
'" & 40
r ;
j''''' : '''' ' 60
100
80
Time, mins.
Figure 8.5 Comparison of pressure differences (mbar) for 10.0 /-lm isopore CPO
and 13.4/-lm nickel SPO membranes without permeate recycle (1200 S'I shear; 0.093 kg m,3 latex feed concentration). 8-9
CHAPTER 8. Results and Discussion (Surface Microfiltration) the membrane itself but depends upon filtration conditions and, during these experiments, the concentration of fine particles within the challenge would be less than in the recycled experiments. It is therefore difficult to define a sin gle critical flux for a particular membrane when studying different experiments. The success of the backflush (at 33 ~m S' I permeate flux) at clearing the membrane pores again suggests reversible fouling. The SPO membrane showed no sign of fouling at a flux as high as 492 ~m S' I (1771 I m'2 h'I), as the average pressure drop during a filtration remained constant. This supports the existence of a much higher critical flux for the slotted membrane, however, as the pore width of the SPO membrane is greater than the diameter of the CPO membrane this improvement in flux is likely to be exaggerated. 8.1.5 Grade Efficiency The results from the previous experiments demonstrate that the slotted pore geometry is probably less prone to fouling than the circular pore, for similar shear and other filtration conditions. However, good permeate flux performance would also be provided by a filter that did not retain any particles. Hence, it was important to investigate the particle removal efficiency of the filters. Clearly, a filter providing cake filtration will show hi gh removal efficiency because of the formation of a secondary filtering membrane; so care has to be exercised in both the experimental investigation and in the interpretation of the data. For these experiments a single pass was used; the column drained under gravity and samples of permeate and challenge suspensions were taken for concentration and particle size measurement as described earlier in Section 7.8.5. Five permeate samples were collected and analysed using the Coulter Multisizer
n.
The Grade efficiency of the CPO membrane can be seen in Figure 8.6, on the next page. There are two sets of data presented: the initial permeate sample (taken over the filtration of the first 43 ml) and an average of the subsequent four samples (43 - 184 ml filtered). Considering the first permeate sample, it is surprising that at least 60 per
cent of particles 1.4
~m
to 8
~
were removed by the I 0.0
~
membrane. Complete
cut-off appears to take place at approximately 8 ~m . After the initial 43 ml had passed through the membrane, particles of all sizes were removed with an efficiency of 99 per cent or higher. This result supports the conclusion that a secondary membrane
8-10
CHAPTER 8. Results and Discussion (Surface Microfiltration)
100
u
v
oSample P1
00 0~
00 0
80 -
~
" Samples P2 - 5 (average)
&'Al'oa:Po
(.)
r:::
0
.~
60
(.)
:::
0'0
Cl)
I: 0
40
;; (.)
Cl)
0
20
()
0 0
10
5
20
15
Particle diameter (d p ), Ilm .
Figure 8.6 Grade efficiency of 10.0 Ilm isopore CPO membrane. Permeate sample PI (up to 43 ml filtered). Average of permeate samples P2 - 5 (43 - 184 ml filtered).
100 -r------------------------------y------------, o Sample P1
-;!.
80 -
" Samples P2- 5 (average)
~
(.)
I: Cl)
'u
:=
o
60 -
o
o iZI "
Cl)
r:::
o
40 -
;; (.)
Cl)
"0
()
20 o O -~~~~~~~~~~_r~--r_~,_~~~--r_~~~
o
5
10
15
20
Particle diameter (d p ), Ilm.
Figure 8.7 Grade efficiency of 13.4Ilm nickel SPO membrane. Permeate sample PI (up to 40 ml filtered). Average of permeate samples P2 - 5 (40 - 179 ml filtered). 8- 11
CHAPTER 8. Results and Discussion (Surface Microfiltration) (or cake) quickly formed . The grade efficiency data for the SPO membrane is shown in Figure 8.7 on the previous page. Once again, the initial collection efficiency (40 ml filtered) and average collection efficiency for the remainder of the experiment (40 -
179 ml filtered) are presented. Collection efficiency increases with patticle size, but the data is scattered with no obvious trend with respect to volume filtered. This result is consistent with a filter on which a secondary membrane did not form . The SPO membrane does not significantly remove smaller particles, which were in the majority in the challenge material. Less than 30 per cent of particles below 5 ).J.m were filtered. An approximately fitted curve to the data indicates that the cut-off size for this membrane type is around 13 ).J.m, which is consistent with the maximum slot width measured using the SEM images (Figure 8.2 on page 8-3). Column samples collected at the end of each grade efficiency experiment were analysed to study the concentration of challenge suspension particles during the filtration. When a membrane is successfully filtering, the concentration of the challenge should be greater than that of the initial feed concentration as particles are retained and, possibly, redistributed by the shear conditions. The initial particle concentration
In
both
grade
efficiency
experiments
was
approximately
8.4 x 106 particles per ml, the final concentrations in the retained suspensions for the SPO and CPO experiments were respectively 10 x 106 and 61 x 106 particles per ml. Hence, it is evident that filtration occurred for both the CPO and SPO membranes but was more efficient in the case of the CPO principally because of the formation of the secondary membrane.
8.2 CONSTANT-RATE FILTRATION WITH 10.0).J.m NICKEL (SPO) MEMBRANE 8.2.1 Membrane Properties A nickel slotted membrane was successfully produced with a slot width comparable to the 10.0 Jlffi circular pores of the polycarbonate isopore membrane. Slot dimensions of the virgin nickel SPO membrane (13.4 x 402 ).J.m) were reduced to 10.0 x 402 ).J.m by application of electroless plating technology described in Chapter Three
8-12
CHAPTER 8. Results and Discussion (Surface Microfiltration)
Figure 8.8 SEM micrograph of nickel slotted membrane with average pore dimensions 10.0 x 402 Ilm; viewed at (a) standard magnification , and, (b) high magnification.
8-13
CHAPTER 8. Results and Discussion (Surface Microfi ltration) (Electroless Nickel Plating Experimental). A consequence of reducing the pore dimensions is to decrease the open area available to flow. In this case, the open area was estimated to be 4.4 per cent (see Section 3.5.3). Figure 8.8 (a) and (b) are SEM micrographs of the nickel-plated membrane with reduced slot dimensions (cf Figure 8.2 on page 8-3).
8.2.2 Effect of Shear Rate on Critical Flux The effect of shear rate on critical flux was studied for the 10.0 f!m nickel slotted membrane in constant-rate filtration experiments similar to those described for the 10.0 f!ffi CPO and 13.4 f!m SPO membranes. The permeate flow was systematically increased until a rapid rise in the pressure drop was observed. A further permeate flux step, which was weB below critical flux confirmed the transition to a fouling regime. There was no back flush between permeate flux steps.
80
o
..: C\l
.0
E
Flux, I'm s" .
+ +
60
f
Q) (J
C
...
!l-
Q)
£ '0 ... :J
40
Q)
f/I f/I
... a.
20
Q)
I
18961
IS6s1
927'-J.
r
0 0
5
10
15
20
25
30
35
Time, mins,
Figure 8.9 Pressure profi le of 10.0 f!m nickel SPO membrane up to the critical flux (600 S-1 shear; 0.020 kg m'3 latex feed concentration).
Figure 8.9 shows the pressure difference across the membrane (in mbar) as a function of filtration time for a constant-rate filtration experiment. The permeate flux 8-14
CHAPTER 8. Results and Discussion (Surface Microfiltration) (10-6 m s-1) is reported above the pressure data. The stirrer was operated at a rotational speed of 400 rpm, which equates to approximately 600
S- 1
shear at the membrane
surface. The initial feed suspension of 0.020 kg m-3 was almost five times more dilute than suspensions used in earlier experiments (Figures 8.3 to 8.5). It was found that low feed concentrations provide a wider operating envelope in which to study critical flux. Below 0.020 kg m-3, critical flux is well defined for even small changes to the controlled stirrer speed. It is of practical interest that feed concentrations of this order are vastly in excess of the Cryptosporidium Parvum oocysts found in the environment. To give some indication of particle numbers 0.020 kg latex per m-3 suspension will contain around 3
X
109 particles per litre. The 10.0 ~m nickel CPO
membrane was cleaned vigorously between experiments using concentrated Decon 90 detergent solution (see Section 7.4). A clean water membrane test before each experiment confirmed the clean state of the membrane; procedural details are reported in Section 7.6. At the initial permeate flux of 868 J.Lm
S- 1
(3125 I m-2 h-' ) the pressure difference
across the membrane was stable increasing by less than 2 mbar during ten minutes of constant flux operation. Stable operation was further maintained when the flux was stepped up to 896 J.Lm
S- 1
(3226 I m-2 h-1) . During the third flux step the pressure drop
increased rapidly suggesting significant membrane fouling. The permeate flux was above the critical flux. The pressure drop increased steadily from 5.0 to 13.9 mbar during the first 8 minutes as latex particles were deposited on the membrane surface. Large slot dimensions relative to the challenge material mean that direct plugging of pores is unlikely, although it is possible that a group of particles could form a bridge over the slot. In all probability the few large particles capable of lodging in the slotted pores are more likely to be sheared from the surface in preference to smaller particles. The presence of straight-through pores eliminates internal fouling of the medium. Particles are deposited on the membrane surface over time leading to formation of a fouling layer. As pores become blocked greater flow through the remaining pores enhance the state of fouling, a situation illustrated in Figure 8.9 by the rapid increase of pressure difference achieving 66 mbar before the run was tenninated. A critical flux exists for the 10.0 J.Lm nickel SPO membrane, and experimental conditions stated, between the highest flux under stable conditions and the successive flux step, which describes a fouling regime, i.e. between 896 ~m
S-1
and 927 J.Lm
S-1
(32261 m- 2 h-1 and 8-15
----------
-
--
--
-
--
- - - - - - - - - - - - -- - - -
CHAPTER 8. Results and Discussion (Surface Microfiltration) 3337 I m-2 h- I ). These respective values are referred to as the ' lower' and 'upper' fluxes within the critical flux range. Since large flux step sizes were avoided the 2
critical flux was specified as the mid-point flux, or 912 !lm S-I (3283 I m- h-
I
).
Reasons for the small rise in the pressure drop during stable flux operation include the influence of changing hydrodynamic conditions and minor particle deposition. According to Darcy's law, when the permeate flux, and hence interstitial velocity through the pores is increased, the pressure drop will increase. A pressure rise of this nature however, although a contributory factor, cannot account for the gradual pressure rise observed during each constant flux run. The peak pressure drop value for example increased from 1.0 to 5.0 mbar between the start of the experiment and the end of the second flux step. Without studying the membrane surface in real-time using a technique such as direct observation through the membrane (DOTM) it is impossible to confirm particle deposition below critical flux (Li et al., 1998, 2000; Mores and Davis, 2001). A finding of this nature would contradict the strong definition of critical flux observed by many authors where no particles deposit below critical flux . It is possible that the blocking of pores moves from a slow process to a very rapid phenomenon when a critical value is exceeded.
r
200 -
..:
D
co
.c
150 -
ai 0 c: Q)
100 -
E
...
Flux, I'm s-' .
--
Q)
... :::J Q)
I/) I/) Q)
... a.
•
833 1 11
:; '1J
of
50 -
1868 1
Critical flux range
1896 1
-
f
f
:t:
1927 1.J
1833 1 :j:
0 lower value
upper value
)""
backflush
-50 0
10
20
30
40
50
Time, mins,
Figure 8.10 Complete pressure profile of 10.0!lm nickel SPO membrane (600
S-I
shear; 0.020 kg m-3 latex feed concentration). 8- 16
CHAPTER 8. Results and Discussion (Surface Microfiltration) The complete pressure profile for the experiment is shown in Figure 8.10, on the previous page, A retum to sub-critical flux operation following the third flux step led to severe membrane fouling. This is illustrated by a steep rise in the pressure drop across the membrane. The recorded pressure, which surpassed the previous highest value in seconds, achieved 200 mbar after a short filtration time. The peristaltic pump was operated at a setting equivalent to the initial flu x step. Comprehensive fouling of the membrane restricted the permeate flux to 833 J.lIIl S' l, compared to a flux of 868 !lm so l achieved before critical flux was exceeded. This would suggest that the fou ling layer was not removed by shear at the membrane surface at zero permeate flow . However, the success of a short backflush at 469 J.lIIl sol implies that membrane fouling was almost completely reversible. The pressure drop approached clean water values immediately after the backflush, but increased steadily thereafter from 4.0 to 6.0 mbar during 5 minutes of sub-critical flux operation.
2000
!
I ';"
rn
1600
!
E
::1.
."
.-".
•
1200
'"")u ~
><
::::s
;;::
800
•
Cij
-
.!:!
. ;: ()
•
J: J:
!
400
•
J:
J:
0 0
200
400
600
800
1000
1200
Shear rate at the membrane surface (Yw)'
1400
1 5. •
Figure 8.11 Critical flux of 10.0 !lm nickel SPO membrane as a function of surface shear (0 - 1200 sol shear; 0.020 kg m'\
8-17
CHAPTER 8. Results and Discussion (Surface Microfiltration) Mid-point critical flux values (10-6 m S-I) for a wide range of shear rates are presented in Figure 8.11, on the preceding page. Tie bars indicate the step size between the lower (stable) and upper (dynamic) permeate flux. The influence of shear rate on critical flux was investigated in a non-sequential order with the aim of eliminating trends in the data caused by possible changes to the stirred-cell or membrane performance over time. The large slotted pores are clearly sensitive to shear at the membrane surface. There was a sizeable increase in critical flux from approximately 346 to 1759 J.lm
S-I
(1246 to 6332 I m-2 h- I ) between low shear regions (100
high shear regions (1200
S-I)
S-I)
and
respectively. A combination of large rectangular slots
and dilute feed concentration resulted in enormous permeate fluxes. The slotted pores were substantially larger than the material being filtered with many particles passing straight through. Large particles, of which there were only 13 per cent or so, had the potential to jam tight in the slot. Particles whose diameter approached that of the slot could sit in the slot resting against the concave sides. 8.2.3 Type of Critical Flux There remains some ambiguity as to the exact definition of critical flux (leri,). Field et al. (1995) defined critical flux as the flux below which a decline of flux with time does not occur; above it fouling is observed. In the strong form of the hypothesis the pressure difference across the membrane is equivalent to the clean water value at the same permeate flux. Alternatively the weak form describes a situation where there is no progressive fouling but the pressure difference is greater than the clean water value. It is important to note that neither statement describes the equilibrium flux (l, ). The reader is referred to Figure 5.5 in Chapter Five (Surface Microfiitration Literature Review) for further explanation. To determine the type of critical flux observed the maximum pressure achieved during each flux step was plotted against permeate flux for all experimental values below the critical flux . Once the critical flux was surpassed there was a dramatic pressure rise caused by particle deposition and cake formation. Attention is drawn to Figure 8.12 on the next page, which shows the pressure difference-permeate flux relationship for constant-rate filtration experiments below the critical value. The range of shear rates studied are categorised according to low shear (0 - 399 s-\ medium shear (400 - 800
S- I)
and high shear (801 - 1400
S-I)
values. The average clean water flux , represented by the solid line on the plot, was
8-18
-
-
-
-
-
-
-
- -
-
-
-
- - - - - - - - - - - - --
- - - - - - - --
- --
-----
CHAPTER 8. Results and Discussion (Surface Microfiltration) calculated from individual clean water flux experiments for the slotted media prior to each filtration experiment. There was a reasonable correlation to a straight-line fit of the clean water data (r2 = 0.872). As seen in Figure 8.12, the pressure difference across the membrane was in most cases above that of the corresponding clean water values. However, the low shear data gave the poorest agreement to the clean water data. As the shear rate increased the data became closer to the clean water measurement. This shows that there was some deposition of particles on the membrane at sub-critical fluxes without membrane fouling . According to the definitions outlined earlier the critical flux observed was of the weak form as the experimental data generally exceeded the clean water values particularly at lower shear values.
12
..: CO
..c
E
ai
u
8
•
... Q)
~ "tl
...::J
6
4
UI UI
...
Q)
a..
•
• • • • ... • • ••• • ... •• ••
Q)
2
Low shear Medium shear High shear - - Average clean water flux
• •
10
I:
•...
•
•
.• . .
....
...
•
..
• • •
•
•
0 0
500
1000
Flux, ~m
1500
2000
1 5- •
Figure 8.12 Relationship between applied flux and pressure difference for filtration with 10.0 !lm nickel SPO membrane compared to pressure difference for same clean water flux. Low shear (0 - 399 S-I), medium shear (400 - 800 S-I) and high shear rates (801 - 1200 S-I) reported.
8-19
CHAPTER 8. Results and Discussion (Surface Microfiltration) 8.2.4 Grade Efficiency The grade efficiency of the 10.0 /lm nickel SPO membrane was determined using a technique described in Section 7.8.5. In a once-through arrangement feed suspension was drained through the membrane and collected for analysis. The permeate flux was maintained at 25 /lm
900 /lm
S-I
S·I
(90 I m·2 h· I), well below the measured critical flux of around
(3240 I m-2 h- I) for the same stirrer speed (providing a shear rate of
approximately 600
S-I)
and feed concentration (0.020 kg m-\
Particle size and
concentration analysis of initial column feed and four permeate samples were provided by a Coulter Multisizer JI incorporating a 70 /lm orifice. The grade efficiency of the SPO membrane can be seen in Figure 8.13. Permeate samples I - 4 were analysed after volumes of 40, 80, 120, and 160 rnl respectively had passed through the membrane. Each sample therefore represented a mean particle size distribution of permeate collected since the previous sample. The first point of note is that collection efficiency increased with particle size. There is again evidence of rejection for particles smaller than the pore width. This is observed in Figure 8.13
Figure 8.13 Grade efficiency of 10.0 /lm nickel SPO membrane. PI
(0 - 40 rnl); P2 (40 - 80 mJ); P3 (80 - 120 rnl); and P4 (120 - 160 mJ).
8- 20
CHAPTER 8. Results and Discussion (Surface Microfiltration) where the collection efficiency for particles below 5 by the larger 13.4
~m
is greater than that exhibited
SPO membrane. Between 40 and 50 per cent of particles 5
and below were removed by the 10.0 per cent for the 13.4
~m
~m
~m
~m
SPO membrane compared to less than 30
slots. The collection efficiency may also improve as more
volume is filtered although such a trend is far from conclusive. The data describes a situation in which a secondary membrane probably did not form on the membrane. 8.2.5 Critical Flux without Surface Shear Critical flux was estimated in the absence of shear at the membrane surface. Back transport of particles away from the membrane could be much reduced without shear at the membrane surface. Both shear-enhanced diffusion and inertial lift are sheardependent mechanisms. In this situation particle back-diffusion across the concentration boundary layer of primarily sub-micron particles is the solitary mechanism to counter deposition on the surface. The feed suspension was mixed well beforehand using the stirrer, which remained in position during the experiment.
80 ,------------------------------------------------ ,
60
12261 · 40 -
it~
20 -
11661 o . _•
12161
~ ~I;(j~
.ail''' !:F14 'In: I. .
+
12261 ~
backfiush -20 -~--~----_.----_.----~----~----._--_.----~
o
10
20
30
40
50
60
80
70
Time, mins.
Figure 8.14 Pressure profile of 10.0
~m
nickel SPO membrane without
shear (0 S-I shear; 0.020 kg m-3 latex feed concentration).
8-21
CHAPTER 8. Results and Discussion (Surface Microfiltration) Figure 8.14, on the previous page, shows the pressure difference (mbar) as a function of filtration time. The initial permeate flux of 166 ~m
S'I
(598 I m'2 h' l) represents a
low value for this membrane. The pressure difference remained stable as the permeate flu x was increased in 10 minute steps to 254 ~m
S' I .
It appears that restricted particle
back-transport was overcome by a permeation flux greater than 254 ~m
S' I
(914 1 m'2 h'l) , A rapid pressure rise indicative of membrane fou ling of about 16 mbar was observed at 264 ~m
S' I
(950 I m'2 h' l) , The existence of a critical flux between
254 ~m
S' I
was further supported by pressure observations during
S'I
and 264 ~m
subsequent operation at sub-critical flux .
2000 ,
~
m
- - Particle deposition model
E
(shear'enhanced effects)
:::i.
..-
•
1500
('Cl Q)
.s::.
-
•
m 0
Q)
1000
• •
::J '0
><
::J
;;:: ('Cl
•
•
500
u
•
~ .;:
(,)
•
•
0 0
•
•
200
400
600
800
1000
1200
1400
Shear rate at membrane surface, (Yw)' S 'l.
Figure 8.15 Critical flux due to shear for 10.0
~m
compared to particle deposi tion model for 5
nickel SPO membrane
~m
latex particles.
8.2.6 Comparison of Particle Deposition Model with Flux Measurements The particle deposition model described in Chapter Six (Surface Microfiltration Theory) assumes that the critical flux is determined by the addition of flu xes away
8-22
CHAPTER 8. Results and Discussion (Surface Microfiltration) from the membrane produced by several different mechanisms. When the sum of these fluxes is greater than the convective flux towards the membrane the flux will be sub-critical. The back transport mechanisms are assumed to be made up from shearenhanced diffusion and lift effects, which are both shear dependent , and diffusion effects at zero shear. These diffusion effects at zero shear were measured experimentally as described in the previous section and a value of 259 J.lm
S-I
(932 I m- 2 h-I ) was determined_ This measurement was then subtracted from the experimental values of critical flux leaving only the two shear dependent back diffusion mechanism values to be calculated. For the latex particle size distribution evaluated the inertial lift back velocity was very small compared to shear-enhanced back-diffusion, and could be neglected. A latex particle mean size of 5 J.lm was used to estimate the shear-enhanced diffusion mechanism. The comparison between the model and experimental critical flux values is presented in Figure 8.15, on page 8-22. As can be seen from this Figure the measured critical flux was much higher than that estimated from the model by about a factor of seven. It would appear that the model does not describe the experimental measurements well for a membrane with slots of 10_0 J.lm. The mass-transfer model requires particles to be substantially retained at the membrane surface, which did not take place with this filter.
8.3 CONSTANT-RATE FILTRATION WITH 4.9 J.lm NICKEL (SPO) MEMBRANE 8.3.1 Membrane Properties The experiments using 10.0 J.lID circular pores and 13.4 J.lffi, and 10.0 J.lm slots gave poor rejection of the latex challenge suspension. Smaller membrane pores and slots were required to facilitate filtration experiments at a size of more practical interest for
Cryptosporidium ParvUln oocysts. Fortunatel y development of a patented electroless nickel plating technology made possible the production of nickel membranes with smaller slot openings. A nickel SPO membrane was prepared with slot dimensions of 4.9 J.lm (width) and 296 J.lm (length). Topographic images of the plated membrane surface and slotted pores are presented in Figure 8.16 (a) and (b), on the next page. The non-uniformity of the nickel slots after electroless deposition is clear to see in
8-23
CHAPTER 8. Results and Discussion (Surface Microfiltration)
Figure 8.16 SEM micrograph of nickel slotted membrane with average pore dimensions 4.9 x 396 J..I.m; viewed at (a) standard magnification , and, (b) high magnification.
8-24
CHAPTER 8. Results and Discussion (Surface Microfiltration) Figure 8.16 (b). Each slot width typically varied by approximately ± I /Lm along its length (sample size 20 slots) with the pore width occasionally becoming even smaller. The faint outline of the original slot gives some indication of the reduction in pore area. The pore area within the expansive metal surface was estimated to be 2.1 per cent. 8.3.2 Wetting of Membranes Preliminary tests using the 4.9 /!ffi nickel SPO membrane showed that there was a potential surface-wetting problem. The flow of dilute feed suspension by gravity through the membrane pores was very slow. Surface wetting had not caused problems for the single use, virgin 13.4 /Lm SPO membrane. The surface of the 4.9 /!ffi nickel membrane was slightl y different having received a nickel-phosphorous coating. In the smaller pores greater capillary pressure had to be overcome before air could be displaced by fluid entering the pore. Drops of water added to the plated membrane surface sat up tightly on the surface exhibiting a steep contact angle. In a series of simple tests, various concentrations of Bardac-2270-E cationic surfactant (n-n-
didecyl-n,n-dimethylammonium chloride) and Decon 90 surface active cleaning solution were contacted drop wise with a test nickel slotted membrane. Observations are reported in Table 8.2, on the following page. The critical surfactant concentration for Bardac-2270-E seemed to be greater than 0.4 g
r1
for satisfactory wetting and
droplet spreading on the membrane surface. Concentrated dosage apart, Decon 90 cleaning solution showed poor wetting capability. Bardac-2270-E solutions prepared using ultra pure water were tested in the stirred-cell column against a clean water control run. The suction-side pressure was monitored continuously whilst solutions were circulated around the column. There was little improvement in filtration performance when the feed suspension was formu lated with 0.4 g
r1
Bardac-2270-E
surfactant solution; the membrane pressure drop was equally sensitive to permeate flow through the pores. In addition there was concern over the use of a weakly acidic surfactant in the stirred-cell rig with a pH of 5.0. Degradation of silicone rubber tubing and corrosion of the membrane itself were distinct possibilities from prolonged use. Surfactant-based feed suspensions had not been used before and were not necessary, nor compatible with track-etched membranes. Indeed use of surfactant in
8-25
CHAPTER 8. Results and Discussion (Surface Microfiltration) Table 8.2 Wetting capabilities of Bardac-2270-E surfactant and Decon 90 cleaning solutions. Surfactant
Concentration
Observations
(2 r1)
Bardac 2270-E Bardac 2270-E Bardac 2270-E
0.1
Well-defined drop on surface. Poor wetting
0.2
Poor wetting. Drop stayed still . Little wetting through to other side Surface wetted but droplet remained for longer. Wetting slow and not very successful. Wetted through to other side Wetting seemed to be fairly good. Not very steep contact angle. Drop could be seen moving at edges. Wetted through to other side well Surface wetted through to other side. Contact angle greater but drop spread over time Surface wetted well. Good contact angle. Wetted underside Quickly Wet surface
0.3
Bardac 2270-E
0.4
Bardac 2270-E Bardac 2270-E Bardac 2270-E Decon 90
0.5
concentrated (no dilution) 0.1
Decon 90 Decon 90
0.5 1.0
Decon 90
3.0
Decon 90
concentrated (no dilution)
1.0
Drops on untreated membrane were similar to water. High contact angle with poor wetting and spreading. Drops on membrane surface did not spread very well Drops sat on membrane with no sign of wetting. Did not wet satisfactorily Drops did not spread well. Very slow - quite steep contact angle Wet surface
the feed suspension could affect particle deposition and change the fouling regime denoted by critical flux . Focus, instead shifted towards treating the membrane with Decon 90 cleaning solution . Decon 90 is an alkaline emulsion of anionic and nonionic surface-active agents amongst other components. A 1.0 g
r'
solution of Decon
90 has a pH of 10.6. According to associated information a 2 - 5 per cent solution is satisfactory for most applications (Website http://www.decon.co.uk). The membrane was contacted with I g r' solution of Decon 90 for 10 minutes prior to a filtration experiment. This had a two-fold effect of removing latex deposits from the previous experiment whilst helping to reduce surface tension in the pores. The cleaning process was enhanced using sonication and by raising the solution temperature to 50 QC. The membrane was rinsed well in ultra pure water before use in the stirred-cell rig to
8- 26
---------
- - - - -- - - - - - - - - - - - - - - - - - - - - - - - - - - - - --
CHAPTER 8. Results and Discussion (S urface Microfiltration) remove residual surfactant from the surface. Followi ng the compl etion of earl y constant-rate filtration experiments the membrane surface was examined under a scanning electron microscope. As seen in Figure 8.16 (a) and (b), on page 8-24, the plated surface was clearly different fro m other nickel membranes. In particul ar there were small areas where the platin g surface appeared to be missi ng, presumabl y corroded. Whe n viewed at high magni ficati on electron microscopy revealed the sharp contrast between the edge of the nickel-phosphorous coating and the region corroded to bare nickel. Electron probe microanalysis (see Section 3.5.4) of the intact and corroded areas of Fi gu re 8. 17 confirmed that the coatin g had received che mical attack. The characteristic phosphorous peak of the nickel plating deposi t (Figure 8.1 8 (a), page 8-28) is miss ing from the corresponding analysis of the corroded area (Fi gure 8.18 (b), page 8-28). From a wider study of the membrane su rface it was clear that the surface corrosion was discovered in its earl y stages as most of the nickelphosphorous coating and as far as could be seen, all the slots, were untouched. The nickel plated surface appeared most susceptible to the aggressive alkaline cleani ng
Figure 8.17 SEM micrograph of 4.9 /lm nickel SPO membrane at high magnification show ing corroded region of nickel-phosphorous coating.
8-27
()
X-RAY: livE::
0 - 20 keU 50::- Pt-E:::-E;t.:
X-RAY: 0::: L i v€::
50 s F.: ema i n j n g:
::c:
0 - 20 keU 50 s Preset:
;J>
50 s F.:ema in i ng:
0::-
...,'Cl tI1 10 00
;d
8.18 (b)
8.18 (a)
(1)
"'";::;'"
0> ::l
0-
t:!
on' 0
"'":!? 0
::l ~
en
"~ 0
(1)
p
~
o· '"'0
:!1 ~
'"'0> 0::. 0
::l
~
<
.9
FS=
~lEt11
:
2K
E..OLtO
ch
kel)
.
-':12= -
11 .2
33
> <
cB
.9 FS= 2K
t'lEt11 :
6.040
ch
kE:1.) 312~
Figure 8.18 Electron microprobe analysis of 4.9 Ilm nickel SPO membrane following Decon 90 treatment (1.0 g (a) nickel-phosphorous coating, and, Cb) region corroded to original membrane surface.
11. 2 > 3'3 et:;
r l cleaning solution);
CHAPTER 8. Results and Discussion (Surface Microfiltration)
environment, which was duly revised to ambient temperature operation. All the nickel slotted membranes were cleaned using Decon 90 cleaning solutions with one exception - that of the virgin nickel SPO membrane.
8.3.3 Effect of Shear Rate on Critical F lux The filtration performance of smaller slotted pores was considered at different surface shear rates. Critical flux was evaluated by monitoring pressure drop across the membrane using flux stepping. Constant-rate filtration experiments were similar to those described for the 10.0 !lm SPO membrane. Figure 8.19 is a pressure profile for constant-rate filtration at medium surface shear (approximately 600 s"). A dilute feed concentration of 0.020 kg m·3 made critical flux detection easier. Permeate flow was increased in small increments (7 !lm s·, average). A ten-minute 'flux-step' duration at constant flux was considered sufficient to monitor critical flux; this receives further attention in a later section. No obvious signs of fouling were observed between 47 and 72 !lm s·, (169 and 259 I m·2 h") permeate flux. Small pressure fluctuations were
120
..:
co E
D
100
, t'
Flux, I'm s·' .
..0
80 .
Q)
0
c: Q)
...
60
-c
40
~
~
::::I
20
...
0
I/) I/) Q)
Il..
"'
@]
+ +
I 71 I
+
00
[ill
a
rn:J
1721 11
F'
+
..-I
+J
backfiush --+ \
-20 0
20
40
60
80
100
Time, mins.
Figure 8.19 Pressure profile of 4.9 !lm nickel SPO membrane (600 s·, shear; 0.020 kg m-3 latex feed concentration).
8- 29
CHAPTER 8. Results and Discussion (Surface Microfiltration) within the ± 2 mbar accuracy of the pressure transducer. After six minutes of constant flux operation at 83 J..I.m s" (299 I m·2 h") a steep pressure rise was interpreted as fouling. Continued operation at low flux confirmed a critical flux between 72 J..I.m s" and 83 J..I.m s" (259 and 299 I m·2 h"). A short, moderate flow backflush briefly returned the pressure drop to near clean water values but the pressure then rose rapidly when operating at a sub-critical flux of 71 J..I.m s·'. This suggests that some of the slots were still blocked with particles. Mid-point critical flux (10.6 m s·,) is reported as a function of surface shear rate (0 1400 s·,) in Figure 8.20. As with large slotted pores, membrane fouling was sensitive to shear rate at the surface. Critical flux increased with increasing shear rate, although flux values were on average fifteen times lower than those recorded for 10.0 J..I.m slots.
Figure 8.20 Critical flux of 4.9 J..I.ID nickel SPO membrane as a function of surface shear (0 - 1200 s· , shear; 0.020 kg m·3 feed suspension) .
8-30
CHAPTER 8. Results and Discussion (Surface Microfiltration) 8.3.4 Assessment of Backflushing The catastrophic consequences of membrane fouling highlight the importance of operating below the critical flux. Latex particles are deposited on the membrane surface and in the pores when the permeate flux exceeds the critical flux. The fouling resistance progressively increases with filtration time and leads to flux decline. There are many cleaning methods available to alleviate membrane fouling and maintain stable flux (See Chapter Five 'Surface Microfiltration Literature Review'). Their effectiveness depends on the type and severity of fouling. Particles loosely resting on the surface or sitting in pores for example are much easier to remove than particles jammed in pores tightly. The suitability of flow reversal to remove latex deposits was assessed following each critical flux experiment. A brief back flush was performed at 178 Ilm
S·I
(641 I m· 2 h· l ) flux for 2 minutes. The results were inconclusive. Despite
an initial return to clean water conditions, the membrane surface quickly fouled in some situations or continued unaffected in others. The post-back flush performance was independent of shear rate. This suggests that only some of the pores were cleared by the use of this backflush procedure. 8.3.5 Repeatability of Critical Flux Analysis Critical flux can be difficult to pinpoint because it is dependent on the factors that influence concentration polarisation such as nature and concentration of the feed, shear rate and membrane characteristics. Even small changes can have a dramatic effect on membrane fouling. In a shear·dependence study, it is essential that the only variable changed from one experiment to the next is stirrer speed. Careful preparation of the stirred-cell rig, feed suspension, and membrane itself, vigorous cleaning methods and adoption of a standard experimental methodology led to greater confidence in experimental results. Higher accuracy was pursued by reduction of the flux step size wherever possible. Selected experimental conditions were duplicated to study the repeatability of critical flux identification. Table 8.3, on the next page, highlights the range of mid-point flux values for a specified number of repeated experiments. Whilst some experimental scatter is unavoidable, there was fairly good agreement between repeated experiments particularly at low shear rates. The spread of data was much wider when the membrane surface is exposed to high shear and there
8-31
CHAPTER 8. Results and Discussion (S urface Microfiltration) was greater flow through the pores, i.e. critical flux range of 55 and 50 ~
S-I
(198
and 180 I m-2 h- I ) at shear rates of 1200 and 1400 S- I respectively.
Table 8.3 Repeatability of 4.9 ilm nickel SPO membrane critical flux experiments. Shear r ate
Number of experiments
S·1
400 600 800 1000 1200 1400
3 2 2 2 3 2
Critical flux r ange (mid-point) ilm S-1 56 -78 70 -78 99 - 102 111-144 102 - 157 160 - 210
250 19 ';"
Ul
200
18
E
::i.
...,""
.-".
.
14
~
><
::J
;:
ca0
-
'': ()
100
9
2
50
0
•11 15 • 5 10 4 • • • • 17• 23• • 21 20
150
7 1 6
••• 200
•
3
•• 24 •• 12
•
16
13
400
600
800
1000
1200
1400
Shear rate at membrane surface (Yw)' S-1.
Figure 8.21 Order of critical flux experiments for 4.9
~
nickel SPO membrane.
8.3.6 Order of Critical Flux Experiments As described earlier the membrane was chemically cleaned after each critical flux measurement. Figure 8.2 1 shows the critical flux results for all the experiments
8-32
CHAPTER 8. Results and Discussion (Surface Microfiltration) carried out using the 4.9
~
slot membrane. The floating numbers on the graph
indicate the sequence of experiments using the same membrane. Critical flux does not appear to be significantly affected by the order of experiments, which suggests that the membrane cleaning procedure was effective. 8.3.7 Type of Critical Flux The relationship between pressure difference across the membrane and permeate flux in comparison with clean water values is presented in Figure 8.22. For each filtration experiment the highest pressure value of each constant fl ux step is plotted up to the measured critical flux. As before, the plotted data is divided into three shear regions describing low shear (0 - 399 S· I), medium shear (400 - 800 S·I) and high shear rates (801 - 1400 S·I). Average clean water values were determined from many clean water
7
..:
ca
•...
Low shear Medium shear High shear - - Average clean water flux
•
6
..c
E
a)
... ;E (1)
4
"tl
3
...:::J (1)
III III
... a..
•
• • •••• •• • •• ... •• ~ .: • ... ...... ... • ..,+ • • • ... •• • ~+ • • ... i • .. I. • ........ • +++ • •
(.)
c::
• •
5
~
2
",
(1)
1 0 0
50
100
Flux, /lm
150
200
S ·l .
Figure 8.22 Relationship between applied flux and pressure difference for filtration
with 4.9 /lm nickel SPO membrane compared to pressure difference for same clean water flux. Low shear (0 - 399 s·'), medium shear (400 - 800 S·I) and high shear rates (801 - 1400 S·I) reported. 8-33
CHAPTER 8. Results and Discussion (Surface Microfiltration) experiments carried out between cleaning and fi ltration stages. It is clear that the critical flux is of the weak form because although there was no long term fouling, the membrane resistance observed during latex particle filtration was greater than that for clean water. Data points appearing below the clean water flux line can be explained only as a combination of experimental scatter and the effect of averaging clean water values since the pressure values associated with clean water operation represent a minimum. The pressure drop is naturally higher for increased flow as a consequence of the physical phenomenon first described by Darcy. The hypothesis which best describes critical flux for both the 4.9 flm and 10 flm slotted pores is that of the weak form. 8.3.8 Grade Efficiency Grade efficiency data for the 4.9 flm nickel SPO membrane is shown in Figure 8.23. Four permeate samples were analysed after filtration of21, 42, 62, and 94 per cent of the initial feed suspension respectively. In determining grade efficiency it is of paramount importance that membrane fouling is avoided wherever possible.
100 0~
>. (.)
s:: .!!! (.)
!EQ) s::
0 :;:;
80 -
'K.
0
~
{;
++++++
+
+
+ + Sample P1
++
X Sample P2
+ ++
{;
20 -
x
++
X
~
40 -
+
{;
'K.td' X
60 -
(.) Q)
0
, ts.
~li~~~~~){~l '" mi ~ ~
'K. t!': xxx >
{; Sample P3
+++
'K. Sample P4
X -tt).+ X +
-F
o0
2
4
6
8
10
Particle dia meter (d p), flm.
Figure 8.23 Grade efficiency of 4.9 flm nickel SPO membrane. PI (0 - 40 m!); P2 (40 - 80 rnJ); P3 (80 - 120 ml); and P4 (120 - 184 rnJ). 8- 34
CHAPTER 8. Results and Discussion (Surface Microfiltration) Consequently the gravity-fed flow was maintained at approximately 6 flm sol (22 I m-2 h- l) or thereabouts, which was well below the critical flux of 70 flm sol (252 I m-2 h- l) for a 0.020 kg m-3 feed suspension and estimated shear rate of 600 S- l. The initial collection efficiency was encouraging despite the trend being slightly lower than expected_ As before particle removal by slotted pores increased as a function of particle size_ There was poor rejection for smaller particles, which was to be expected for a fllter exhibiting 4_9 flm slotted pore openings_ A collection efficiency of 74 per cent for particles 5 flill and less was achieved. This represents a significant improvement ahead of the larger 10.0 and 13.4 flm SPO membranes discussed earlier. And, as intimated with the 10_0 flm SPO membrane the rejection seemed to improve with filtration volume_ The remaining three permeate samples collectively showed good collection efficiency after further filtration_ A rejection in excess of 90 per cent for particles diameters of 3 flill and greater was observed when half the feed suspension had been filtered _ The slots were successful at consistently filtering approximately 95 per cent of particles larger than the slot size_
30
..:
D
C'O
.Q
E
a;
Flux, I'm
1 5- ,
20
u
112 1
t:
... Cl)
Cl)
;;: -c
m
...::::J Cl)
1/1 1/1
...
": / ;.!tCil
10
0
Cl)
backflush
a..
+
-10 0
20
40
60
80
Time, mins_
Figure 8_24 Pressure profile of 4_9 flill nickel SPO membrane without surface shear (0 sol shear; 0_020 kg m-3 latex feed suspension)_
8- 35
CHAPTER 8. Results and Discussion (Surface Microfiltration) 8.3.9 Critical Flux without Surface Shear The effect of particle diffusion on critical flux was investigated without shear at the membrane surface. The feed suspension was well mixed prior to filtration. The stirrer rod remained in position throughout the experiment. The pressure difference across the membrane is plotted against filtration time in Figure 8.24, on the previous page. The peristaltic pump was operated at a set flow rate for 10 minutes. Permeate flux was increased stepwise by 6 Ilm
S· I
(22 I m·2 h-I) until stable pressure operation was no
longer possible. With no shear to re-suspend deposited particles, particle diffusion was the only mechanism working against forward particle transport. A limited resistance to particle deposition was demonstrated by a low critical flux in the region 18 - 24 Ilm
S· I
(65 - 86 I m-2 h· I) . This was considerably lower than the shear-
independent critical flux for 10 Ilm slots (254 - 264 !lID
S-I)
where many particles
smaller than the slots passed through with the permeate flow .
8.4 CONSTANT-RATE FILTRATION WITH 5.0 Ilm PC ISOPORE (CPO) MEMBRANE 8.4.1 Membrane Properties Polycarbonate isopore membranes with circular pores of 5.0 Ilm diameter were evaluated against the 4.9 !lID slots. The topography of the membrane was examined using a scanning electron microscope and sample micrographs are shown in Figure 8.25 (a) and Cb), on the following page. The nuclear track-etched membranes have structures composed of a series of predominantly cylindrical pores dispersed in what appears to be a random manner. Single, discrete pores are of an almost uniform diameter throughout the membrane whilst doublets have a variable shape in both their axial and radial directions. A pore size distribution of +0 % to -20 % and open area of 8 per cent are quoted in the manufacturer's literature for this isopore membrane. 8.4.2 Effect of Shear Rate on Critical Flux Comparative constant-rate filtration experiments were performed for 5.0 Ilm polycarbonate track-etched membranes having circular pores. Critical flux was estimated over a range of shear rates for 0.020 kg m-3 latex feed suspension using the flux stepping technique. Figure 8.26, on Page 8-38 is a pressure profile for constant8-36
CHAPTER 8. Results and Discussion (Surface Mi crofiltration)
Figure 8.25 SEM micrograph of 5.0 /lm polycarbonate isopore membrane with circular pore openings; viewed at (a) low magnification, and, (b) high magnification.
8-37
CHAPTER 8. Results and Discussion (Surface Microfiltration) rate filtration under conditions of medium shear (600
S-I).
The plot is comparable to
Figure 8.19 (on page 8-29) for the 4.9 !lm nickel SPO membrane. The track-etched membrane with circular pores fouled at 45 !lm
S-I
(162 I m-2 h- I) , which was almost
half the critical flux reported for the 4.9 !lm nickel slots (78 !lm
S- I).
Identification of
critical flux proved to be particularly difficult for track-etched CPO membranes where low flux operation resulted in a small pressure drop rise of 10 mbar in the fouling domain. A new membrane was utilised for each experiment. The membranes were coated with a hydrophilic layer and wetted easily.
30
..::
D
E
Flux, ~m s·' .
20
Cl) (.)
c
... Cl)
~ 'tI
136 1
10
00
D..[]
... Cl)
::::J
I/) I/) Cl)
*~
*
0
... a..
backflush _ -10 0
20
40
60
100
80
Time, mins.
Figure 8.26 Pressure profile of 5.0 ).lm polycarbonate isopore CPO
membrane (600
S- I
shear; 0.020 kg m-) latex feed concentration).
Figure 8.27, on the next page, indicates the relationship between critical flux and shear rate at the membrane surface for 5.0 ).lm track-etched CPO membrane. The potential range (tie bars) and average values (data point) of critical flux are plotted. The critical flux increased with shear rate in a somewhat disjointed, stepwise fashion. There are three distinct regions where critical fl ux seemed to be constant with shear: 39 ).lm
S-I
up to 400
S-I,
45 ).lm
S-I
between 600 and 1000
S·I
and 50
).lID S-I
and
8- 38
CHAPTER 8. Results and Discussion (Surface Microfiltration) upwards beyond 1200 S·l. The pressure drop often increased to 300 mbar during a back flush giving some indication of the severity of fouling deposits.
80
,
~
Ul
E
60
::i.
-::.
'"
~
><
! ! ! !
40
::s
;;::
! !
!
!
~u
Cij 0
'';::; '':;
20
U
O+---~r----.----~----~--~-----,--~
o
200
400
600
800
1000
1200
Shear rate at membrane surface (Yw)'
1400
S -1_
Figure 8,27 Critical flux of 5.0 ~m polycarbonate isopore CPO membrane as a function of surface shear (0 - 1200 S· l shear; 0.020 kg m-\
8.4,3 Type of Critical Flux Figure 8.28, on the following page, shows the pressure difference achieved prior to critical flux during constant-rate filtration of latex challenge material. Studied at varying surface shear rates (0 - 399 S·l, 400 - 800 S·l, and 801 - 1400 so l), experimental values report the maximum value of the oscillating pressure profile during each constant flux step. Clean water pressure values (average) measured for several 5.0
~m
isopore membranes are plotted as a solid line. It is seen that all the experimental pressure values were consistently above the corresponding values for clean water. According to the definitions chosen to describe critical flux this represents the socalled 'weak' form of critical flux. Regardless of membrane type or pore geometry, membranes having 10.0
~m,
4.9
~m
slotted and 5.0
~
circular pores all exhibited
weak critical flux for latex particle filtration . 8-39
CHAPTER 8. Results and Discussion (Surface Microfiltration)
7
..:
nI
•..
6
,Q
E aI t)
r:
... Q)
E ":s...
-
Low shear Medium shea r High shear
•
-
Ave rage clean water flux
5
..
4 3
Q)
III III
2
D..
1
...
•
•
-
..
. • • •. • • •. • •
•.. ...•• •• •• • . •
•
Q)
0 0
10
20
30
Flux, f..lm
40
50
60
1 5- •
Figure 8.28 Relationship between applied flux and pressure difference for filtration with 5.0 f..lm isopore CPO membrane compared to pressure difference for same clean water flux. Low shear (0 - 399
S-I),
medium shear (400 - 800 S-I) and high shear rates
(801 - 1400 S-I) reported.
8.4.4 Grade Efficiency A comparative grade efficiency experiment was carried out for the circular pores of the 5.0 !-lm isopore membrane. The experimental procedure was generally as stated previously and the reader is directed to Section 7.8.5 for further details. This time five permeate samples were collected at filtration intervals of 33,63,95, 126, and 175 ml drained through the membrane under gravity. 0.020 kg m-3 latex feed suspension was used and surface shear was generated by the motion of a stirrer rotating at 400 rpm (600
S-I
calculated shear). A low permeate flux of 11 !-lm
comfortably less than the 45 !-lm
S-I
S-I
(40 I m-2 h- I ) was
(162 I m-2 h- I), critical value. As seen in Figure
8.29, on the next page, rejection increased sharply with particle size for all samples until complete cut-off was achieved above 4 !-lm. This is equivalent to the lower
8--40
CHAPTER 8. Results and Discussion (Surface Microfiltration) confidence limit for manufacturer's pore size distribution, i.e. 5 ± 1.0 f.1m. There appeared to be a subtle improvement in the rejection observed after 95 ml had been filtered (permeate sample four onwards). In contrast to the 10.0 f.1m isopore membrane there was no suggestion of secondary membrane formation although the larger pores were challenged with four-and-a-half times more particles. Neglecting the initial permeate sample, the 4.9 f.1m slotted pores generally outperformed the circular pores when filtering smaller particles. 100
'" f,if< ~'1>-:;(-*-'Tlj[ ..
'''~
Os
~iIi
~ 0
~
Iv ~
?-){.~
80 -
(.)
x
:f. ~ A
60
0 :.::;
A SampleP3
:1St: ~+
t XI-
:f.~
40
(.)
.9! 0
x Sample P2
xl-
lEQ) c:
+ Sample P1
:f. A"#
c: Q)
(.)
+
20 -
U
+
:f. Sample P4
Sample P5
dIf; :I: A
O +---~---''---~---r----'---'----r----r---~--~
o
2
4
6
8
10
Particle diameter, f.1m.
Figure 8.29 Grade efficiency of 5.0 f.1m polycarbonate isopore CPO membrane.
PI (0 - 33 rnI); P2 (33 - 63 rnI); P3 (63 - 95 ml); P4 (95 - 126 ml); PS (126 - 175 rnI).
8.4.5 Critical Flux without Surface Shear
Figure 8.30, on the next page, is a pressure profile recording constant-rate filtration without surface shear for the 5.0 f.1ID
epo membrane. Permeate flux was steadily
increased until a rise in pressure drop intimated a critical flux between 36 and 42 f.1ID s-' (130 and 151 I m-2 h-'). In contrast to the 4.9 f.1ID nickel SPO membrane, there was little variation between critical flux at zero shear and that of shear rates up to 400 s-'. This would suggest that additional contributions from shear-dependent
8-41
CHAPTER 8. Results and Discussion (Surface Microfiltration) back transport mechanisms had little additional effect to diffusion mechanisms (see model at very low shear). The 4.9 flm nickel SPO membrane was much more sensitive to shear at the membrane surface than the equivalent CPO membrane.
30
..:
D
.c
Flux, ~m
5" .
E
Q) (J
20
cQ)
...
~ '0
...::J Q)
[ill
10
00
t/I t/I
... Q)
11.
0 0
20
40
60
80
100
120
Ti me, mins,
Figure 8.30 Pressure profile of 5.0 flm polycarbonate isopore CPO membrane without shear (0
S,I
shear; 0.020 kg m,3 latex feed concentration).
8.5 COMPARISON OF CRITICAL FLUX FOR 5 flm SPO AND CPO
MEMBRANES In Figure 8.31, on the following page, the rate of increase of critical flux for progressively greater shear was clearly more pronounced for slotted pores. The critical flux at zero shear has been subtracted from the measured values of critical flux for each of the two membranes tested. The SPO membrane typically fouled at higher permeate flux than the CPO membrane. This was most significant at high shear rates. In contrast the circular pores fouled at comparatively low permeate flow irrespective of surface shear; LV varied by only 16 fl m
S,I
(58 I m,2 h,l) for all shear rates. The
mechanisms by which each membrane fouls are clearly different. Slotted pores can
8-42
CHAPTER 8. Results and Discussion (Surface Microfiltration)
250
•
"T
III
E
::l.
A
4.9 !lm nickel SPO membrane 5.0 !lm isopore CPO membrane
200
.: nI Cl)
~
III 0
150
.... Cl)
::J '0
100
)(
I
• • •
A
A
A
600
800
1000
::J
. I;:
nI
u .;: U
•
50
•••
0 0
200
•
I
400
• • • •• • A
•
A
1200 1
Shear rate at membrane surface (Yw)' 5 -
1400
.
Figure 8.31 Comparison of critical flux due to shear for
4.9 !lm nickel SPO and 5.0 !lm isopore CPO membranes.
maintain higher permeate flux at medium and high shear where contribution from shear-dependent back-transport mechanisms is significant. This primarily concerns the shear-enhanced diffusion mechanism although inertial lift becomes significant at higher shear. The independence of shear suggests that perhaps the membrane pores are being blocked and that flux for the CPO membrane is controlled by particle-pore interactions. With no shear at the surface, the circular pores outperformed their slotted counterparts, fouling at twice the permeate flux. This result may reflect the higher open area of the CPO membrane, which could affect the entry of latex particles into the pores. It suggests that the mechanism of fouling is probably different to that described by the model used in this thesis and that particle-pore interactions are also an important phenomena to consider.
8-43
CHAPTER 8. Results and Discussion (Surface Microfiltration) 8.5.1 Effect of Time on Critical Flux Analysis Extended filtration experiments were conducted to determine the effect of time on filtration below critical flux for SPO and CPO membranes. In the first experiment involving the 4.9 ~m slotted membrane the flux was fixed at 84 ~m
S-I
2
(302 I m- h-
for 140 minutes, which is below the estimated critical flux of 104 ~m
I
)
S-I
(3741 m-2 h- I). An experiment at 0.10 kg m-3 concentration was found to foul rapidly so the feed concentration in the stirred-cell was replaced with 0.010 kg m-3 latex suspension. The rotational stirrer speed was held at 400 rpm. Figure 8.32 shows the pressure history over the prolonged operating period. There was a slight increase in pressure difference from 1.8 to 4.0 mbar after more than 2 hours filtration. In fact the pressure was almost unchanged over the first 40 minutes. It is apparent from the results that no membrane fouling occurred. 30 r-------------------------------------------~
..: Cl!
.c
E
ai o c::
20
~
~ "C ~
::J
10
III III
~
a..
o o
20
40
60
80
100
140
120
Time, mins.
Figure 8.32 Extended filtration of 4.9 ~m nickel SPO membrane at 84 ~m for 140 minutes (600
S-I
S-I
shear; 0.01 kg m-3 latex feed concentration).
The filtration performance of isopore membranes with 5.0
~m
circular pores was
assessed over the same filtration time. 0.10 kg m-3 latex suspension provided the challenge material. As before the feed suspension was stirred at a constant rotational 8-44
CHAPTER 8. Results and Discussion (Surface Microfiltration) speed of 400 rpm. During the experiment, the permeate flux was maintained at 7 etm S-I (25 I m-2 h- I), i.e. half the critical value. As seen in Figure 8.33 the pressure difference across the membrane was stable fluctuating only slightly between 2.5 and 3.3 mbar. According to these and the results of other related experiments flux step durations of 10 minutes (minimum) were considered acceptable. All critical flux investigations were carried out according to this criterion.
30 ,-------------------------------------------, ..: Cl!
.Q
E Cl>
(.)
20
c
~
~ "0 ~
::::I
10
I/) I/)
~
D.
o o
20
40
60
80
100
140
120
Time, mins.
Figure 8.33 Extended filtration of 5.0 etm isopore CPO membrane at 7 et ms-I for 140 minutes (600 S-I shear; 0.10 kg m-3 latex feed concentration).
8.6 CONST ANT-RATE FILTRA TION WITH 4.1 etm NICKEL (SPO)
MEMBRANE The fouling characteristics of smaller slotted pores were evaluated for similar stirredcell conditions. A nickel SPO membrane was produced by electroless nickel plating with an average slot width of 4.1 etm and slot length of 382 etm. The open area was estimated to be 1.8 per cent. Micrographs of the membrane surface showing the small slotted pores can be seen in Figure 8.34 (a) and (b) on page 8-46. As is the case with
8-45
CHAPTER 8. Results and Discussion (Surface Microfiltration)
Figure 8.34 SEM micrograph of nickel slotted membrane with average pore dimensions 4.1 x 382 Ilm; viewed at (a) standard magnification, and, (b) high magnification .
8-46
CHAPTER 8. Results and Discussion (Surface Microfiltration) all plated slots the width varies along the length. At higher magnification it is seen that despite the considerable reduction of slot size the plating bath delivers a uniform nickel-phosphorous deposit without deposition across the slots. 8.6.1 Effect of Shear Rate on Critical Flux
It is seen in Figure 8.35 that the smaller slots, as is the case with larger slotted pore openings, are sensitive to shear rate at the surface. The permeate flux values are similar to the 4.9 Ilm slotted pores, but again much smaller than the large 10.0 Ilm slots. There appears to be greater variation between repeated experiments when the 4.1 Ilm membrane was used particularly at lower shear rates. The critical flux at zero shear was estimated at 32 Ilm
S- I
(lIS I m-2 h-'), a value which lies between
measurements for 4.9 Ilm slots and 5.0 J.lm circular pores respectively.
160
! !
.tn
~
120
E
::i
-.,.
~~ ~
><
I
80
I
•
::J
!
;: (ij 0
:;:: .;:
40,
(.)
.
~
I
!
0 0
200
400
600
800
1000
1200
1400
Shear rate at membrane surface (Yw)' S·l.
Figure 8.35 Critical flux of 4.1 Ilm nickel SPO membrane as a function
of surface shear (0- 1200
S- I
shear; 0.020 kg m·3 latex feed concentration).
8-47
CHAPTER 8. Results and Discussion (Surface Microfiltration)
8.7 COMPARISON OF MODEL WITH EXPERIMENTAL MEASUREMENTS The particle deposition model described in Chapter Six (Surface Microfiltration Theory) has been applied to the experiments using the 4.1 !lm and 4.9 Ilffi slotted nickel membranes and the 5.0 !lm circular pore membrane. The mean particle size used was 5 !lm. As seen from Figure 8.36 the model predicts the effect of shear on the slotted membrane reasonably well. However the effect of shear on the circular pore membrane is predicted very poorly. This difference in behaviour again suggests a change in controlling mechanisms between circular pores and slots or perhaps the effect of changing the filter material. According to Figure 8.23 (page 8-34), the 4.9 !lm SPO membrane provided good rejection of the 5 !lm latex particles considered in
the deposition model. The concentration in the boundary layer is sufficient to establish a concentration gradient necessary for back-diffusion to take place and the
250
•
';" U)
E
:::l.
.:
4.9 I'm nickel SPO membrane 4.1 I'm nickel SPO membrane 5.0 I'm isopore CPO membrane - - Particle deposition model (shear-enhanced effects)
•...
200
III
Q)
J:
U)
0
150
• • •t • • • •• •
::J
"
><
100
::J
I;:
III ()
:p
•
•
Q)
• • • •• • •
50
.;: 0
0 0
200
400
•
...
...
...
600
800
1000
... ... 1200
Shear rate at membrane surface (Yw)'
1400
S -1.
Figure 8.36 Comparison of critical flux due to shear for 4.9 !lm and 4.1 !lm nickel slotted, and 5.0 !lm polycarbonate isopore membranes
with particle deposition model for 5 Ilffi latex particles. 8-48
CHAPTER 8. Results and Discussion (Surface Microfiltration) mass- transfer-limited model shows good correlation with the experimental data. That the circular pores also show good rejection without forming a secondary membrane suggests the existence of additional
particle-membrane mechanism(s) were
controlling particle deposition. To examine the effect of material on filtration behaviour a 4.8 fJ.m nickel circular pore membrane was then tested.
8.8 EFFECT OF SURFACE SHEAR ON 4.8 fJ.m NICKEL (CPO) MEMBRANE A new nickel membrane with 'circular' pores was manufactured by electroless nickel plating. The membrane pores viewed using optical microscopy are shown in Figure 8.37 (a) and (b), on the next page. Of particular interest is the inset of Figure 8.37 (b), which shows the original complex form of the pores, i.e. hexagonal shaped. The noncircular pore geometry was accommodated in pore size analysis by inscribing a circle within the pore and reporting the smallest dimension. The pore size and open area of the virgin substrate were estimated to be 21.1 fJ.m and 5.9 per cent respectively. Smaller and more rounded pores emerged from the plating bath. A four-fold reduction in pore size to 4.8 fJ.m limited the flow area to 0.3 per cent. Constant-rate filtration experiments were comparable to those described for similar sized slotted and circular pores. In summary a 0.020 kg m-3 latex suspension was filtered at various stirrer speeds which translate to a surface shear range of 0 - 1400 S-I. The filtration flux was stepped up gradually until the membrane fouled. The critical flu x at zero shear was found to be as low as 3 fJ.m
S- 1
(11 I m-2 h-\ The
open area of the filter was approximately seven times lower than that of the SPO membrane and critical flux was on average six times lower. This suggests again that the effect of open area is important for the CPO nickel membrane as well. The comparison between the model and experiments presented in Figure 8.38, on page 851, shows that the nickel CPO membrane does not appear to obey the shear model being applied. The indications are that the critical flux through the slotted membrane follows the shear model. However, it seems that critical flux of circular pore membranes are determined more by the particle-pore interactions.
8-49
CHAPTER 8. Results and Discussion (Surface Microfiltration)
Figure 8.37 Optical microscope image of 4.8 Ilm nickel membrane with 'circular' pores; viewed at (a) standard magnification , and, (b) high magnification. Original hexagonal pore shape shown (inset).
8-50
CHAPTER 8. Results and Discussion (Surface Microfiltration)
250
•
-.;
4.9 ~m nickel SPO membrane 4.1 ~m nickel SPO membrane 5.0 ~m isopore CPO membrane 0 4.8 ~m nickel CPO membrane - - Particle deposition model (shear-enhanced effects)
m
E
:::t
...-
A
•
200
nI Q)
..c:
m
150
0
:::J 00
><
100
:::J
co::
nI t.l
•
•
Q)
50
A
:;:::
A
•
I
.;: ()
0 0
200
• •
•
Q
~
~
400
600
800
I
A
••
• A
A
0
8
•
1000
•
• • 1200
Shear rate at membrane surface (Yw)'
1400
1 5- •
Figure 8.38 Comparison of critical flux due to shear for 4.9 f.lm and 4.1 f.lm nickel SPO membranes, 5.0 f.lm polycarbonate isopore CPO membrane, and 4.8 f.lm nickel CPO membranes together with particle deposition model for 5 f.lm latex particles.
An attempt was made to reduce the open area of the slotted membrane by blocking off
pores with PV A glue. Unfortunately these results were not conclusive and are described in Appendix J.
8.9 SENSITIVITY OF DEPOSITION MODEL TO PARTICLE SIZE The effect of particle size on the critical flux at different shear rates was calculated for particles ranging from 2 to 5 f.lID. As can be seen from Figure 8.39, on the next page, the particle deposition model is sensitive to particle size, which is to be expected as shear dependent mechanisms of shear-enhanced diffusion and inertial lift are strongly influenced by particle radius. Reducing the particle size in the model from 5 to 2 f!m lowers the critical flux due to shear from 200 to 50 f!m
S· I
(720 to 180 I m·2 h- I) at the
upper shear rate. The poly-disperse latex feed suspension used in stirred-cell 8- 51
CHAPTER 8. Results and Discussion (Surface Microfiltration) experiments with particles distributed between I and 20 to a mono-size suspension of 5
~m
~m
was modelled according
particles (equivalent to djo). It is interesting that
the model shows a good fit with the slotted pore data for 4 even for an unrepresentative particle diameter of 2
~m
~m
particles. However,
there is poor correlation
between the model and circular pores reinforcing the hypothesis that additional particle-membrane interactions are unaccounted for. A deposition model based on mono-size particles enables comparison of variable parameters such as particle size, shear rate, and particle feed concentration with experimental data without the added complication and uncertainty of a poly-disperse model.
250.---------------- -- - - - - - - - - -- -- - - - - - - - - - - - .
••
~
' 1/1
E
::I.
.:
4.9 ~m nickel spa membrane 4.1 ~m nickel spa membrane 5.0 ~m isopore epa membrane 4.8 ~ m nickel epa membrane Model v, (5 ~m particles)
•o
200
res
Q)
.J::. 1/1
.2
150
Model v, (4
~m
particles)
v, (3
~m
particles)
Model v, (2
~m
particles)
Model
• •
100
-.-
50
o
200
400
600
800
1200
Shear rate at membrane surface (r w )'
1400
S-l _
Figure 8.39 Effect of particle size on critical flux for the particle deposition
model. Experimental data for all smaller sized SPO and CPO membranes plotted on same axes.
8- 52
CHAPTER 8. Results and Discussion (Surface Microfiltration) 8.10 EFFECT OF FEED CONCENTRATION ON CRITICAL FLUX In a series of experiments carried out using the 4.9
~m
nickel SPO and 5.0
~m
polycarbonate isopore CPO membranes the feed concentration was varied between 0.0035 and 0.4 kg m·3 Critical flux was determined using constant-rate filtration experiments as described in Section 7.8.4. A constant stirrer speed of 400 rpm was employed to provide a surface shear rate in the region of 600
S·l
The nickel
membrane was cleaned under sonication between experiments. It is worth mentioning that the open areas of SPO and CPO membranes were determined as 2.1 per cent (measured) and 8 per cent (manufacturer's supplied data) respectively. In Figure 8.40 the influence offeed concentration on critical flux is presented for circular and slotted geometries. The critical flux behaviour indicates that at this shear rate the slotted membrane consistently outperforms the circular pore membrane up to 0.4 kg m-3 This greater filtration capability of the slotted pores is observed in spite of a four-fo ld
200
•...
•
,
~
III
4.9
~m
nickel SPO membrane
5.0
~m
isopore CPO membrane
150
E
...
:::1.
...,.""
•
~
><
••... ~
100
•
::J
. E;:
..... . .
iii (,)
50
';:
...
0
0 0.0
.......•...
•
•
......•
•...
0.1
0.2
0.3
... 0.4
Initial feed concentration, kg m-3 •
Figure 8.40 Comparison of critical flux as a function of feed concentration for 4.9
~m
nickel SPO and 5.0
~m
polycarbonate isopore epo membranes.
8-53
CHAPTER 8. Results and Discussion (Surface Microfiltration) deficit in comparative flow area. From the previous experimental results at higher shear rates this improvement might be expected to be greater whilst at lower shear rates the circular pore membrane may give a better critical flux performance. The model follows the appropriate shape for a diffusion-controlled mechanism. At low concentratIOns a high convective flux is required to establish a concentration gradient at the surface whereas for high concentration this is not necessary. From data illustrated in Figure 8.40, on the previous page, the optimum feed concentration was identified for use in stirred-cell experiments reported earlier. Laboratory-scale work of this nature provides critical design information useful for the application of slotted membranes as a classifier.
8.11 CONCLUSIONS Millipore isopore membranes with 10.0 ~m circular pores were compared to a nickel membrane having slot rumenslOns of 13.4 x 402 ~m. Critical flux (len') was evaluated using a 'flux-stepping' method of constant-rate filtration. The CPO membrane filtered by fOrmIng a cake on the membrane surface whereas the nickel SPO membrane seemed to behave in the manner of a surface filter with no suggestion of cake deposition. The experimental results suggest that slots give better fluxes but this is not certain because of the pore size difference between the membranes. A nickel SPO membrane with smaller slot dimensions of 10.0 x 402
~
was
produced by a patented electroless nickel plating process. The 10.0 ~m slots exhibited a high dependency on surface shear at a lower feed concentration; the lower concentration is of more practical significance. The permeate flux data below critical flux was found generally to give higher transmembrane pressure drops than the corresponding clean water. Tlus suggests particulate deposition below critical flux although the grade efficiency indicated transmission of particles up to 10
~
in size.
The slotted membrane appeared to be cleaned effectively by a moderate backflush. Comparison of the particle deposition model introduced in Chapter Six (Surface Microfiltration Theory) and experimental measurements showed that the critical flux due to shear was much greater than predicted. It is possible that these high fluxes were caused by sub-critical transmission of latex particles.
8-54
CHAPTER 8. Results and Discussion (Surface Microfiltration) Constant-rate filtration experiments were performed using smaller membrane pores and slots. A nickel SPO membrane produced with 4.9 Ilm slots by electroless nickel plating was compared with the 5.0 Ilffi circular pores ofMIilipore isopore membranes. Initial surface wetting problems caused by the nickel-phosphate coating were resolved by regular cleaning with Decon 90 solution. The critical flux of the 4.9 Ilffi nickel SPO membrane was highly dependent on shear rate whilst that of the 5.0 Ilffi isopore CPO membrane was almost unaffected by surface shear. Backflushing only partially restored the flux of the slotted membrane. Both membranes gave lower fluxes than the clean water fluxes when operating below the critical flux. This is termed the weak form of critical flux. The partIcle deposition model gave good agreement with the data measured for the 4.9 Ilm nickel SPO membrane. The measured flux for the 5.0 Ilffi isopore CPO membrane was much lower than the model predicted. This is thought to be due to an additional pore-particle interaction for the CIrcular pores. Tests using a nickel membrane with circular pores of 4.8 Ilffi diameter also suggested that there were pore-particle interactIOns, although the open area was much lower than the 5.0 Ilffi isopore CPO membrane. The critical flux for the 4.9 Ilm nickel SPO membrane consistently outperformed the 5 0 Ilffi isopore CPO membrane for latex feed concentrations up to 0.4 kg m·3.
8-55
CHAPTER 9. Conclusions and Recommendations
CHAPTER NINE CONCLUSIONS AND RECOMMENDATIONS 9.1 CONCLUSIONS The main project aim reported by this thesis was to investigate filtration performance of surface microfilters and to compare the geometry of the pores of such filters. It is believed that the phenomenon of critical flux is more likely to be of practical use when using surface filters, rather than conventional membrane filters relying on depth filtration mechanisms for their pore rating. This is due to the polydispersity of most waters to be filtered: although large particles may be retained on the surface of any type of membrane filters, there is likely to be other colloidal or dissolved components that will become lodged within a depth-type filter. This leads to declining flux for a fixed pressure drop, or increasing pressure for a fixed permeate rate. However, true surface filters do not have an internal matnx for the colloidal, or dissolved, particles to adhere to. Thus, they are not as badly affected by the polydlspersity of real process waters. In the absence of particle-pore interaction, including pore plugging, a particle challenging a true surface lll1crofilter will either be retained or pass through into the permeate. If the particle is retained, there is the possibility that a secondary membrane may be formed which Itself then becomes the filtering membrane surface. Under these conditions the filtration is no longer defined by the properties of the original membrane; the properties of the particle-particle interaCtion become most significant. All these types of behaviour were witnessed during the course of this project. Constant-rate filtration experiments with a slotted membrane having a slot width of 13 lUll, and the standard challenge suspension with particle sizes up to 22 lUll, showed considerable passage of challenge particles into the permeate. The cntlcal flux was many thousand litres per metre of filter area per hour, the process was very stable and when fouled (at very high permeate flux) the fouling was easily reversed by a backflush. Thus, the challenge particles did not interact with the membrane surface to a significant degree. Obviously, particle retention was not good with this filter and a shallow grade efficiency curve resulted, with 100 per cent particle retention close to the measured filter size of 13 1J.Ill. However, it is important to note that the particles close to the apparent pore size of the filter (Le. slot width) did not cause catastrophic
9-1
CHAPTER 9. Conclusions and Recommendations fouling, or reduction in the flux, as has been reported before. This could be attributed to the novel pore geometry used in this study. An innovative e1ectroless nickel plating technique has been developed during this work to manufacture high-quality slotted surface microfilters with pore widths as narrow as 4 !lm. Development and implementation of the production process proved to be a major challenge because of the sensitivity of complex process chemistry to various operating parameters. Comparisons between the circular pore opening (Cpa) and the slotted pore (Spa) geometry were made with 10 Jlffi pore diameter membranes. Under similar operating conditions the isopore cpa filter showed signs of forming a secondary membrane very easily: the grade efficiency curve very quickly approached that of 100 per cent retention for all particle sizes below the pore size. This provides a very effective filtration barrier to the passage of particles, but shows that classification of particles is not possible under these operating conditions with the cpa type of membranes. Further comparisons were possible with a 5.0 Jlffi isopore cpa membrane and a 4.9 Jlffi nickel slotted membrane. At this pore size the isopore cpa filter exlublted a critical flux that appeared to be almost independent of the applied shear at the filter surface. By contrast, for the 4.9 Jlm slots, the critical flux was highly dependent on surface shear. The data presented within the thesis can be used to determine operating conditions appropriate for fractionation, or classification, of suspended particulate material using this type of filter. For classification, It is important that a secondary membrane is not allowed to form; hence the flux must be below that of the critical flux. Values above this will result in particle deposition, fouling and secondary membrane formation. The design of slotted Inlcrofilters as sub-lO micron classifiers for the specific material tested IS possible using graphical data presented, for grade efficiency (Figure 8.23), and critical flux as a function of solid concentration (Figure 8.40) and surface shear (Figure 8.38). The mathematical model will provide classifier throughput data for different operating conditions. The difference in operating characteristics between the nominal 5 !lm cpa and spa membranes suggests that the two filters are subject to different particle fouling mechanisms or phenomena. The isopore cpa membrane is not shear dependent and it is likely that the fouling is particle-membrane related. ThiS could be explained by a filter where the particles plug the pores very strongly and the shear is insufficient to
9-2
CHAPTER 9. Conclusions and Recommendations remove them from the surface. The nickel SPO membrane is influenced by the surface shear and an appropriate filtration model needs to allow for this. FiltratIOn modelling looked at the colloidal interactIOn of the particles and the membrane, and various other interactions that are reported in the literature. However, in discussing one of the more well-known microfiltration models, it is not very satisfactory to apply a model of van der WaaI' s attraction force integrated over infinite distance from where the interaction is supposed to occur. It would be possible to make the model fit, by a selective application of empirical constants, but it would have limited physical meaning. It is well known that van der WaaI's force is a shortrange force; so integrating over a large distance is not appropriate. A different approach was adopted based on conventional mass transfer correlations and using previously published exponents for the relation between the Schmidt, Sherwood and Reynolds numbers. This provided a very close fit to the experimental data for the nominal 5 J.UI1 slotted membrane. Thus, it should be possible to apply this model to determine the critical flux at operating conditions not investigated within this work. The conventional mass transfer type of model is founded on there being a concentration gradient of particles close to the membrane surface. It is the convective transport of particles towards the membrane surface that is balanced by the diffusional mass transfer of particles away from the surface. When these are just in balance the critical flux is achieved. An increase in the flux will cause more particles to arrive at the membrane leadmg to particle deposition and fouling. Thus, by this mechanism the critical flux reported here is of the so-called weak form; i.e. less than the flux found when filtering water only at the same condItion of transmembrane pressure. The alternative, strong form, occurs when the shear is sufficient to remove all fouling from the surface of the filter, including the concentration gradient at the surface. The
strong form is normally only observed when very large particles are filtered on membranes with much smaller pores, and in the absence of any other fouling material, i.e. no polydispersity of the challenge material.
9-3
CHAPTER 9. Conclusions and Recommendations 9.2 RECOMMENDATIONS FOR FURTHER WORK During the lifetime of this project a filter appropriate for the complete rejection of very small parasitic material, such as Cryptosporidium Parvum oocysts, was not available. Further development of the electroless nickel plating process, assisted by results from earlier sections of this thesis, has led to the consistent manufacture of nickel slotted media with slot widths down to 2 /lm. Polystyrene latex challenge material can now be specified as larger, well-defined spherical particles of 3 to 4 /lm diameter. Thus, the filtration performance of smaller nickel slotted (Spa) and isopore circular pore (Cpa) membranes can be compared for the removal of monosize latex particles as analogues for small parasitic material. A new generation of sub-micron membranes with slotted openings are currently under development. These small 'test' samples produced by a laser etching process can be evaluated against a variety of commercially available pore sizes in the isopore cpa membrane range. A more detailed model of partIcle retention at the membrane surface would result from a population balance conducted on the material challenging the filter and passing through it. The latter is aclueved from the grade efficiency curve. Thus, It would be possible to determine the concentration of particles of each size grade at the surface and this could be used in the mass transfer model to predict the cntical flux under given shear conditions. Particle-pore interaction was clearly evident with the isopore epa membrane and it may be present with the nickel spa membrane. Tests for this were attempted at the time, but with only limited success. In recent years two authors have suggested the direct measurement
of membrane-particle interaction using Atomic Force
Microscopy, with a particle of the challenge type stuck to the AFM cantilever arm approaching the membrane in the same liquid phase as the tests. This appears to be an elegant method for direct measurements of the interaction forces. The advances to rectangular pore dimenSIOns and latex particle sizes have broadened the scope for such an investigation. However, it is recommended that several tests should be conducted with the epa membranes at different stages of wetting, as the commercial track-etched filters are coated in surfactant to assist in their wetting. Hence, the particle-membrane Interaction force will be different for a fresh and used membrane. The slotted filters did not have any surfactant added during the filtrations.
9-4
CHAPTER 9. Conclusions and Recommendations From a practical point of view, it would be reasonable to assume that the open filtering area should have some influence on the filtration properties of the slotted microfilters. Currently, the open filtering area is low: between 0.5 and 5 per cent of the total area. Small 'test' samples of limited area prepared by a laser etching development process have a much higher open area and should be investigated. AdditIOnally, new substrate materials with higher starting open area are desirable to provide electroless deposited filters with higher open areas. However, It IS worth noting that the low open area does not appear to influence the validity of a mass transfer model for the 5 JUI1 nickel slotted membranes, where the defining conditions appear to be away from the membrane surface. Although this has been shown to be the case in this study it remains to be seen if this is a general result.
9-5
REFERENCES
REFERENCES Adin, A., Biswas, P. and Dutary, G., (2000) "Cryptosporidium transport in filter medium", in Proceedings of the Eighth World Filtration Congress, Brighton, The FiltratIon Society and Elsevier, pp.347-350. Aimar, P. and Howell, I.A., (1989), "Effects of concentration boundary layer development on the flux limitations in ultrafiltration", Chem. Eng. Res. Des., 67, pp.255-261. Aleksinas, M.I., (1990), ''Troubleshooting electroless nickel plating solutions", (Chapter 3 in Electroless plating: fundamentals and applications, erutors, Mallory, G.O. and Hadju, I.B.), American Electroplaters and Surface Finishers Society, Florida, pp. 101-109. Al-Malack, M.H. and Anderson, G.K., (1997), "Crossflow microfiltration with dynamic membranes", Water Res., 31, pp. 1969-1979. Altena, F.W. and Belfort, G., (1984), "Lateral migration of spherical particles in porous flow channels: application to membrane filtration", Chem. Eng. Sci., 39, 2, pp.343-355. Altena, F.W., Weigand, R.I. and Belfort, G., (1985), "Lateral migration of spherical particles in laminar porous tube flows: application to membrane filtration",
Physicochem. Hydrodyn., 6, pp.393-413. Altmann, I. and Ripperger, R, (1997), "Particle deposition and layer formation at the crossflow microfiltration", J. Membr. Sci., 124, pp.119-128. Atkins, P.W., (1998), Physical Chemistry, 6th ed., Oxford University Press, Oxford. Bacchin, P., Aimar, P. and Sanchez, V., (1995), "A model for colloidal fouling of membranes", AIChE J., 41, pp.365-376. Bacchin, P., Aimar, P. and Sanchez, V., (1996), "Influence of surface interaction on transfer during colloid ultrafiltration", J. Membr. Sci., 115, pp.49-63. Bartlett, M., Bird, M.R. and Howell, l.A., (1995), "An experimental-study for the development of a qualitative membrane cleaning model", J. Membr. Sci., 105, pp. 147-157.
I....
REFERENCES Belfort, G., (1989), "Fluid mechanics in membrane filtration: recent developments", J. Membr. Sci., 40, pp.123-147.
Belfort, G., Davls, R.H. and Zydney, A.L., (1994), ''The behaviour of suspensions and macromolecular solutions in cross-flow microfiltration", J. Membr. Sci., 96, pp. 1-58. Bhattacharjee, C. and Datta, S., (1996), "Analysis of mass transfer during ultrafiltration of PEG-6000 in a continuous stirred cell: effect of back transport", J. Membr. Sci., 119, pp.39-46.
Bhattacharjee, C. and Datta, S., (1997), "A mass transfer model for the prediction of rejection and flux during ultrafiltration of PEG-6000", J. Membr. Sci., 125, pp.303-310. Blake, N.J., Cumrning, lW. and Streat, M., (1992), "Prediction of steady state crossflow filtration using a force balance model", J. Membr. Sci., 68, pp.205-216. Blatt, W.F., Dravld, A., Michaels, A.S. and Nelsen, L., (1970), "Solute polarization and cake formation in membrane ultrafiltration: causes, consequences and control techniques" (m Membrane Science and Technology, editor, Flinn, lE.), Plenum Press, New York, pp.47-99. Bowen, W.R., Calvo, J.l and Hemandez, A., (1995), "Steps of membrane blocking in flux decline during protein microfiltration", J. Membr. Sci., 101, pp.153-165. Bowen, W. R., Hilal' N., Lovitt, R. W. and Williams, P. M., (1996) "Atomic force microscope studies of membranes: surface pore structures of Anopore and Cyclopore membranes", J. Membr. Sci, 110,2, pp.233-238. Bowen, W.R. and Jenner, P., (1995), ''Theoretical descriptions of membrane filtration of colloids and fine particles: An assessment and review", Adv. Colloid Interface
Sci., 56, pp. 141-200. Bowen, W.R. and Sabuni, H.A.M., (1992), "Pulsed electrokinetic cleaning of cellulose nitrate microfiltration membranes", Ind. Eng. Chem. Res., 31, 2, pp.5l5-523. Bromley, A.J., Holdich, R.G. and Cumrning, lW., (2002), "Particulate fouling of surface rnicrofilters with slotted and circular pore geometry", J. Membr. Sci., 196, pp.27-37.
REFERENCES Buffham, B.A. and Cumming, I.W., (1995), "Prevention of particle deposition in crossflow microfiltration", Trans. IChemE, 73, pp.445-454. Capaccio, R., (1990), "Wastewater treatment for electroless plating", (Chapter 20 in Electroless plating: fundamentals and applications, editors, Mallory, G.O. and
Hadju, I.B.), American Electroplaters and Surface Finishers Society, Florida, pp.519-528. Carrere, H., Blaszkow, F. and Roux de Balmann, H., (2001), "Modelling the clarification of lactic acid fermenation broths by cross-flow microfiltration", J. Membr. Sci., 186, pp.219-230.
, Chan, R. and Chen, V., (2001), ''The effects of electrolyte concentration and pH on protein aggregation and deposition: critical flux and constant flux membrane fIltration", J. Membr. Sci., 185, pp.l77-192. Chang, DJ., Hsu, F.C. and Hwang, S.1., (1995), "Steady-state permeate flux of crossflow microfiltration", J. Membr. Sci., 98, pp.97-106. Chang, DJ. and Hwang, S.I., (1994), "Unsteady-state permeate flux of crossflow microfiltration", Sep. Sci. Technol.,29, 12, pp.l593-1608. Chedda, P., Grasso, D. and van Oss, CJ., (1992), "Impact of ozone on stability of montmorillorute suspensions", J. Colloid Interface Sci., 153, 1, pp.226-236. Chellam, S. and Wiesner, M.R., (1992), "Particle transport in clean membrane filters in laminar flow", Environ. Sci. Technol., 26, pp.1611-1621. Chellam, S. and Wiesner, M.R, (1998), "Evaluation of crossflow filtration models based on shear-induced diffusion and particle adhesion: ComplicatIOns mduced by feed suspension polydispersivity", J. Membr. Sci., 138, pp.83-97. Chen, V., Fane, A.G., Madaeni, S. and Wenten, I G., (1997), "Particle deposition dunng membrane filtration of colloids: transition between concentration polarization and cake formation", J. Membr. Sci., 125, pp. 109-122. Christenson, H.K., (1988), "Non-DLVO forces between surfaces - solvation, hydration and capillary effects", J. Dispersion Sci. Technol., 9, 2, pp.l71-206. Chuang, CJ. and Hsu, C.W., (1998), "Hydrodynamics in a flat-plate crossflow filter and particle trajectory with electric fields imposed on the filter", J. Chem. Eng. Jpn., 31, pp.407-416.
REFERENCES Chun, M.S., Chung, G.Y. and Kim, J.J., (2001), "On the behaviour of the electrostatic colloidal interaction in the membrane filtration of latex suspensions", J. Membr.
Sci., 193, pp.97-109. Cohen, R.D. and Probstein, R.F., (1986), J. Colloid Interface Sci., 114, pp.194Colaruotolo, J. and Tramontana, D., (1990), "Engineering applications of electroless nickel", (Chapter 8 in Electroless plating: fundamentals and applications, editors, Mallory, G.O. and Hadju, J.B.), American Electroplaters and Surface Finishers Society, Florida, pp.207-227. Colton, C.K. and Smith, K.A., (1972), "Mass transfer to a rotating fluid: Part IT. Transport from the base of an agitated cylindrical tank", AlChE J., 18, pp.958967. Connell, H., Zhu, J. and Bassi, A., (1999), "Effect of particle shape on crossflow filtration flux", J. Membr. Sci., 153, pp.121-139. Cox, R G. and Brenner, H., (1968), "The lateral migration of solid particles in Poiseuille flow - I theory", Chem. Eng. Sci., 23, pp.147-173. Davis, R.H., (1992), "Modeling of fouling of cross-flow microfiltration membranes",
Sep. Pun/. Methods, 21, pp.75-126. Davis, R. H., (1992a), "Microfiltration: Definitions", (Chapter 31 in Membrane
handbook, editors, Ho, W.S.W. & Sirkar, K.K.), Van Nostrand Reinhold, New York, pp.457-460. Davis, R.H., (1992b), ''Theory for crossflow microfiltration", (Chapter 33 in
Membrane handbook, editors, Ho, W.S.W. & Sirkar, K.K.), Van Nostrand Reinhold, New York, pp.480-505. Davis, R.H. and Birdsell, S., (1987), "Hydrodynamic model and experiments for crossflow microfiltration", Chem,!)ng. Commun., 49, pp.217-234. Davis, R.H. and Leighton, D.T., (1987); "Shear-mduced transport of a particle layer along a porous wall", Chem. Eng. Sci., 42, 2, pp.275-281. Defrance, L. and Jaffrin, M.Y., (1999), "Comparison between filtratlons at fixed transmembrane pressure and fixed permeate flux: application to a membrane bioreactor used for wastewater treatment", J. Membr. Sci., 152, pp.203-21O.
REFERENCES Dennis, J.K. and Such, T.E., (1993), Nickel and chromium plating, 3rd ed., Woodhead Publishing I1mited, Cambridge, UK. Dickenson, C.T., (1997), Filters and Filtration Handbook, 4th ed., Elsevier Advanced Technology, Oxford, UK. Drew, D.A., Schonberg, J.A. and Belfort, G., (1991), "Lateral inertial migration of a small sphere in fast laminar flow through a membrane duct", Chem. Eng. Sci., 46, 12, pp.3219-3224. Droste, R L., (1997), Theory and practice of water and wastewater treatment, John Wiley & Sons, Inc., Canada. Duffek, E.F., Baudrand, D.W. and Donaldson, J.G., (1990), "Electroless nickel applications in electronics", (Chapter 9 m Electroless plating: fundamentals and applications, edItors, Mallory, G.O. and Hadju, J.B.), American Electroplaters
and Surface Finishers Society, Florida, pp.229-259. Eckstem, E.C., Bailey, D.G. and'Shapiro, A.H., (1977), "Self-dIffusion of particles in shear flow ofa suspensIOn", J. Fluid Mech., 79, Part 1, pp.191-208. Fane, A. G., (1996), "Membrane for water production and wastewater reuse", DesalinatlOn, 106, pp. 1-9.
Feke, D.L., Prabhu, N.D., Mann Jr., J.A. and Mann
rn, J.A., (1984), "A formulatIOn
of the short range repulsion between spherical colloidal particles", J. Phys. Chem., 88, pp.5735-5739.
FIeld, R.W., Wu, D., Howell, J.A. and Gupta, B.B., (1995), "Critical flux concept for microfiltration fouling", J. Membr. Sci., 100, pp.259-272. Flhppov, A., Starov, V.M., Lloyd, D.R., Chakravarti, S. and Giaser, S., (1994), "Sieve mechanism of microfiltration", J. Membr. Sci, 89, pp. 199-213. Flscher, E. and Raasch, J., (1985), Cross-flow filtratIOn", Ger. Chem. Eng., 8, pp.211216. Fradm, B. and Field, R.W., (1999), "Cross-flow microfiltration of magnesium hydroxide suspensions: determination of critical fluxes, measurement and modelling offouling", Sep. Purif. Technol., 16, pp.25-45.
REFERENCES Fricker, C. and Clancy, J., (1998), "Crypto's protocol prospects", WQI, May/June, pp.lI-14. Friedman-Huffman, D. and Rose, J., (1998) "Emerging waterbome pathogens", WQI, NovlDec, pp. 14-18. Gan, Q., Howell, J.A., Field, R.W., England, R., Bird, M.R. and McKechinie, M.T., (1999), "Synergetic cleaning procedure for a ceramic membrane fouled by beer microfiltration", J. Membr. Sci., 155, pp.277-289. Gawnlov, G.G., (1979), Chemical (electroless) ruckel plating, Portcullis Press Ltd., Redhill, UK. Gesan-Guiziou, G., Boyaval, E. and Daufin, G., (1999), "Critical stability conditions in crossflow microfiltration of skimmed milk: transition to irreversIble deposition", J. Membr. Sci., 158, pp.211-222. Goodwin, J.W., Heam, J., Ho, C.C. and Ottewill, R., (1974), "Studies on the preparation and characterisation of monodisperse polystyrene lattices", Colloid
Polym. Sci., 252, pp.464-471. Gorbunova, K.M. and Nikiforova, A.A., (1963), Physicochemical principles of nickel plating, Israel Program for Scientific Translations Ltd., Jerusalem. Green, G. and Belfort, G., (1980), "Fouling of ultrafiltration membranes: lateral migration and the particle trajectory model", Desalination, 35, pp.129-147. Gregory, J., (2000), "Monitoring water filtration processes for optimal particle removal", in Proceedings of the Eighth World Filtration Congress, Brighton, The Filtration Society and Elsevier, pp. 107-11 O. Gupta, B. B., Howell, J. A. and Wu, D., (1995), "A helical baffle for cross-flow rrucrofiltration", J. Membr. Sci., 102, pp. 31-42. Gutrnan, R.G., (1987), Membrane filtration: the technology of pressure-driven crossflow processes, lOP Publishmg, Bristol, UK. Hajdu, J., (1990), "Surface preparation for electroless nickel plating", (Chapter 7 in
Electroless plating: fundamentals and applications, editors, Mallory, G.O. and Hadju, J.B.), American Electroplaters and Surface Firushers Society, Florida, pp.193-206.
REFERENCES Harmant, P. and Aimar, P., (1998), "Coagulation of colloids in a boundary layer dunng cross-flow filtration", Colloids Surf. A,138, pp.217-230. Heran, M. and Elmaleh, S., (2001), "Microfiltration through an inorganic tubular membrane with high frequency retrofiltration", J. Membr. Sci., 188, pp.181-188. Hermia, I., (1982), Trans. IChemE, 60, p.193. Hilal, N., Bowen, R. W., Lovitt, R. W. and Wright, C. I., (2002) "(Bio)fouling of polymeric membranes: Atomic Force Microscope Study", Journal of
Engineering in Life Sciences, 2, 5, pp.131-135. Hillis, P., (1997),
"Full-scale
application of membrane microfiltration in
NorthWestWater: Huntingdon stage 4, provision of an 80 Mid plant",
Desalination, 113, pp.267-272. Ho, B.P. and Leal, L.G.; (1974), "Inertial migration of rigid spheres in twodimensional unidirectional flows", J. Fluid Mech., 65, pp.365-400. Ho, C.C. and Zydney, A.L., (1999), "Theoretical analysIs of the effect of membrane morphology on fouling during microfiltration", Sep. Sci. Technol., 34, 13, pp.2461-2483. Hodgson, P.H., Pillay, V.L. and Fane, A.G., (1993), "Visual study of crossflow micro filtration wIth inorganic membranes: resistance of biomass and particulate cake", in Proceedings of the Sixth World Filtration Congress, Nagoya, pp.607610. Hogg, R., Healy, T.W. and Fuerstenau, D.W., (1966), "Mutual coagulation of colloidal dispersions", Trans. Faraday Soc., 62, pp.1633-1651. Holdich, R.G., Cumrning, I.W., Awang, A.R.B. and Bromley, A.I., (2000), "Seeded microfiltration for base metal removal from effluents", in Proceedings of the
Eighth World Filtration Congress, Brighton, The Filtration Society and Elsevier, pp.355-358. Holdich, R.G., Cumrning, I.W. and SIDlth, I.D., (1998), "Crossflow microfiltration of oil in water dispersions using surface filtration with imposed fluid rotation", J.
Membr. Sci., 143, pp.263-274. Holdich, R.G. and Zhang, G.M., (1992), "Crossflow microfiltration incorporating rotational fluid flow", Trans. IChemE, 70, Part A, pp.527-536.
REFERENCES Holdich, R.O. and Zhang, O.M., (1995), "Internal Fouling of Microporous Cross-flow Filtration Membranes with Dilute Latex Suspensions", Chem. Eng. J., 60, pp.31-37. Holland, F.A. and Bragg, R, (1995), Fluid Flow for Chemical Engineers, 2nd ed., Edward Arnold, London, UK. Houi, D. and Lenormand, R., (1986), "Particle accumulation at the surface of a filter", Filtr. Sep., pp.238-242.
Howell, I.A., (1995), "Sub-critical flux operation of microfiltration", J. Membr. Sci., 107, pp. 165-171. Howell, I.A., Field, R.W. and Wu, D.x., (1993), "Yeast-cell microfiltration - flux enhancement in baffled and pulsatile flow systems", J. Membr. Sci., 80, pp.5971. Huisman, I.H., Vellenga, E.D., Tragardh, o. and Triigardh, C., (1999), "The influence of the membrane zeta potential on the critical flux for crossflow microfiltration of particle suspensions" (short communication), J. Membr. Sci., 156, pp.l53158. Hunter, RI., (2001), Foundations of colloid science, 2nd ed., Oxford University Press, Oxford, UK. Hwang, KJ., Yu, Y.H. and Lu, W.M., (2001), "Crossflow microfiltration of submicron microbial suspension", J. Membr. Sci., 194, pp.229-243. Ishii, K. and Hasimoto, H., (1980), "Lateral migration of a sphencal particle in flows in a circular tube", J. Phys. Soc. Jpn., 48, pp.2144-2153. Jacangelo, 1.0., Rhodes Trussel, R and Watson, M., (1997), "Role of membrane technology in dnnking water treatment in the United States", Desalination, 113, pp.l19-127. Kaufmann, T.O. and Leonard, E.F., (1968), "Mechanism of Interfacial mass transfer in membrane transport", ibid., 14, pp.421Kawakatsu, T., Nakajima, M., Nakao, S.I. and Kimura, S., (1995), "Threedimensional simulation of random packing and pore blocking phenomena during microfiltration", Desalination, 101, pp.203-209.
-----------------
-
-
-
-------------------
REFERENCES Khilar, K.C. and Fogler, H.S., (1998), Theory and applications of transport in porous media: Volume 12, Migration of fines in porous media, Kluwer Academic, Dordrecht, The Netherlands. Kosvintsev, S., Holdich, R.G., Cumming, lW. and Starov, V.M., (2002), "Modelling of dead-end microfiltration with pore blocking and cake formation", J. Membr. Sci., 208, pp.l81-192.
Kuberkar, V.T. and Davis, R.H., (2001), "Mlcrofiltration of protein mixtures with crossflushing or backflushing", J. Membr. Sci.,183, pp.I-14. Kuczma Jr., J.J., (1990), "EqUipment design for electroless nickel plating", (Chapter 5 in Electroless plating: fundamentals and applications, editors, Mallory, G.O. and Hadju, J.B.), American Electroplaters and Surface Finishers Society, FlorIda, pp. 139-167. Kuiper, S., van Rijn, C. J. M., Nljdam, W. and Elwenspoek, M. C., (1998), "Development and application of very high flux microfiltration membranes", J. Membr. Sci.,150, pp.I-8.
Kwon, D. Y. and Vigneswaren, S., (1998), "Influence of particle size and surface charge on critical flux of crossflow microfiltration", Water Sci. Technol., 38, 4-5, pp.481-488. Kwon, D.Y., Vigneswaran, S., Fane, A.G. and Ben AIm, R., (2000), "Experimental determination of critical flux in cross-flow microfiltration", Sep. Purif. Technol., 19, pp.l69-181. Leighton, D.T. and Acrivos, A., (1986), "Viscous resuspension", Chem. Eng. Sci., 41, pp. 1377-1384. Leighton, D.T. and Acrivos, A., (1987), "Measurement of the shear induced coefficient of self-diffusion in concentrated suspensions of spheres", J. Fluid Mech ,177, p.l09-131.
Leonard, E F. and Vassiheff, C.S., (1984), "The deposition of rejected matter in membrane separation processes", Chem. Eng. Commun., 30, pp.209-217.
LI, c.L., Zhao, HX., Tsuru, T., Zhou, D. and Matsumura, M., (1999), "Recovery of spent electroless nickel plating bath by electrodialysis", J. Membr. Sci., 157, pp.241-249.
REFERENCES Li, H., Fane, AG., Coster, H.G.L. and Vlgneswaran, 5., (1998), "DIrect observation of
particle
deposition
on
the
membrane
surface
during
crossflow
microfiltration", J. Membr. Sci., 149, pp.83-97. Li, H., Fane, AG., Coster, H.G.L. and Vigneswaran, 5., (2000), "An assessment of depolarisation models of crossflow rnicrofiltration by direct observation through the membrane", J. Membr. Sci., 172, pp. 135-147. Li, H.Y., Bertram, C.D. and Wiley, D.E., (1998), "Mechanisms by which pulsatile flow affects cross-flow microfiltration", AIChE J., 44, pp.1950-1961. Lojkine, M.H., Field, R.W. and Howell, J.A., (1992), "Cross-flow microfiltration of cell suspensions: A review of models with emphasis on particle size effects", Trans. IChemE, 70, pp.149-164.
Lopez, R. V., Elmaleh,
s.
and Gaffor, N., (1995), "Cross-flow microfiltration of
hydrocarbon emulsions", J. Membr. Sci., 102, pp.55-64. Lu, W.M. and Hwang, K.J., (1995), "Cake formation in 2-D cross-flow filtration", AIChE J., 41, pp. 1443-1455.
Lu, W.M. and Iu, S.C., (1989), "Selective particle deposition In cross-flow filtration", Sep. Sci. Technol., 24, pp.517-
Ma, H., Hakim, L.F., Bowman, C.N. and Davis, R.H., (2001), "Factors affecting membrane fouling reduction by surface modification and backpulsing", J. Membr. Sci., 189, pp.255-270.
MacConnick,
T.,
(1999),
"Cryptosporidium -
driving
the
shift
towards
microfiltration?", Filtr. Sep., Jan/Feb, pp.16-19. Mackley, M.R. and Sherman, N.E., (1992), "Cross-flow cake filtration mechanisms and kinetics", Chem. Eng. Sci., 47, pp.3067-3084. Madaeni, S.S., (1997), ''The effect of operating conditions on critical flux in membrane filtration of latexes", Trans. IChemE, 75, Part B, pp.266-269. Madaeni, S.S. and Fane, AG., (1996), "Microfiltration of very dilute colloidal mixtures", J. Membr. Sci., 113, pp.301-312.
REFERENCES Madaeni, S.S., Fane, A.G. and WIley, D.E., (1999), "Factors influencing critical flux in membrane filtration of actlvated sludge", l. Chem. Technol. Biotechnol., 74, pp.539-543. Madec, A., Buisson, H. and Ben Aim, R., (2000), "Aeration to enhance membrane critical flux", In Proceedings of the Eighth World Filtration Congress, Brighton, The Filtration Society and Elsevier, pp. 199-202. Mallory, G.M., (1990a) ''The fundamental aspects of electroless nickel plating", (Chapter 1 in Electroless plating: fundamentals and applications, editors, Mallory, G.O. and Hadju, J.B.), American Electroplaters and Surface Finishers Society, Florida, pp.1-56. Mallory, G.M., (1990b), "The electroless nickel plating bath: effect of variables on the process", (Chapter 2 in Electroless plating: fundamentals and applications, editors, Mallory, G.O. and Hadju, J.B.), American Electroplaters and Surface Finishers Society, Florida, pp.57-99. Malone, D.M. and Anderson, J.L., (1977), "Diffusional boundary layer resistances for membranes with low porosity", A/ChE l., 23, pp. 177Maron, S.H. and Prutton, C.F.• (1959). Principles of physical chemistry, 3rd ed., The Macmillan Company, New York. Matsumoto, Y., Mtwa, T., Nakao, S. and Kimura, S., (1996), "Improvement of membrane permeation performance by ultrasonic microfiltration", J. Chem. Eng. Jpn., 29, pp.561-567. McDonogh, R.M., Bauser, H., Stroh, H. and Chmiel, H., (1990), "Concentration polansation and adsorption effects in crossflow ultrafiltration of proteins", Desalination, 79, pp.217-231. McDonogh, R.M., Bauser, H., Stroh, N. and Grauschoph, U., (1995), "Experimental in situ measurement of concentration polarisation during ultra- and rnicrofiltration of bovin serum albumin and dextran blue solutions", J. Membr. Sci., 104, pp.51-63. McDonogh, R.M., Fane, A G. and Fell, C.J.D., (1989), "Charge effects in the crossflow filtratlon of colloids and particulates", J. Membr. Sci., 43, pp.69-85.
REFERENCES McDonogh, R.M., Fane, A.G. and Fell, C.J.D., (1998), "Influence of polydispersity on the hydraulic behaviour of colloidal fouling layers on membranes Perturbations on the behaviour of the 'ideal' colloidal layer", J. Colloids Surf., A, 138, pp.231-244. Metal Finishing Guide Book and Directory, (1994). Metsamuuronen, S., Howell, J. and Nystrom, M., (2002), "Critical flux in ultrafiltration of myoglobin and baker's yeast", J. Membr. Sci., 196, pp.13-25. Metsamuuronen, S. and Nystr6m, M., (2ooo), "Critical flux measurements by constant pressure method", in Proceedings of the Eighth World Filtration
Congress, Brighton, The Filtration Society and Elsevier, pp.203-206. Milhpore Product Bulletin, (1992), "Isopore track-etched membrane filters", Millipore Corporation, USA. Millward, H.R, Bellhouse, B.J. and Walker, G., (1995), "Screw-thread flow promoters - an experimental-study of ultrafiltration and microfiltration performance", J. Membr. Sci., 106, pp.269-279. Mir, L., Michaels, S. L, Goel, V. and Kaiser, R, (1992), "Crossflow Microfiltration: ApplicatIOns, Design and Cost", (Chapter 35 in Membrane handbook, editors, Winston Ho, W. S., and Sirkar, K. K.), Van Nostrand Reinhold, New York. Mitchell, B.D. and Dean, W.M., (1984), ''Theoretical effects of macromolecules concentration and charge on the rejection coefficients rigid macromolecules in track-etched membranes", J. Membr. Sci., 19, pp.75Mores, W.D., Bowman, C.N. and Davis, RH., (2ooo), ''Theoretical and experimental flux maximization by optimisation ofbackpulsing", J. Membr. Sci., 165, pp.225236. Mores, W.D. and Davis, RH., (2001), "Direct visual observation of yeast deposition and removal during microfiltration", J. Membr. Sci., 189, pp.217-230. Mugnier, N., Howell, J.A. and Ruf, M., (2ooo), "Optinusation of a back-flush for zeohte microfiltration", J. Membr. Sci., 175, pp.149-161. Nlcolas, S., Balannec, B, Beline, F. and Bariou, B., (2ooo), "Ultrafiltration and reverse osmosis of small non-charged molecules: a comparison study of
-
---------------------------------------------------------------~
REFERENCES rejection in a stirred and an unstirred batch cell", J. Membr. Sci., 164, pp.141155. Nunes, S.P. and Peinemaun, K.V. (editor), (2001), Membrane technology in the chemical industry, Wiley-VCH, Verlag GmbH, Weinheim. Opong, W.S. and Zydney, A.L., (1991), "Diffusive and convective protein-transport through asymmetric membranes", AIChE J., 37, pp. 1497Otis, J.R., Altena, F.W., Mahar, J.T. and Belfort, G., (1986), "Measurement of single spherical particle trajectories with lateral migration in a slit with a porous wall under laminar flow conditions", Exp. Fluids, 4, pp.l-lO. Owen, G., Bandi, M., Howell, J.A. and Churchouse, S.J., (1995), "Economic assessment of membrane processes for water and waste water treatment", J. Membr. Sci., 102, pp.77-91.
Ozekin, K. and Westerhoff, P., (1998), "Bromate formation under Cryptosporidium inactivation conditions", WQI, May/June, pp.16-17. Persson, K. M. and Nilsson, J. L., (1991), "Fouling Resistance Model", Desalination, 80,2-3, pp.123 -138. PMD (UK) Ltd, (1993), Process Data Sheet: Ni-Shield M Electroless Nickel, Issue 3, PMD (UK) Ltd., Coventry, UK. PMD (UK) Ltd, (1998), Process Data Sheet: Ni-Shield M Electroless Nickel, Issue 4, PMD (UK) Ltd., Coventry, UK. PMD (UK) Ltd, (1998), Safety data sheet: PMD Ni-Shield MMU, PMD (UK) Ltd., Coventry, UK. Porter, M.C., (1972), "Concentration polarisation with membrane ultrafiltration", Ind. Eng. Chem. Prod. Res. Develop., 11, pp.234-248.
Raasch, J., (1987), "Leading features of crossflow filtration", in La place des techniques a membrane en traitement et epuration des eaux industrielles, IFrS,
Agen, France, Nov, pp 43-56. Rautenbach, R. and Schock, G., (1988), "Ultrafiltration of macromolecular solutions and cross-flow microfiltration of colloidal suspensIOns. A contrIbution of permeate flux calculations", J. Membr. Sci., 36, pp.231-242.
REFERENCES Redkar, S., Kuberkar, V. and Davis, R.H., (1996), "Modeling of concentration polarization and depolarisation with high-frequency backpulsing", J. Membr. Sci., 121, pp.229-242.
Rtedel, W., (1991), Electroless nickel plating, Finishing Publications Ltd., Stevenage,
UK. Roessler, N., (1998), "Cryptosporidium removal: Control of Cryptosporidium in bottled water using cartridge filtration systems", Flltr. Sep., JanlFeb, pp.37-39. Romero, C.A. and Davis, R.H., (1988), "Global model of crossflow microfiltration based on hydrodynamic particle diffusion", J. Membr. Sci., 39, pp.157-185. Romero, C.A. and Davis, R.H., (1991), "Experimental venfication of the shearinduced hydrodynamic diffusion model of crossflow microfiltration", J. Membr. Sci., 62, pp.249-273.
Ruckenstein, E. and Prieve, D.C., (1976), "Adsorption and desorption of particles and their chromomatographic separation", A/ehE J., 22, 2, pp 276-283. Rushton, A., Ward, A.S. and Holdich, R.G., (1996), Solid-liquid filtration and separation technology, Wiley-VCH, Weinheim. Rushton, A., Ward, A.S. and Holdich, R.G., (2000), Solid-liquid filtration and separation technology, 2nd ed., Wiley-VCH, Weinheim. Rushton, A. and Zhang, G.S., (1988), "Rotary microporous filtration", Desalination, 70, 1-3, pp.379-394. Schonberg, J.A. and Hinch, E.J., (1989), "Inertial migration of sphere in Poiseuille flow", J. Fluid Meeh., 203, pp.517-524. Segre, G. and S!lberberg, A., (1962), "Behaviour of macroscopic rigId spheres in Poiseuille flow", J. Fluid Meeh., 14, pp. 115Sethi, S. and Wiesner, M.R., (1997), "Modeling of transient penneate flux in crossflow membrane filtration incorporating multiple particle transport mechanisms",
J. Membr. Sci., 136, pp. 191-205. Shaw, DJ., (1996), Introduction to colloid & surface chemistry, 4th ed., ButterworthHeinemann, Oxford, UK.
'
REFERENCES SiHassen, D., OuldDris, A., Jaffnn, M.Y. and Benkahla, Y.K., (1996), "Optimisation of an Intermittent cross-flow filtration process of mineral suspensions", J. Membr. Sci., 118, pp.l85-198.
Silva, C.M., Reeve, D.W., Husain, H., Rabie, H.R. and Woodhouse, K.A., (2000), "Model for flux prediction in high-shear micro filtration systems", J. M embr. Sci., 173, pp.87-98.
Sirkar, K., Puri, P. and Tonkovich, A.L.Y., (1999), "Membrane Technologies", (Chapter 3 in Emerging Separation and Separative Reactions Technologies for Process Waste Reduction, editors, Radecki, P.P., Crittenden, J.C., Shonnard, D.
R. and Bulloch, J.L.), Center for Waste Reduction Technologies, American Institute of Engineers, New York. Smith, ID, (1998), The surface filtration of 011 in water disperSIOns With imposed fluid rotation, Ph.D. Thesis, Loughborough University. Smith, K.A., Colton, C.K., Merrill, E.W. and Evans, L.B., (1968), "Convective transport in batch dialyser: determination of true membrane permeability from a single measurement", Chem. Eng. Prog. Symp. Ser., 12, 84, pp.45-58. Sondhi, R. and Bhave, R , (200 1), "Role of backpulsing In foulIng minimization In crossflow filtration with ceramic membranes", J. Membr. SCI., 186, pp.41-52. Sondhl R., Lin, Y.S. and Alvarez, F., (2000), "Crossflow filtration of chromIUm hydroxide suspensions by ceramic membranes: foulIng and its minimisation by backpulsing", J. Membr. Sci, 174, pp.l11-122. Suzuki, M. and Oshima, T., (1985), "Co-ordination number of a mUlti-component randomly packed bed of spheres with size distnbution", Powder Technol., 44, pp.213-218. Tarleton, E. S. and Wakeman, R .J., (1993), "Understanding flux decline in crossflow microfiltration: Part I. Effects on particles and pore size", Trans. IChemE, 71, Part A, July, pp.399-410. Tarleton, E. S. and Wakeman, R. J., (1994a), "Understanding flux decline in crossflow microfiltration: Part IT. Effects of process parameters", Trans. IChemE, 72, Part A, pp.431-440.
REFERENCES Tarleton, E. S. and Wakeman, R. J., (1994b), "Understanding flux decline in crossflow microfiltration: Part ill. Effects of membrane morphology", Trans. IChemE, 72, Part A, pp.521-529.
Tassopoulus, M.J., O'Brien, J.A. and Rosner, D.E., (1989), "Simulation of microstructure/mechanism relationships in particle deposition", AlChE J., 35, pp.967-980. Turker, M. and Hubble, J., (1987), "Membrane fouling in a constant-flux ultrafiltration cell", J. Membr. Sci., 34, pp.267van Oss, C.J., (1994), Interfacial forces in aqueous media, Chapter I, Marcel Dekker Inc., New York, . Vasseur, P. and Cox, R.G., (1976), "Lateral migration of spherical particles in twodimensional shear flow", J. Fluid Mech., 78, pp.385-413. Vassilieff, C.S., (1992), "Convective model of cross-flow microfiltration", Adv. Colloid Interface Sci., 40, pp. 1-36.
Vllker, V.L., Colton, C.K. and Smith, K.A., (1981), "Concentration polarisation
ID
protein ultrafiltration. Part I: An optical shadowgraph technique for measuring concentration profiles near a solution-membrane interface", AIChE l., 27, 4, pp.632-637. Visser, J., (1972), "On Hamaker constants: a comparison between Hamaker constants and Lifshitz - van der Waals constants", Adv. Colloid Interface Sei., 3, pp.331363. Wakeman,
R.J.,
(1994),
"VIsualisation
of cake
formation
in
crossflow
microfiltration", Trans.IChemE, Part A, 72, pp.530-540. Wakeman, R.J. and Sabri, M.N., (1996), "Utilizing pulsed electric-fields in cross-flow Ill1crofiltration oftitania suspensions", Chem. Eng. Res. Des., 73, pp.455-463. Wakeman, R.J. and Tarleton, E.S., (1986), "Membrane foulIng prevention in crossflow microfiltration by the use of electric fields", Chem. Eng. Sci., 42, pp.829-842. Wakeman, R.J. and Tarleton, E.S., (1991), "An experimental study of electroacoustic crossflow microfiltration", Chem. Eng. Res. Des., 69, pp.386-397.
REFERENCES Wakeman, RI. and Williams, C.J., (2002), "Additional techniques to improve microfiltration", Sep. Purif. Technol., 26, pp.3-18. Wang, Y.Y., Howell, I.A., Field, R.W. and Wu, D.x., (1994) "Simulation of crossflow filtration for baffled tubular channels and pulsatile flow", J. Membr. Sci., 95, pp.243-258. Weigart, T., Altmann, 1. and Ripperger, S., (1999), "Crossflow electrofiltration in pilot scale", J. Membr. Sci., 159, pp.253-262. Wenten, I.G., (1995), "Mechanisms and control of fouling in cross-flow microfiltration", Filtr. Sep., 32, pp.252-253. Whorlow, R.W., (1992), Rheological Techniques, 2nd ed., Ellis Horwood, Chichester, UK.
Wiese, G.R and Healy, T.W., (1970), "Effect of particle size on colloid stability", Trans. Faraday Soc., 66, pp.490-499. Wiesner, M R. and Chellam, S., (1992), "Mass transport considerations for pressuredriven membrane processes", J. American Water Works Assoc., 84, pp.88-95. Wilkes, 1.0., (1999), Fluid mechanics for Chemical Engineers, Prentice Hall PTR, London, U.K. Wu, D., Howell, I.A. and Field, RW., (1999), "Critical flux measurement for model colloids", J. Membr. Sci., 152, pp.89-98. Yamazaki, I.M., Paterson, R and Geraldo, LP., (1996), "A new generatIOn of tracketched membranes for micro filtration and ultrafiltration. Part I. Preparation and characterisation", J. Membr. Sci., 118, pp.239-245. r
Yoon, S.H., Lee, C.H., Kim, K.J. and Fane, A.G., (1999), ''Three-dimensional simulation of the depOSItion of multi-dispersed charged particles and prediction of resulting flux during cross-flow microfiltration", J. Membr. Sci., 161. pp.720. Zydney, A.L. and Colton, C.K., (1986), "A concentration polarization model for the filtrate flux in cross-flow microfiltration ofparticulate suspensions", Chem. Eng. Commun., 47, pp.1-21.
REFERENCES
TABLE OF JOURNAL ABBREVIATIONS
Journal abbreviation Adv. Colloid Interface Sci. AIChEJ. Chem. Eng. Commun. Chem. Eng. J. Chem. Eng. Prog. Symp. Ser. Chem. Eng. Res. Des. Chem. Eng. Sci. CollOId Polym. SCI. Colloids Surf. A Environ. Sci. Technol. Exp. Fluids Filtr. Sep. Ind. Eng. Chern. Res. J. American Water Works Assoc. J. Chem. Eng. Jpn. J. Chem. Technol. Biotechnol. J. CollOId Interface Sci. J. Dispersion Sci. Technol. J. Fluid Mech. J. Membr. Sci. J. Phys. Chem. J. Phys. Soc. Jpn. Miner. Eng. Powder Technol. Sep. Purif. Methods Sep. Punf. Technol. Sep. Sci. Technol. Trans. Faraday Soc. Trans. IChemE Water Res. Water SCI. Technol.
Journal title Advances in Colloid and Interface Science AIChE Journal Chemical Engmeering Communications Chemical Engmeering Journal Chemical Engineering Progress Symposium Series Chemical Engineenng Research and DesIgn Chemical Engineenng Science Colloid and Polymer Science Colloids and Surfaces A Environmental Science and Technology Expenments in Fluids Filtration and Separation Industrial and Engineenng Chemistry Research Journal of the American WaterWorks Association Journal of Chemical Engineering of Japan Journal of Chemical Technology and Biotechnology Journal of Colloid and Interface Science Journal of Dispersion Science and Technology Journal of Fluid Mechanics Journal of Membrane SCIence Journal of Physical Chemistry Journal of the Physical Society of Japan Minerals Engmeering Powder Technology Separation and Punfication Methods Separation and PunficatIon Technology SeparatIon Science and Technology Transactions of the Faraday Society Transactions of the InstitutIon of Chellllcal Engmeers Water Research Water Science and Technology
NOMENCLATURE
NOMENCLATURE
A
External membrane surface area
m2
All
Particle-particle Illteractlon constant
dlmensionless
All
Dispersion medium-dispersion medium interaction
dimensionless
constant Am
Effective Hamaker constant between identical particles
dimenslonless
and dispersing medium
a
Stirrer (cone) radius
m
b
Active membrane radius
m
or, dimensionless distance from wall
dimensionless
C
Particle concentration
kgm-3
Cb
Concentration of particles III bulk suspension
kgm-3
Cg
Concentration of particles in gel layer
kgm-3
Cw
Concentration of particles at membrane surface
kgm-3
c
Molar concentration of electrolyte
D
Diffusion coeffiCient
g mol rl m 2 S-1
DB
Particle diffusion coefficient
m 2 S-1
D,
Shear-enhanced diffUSIOn coefficient
m 2 S-1
d,
Column internal diameter
m
dh
Hydraulic mean diameter of channel entrance
m
dm
Membrane diameter
m
do
Column outside diameter
m
dp
Diameter of particle
m
dso
Mean particle size
m
e
Elementary charge of electron
C
g
AcceleratIOn due to gravity
m S-2
H
Ratio of separation distance to particle radius
dimensionless
h
Separation distance between particles
m
ho
Minimum eqUilIbrium distance
m
J
Permeate flux
m S-1
Jent
Cntical flux
ms-I
Je
EqUilIbrium flux
ms-I
NOMENCLATURE
k
Mass transfer coefficient
ms-I
kB
Boltzmann constant
JKI
kd kg
Mass transfer coefficient for particle diffusion
ms-I
Mass transfer coefficient for shear-enhanced diffusion
ms-I
L
Membrane length
m
m
Stirrer (cone) height
m
NA
Avogadro's constant
gmor l
n
Membrane thickness
m
no
Particle number concentration
numberm-l
LJPm
Pressure difference across the membrane
Nm-2
M>TMP
Transmembrane pressure drop (TMP)
Nm-2
Q
Column feed volume to overflow level
ml
q
Total feed flow rate
m 3 S-I
R'/
Sum of irreversible and reversible fouling terms
m-I
R"
Resistance due to irreversible fouling which is not
m-I
removed when the is pressure released
Rm R,
Clean membrane hydraulic resistance
m-I
Resistance due to reversible fouling and concentration
m-I
Polarisation
rm
Membrane (full) radius
rp
Particle radius
m
s
Axial gap between stirrer (cone) tip and membrane
m
T
Average feed suspension temperature
K
t
FlItratlon time
s
Umean
Mean tangential velocity of flow over membrane
ms-I
V
Total volume of permeate
ml
VAB V'AB
Acid-Base (AB) interaction energy
J
Hydrophobic interaction
J
VB
Potential barrier between particles induced by surface
m
interactions
VBR VDLR
Born repulsive interaction energy
J
Electrostatic double-layer repulsive interaction energy
J
VLVA
London-van der Waals attractive interaCtion energy
J
NOMENCLATURE
VI
Total interaction energy
J
Velocity of body
m S-I
Mean velocity of fluid
m S-I
Effective back transport velocity by particle diffusion
m S-I
Superficial permeation velocity
m S-I
Gravitational settling velocity
m S-I
Effective back transport velocIty by interaction
m S-I
enhanced migration VI
Effective back transport velocity by inertial lift Effective back transport velocity by shear-enhanced diffusion
VtotaJ
Total back transport velocity (VI not considered)
m S-I
w
StlITer shaft diameter
m
y
Column height to overflow
m
z
Charge number of electrolyte
dimensionless
NOMENCLATURE
Greek Symbols
Yw
(gamma)
Shear rate at membrane surface
S-I
0
(della)
Boundary layer thickness
m
e
(epsilon)
Void fraction (porosity)
dlmenslonless
er
(epsilon)
Relative permittivity of water (dielectric constant)
dimensionless
t;.,
(epsilon)
PermittivIty of water
Fm-I
t{)
(epSIlon)
Permittivity of free space (in a vacuum)
Fm-I
(
(zeta)
Zeta potential of particle
V
B
(thela)
Stirrer (cone) angle
radians
I(
(kappa)
Conductivity of solution
Srn-I
I(D
(kappa)
Debye reciprocal length (thickness of double layer)
m-I
Am AO..
(Lambda)
Molar conductivity of electrolyte
S m 2 gmorl
(Lambda)
Limiting molar conductIVIty
S m 2 gmor l
A.
(lambda)
Decay length of liquid molecules
m
Jl
(mu)
Fluid viscosity
kg m-I S-2
V
(nu)
Kinematic viscosity
m 2 S-I
Pt
(rho)
flUId density
kgm-3
Pp
(rho)
Particle density
kgm-3
(j
(Slgma)
Atomic colhsion diameter
m
?
(phI)
Volume fraction of particles
dlmenslOnless
\fIo (pSI)
Stem potential
V
n
(Omega)
Stirrer rotational speed
revs min- I
(iJ
(omega)
Rotational (angular) velOCIty of stirrer
radians S-I
(Greek names for the indIVIdual Greek letters appear in brackets)
APPENDICES
APPENDICES
APPENDIX A
Electroless NIckel Plating Experimental Details and Results
APPENDIX B
Electroless Nickel Plating Procedure for the Production of Metal Mlcrofilters
APPENDIX C
RadIal and Tangential Shear Rate Analysis of Stirred-cell Operation
APPENDIX D
SupportIng Calculations for Particle DepOSItion Model
APPENDIX E
Pdesol™ Model for Interaction Enhanced Migration
APPENDIX F
Stirrer Calibration
APPENDIX G
Pressure Transducer CalIbration
APPENDIX H
Pore Size Distributions of Nickel SPO and CPO Membranes '
APPENDIX I
Constant-rate Filtration Experimental DetaIls and Results
APPENDIX J
Relative Importance of Membrane Open Area to CntIcal Flux' .
APPENDIX K
Author's PublicatIons
APPENDIX A. Electroless Nickel Plating. Experimental Details and Results
APPENDIX A ELECTROLESS NICKEL PLATING: EXPERIMENTAL DETAILS AND RESULTS KEY TO EXPERIMENTAL Metal Microfilters NiCPO-17
Nickel screen: l7.l fl1I1 (average) circular pores
NICPO-39
Nickel screen: 39.8 fl1I1 (average) circular pores
NiSPO-13
Nickel screen: 13.4 fl1I1 (average) slotted pores
SSPO-29
Stainless steel mesh: 29.2 (average) fl1I1 square pores
Solution Agitation AS
Air sparge
OHS
Overhead stirrer
PHC
High capacity peristaltic pump
PLC
Low capacity peristaltic pump
Example
PHC 30 (4) R
High capacity pump, setting 30, (4 mm tubing thickness), flow reversed
Pre-treatment PT (I - 16)
Pre-treatment schedule (see Section 3.4)
El
Electroless nickel plating
E2
Nickel electroplating
E3
Nickel electroplating with polarity reversal
N
Number of electrodes
Replenishment (%)
Percentage solution replenishment
F
Solution filtered before plating (0.8 fl1I1 membrane)
A-I
Table A.1 Experimental conditions for pre-treatment and electroless nickel plating of 17.1 Jlm nickel epo screen .
-... l!!
a X
C GI
W
E
.~
GI
Il.
x
w
1 3 4 5 7 10 12 19 24 27 33 35 38 40 41 42 43 44
..
UI
-'"
"
GI
GI
a;
c
Qi
a:
21/12198
11/02199 16/02199 19/02199 22102199 25/02199 01/03/99
02103/99
C GI
C GI
" c Co GI._ Ea; .~ E
E a;
;:E
GI
00
......
GI ...
Il.O
x-
w.5
:::1:::1
Il.
'
-
New solution and compact no Plating with downslzed riQ New Sartonous filter holder Longer HCI stage Agitation with overhead stirrer 4 samples (23, 33, 43, and 52 mins) Assess plating solution Effect of cr in Ni bath Assess platlnQ solution Lower temperature Low pump flow ' ,
-
'
PT4
,
,
Tubular flow comparable HCI deQass Introduced ,
No Na2C03, buffer solution Plating solution pushed upwards
,
,
0
.~
'"
:!:: Cl Cl;
a:
ml TrouQh F 1800 F 1800 1800 1400 1400 1400 1400 1400 1400 1400 1400 1400 1400 1400 F 1400 1400
E
;:
;:
Cl
'Il."
Cl C
,
100 , ,
50
29
PHC30 R PHC30 R PHC30 R PHC30 R OHS OHS OHS OHS OHS OHS PHC20 R PHC 12 R PHC 13 R PHC 14 (4) R PHC 14 4 PLC70 4 PHC10 4 PHC 10 (4) R
ci. E
.. GI
.c a;
Jl
E
a;
0-
ii:
~
mins 35 75 45 106 82.5 20 23 30 60 30 60 85 90 60 60 110 60 60
84.6 78.4 79.0 78.5 70.0 80.0 84.0 86.2 84.8 85.2 79.4 78.5 78.0 79.0 81.0 82.5 82.5 83.5
:::I
%
,
GI
0
C
GI
Ul>
- -
C
E .c UI '2 GI ii.
5 GI
l!!
- ,
23/10/98 03/11/98 05/11/98 06/11/98 16/11/98 20/11/98 26/01/99 01/02199 03/02199 09/02199
...
push push push push
-
, '
,
push
°c
Table A.I (continued) Experimental conditions for pre-treatment and electroless nickel plating of 17.1 J.U1l nickel epo screen.
..
III
15.
C
Cl)
E
·cCl)
r::L
><
w
><
" c Co
"C
E"tU ·c E CI) ..
CI)_
Cl) Cl)
Cl)
"tU
c
"tU
r::LO
Gi
><-
woE
a:
.. .. ..
.. C
Cl)
C
E .c III CCl) ii. Cl)
Cl)
E "tU Cl)
•
Cl)
D.
C
Cl)
~E :::l :::l 00
a:
m>
02103/99 04/03/99 04/03/99
09/03/99 09/03/99 09/03/99 22103/99 24/03/99 25/03/99 29/03/99 30/03/99 31/03/99 08/04/99 09/04/99 13/04/99 13/04/99 14/04/99
44 44 ~
,
~
48
51 53 53 53 53 ,
61 61
Na2C03 run Delonised water in place of plating solution Delonised water In place of plating solution no Na2CO a treatment stage Na2C03 treatment stage Plating solution pumped occasionally Pump on-off (60 son, 60 s off) Pump on-off (40 son, 300 s off) Sample left until bubbles observed Pumping activated after 10 mlns Pumping commenced after 5, 15 minutes
,
'~
,
, ~
1400 1400
50 50
F
~
~
,
1400 Aim to completely plate sample Plating solution stirred Air sparge Introduced, 40 minute run time Air sparge, 20 minute run time Air sparge, 30 minute run time
Cl)
1:i ca
:0=
0
C
0
~
E Cl C
r::L
:!::
E
:0=
<
D.
ii:
Cl
ca
:::l
%
ml 45 46 47 48 49 50 51 52 53 55 56 58 59 60 61 62 63
C
29 ,
~
PHC 10 4) R PHC10 4) R PLC 40 4) R PLC 40 (4 R PLC max 4) PLCmax 4) PHC 14 4) PHC 14 4 PHC10 4 PLC 70 4 PLC 70 4) PLC 70 (4) R PLC 70 (4) OHS AS AS AS
push push push push
,
push
mins 60 40 120 120 120 120 120 120 120 80 110 70 107 60 40 20 30
ci. E Cl)
..
.c "tU
.c
~ °C 850 84.3 85.0 845 84.5 84.5 84.5 84.8 84.9 84.6 85.2 84.6 84.5 86.3 85.5 85.5 86.0
Table A.I (continued) Experimental condItions for pre-treatment and electroless nickel plating of 17.1 /1IIl nickel epo screen.
..: Q)
.. ...
Q.
cQ)
>< Q)
E
"Cl
'i: Q)
a.
><
W
GI
Q)
"la
C
"la
Gi CC
..
..
UI
Co GI._
E"Ia 'i: E Q) ... a.o
E "la Q)
... "';'
lI!
>
D.
Q)
14/04/99 15/04/99 15/04/99 15/04/99 15/04/99 15/04/99 19/04/99 19/04/99 04/05/99 04/05/99 04/05/99 04/05/99 05/05/99 05/05/99 05/05/99 05/05/99
61 61 65 65 65 65 65 65 72 72 72 65 76 76 76
Air sparge 25 mlns plating time) Air sparge 10 mins plating time) Airsparge 40 mlns plating time} Air sparge (20 mins plating time) Air sparge (50 mlns plating time) Air sparge 30 mins plating time Air sparge 60 mlns plating time Airsparge 45 mins plating time No air sparge 40 mins plating time No air sparge 20 mlns plating time No air sparge 60 mlns plating time No air sparge (50 mins plating time) Air sparge (60 mins plating time) Air sparge (50 mlns plating time) Air sparge (45 mlns plating time) Air sparge (30 mlns plating time)
;:lE
00 (J»
CC
5Q)
0
C 0
..
a.. i
::I ::I
Q)
~
~
'-::~~ ~
~ III a.
c ~
::I
D.
AS AS AS AS AS AS AS AS no AS no AS no AS no AS AS AS AS AS
Q)
E ;:l
E
%
~
~
C
E
.c UI '2 Q)
ml 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79
ci.
C
cQ)
.. c
Cl
....
E Q)
.c III oD
ii:
~
mlns 25 10 40 20 50 30 60 45 40 20 60 50 60 50 45 30
84.8 80.0 80.0 800 800 79.6 80.5 79.3 79.3 80.0 79.9 80.0 79.6 79.6 79.3 79.8
°C
Table A.2 Observations made during pre-treatment and electroless nickel platmg of 17.1
-
II
'"
Q.
~
C CII
e ;: CII
0.
.ll
1
"CII ."
."
~
Q
g>",
C'" CIIC
.C ~o
e.2
.!!;
c. ..
-. ~~ CII
Q;
a:
a.o
.!!.o
U .l< '"
11.10
23/10/98
~
3 4
..
.
.
~
,
5
06/11/98
,
,
,
,
16/11/98
10
20/11/98
12
21/12198
,
,
,
,
.!!~.a
a.e
11.1 CO
o g! a._
,
Pressure rose qUickly In 15mlns Pump reversed and retumed pressure dropped
~
'.
,
z
Clu
"
,
reversing pump reduced pressure took a while. Pressure increased gradually No bubbles observed downstream of sample Pressure Increase on the suction Side was slow Initially. then static. Pressure was slow to recover after flow reversal. Magnetic stirrer was activated after 50 minutes and plating reaction observed Small bubbles observed leaving sample after 60 minutes Small bubbles were seen on surface Immediately after Immersion small bubbles were COPIOUS observed on surface and nSlng up through plating bath
CIIO
Whatman filler holder used. When the flow was reversed the tubing was forced off at the pressure transducer. Problems keeping the holder in the desired posrtlon thereafter Blue filter holder used NI plating was minimal
,
,
Blue filter holder was used with a poor seal. Rate of pressure Increase was small dunng plating experiment. Pore size of 20 microns measured New transparent Sartonous filler holder was utilised The penstalllc tubing filled With bubbles at times dUring the expenment. A nng around the sample outer edge was noticeable after the expenment. Plating on the sUlface was not Uniform With some pores untouched
, ,
t
..
,
,
.'
~
_C
ION
,-
.
"
. 7
e- i:
~CIICII
,
03/11/98
05111198
-
0
0.
-0
-
.' ,
CD.!!:
ei:
u",
[
CPO screen.
m '"" ~=-mu . -e '0 .." ,,et: .; :!le 00
0.5 c m~ 0
- c."
~CII
';'CII CII'" _.0
-. iilc
[~
Ill",
~m nickel
,
,
Overhead stirrer Introduced. Complete plating of most pores Pore size determination uSing microscope and gratlcule revealed pore size reduction after plallng to 14 microns High temp meant that even first sample was completely plated
Table A.2 (continued) Observations made during pre-treatment and electroless nickel plating of 17.1 IlID nickel epo screen.
-.. .. -..
ol
I:
E
0;:
0.
III
C
19 24
01/02199
-..,.'" 0.
.
1ii a;
a:
"'",
CUI .. I:
E.!:!
§.::::-
- . - ....
~'E
c. ..
.
u", .!!.c
,/,"
"'N
.!! .!!.a
c.:S
1111:0
0>
c.~
'0"
1!!
0.
0
.;
z
.:g -.
UI -I:
.. E E "0 Clu
i
. .
.
.
"
~
.
,
,
" ~
. ,
1 "u~vv
Hel .~!age sOnicated
16/02199
wasn't
"
1,
.
.
i
c'
~
T
,
.
. . ~
I
seen nSlng up suction side tUbing. Pressure dropped after flow reversed. Suction Side pressure >400mbar Pressure drop ~h~!!~r ,With stirrer on low setting Small bubbles seen nSlng up to surface. Bubbles were nSIn~ up tubln!) from filter holder Flow 'U""'UO~ b"efl~ after 12 mlns and stepped up to 15 setting Sample surface coated In black depOSit, which came off after sonication. May've been related to ,step Few signs of :b~~~led'~lany! Problems with tubing I pump.
~
155
164
20
to be plated on visual Inspection but -:0-.", I otherwise Some holes completely blocked. Many have coating around the pore suggesting flow Impeded
,
,
"
"
"
,.
,
.
.
.
.
0,
~
19.1
.
~
186
10
~g,:,~,~ ~v~" ,~J to have ou"gud finish similar to that of a plated Item. Microscope confirmed very little plating.
,
• Tu~~ng
,~
T
"
Surface reacting less 'on removal
41
•
,
Sample completely plated Left In solution too long
"
"
~
T
" "
,
40
-"'u.
~::::-
..;: .0 -0 ....
-0
u"" ..
1110
§.
~
33
38
- c." 2 - C!
2~
27
35
CI).5 c CD.!:::
",_0
.!!!;:
~~ '" .. .c c.o
...
"''''
.El: -0
.
~
.~
0,
.
filled with air InllIa!ly, vu' ,u,u~, "'~ filter pnor to use and '~~lIlng .t~blng with Pores were . - 15
Table A.2 (continuet!) Observations made during pre-treatment and electroless nickel plating of 17.1 IlIll nickel epo screen.
..:
-" !
.... -;1:.. ..,.... --'" 1= 15.
I:
... -., c'" E
'l:
"-
w
42
25/02199
43
01/03/99
44
02103199
Cl ..
Eo!! ",os
a:
.,.5 .. _
-.
CD..!;:: !! -"- i:
100
0.=
wo
,
~
.. \
45
02103/99
44
46
04/03199
44
" ~
"
.
WI:O
11.:5-
~:::-
.-"u'" .,et: ~o ION
.!~.a
.!!!.c
~.a
11.0
'" -" .
~c -0
U.l< ..
U ..
§.
§.=-os
-
e~
!i:
'ii
Cl .. C 0
.El:
o " Q,~
..i!! 0
"-
'0 .; z
04/03199
48
09/03/99
44
~.
.
.
" ~
..
. 0
. 49
09/03199
48
Na2C03 treatment Contact for 2 mInutes and qUIckly put In holder
.. ,
, "
.0
.
.
"
.
.
~
.
.
.
, ~
..
.
,
. ~
.
~
~
0
,",u
,
~
No Na2C03 treatment
E E
FIsh tank removed. Two 1kW heaters utIlised
~
47
~
2! ..
PlatIng solutIon filtered after nrckel metal found In bath
Few sIgns of reactIon on the sample surface Very little change In pressure, apart from when It reached 60 mbar. Pressure dropped back to 0 mbar Pressure remaIned low throughout, le. -0.06 mbar (suggestIng suctIon) • Flow reversed every 5 mInutes Pressure dId not rise above 0 06 mbar "
.. asi
"
. ,
PlatIng result sImIlar to expenment 44. Sample appearance was dIfferent WIth a darker surface possIbly as a result of sodIum carbonate treatment Sample was Inrtlally sonrcated In delonrsed water for 5 mInutes. Water pushed up through sample Delonrsed water replaced platIng solutIon. Flow reversed every 5 mInutes, pump settIng 30 When hlter holder was constructed In buffer solutIon, the o-rrng slipped. Area behInd o-nng plated non unrformly. The sIze of pores which expenenced flow were only sllQhtly reduced to - 15 mIcrons SodIum carbonate treated sample dId not seem to be tamlshed as usual. Pore sIze reductIon was mlnrmal, i e. - 18 mIcrons. Only slight darkenrng around the Dores. Platlng~ observed where o-nng rests
Table A.2 (continued) Observations made during pre-treatment and electroless nickel plating of 17.1 IlIIl nickel epo screen.
-.,... - . -.,., =~
oi
-., ~
E
"Cl
Co
..!!
., .,
"C
10
.ll
C
50
09/03199
51
22103199
52
24/03/99
53
25/03/99
. ., ..
"' .se -0
Cc
c.
c
"0
E-
- c."
!o.,
c.O
1110
_.0
.!! .!:!.a I1Ico
.b" ~.o
a:
HCI heated on 0, 5, 11 minutes. Sample effervesced. Reaction I," occurred when added to Na,CO, solution
51
,
,
,
Bubbles observed nSlng from surface towards end of plating expenment.
Bubbles observed on NI sample. Other than initially, there were few bubbles nSlng through solullon , dunng pumping Pressure rose to 20 mbar between 10 - 20 minutes (without pumping) Small bubbles observed nSlng up from surface of sample without " oumOlna Sample reacted after 60 s.
,,'
"
2- ~ _Gl., u", ..
"
"
,
"'''0 en" 4D.! ::
.. cc
.!;: c. .. 2~ U.,
. . .- . ca5i .§.~
-s i!!c
-0
"N
§.
~e
"u
., ~
0
Co
00 Q.c
Q.~
"0 .; Z
Clu
36
39
20
142
144
20
little pressure Increase when pumping every 10 minutes. O-nng became stuck to the NI sample (never previously observed) Plating was random NI sample left in selutlon fer too long Surface appears different with a nice finish and shlnv reverse Side Old not seem to be as much actiVity as preVIous experiment. Flow reversed on 28 minutes. Few holes were completely blocked
7.1
7.8
20
, ,
"
\ "
, 55
29/03199
'
"
53
"
"
"
"
,
,
"
"
, ,
, ,
,
,
,
"
, "
,
Surpnsed to see pressure remain high after backllush. Flow reversed after 32 minutes. Lots of bubbles observed passing through sample on retum to normal pumping. Pressure dropped to zero
~E
2!t: o GI
~~
7.1
7.8
20
~E
"0
Pump direction reversed for 40 seconds after 30, 60 and 90 minutes Small bubbles leaVing surface ceased when pumping commenced. Bubbles remained when pump activated on 60 minutes Pump started In reverse after 5 minutes and solution pumped up (setting 10) through sample. A lot of gas was pulled through tubing when pump in reverse mode. When pump stopped after 8 minutes, there was no sign of bubbles Pump LC Introduced after 25 minutes (problems encountered with Pump HC), settlng40 Stepped upJe setting 70 after 30 minutes Pressure 21 - 23 mbar at end of experiment
Table A.2 (continued) Observations made during pre-treatment and electroless nickel plating of 17.1 JUIl nickel epo screen.
...: l!!
-" c E
-.
".."
W
c
56
58
59
-..
CD
.Ec Ojo
.,Cc
m[:
~!
.!! o.a;
-OJ
Gi
&~ ~.D
u". en
U ..
.!!.!:!.a
.!!.D
0..0
30103/99
53
Sample remained In Na"C03 and buffer solutions for longer than olanned
31/03199
53
Wo
,
53
wco
,
"
, ,
.
,
,
60 61
09/04199 13/04199
,
"
, 62
13104199
61
Pump activated after 5 minutes. Flow reversed initially to remove air Irom tubing Bubbles observed nsing through solution No sign of bubbles. No nse In , pressure Little sign of plating Inlbally Few bubbles observed on surface Pump started In reverse after 5 minutes Bubbles passed through tube from sample
[
-0
_IU
CD .~ ~
ei:
.,. ~
~e
.. N
IOu
00 0..:5-
o ~ 0.._
0
0
0
z
Reaction observed on surface after 30s.
36
38
20
9.1
99
20
,
,
,
,
I, ,
, , 0
0
63
14/04/99
61
,
, ,
.'
.
,
,
,
,
0
,
0
,
,
.. ."
_c
~E
~E
"0 Cl ..
All pores had received plating
"
, ~
[~
11-sc
-. -e-....c:
1i;
a:
08/04/99
m., cni 2
.. C E.2
'C
~
m.,
1: ..
,
O-nng seal was questioned. Non-unrtonm plating noted with some areas of sample completely plated Pump direction changed after tubing filled. Lots of small bubbles passed through at once. Pump was operated fro 2 minutes wrth little actiVity. When pump stopped more bubbles were noticed on surface than previously. Pressure had nsen to 20 mbar at end of expenment. All holes have been plated. Holes directly below o-nn~ were untouched NI sample was completely plated Compressor generated many bubbles, which could Interfere with nickel plating process. All pores seem to have received platln~ Dark areas on surface after plating were similar to scorch marks. Surface coating was noticeably smooth. Greater un"onmly of plallng observed than results from pumped expenments. Most holes plated Difficult to detenmne pore size uSing microscope and gratlcule Wide range of pore sizes from very small pores up to 10 microns In diameter The two lines across centre of sample correspond to most slgmflcant plating
Table A.2 (continued) Observations made during pre-treatment and electroless nickel plating of 17.11lm nickel CPO screen .
..:
-" l!!
E
"c.
.ll
-. IQ
Q
64 65 66
1:.,
Ill.,
"c:
.Ec:
".!!IQ ~~
c.IQ 2~
c.
c:
"C
-.,.."
a;
a:
-0
Ill.,
.!!;:;
or'" ",.,
u., -"
1l~CI) _ _ .<:I
1110
lIIc:o
.!!.<:I
11.0
~
-0
",u.,
00
7~
~
,
. ..
.,
l1.e
,. '
t!t:
., -
11.~
z
~:::-IQ .,u o~
"c.0 ~
'iii ~E
0
~E "'0 Clu
0
, ,
,
,
b~ih:u",v. vU l~wl~g
"
,
.
. .
.
.
,
.
"''''
.
,
-
,
-
W"~~,O
,
.
,
I Dore
,
,
.,
-' '
.
0
,
.
I
.
.
-. .
. -
.
.
"
,
.
..
,
.
,
Part of:
,
I
-
,
was black after,
holes had
- ,
-
"
. , NIPlatlnQ
,
-
Sl:~st
(shinY side) was.
o
0
,
,
.
, 0
,
.
the PI~:~~~
NI
. Most holes seem to have Dlated
, ,
,
.
'
--
,
78 79
.
.
. '
~
.,
§.
"
,
72 72 72 65
~
.
,
65 65 65 65 65
"N "'-
,
'
.
69 70 71 72 73
-s
~c:
-CD.!! c.IQ
.
61 61 65
67 68
:1.::::-
= 2-
E.2
~.<:I
E
en.E c0 .,_
.
Table A.3 Experimental conditions for pre-treatment and electroless nickel plating of 13.4 J.I1I1 nickel SPO screen.
...: GI
-
GI
"1ii
-
)(
a.
GI
W
C
34 36 37 54 57 80 81 82 83 84 85 86 87 88 89 90 91
09/02199 12102199 15/02199 25/03/99 31/03/99 12105/99 13/05/99 13/05/99 14/05/99 20/05/99 25/05/99 25/05/99 25/05/99 25/05/99 26/05/99 26/05/99 26/05/99
E 1ii
E1ii .;: E GI .. 0.0 )(-
GI
1ii
GI
GI·-
GI
E .;:
C
c Co
Q.
C
)(
-
en
1!!
a;
1!!
..
';' GI
w.:
CC
ll.
-
GI
GI
>-
Cl
c
:;::;
«I
ii:
:;::;
c
GI
Cl c ~E ::::J ::::J :; 00 ll.
-
~
-
80 80 80
Electroplating Electroplating
-
:
,
-
-
85 85 85 85 85 85
test ni bath life
_
c
, w
.
,
:
:
~
. - -
PT 4 PT 4 PT 10 PT 10 PT10 PT15 PT15 PT15 PT15 PT15 PT15 PT15 PT15 PT15 PT15 PT15 PT15
El El El,E2 E2 E2 El El El El El El El El El El El El
GlS
Glo
SGI 01ii :;::;E
GI Cl «I
..0 ..
::::J
0
::::JGI
Eti
Za;
>
0
mins
.. " ".....
'Olll
c
1!!
11»
ml ~
-.. -
E
Q.
mA
ti~
mm
V
15 15 60
230 230 230
1
0.69 0.65 0.64
IN
'
-
-,
.
.
-.
: ,
45
ml 1800 F 1800 1800
c GI E .c en '2
~
GI
E
C
:;::;
GI
"ia
Cl
GI
'61
0
-
Q.
cc
c ~
0 0
ci. E
GI
.c 1ii ..0
~
ii:
OHS OHS OHS
mins °C 45 79.0 60 80.3 60 79.2
%
nla
-
1
nla nla
-
::::J::::J
wC/) 00 11»
J!GI
nla nla 650 650 650
-
1800 1800 1800 1800 1800 1900 F 1900 1900 1900 1900 F 1900 1900
--
16.7 100
AS AS AS AS AS AS AS AS AS AS AS AS
,
-
30 60 60 50 60 60 60 60 60 60 60
79.0 79.6 80.2 79.7 78.5 79.5 79.9 80.0 79.7 79.9 79.9 80.0
Table A.3 (continued) Experimental conditions for pre-treatment and electroless nickel plating of 13.4 flII1 nickel SPO screen.
..:
I!!
a><
C
QI
QI
E
"0
'': QI
I:\.
><
W
92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110
-
III
QI
QI
1U
Q
26/05/99 26/05/99 21/06/99 21/06/99 23/06/99 23/06/99 23/06/99 23/06/99 28/06/99 28/06/99 28/06/99 28/06/99 28/06/99 28/06/99 30/06/99 30/06/99 30/06/99 30/06/99 30/06/99
1U a; a:
"Ea QI.-
><-
W.:
106 106 106 106
Cl
c
0j ii:
I:\.
,
,
,
No pre-treatment No pre-treatment No pre-treatment No pre-treatment No pre-treatment No pre-treatment No surface activate No surface activate No surface activate No surface activate No surface activate
PT15 PT15 PT15 PT15 PT15 PT15 PT15 PT15 none none none none none none no Hel, no no Hel, no no Hel, no no Hel, no no Hel, no
QI
I:\.
1;...
94
100 100 100 100 100
...>-
I!!
85 85
96 96 96
QI
QI
1U
E1U 1:\.0
'': E QI ...
C
cQI E
E2 E2 E2 E2 E2
aQl
E
...
;:;
;:;E cCl ::1::1
E1 E1 E1 E1 E1 E1 E1 E1 E1 E1 E1 E1 E1 E1 E1 E1 E1 E1 E1
cQI ·cQI 01U 2E ::1::1 Eo o~ Q. ::I QI QI 00 a: Za; .l!!QI Will VI>
... 0
00
i
ml
mins mA
::I ()
>
......
mm
V
,
..
.
,
.
;
,
'
,
, ,
. '
,
III
"'"0
QI 0 ..c ...
Cl la
I!! ...
m> ii:
Qla '5:3 "0.-
QI
C
E .c
,
rnl 1900 1900 350 350 650 650 650 650 650 650 650 650 650 650 650 650 650 650 650
QI
0
;:;
;:; la
Cl C
01i ii:
:l:: Cl
% ,
100 100
,
., 100 ,
.
.
E
C
AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS
r:i. E
....c QI
1U
..c
~
mins
·C
60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60
80.3 80.0 80.1 79.7 80.3 79.8 80.2 800 79.8 805 80.2 80.2 80.0 80.3 79.3 79.8 80.2 80.3 80.0
Table A.3 (continued) Experimental conditions for pre-treatment and electroless nickel plating of 13.4l1m nickel SPO screen .
... 4l
Q. >< GI
C
GI
E .;: GI
e. ><
W
111
GI
"1ii
c
a:
GI
1ii
Qj
01/07/99
112 01/07/99 111 113 01/07/99 111 114 01/07/99 111 115 01/07/99 111 116 07/07/99
,
117 07107/99 116
..
..
U/
"25 GI._ E1ii .;: E GI ..
e.o >
..
"';'
c
GI
GI
e.
>-
.. ..
Cl
GI
GI
no HCI, E2 no HCI, E2 no HCI, E2 no HCI, E2 no HCI, E2
,
0
~
::::I
E
~
Cl C
~ ii:
-.. .. C
GI
Polarity reversed
-U/
.. ".... e .- ~~ Gle.
OGl ".. GlO .c
GI
Cl
ca
Gl5
1U Eo u ca ::::IGI -GI zQj wu/ mm 2N 35
i
0
El, E2
ml mins mA 600 10 650
V 0.84
El,E2
600
10
650
0.84
2
40,35
650
El,E2
600
10
650
0.8
2
40,35
650
El,E2
600
10
650
0.78
2
30,35
650
El,E2
600
10
650
0.77
2
35,30
650
El, E3
600 1,10 650
2
35,40
El,E3
600 1,10 650
2
.
El, E3
600 1, 10 650
0.84, 0.83 0.86, 0.80 0.78,
2
40,35
ii:
a..
c
GI
en
::::I
(.)
,
118 07107/99 116
d.
C
c GI E 1ii
~
ad
E
GI
.c
c
U/
..
'cGI
0
1ii
Q.
~
GI
a:
E
~
..
E GI
.c
c
1ii .c
ii:
~
Cl
:;
ml mins ·C % 650 100 AS 60 80.0
AS
60
80.0
AS
60
80.0
AS
60
80.0
AS
60
80.0
650 100 AS
40
80.8
650
AS
40
80.0
650
AS
40
79.5
,
,
,
Table A.4 Observations made during pre-treatment and electroless nickel plating of 13.4 IJlIl nickel SPO screen. \
.,
...: l!! c
-" E
;:
Q. ><
-
.ll
c
34
09/02199
37
" C E.2
.Ec Ojo
.
u .. -" .!!.D
Wo
12102199 .
15/02199
54 57 80 81
82 83
0
-
e0
z
a..
l!! 0 a.
.!!
0
.;
~
~E
~E
0
.;
"0 Clu
z
a..
,
"
"
"
.
.
.
"
,
"
"
",
"
;
.
;
r
.
25/03/99
·" .
· .
"
"
12105199
Small bubbles on surface of sample after 8M Hel heated After 10 mlns contacting wIth hot Hel, aCId tumed pale qreen
13105/99
13105/99 14105/99
,
"
·
"
.
",
" "
. "
.
,"
.
.·
"
"
No bubbles InitIally on sample. Effervesced after " 15 mlns No sIgn of bubbles on sample after 30 s. Sample dId not effervesce
"
"
• 13.4
15/6
.
,
,
"
"
:
.
"
Fresher solutIon used from expenment NBW1176 "
"
;
3925
"
"
" .
ShadIng on sample surface was not uniform. Some dark and lighter areas Acetone sprayed over surface and contacted wIth sample
.
.
Effervesced ImmedIately on introductIon
" "
.
"
"
20/05/99
25/05/99
"
"
"
"
"
.
"
"
"
85
-
"j
. Qii
III
c
No bubbles observed
"
84
..
"-
-'"
.c
l!! a.
0
"
more actIvIty than for steel mesh
.,
.,
.c
.
~
!!
Surface covered WIth small bubbles In NaOH. Bubbles coated surface In Hel Bubbles on filter surface. Hel tumed green
"
en""''' -.,
[
"-
E
.,cc 0 CD.!!::: - a.'" 2 - c: u.>< .. .!~.D wco
c.~
~.D
"
36
"'.,
c .,
" "1ii" ~~ or" " Gi a: 0..0
" "'" a.
-
15
. "
"
, "
,
Table A.4 (continuetl) Observations made during pre-treatment and electroless nickel plating of 13.411ID nickel SPO screen.
..:
-" c:: E c
'"
Q,
.:l
ii
...
-
10
87
25105/99
.5c:: -0 .!2; a. to
mm en:;.2 .! "ii «i !!- ~
0.,
0.>< ..
wo
wc::o
a:
,
~.c
a.o
" ,
25105/99
"
89
26105199
90
26/05/99
91
26/05199
92 93 94 95
,
"
"
'
26/05199 , 26/05/99 21/06/99 21/06/99 ,
,
96 97 98
23106199 23/06/99 23/06199
.,c::c::
~~ !!~ -"'", '~" -" -" .!!!.!:!.c do" .!!.c
c
25/05199
m ..
liic:: E.2
lil
.
86
88
-.
m
i!!
"
00
§.
.,
§.
...'i s=
-
s=
i!! 0 a.
m
-
..
c:: .!!
0
~
CD ~
0
0 Do
0 Do
Z
.,
-.,
i!! 0 a.
-
_c::
to" ~E
0
:!! E "'0
0
Clo
Z
No bubbles seen before AS Introduced (30 s)
133
15/5
3860
15
No effervescence In first 40 s before air sparge No effervescence In first 40 s before air sparge No effervescence In first 40 s before air sparge Effervescence around sample after 10 s In bath Effervescence around sample after 10 s In bath Reaction almost observed Immedlatelv after addition Sample effervesced Instantlv No sign of bubbles before air spar(le Effervesced Immediately. Sample stili reacting after 60 mlns Effervesced Immediately Effervesced Immediately Effervesced Immedlatelv
128
15/5
3963
15
Electroless bath held at 80 ·C for 60 mlns before thiS expenment. Air sparge knocked shghUy at an angle. Sample moved back and to ,
146
4068
5/5
5 "
136
, "
3768
5
,
5
"
{'
r
;
'"
' ,
"
"
"
" "
,
;
5
"
"
393.4
5
"
,
,
" ,
613
"
"
"
"
f
,
,
Tall400ml glass beaker. Small slntered plastiC aerator "
56
10/6
3795
20 "
"
,
" , "
Tall 650ml !llass beaker. Retum to slntered (llass air spame "
" 0
.
.
Table A.4 (continued) Observations made during pre-treatment and electroless nickel platmg of 13.4 ~ nickel SPO screen.
-.,.,..
...:
-., ~
a.
c E
.,a. ;:
w
..
c
99 100
23/06/99 28106/99
101
28/06/99
102
28106/99
103
28106/99
104 105 106
CD
0;
28106/99
~"
,
,
a. ..
2 [:
.
,
108
30/06/99
,
109
30/06/99
.
~.
cn:;.2 .!! D. 1ii
2- [:
wo
wco
.,
Effervesced Immedlatelv No bubbles Inlllally. Effervesced lightly at end of exoenment No reactIon dunng fIrst mInute
"
In In
00
In In
~
,
In In ,
,
'. .
,
~
.
,
.
~
,
[
.,.
[ .:
0
c .!!
.
"'3: S£
~
a. 0
CD ~
.
0
a.
Z
.,
Sample was streaked (honzontal) after electroless solutIon ,p
0
.
.
,
Few bubbles seen after 60 s and at the end Many bubbles on surface after 60 s. Effervesced at the end Bubbles observed after 20 s and at end Effervesced dunng fIrst mInute. Many bubbles No sIgn of bubbles Inlllally. DIffIcult to observe at the end No sIgnIfIcant reactIon observed before aIr sparge Introduced No reactIon seen In fIrst 60s. No obVIOUS effervescence at end No effervescence In 60 s. Few bubbles wItnessed at end No sIgn of bubble In fIrst 60 s. Bubbles seen at end Surface effervesced VIgorously Instantly after addItIon and at the end
l'!E "0 Clu
ci
0
.
~E
0
~
Z
., ca;
~
Cl
ci
0
.,
,
,
~,
,>,
0
0,
;
~~
"
.
"
. ,
,
.
.
~
FIrst sIgn of "sIgnifIcant" plating
..
,
"
• 0
,,
,
Cl.,
.. Cc
u-'" ., u .. -" -"" .!!.a '!!.!:!.a
•
30/06/99
01/07/99
.!! ..
-0;
Bnefly dIpped buffer solutIon Bnefly dIpped buffer solutIon Bnefly dIpped buffer solutIon Briefly dIpped buffer solutIon Briefly dIpped buffer solution Bnefly dIpped buffer solution
107
111
-0
E.!!
~.a
30/06/99
30/06/99
Cl.,
!: c
"~ ~~ Qj ". a:
28106/99
110
E., ., c
..
o •
.
.
,
.
00
,
~
,
,
. 19
.
15/5
,
,
,
87
1515
3693
15
1.0
1515
595
15
-0
"
,
•
0'
.
-
"
0'
,
,
~,
Table A.4 (continuetl) Observations made during pre-treatment and electroless nickel plating of 13.4 Jlffi nickel SPO screen.
"' . "., .se "'''' ~~ -2-. . -" .b'" . '"
~
-. !
.. c
c E
112 113
E.2
-0 5; 0.", 2 ~
1;"
~.o
0.0
-.a 1110
-
f
f
en.5 c ",_ 0 cv.!!:
0. '"
~
.!!~Jl III co
" More vigorous than other
~
15 15
(most to date). Stili reacting at the
96 _15/5_ 3743 79 15/5 3983 3 5 15/5
at start
6.4
1515
polanty
7.1
15/5
polanty
I
I
~
20
lonthe IIIUIIOO
1515 10/5
"0
(!Iu
392.0
observed
v ",vu ••
28 69
1!~ .. E CE
'0 .; z
'0 .; z
u ... '"
~
8-
8-
polanty
end
117
u"u,,_o
118
u"u"_o
.......
~~
,
11
Table A.S Experimental conditions for pre-treatment and electroless nickel plating of 29.211m stainless steel mesh.
..
...: Cl
...
1/1
a. >< Cl
1: Cl
E .;: Cl a.
><
W
'tI
Cl Oi
Q
Cl Oi Gi
a:
" c Co CI._
EOi .;: E Cl ... a.o
>
..
C Cl E Oi Cl
Cl
£:
...
tII
C
";'
~ ii:
...Cl
D.
:::J
g C
0 :;:: :::J
'0 (J) ml
10 13 15 16 17 18 20 21 22 23 25 26 28 29 30 31 32
20/11/98 15/01/99 20/01/99 21/01/99 22101/99 25/01/99 28/01/99 28/01/99 29/01/99 01/02199 01/02199 02102199 03/02199 04/02199 04/02199 08/02199 08/02199
electroplatlnQ study electroplating study electroplatlnQ study electroplatlnQ study electroplating study electroplatlnQ study electroplatlnQ study electroplating study short Na2C03 contact no Na2C03 treatment quick dip in Na2C03 10 mlns in Na2C03 no Na2C03 dip In deionised water 15 mlns In Na2C03 buffer solution introduced no buffer solution
PT6 PT6 PT6 PT6 PT6 PT6 PT6 PT8
, "
El El,E2 El,E2 El,E2 El,E2 El,E2 El,E2 El,E2 El,E2 El,E2 El,E2 El,E2 El,E2 El,E2 El,E2 El E2 El,E2
600 600 600 600 600 600 600 600 600 600 600 600 600 600 600 600
-
Cl E
Cl E Cl E
:;:: tII
C
i
ii:
.. .. ... C
I!! :::J
(,)
mins mA 15 15 15 15 15 15 8 15 15 15 15 15 15 15 15 15
231 230 230 230 231 231 460 230 230 230 230 230 230 230 230 230
Cl
tII
III
'0
>
V
C Cl E
:::J
5 -1/1
'0
OCl > Cl 'tI .- ... 'tI C 0 Clo oOi ,g ... :;:: 'l:)~ E'l:) :::J
..c 1/1 '2
... ...
..!!!CI WI/I
:::JCI
zGi
mm
0.75 30 0.75 35 0.69 20 0.70 25 0.71 1 0.69 1 0.98 0.76 30 0.77 50 0.74 ," 0.72 ' " 0.73 45 0.67 0.70 0.68 0.73
Cl)
a.
'0 (J)
Cl)
a: ml
1 1 1
2N 1 1 1 1 1
",
' ,"
"' " ,
1 1
.. E Cl
C
:;::
~
C
,g
ii:
~
0
tII
:!:: tII
i
OHS OHS OHS OHS PHC30 OHS OHS PHC OHS OHS OHS OHS OHS OHS OHS PHC20 R OHS
mins 40 40 60 90 56 60 60 90 40 60 60 60 60 60 60 60 60
<
%
1800 0 1800 0 1800 F 100 1800 F 0 1800 F 0 0
1800
ci.
Cl E
..c Oi
°C 86.0 85.9 86.2 85.8 85.7 85.4 85.8 84.0 84.5 85.9 855 852 85.0 85.1 85.1 82.0 80.0
Table A.6 Observations made during pre-treatment and electroless nickel plating of 29.2 !1IIl stainless steel mesh .
-.,."
..:
-., ! c
E
Q.
""<>. !:l
-.
10
20/11/98
01
Q
."
~ a;
a:
m .. -=c
CUI ., C E.!!
'D.:::
~.Q
u" .,rn -.Q
e~
~~ 01 UI
-
,
15/01/99
,
No bubbles observed on surface
Steel mesh surface was covered in tmy bubbles. Cathode seemed to be pushed away from NI anode
Small bubbles were noted on the surtace.
"
,
15 16 17
20/01/99
'
,
22101199 25/01199
20
28101199 28/01199
White precipitate formed after Na2C03 White precIpitate formed after Na2C03 White precIpitate formed after Na2C03 "
' "
22 23
,
White precipitate formed after Na2C03
,
,
29/01199 01/02199
,
, ,
,
... ~
~
.
!0
Q.
0
.;
z
,
, "
, ,
,
,
,
,
130
150
UI
_c I! " "E C E " 0 C-'u
Steel mesh did not seem to react m NI plating bath. Microscope analysIs indicated unsuccessful DlatmD Overhead stirrer speed set low to prevent damage to stainless steel sample. Difficult to identify whether overhead stirrer proVided satisfactory solution aDltatlon
24 '
21/01/99
18
21
Whrte, cloudy precipitate observed when electroplated sample placed In Na2C03 solullon
00 ... :5-
Wco
"
13
01-
!!.!:!.c
, ,
"
-Ulu. 01;: ., o >
~::::-
NC -0 "N
Q."
u"" UI
Wo
<>'0
§.
§.::::-
S =~; e- r: -".,
100
-1ij
... .,-
m ..
en.5
, '
,
,
," ,
.
,
"
" ,
,
,
,
, ,
, ,
,
"
"
,
'
,
,
,
,
,
.,
"
.
4
Old not plate well
No activity observed Steel mesh curted towards anode (50 mm ~P' 20 mm bottom seoarallon Surface reaction was more VlOOroUS than usual
,
, ,
,
,
Lots of small bubbles observed m electroless solution around steel mesh
80
75
20
Many pores plated
"
,
,
"
Table A.6 (continued) Observations made during pre-treatment and electroless nickel plating of 29.2 flITI stainless steel mesh .
., -., 5ic
..:
-" " ~
ii
C
><
"
E
"C
;:
">< W
C
25
01/02199
26
02102199
0.
28 29
~
03/02199 04/02199
30
04/02199
31
08/02199
32
08/02199
~
'ii
lE:
"'., 0. ID
~'" 11.0
.
,
cC
White precipitate formed after Na,CO.
.
No sign of white In precIpitate water nnse
I;
,,(; (;
.!.!:!.c Wco
Wo
,
.0 ., N
U.l<
,!!'"
,
N C
-.,.,co
e~ -., u.,
-r" .,en
§.=-
w.E c Ult;.2 CD 'tU '0 0. ~ ~-
-0
.!;::
1ii~ :!!
-. ,,-
"'''
.!:c
E.2
.
, '0
: . Steel mesh cu~ed up (50 mm top, 20 mm bottom separallon)
::
•
,
...
.
,
.
Separallon between electrode and sample (40 mm too. 20 mm bottom)'
' c'
,
c
•
:
!~
11.:5-
o ~ 11.-
135
13.0
..
.
90 202
11 3 204
. .
•
,
§. If=-ID ., U
.
.
Steel mesh cu~ed
. '
.!!
- ." 0
0.
0
.;
_C
E
~
~E
z
"0 Clu
20
NI electroless plating considerably less effective Sample did not plate well. Bottom and Sides of plating bath coated In metal. Thin sheet of foil found In beaker
20 20
,
;
0
Solution full of bubbles.
,
ID
:!!
Solution full of bubbles. Flow reversed ea~y on and bubbles seen leaVing sample (thought to have come from tubing)
.
"
.
'
.
c
• •: '
.
~I
,0
Stirrer, bath Sides and bottom covered in metal. Sample did not appear damaged ApprOXimately 4 minutes to prepare sample In holder, In buffer solution. Pressure 7·15 mbar
.
. '0'
.
-
- - - - - -
--
----------
APPENDIX B. Electroless Ni Plating Procedure for Production of Metal Microfilters
APPENDIXB ELECTROLESS NICKEL PLATING PROCEDURE FOR THE PRODUCTION OF METAL MICROFILTERS
The following procedure outlining the pre-treatment and electroless nickel plating of metal screen material was developed in the Chemical Engineering Department, at Loughborough University. The pre-treatment conditions are more vigorous than those described in Chapter 3 (Electroless Nickel Plating Experimental), and additionally; the metal substrate is sonicated during electroless nickel deposition.
(i)
Degreasing and surface cleaning - use acetone to remove the traces of organic compounds adsorbed by the metal surface. The best and most effective results are experienced in an ultrasonic bath with 3 to 5 minutes contact time.
(ii)
Clean water rinse - rinse the sample twice in filtered water under sonication.
(iii)
Removal of organic contaminants - contact the sample with CitriC aCid (2 % by mass) in an ultrasonic bath for 5 minutes.
(iv)
Clean water rinse - rinse the sample twice in filtered water using sonication.
(v)
Alkaline degreasing - contact the sample with 5M sodium hydroxide solution for 10 to 15 minutes in an ultrasonic bath preheated to 50°C.
(vi)
Clean water rinse - rinse the sample three to four times with filtered water.
(vii)
Acid pickling - 8M HCI acid pickling of sample in preheated (50 GC) ultrasonic bath to remove exposed surface layers. The duration of the pickling stage is defined by several parameters, which are beyond the operator's control including: surface smoothness, sample thickness, state of nickel (annealed or not), quality of nickel, and so on. Consequently the picklmg bme for different screens can vary by 3 to 5 times and typically lies between 2 and 20 minutes. The pickling is considered successful when hydrogen bubbles are observed rising from all sites on the screen surface.
B-1
APPENDIX B. Electroless Ni Plating Procedure for Production of Metal Microfilters (viii) Alkaline wash - place the sample immediately in 2 % w/w sodium carbonate solution for 5 minutes with cleaning from bubbles by ultrasonic action. (ix)
Sample transfer - add the sample to a dummy nickel plating solution and leave for 2 to 4 minutes.
(x)
Electroless nickel plating - submerge the sample in the electroless nickel plating solution (located in an ultrasonic bath) for between 60 and 90 minutes depending on the desired pore size. The solution temperature should be controlled at 83 °C.
(xi)
Clean water rinse - finally, wash the sample in filtered water under sonicatIOn.
B-2
APPENDIX C. Radial and Tangential Shear Rate Analysis of Stirred-cell Operation
APPENDIXC RADIAL AND TANGENTIAL SHEAR RATE ANALYSIS OF STIRRED-CELL OPERATION Design of the stirrer operated in the stirred-cell was based on a cone-and-plate constant shear viscometer (rheometer). Also
known
as
the
' Weissenberg
rheogoniometer' , the cone-and-plate viscometer was developed and perfected in the 1950s and 1960s by Professor Karl Weissenberg, for measuring the viscosity of liquids (Wilkes, 1999). The principle of operation applied to the stirred-cell is illustrated in Figure C.I . An upper shallow-angled cone rotates at a constant angular rate
IV,
and in close contact with the lower stationary membrane surface. As the cone
rotates steadily the fluid beneath is sheared with every element describing a horizontal path, and causing liquid in the gap to move in concentric circles. The angle between the faces, Bradians, is typically (Holland, 1995) between 0.5 0 and 40 (0.0087 and 0.07 radians). In contrast to the original design there is flow normal to the cone passing through the lower (membrane) surface.
a
I.
cv
~
1
jj
CO
jj y
(
) Membrane
··················l ······· r·······l ················· 1 1
Figu re C.l Cone-and-plate viscometer theory applied to stirred-cell. C- I
APPENDIX C. Radial and Tangential Shear Rate Analysis of Stirred-cell Operation The tangential velocity component, v, varies linearly from zero at the lower membrane surface to the speed of the cone at the cone's surface. At a radial distance r,
the cone's tangential speed is OJ r and the distance between the faces is
tangential shear rate,
e r. The
r",.",;" ' is given by Equation C.l.
Ymngcnlial
av az
This equation shows that the shear rate
OJr
= er IS
=
OJ
e
(C. I)
constant throughout the fluid and is
independent of radial position. C.I SHEAR RATE ANALYSIS OF STIRRED-CELL Consider the operation of the stirred-cell shown in Figure C.l on the previous page. The stirrer cone rotates at high speed in very close proximity to the membrane, generating shear on the surface through its action. Unlike the cone-and-plate viscometer, a body of fluid travels slowly down the column and passes through the permeable lower surface. The stirrer operates at a rotational velocity, OJ, in radians s". A simple mathematical model has been proposed to investigate the significance of the radial shear component compared to tangential shear for stirred-cell operation. Incremental radial shear rates of fluid between the stirrer cone and the lower membrane surface were evaluated. Since the radial shear of fluid was not considered at extremities of the 'active' membrane surface, i.e. r > a, the physical boundary of the model was defined as r
= a. The fluid volume between the stirrer cone and lower
membrane surface was divided into a large number of three-dimensional concentric rings, radiating out from the cone tip, i.e. two concentric cylinders defined each ring. The height of each ring represented the separation distance between the cone and some datum level (the cone tip). The axial distance between the cone tip and membrane surface, s, was typically 0.001 m. Since a large number of concentric rings were specified the ring width in the radial direction - r was very small and could be described as 8 r . In this calculation, one thousand rings were evaluated, i.e. N = 1000. The radial shear rate analysis commenced at the outermost n'h ring, or n = 1000. A thin slice of the ring circumference (0.1 %) was studied during initial calculations.
C-2
APPENDIX C. Radial and Tangential Shear Rate Analysis of Stirred-cell Operation Figure C2 is a diagrammatic representation of ring n and the resultant thin ' analysis' strip with dimensions
or (width), Oh (height between cone and datum) and oz (chord
length). Note that dimensions in the Figure have been grossly exaggerated to aid the explanation.
y z
iLr .'
ring n-J
MEMBRANE
ring n
Figure C.2 Evaluation of radial shear rate.
The radial shear rate was calculated at the midpoint of the strip, 8h , from Equation C.2; a shear formula for flow in a rectangular cross section where v is the mean tangential velocity (Ruston et al., 1995):
6v
(C2)
8h
The radial flow of fl uid leaving the
nth
ring, and entering the adjoining (n-l )th ring,
was resolved from a balance of radial flow entering the
nth
ring and flow passing
through the membrane within that ring. Incremental radial shear rate acting in the range 0 < r < a was thus investigated using an iterative calculation process developed using an MS Excel™ spreadsheet. The tangential (rotational) shear rate was estimated to be 600.0 S·l , for a stirrer speed of 400 rpm by application of Equation CJ below.
C-3
APPENDIX C. Radial and Tangential Shear Rate Analysis of Stirred-cell Operation
liJ
rlang,,"..!
(C.3)
= B
In Figure C.3 the radial shear rate, i'md,,' , (S·I) IS reported as a function of radial distance, r, in the range 0 < r < a. It is seen that the radial shear rate calculated for each ring was generally very small compared to 600 s·1 (typical) tangential shear rate, that is with the notable exception for fluid in close proximity to the cone centre. This outcome suggests that the effect of fluid flowing in a radial direction between the cone and lower surface does not sigmficantly compromise the constant shear coneand-plate viscometer model when applied to the stirred-cell. A detailed worked example for calculation of radial shear is presented in the following Section.
10' ~
'(1) ~
.
0;
:;; ~ .......
-.
103
Cl)
f!
1"11
Cl)
.s:. (1)
~
1()2
"t:I 1"11
a:
101 0
5
10
15
20
Radial distance from stirrer cone centre (t), mm.
Figure C.3 Radial shear rate, i'mm..!' (S·I) calculated as a function of radial distance from centre of the stirrer cone (mm) using simple radial shear model.
C-4
APPENDIX C. Radial and Tangential Shear Rate Analysis of Stirred-cell Operation C.l.l Calculation of Radial Shear Rate
In the following worked example, the radial shear rate is calculated for the 500th ring using a simple shear rate analysis model introduced in the previous secUon. The fluid volume between the cone surface and an arbitrary datum was divided into N = 1000 concentric rings. The datum was specified at the same level as the stirrer cone tip. All relevant physical measurements and operating parameters are presented in Table C.I below. Calculated values are quoted to 2 decimal place accuracy unless stated otherwise. The notation used in the context of the calculations set out here is explained in Section C.2 Nomenclature. Table C.l Equipment specifications and operating conditions. SUrrer (cone) angle Stirrer (cone) radius Active membrane radIUS Axial distance between cone ti p and membrane Stirrer rotational speed Total system flow rate
The rotational stirrer velocity,
liJ,
(6) (a) (b) (s)
(!l)
Q
0.Q70 0.0185 0.0205 0.001
radians m m m
revs min400 7 3.33 x 10- m s-
was calculated, in radians S·l, using Equation CA,
liJ=
2n-Q
60
(CA)
where, n. is the stirrer rotaUonal speed measured in revolutions per nunute (rpm). At a typical stirrer operating speed of 400 rpm the corresponding value of liJ was calculated as 41.9 radians S-l.
liJ =
2xn-x400
60
=
.
41.9 radians S·l
Now, the cross-sectional area of the stirrer cone, As, was evaluated using Equation C.5. (C.5)
C-5
APPENDIX C. Radial and Tangential Shear Rate Analysis of Stirred-cell Operation A, =
Tt: X
0.0185 2
With the exception of high flow rates experIenced by the 10.0 IJll1 nickel SPO membrane. a representative flow rate of fluid through the membrane. Q. was 20 ml min· 1 (or 3.33 x 10.7 m3
S·I).
The velocity of fluid through membrane. vI> was
calculated as 3.08 x 10-4 m S·I from Equation C.6.
v, =A,Q-
(C.6)
The radial distance from the cone tip to the outside edge. ro. of ring n = 500 was determined by Equation C.7.
n r = a-
N
o
(C.7)
rO = 0.0185 x 500 = 9.25xl0·3 m (9.25 mm) 1000
In this case. the radial distance. ro. was equal to half the stirrer radius. Le. 0.00925 m.
SiIrularly the radial distance from the cone tip to the inside edge. r" of ring n = 500 was resolved using Equation C.8.
r,
r,
= 9.25xl0·3
-
=r
0
-Or
(C.8)
1.85xl0·5 = 9.23xlO·3 m (9.23 mm)
C-6
APPENDIX C. Radial and TangentIal Shear Rate Analysis of Stirred-cell Operation Equation C.9 provided the cross-sectional area, A" of ring n =500.
(C.9)
From Equation C.IO the ring height, ah, between the cone surface and datum level was specified. The height difference between inside, M" and outside edges, aho, of the nng was marginal as a consequence of the narrow radial width of each ring.
lih = r tan B
(C.1 0)
lih, = r, tan B
liho = ro tan B liho
= 9.25xlO·3
lih,
x tan (0.070)
liho = 6.47x10 4 m (0.647 mm)
= 9.23xlO·3 x
tan (0.070)
lih, = 6.45x104 m (0.645 mm)
A thin slice of the circumference (0.1 %) was consIdered in the calculations. The ' , chord length of the strip, &, was calculated USIng EquatIOn C.lI, where x is the fraction of the ring cIrcumference.
(C.l I)
&=21irx
&, = 21ir, x
&0 = 21iro x &0 = 2 x 1i
X
9.25xlO·3 x 0.001
&0 = 5.8IxlO·s m (0.0581 mm)
.<:. U4,
=2 x
1i
X
9.23xlO·3 x 0.001
&, = 5.80xlO·s m (0.0580 mm)
C-7
APPENDIX C. Radial and Tangential Shear Rate Analysis of Stirred-cell Operation The cross-sectional areas for outside and inside faces of the analysis strip (h - z plane) were investigated using Equation C.12.
A=Ohx&
A. Ao
= 6.47xlO"
x 5.81xlO·s
(C. 12)
= Oh,
x Oz,
A. = 6.45 x 10"
x 5.80xlO·s
The volumetric fluid flow into the strip, qm, under investigation (ring 500) is the same as flow out of the previous ring 501. In the case of n = N, ramal flow into the outermost ring can be derived from the total system flow multiplied by the fraction of the circumference. The volumetric flow leaving the analysis strip of ring 501, and hence entering strip 500, was calculated using the MS Excel™ Spreadsheet, as
qm = 8.35xlO·1I m 3 s·'
(5.01xlO·3 mlmin·'). The velocity of fluid,
mto the analysis strip of ring 500 was identified as 2.22 x 10-3
ID S-1
V m,
passing
using Equation
C.B.
(C.B)
vm
=
8.35 X 10-11 3.76 X 10-8
= 2.22 X 10-3
ID s-'
Equation C.14 was used to determine the flow lost through the membrane surface within the analysis strip, i e. 3.57 x 10- 13 m3 s·l.
(C.l4)
C-8
APPENDIX C. Radial and Tangential Shear Rate Analysis of Stirred-cell OperatIOn
qm = 1.16xlO-6
X
3.08xlO"" x 0.001
Likewise, the velocity of flow through the membrane,
Vm ,
was calculated using
Equation C.15,
(C.15)
where, Oi is the average chord length of the inside and outside edges of the analysis strip. This geometrical distance was evaluated using Equation C.16 below.
(C.16)
Introducing respective values for OZo and Oz, and solving for Oi, gives:
s
s
Oi = 5.8lxlO- + 5.80xlO- = 5.8lxI0-s m (0.058lmrn) 2
The velocity of flow through the membrane was 3.32 x 10-4 m
S·l.
3.57xlO·13
vm = -1.-85-:-x-IO~-5;--x-5-.8-l-x-"'1-0.7s
C-9
APPENDIX C. Radial and Tangential Shear Rate Analysis of Stirred-cell Operation The radial volumetric flow out of ring 500, qour. was calculated from the difference between the flow into the strip, q,n. and the flow through the membrane in that strip,
qm (Equation C.17).
(C. 17)
qo",
= 8.35xlO· 1l
-
3.57xlO·13
The velocity of flow leaving this analysis strip was realised from Equation C.18.
v
=
qout ~
out
vou ,
=
8.31 X 10.11 3.74xlO·8
The mid-point radial shear,
r
nm.J'
was determined from application of Equation C.2.
.
r rawal
Where,
&i
(C.lS)
6ii
=
oh
(C.2)
represents the mid-point strip height (Rushton et al., 1996). The mean
velocity, ii, was calculated according to Equation C.19.
_
v=
(v,.
+ v. u,) 2
(C. 19)
C-lO
APPENDIX C. Radial and Tangential Shear Rate Analysis of Stirred-cell Operation 3
3
v = (2.22xlO· + 2.22 X 10. ) = 2.22xlO·3
ms"
2
The strip height at the mid-poInt, defined as
oh,
was given by the following
expression,
oh = Oh. + Oh, 2
(C.20)
&i = 6.47xlO" + 6.4SxlO" 2
&i = 6.46xlO" m (0.6Smm)
Finally, the radial shear rate at the mid-point of ring n = SOO was estimated as 20.6 S·I.
y"""..
6 x 2.22 X 10.3 = 6.46xlO"
It.. = 20.62 s·'1 ID"
C.l.2 Calculation of Tangential (rotational) Shear The tangential velocity of fluid, Vo. was calculated as 0.39 m S·1 using Equation C.21,
(C.21)
9.2SxlO·3
Vo
= 41.9
Vo
= 0.39 m s·'
X
C-ll
APPENDIX C. Radial and Tangential Shear Rate Analysis of Stirred-cell Operation The tangential shear rate of fluid between the cone and lower surface,
t tongeonol
was
calculated from the ratio of angular velocity and stirrer cone angle (Equation C.3). For the ring under investigation, i.e. ring 500, the tangential shear rate was estimated to be 600.0
S·I;
a value which is approximately thirty times greater than the radial shear
rate.
r
tangentlaJ
.
=()
(C.3)
41.9
r "'gonnol = 0.070
It...
g",.oI
= 600.0 s·'1
C-12
APPENDIX C. Radial and Tangential Shear Rate Analysis of Stirred-cell Operation
C.2 NOMENCLATURE FOR SHEAR RATE ANALYSIS
2
A
Cross-sectional area of radial face (y - z plane)
m
Am
External membrane area
m2
Ar
Ring cross-sectional area (r - plane)
m
As
Area dIsplaced by stirrer (cone) motion
m2
a
Stirrer (cone) radius
m
b
Active membrane radius
m
h
Rmg height (distance between cone surface and datum).
m
2
(y - axis)
ii
Mean ring height (at mid-point)
N
Total number of rings
n
SpecIfic ring number
Q
Total feed suspension flow rate
m 3 S·I
q
RadIal volumetric flow
m 3 S·I
qm
Radial volumetnc flow through membrane (within nng)
m 3 SI
r
Radial dIstance from cone tip (r - axis)
m
s
Axial gap between stirrer (cone) tip and membrane
m
v
Radial flUId velocity
m S·I
vI
SuperficIal velocIty of fluid through membrane
m s"1
Vm
VelocIty of fluid through membrane (within ring)
m S·I
V
Mean radial fluid velocity (at midpoint)
m S·I
Vo
Tangential velocity
m S·I
x
Fraction of ring circumference considered
z
Analysis strip (tangential) chord length (z - axis)
m
Z
Mean chord length (at midpoint)
m
m
Greek symbols
RadIal shear rate
t
tangentJa]
Tangential shear rate C-I3
APPENDIX C. RadIal and Tangential Shear Rate Analysis of Stirred-cell Operation
(J
Stirrer (cone) angle
radians
n
Stirrer rotational speed
revs min· 1
Rotational (angular) velocIty of stirrer
radians S·I
Subscripts
i
Ring inner (radial) edge
o
Ring outer (radIal) edge
in
Radial transport into ring
out
Radial transport out of ring
C-14
APPENDIX D. Supporting Calculations for the Particle Deposition Model
APPENDIXD SUPPORTING CALCULATIONS FOR THE PARTICLE DEPOSITION MODEL Supporting calculations for the particle deposition model introduced in Chapter Six (Surface Microfiltration Theory) are presented here. Worked examples for all transport mechanisms acting towards and away from the membrane surface are set out for monosized spherical latex particles. The mean particle size of the latex challenge material was chosen as a sensible basis for model development. All experimental measurements including physical properties, and equipment specifications etc., used in the calculations are reported in Table 0.8 (see Section 0.8). Calculated values are quoted to one decImal place unless otherwise stated.
D.I SEDIMENTATION (vd An example calculation of gravitational settling velocity,
VG'
is presented for 5 IIl1l
diameter latex particles. The physical properties required for calculation of the sedimentation velocity towards the membrane surface are summarised m Table 0.1 below. Table D.I Sedimentation calculation data. (rp) 2.5 x 10-6 Particle radius Particle density (pp) 1050 Fluid density (Pr) 998 Acceleration due to gravity (g) 9.81 Fluid viscosity ()1) 0.001
m kgm· kgm· ms· . kgm· s
The gravitational settling velocity for spherical particles, vc, was calculated usmg Stokes' law:
(0.1)
0-1
APPENDIX D. Supporting Calculations for the Particle Deposition Model Substituting Table D.1 values Into Equation D.1 and solving gives a gravitational settling velocity towards the membrane, VG, of7.1 x 10.7 m S·I.
2 vG
=
X
(2.5xlO-6)2 x (1050-998) x 9.81 9 X (1 X 10.3)
An average penneate flow rate for constant rate filtration experiments was evaluated as 20 ml min· I , or J =3.3x10·7 m 3 s·'. The superficial penneation velocity, vf, was found to be 1.9 x 10"" m S·I.
3.3xlO·7 = A = 1.7xlO·3 J
vI
The average penneation velocity was therefore some three orders of magnitude greater than the settling velocity for a 5 Ilm particle. D.2 MASS TRANSFER COEFFICIENT The general mass transfer correlation for transport to a stationary membrane surface in a stirred-cell is given by:
Sh = A (Re)m (SC)033
(D.2)
There are many mass transfer coefficients available in literature for stirred-cell geometries. Smith et al. (1968) reported experimental measurements for the mass
D-2
-----------------
APPENDIX D. Supporting Calculations for the Particle Deposition Model transfer coefficient in the laminar boundary layer. The constant parameters A and exponent m were evaluated as A
=0.285 and m =0.55 respectively in this region.
Sh = 0.285 (Re)oss (SC)033
(D.3)
The mass transfer coefficient is applicable to a specIfic stirred-cell geometry and membrane permeability. Correction factors proposed by Smith et al. (1968) allow adjustment of the experimentally determined parameter A for the stirredcell/membrane system. The coefficient A depends on the stirrer and membrane geometries relative to one another and the membrane permeability described by the Sherwood number at the wall (Shw), i.e. EquatIOn DA.
(DA)
The correction factors are applied to the mass transfer correlation as multipliers of the coefficient A, and the complete mass transfer correlation for the laminar regime becomes:
Sh = a(b/a)/3(s/a)
(Sh w ) A (Re)m (SC)033
(D.5)
Where, a(b/ a)
Corrects for the ratio of (active) membrane radius to stirrer (cone) radius
/3 (s/a)
Corrects for the ratio of axial gap distance to (active) membrane radius Accounts for the relative membrane permeabIlity
The necessary stirrer and membrane geometries of the stirred-cell system required for use with the correction factors are presented in Table D.2. Values in square brackets refer to system dimensions reported in the original study (Smith et al., 1968).
0-3
APPENDIX D. Supporting Calculations for the Particle Deposition Model
Table D.2 Stirrer and membrane geometries for use with correction factors. Stirrer (cone) radius (a) 0.019 [0028] m Active membrane radius (b) 0.021 [0.032] m Axial gap between stirrer ti and membrane (s) 0.001 [0.0032] m
The correction factor a (bl a) is presented in Figure 9, in Smith et al. (1968) as a function of b I a . The graph is approximately linear between 1.0
< bI a < 1.3 giving
rise to Equation D.6.
a(~) = 2.449 - 1.30S(~)
1.0 < bla < 1.3
(D.6)
Introducing values for the stirrer (cone) radius and membrane (active) radius into Equation D.61eads to the following expression:
a(b) = 2.449 _ 1.30S(0.020S) 0.018S a
a ( ~ ) = 2.449 - 1.30S (1.1 08) = 1.00
The correction factor
a (bl a)
was identical to that used in the original experiments
by Smith et al. (1968). Similarly, the correction factor
f3 (s I a)
was approximated by
Equation D.7 between the limits O.OS < sla < 0.20. The relationship between
f3 (sI a)
and s/a is considered across a broader range in FIgure 10, in Smith et al.
(1968).
00S
D-4
APPENDIX D. Supporting CalculatJons for the Particle Deposition Model The ratio s/a was calculated from measured values in Table 0.2. The axial gap between stirrer tip and membrane surface, s, was consistently set at the distance recorded above.
fJ (.:..) = 1.092 _ 0.820 ( a
0001 ) 0.0185
fJ ( : ) = 1.092 - 0.820 (0.054) = 1.05
The ratio of axial gap to membrane (active) radius was approximately half that used in the original study, i.e. 0.054 in this study compared to 0.118 previously. The correction factorctl (Sh w ) for the membrane permeability can be neglected because, unlike the system studied by Smith et al., there
IS
no resistance on the downstream
side of the membrane to consider.
The coefficient A specified in EquatJon 0.3, was multiplIed by correction factors for cell geometry only.
A = 0.285 x 1.0 x 1.05 A
= 0.299
The corrected mass transfer coefficient (Equation 0.8) was appropriate for use with the stirred-celUmembrane system described in Chapter Seven (Surface Microfiltration Experimental).
Sh = 0299 (Re)o55 (SC)033
(0.8)
Dimensionless Sherwood (Sh), Reynolds (Re) and Schmidt (Sc) numbers are represented by,
0-5
APPENDIX D. Supportmg Calculations for the Particle Deposition Model
Sh=kb D
wb 2 Re=--
v
Sc = -
v
D
(D.9) to (D.ll)
Equation D.12 culminates from the substitution of Equations D.9 to D.ll into Equation D.S.
k;
=0.299
2)055 ( ; )033 w: (
(D. 12)
Addttionally this equation can be easily rearranged to describe the mass transfer coefficient, k.
k =0.299
( ~ )(
w:2)055 (; )033
(D.12a)
The rotational (angular) velocity of the stirrer is related to shear rate (at the membrane surface) by the relationship:
W
rw
= - 7r 45
(D.13)
Rewriting the mass transfer coefficient in terms of shear rate and simplifying leads to Equations D.14 and D.14a respectively.
k =0.299
_
. 2 )055 ( )033 ~ r~:v7r ( )(
k -0.069
;.
(Diw055b11V033J bv
055 D 033
(D. 14)
(D.l4a)
D-6
------
APPENDIX D. Supporting Calculations for the Particle DepositIOn Model And, ultimately the 'working' mass transfer coefficient, Equation D.15, is established.
_
k -0.069
(D067 twO" bOt) v
(D. 15)
022
The validity of this coefficient is restricted to the laminar regime, or more precisely Re < 32000. Reynolds number calculations outlined below indicate that for the stirred-cell system descnbed in Chapter Seven, Equation D.15 should only be employed below a shear rate of 1000 s·\. For a shear rate of 1000 s·t, the rotational (angular) velocity is:
co = 1000 x 1C = 69.8 rad s·\
45
The ratio of inertial forces to viscous forces (Reynolds number) is calculated from Equation D.I O.
Re = cob V
2
= 69.8
x (0.02IY = 30782
Ix 10.6
Back transport away from the membrane surface by the dIffusion mechanism was evaluated using Equation D.16,
J=V=kln(~:)
(0.16)
wluch sImplifies to the following equation when k is substituted from Equation 0.15.
v =0.069
D067 r ~22 bOt) In (C) --!!.. 055
(
v
Cb
(0.17)
D-7
-
APPENDIX D. Supporting Calculations for the Particle DepOSItion Model
D.3 PARTICLE DIFFUSION (Vd) An example calculation of the effective back transport velocity due to particle diffusion, Vd, is presented for 5 I1I11 diameter latex particles. A shear rate of 600 S·I at the wall IS representative of the average operational value. Physical properties and operational values reqUIred for calculation of the particle diffusion back-transport velocity are summarised in Table 0.3 below. Table D.3 Particle diffusion mechanism calculation data. Boltzmann constant (kB) 1.381 x 10.23 Average feed suspension temperature (T) 297.3 Auid viSCOSity (P) 1 x 10.3 (rp) 2.5 x 10.6 Particle radIUs Stirrer rotational speed (ll) 400 Membrane radius (b) 2.05 x 10.2 Kinematic vIscosity (v) 1.0 x 10.6 Concentration at membrane surface (Cw) 525 Concentration of particles in bulk (Cb) 2.0 x 10.2
JK' K kg m·' s·, m revs mm·' m m" s·, kgm· j kgm· j
Particle diffusivity, DB, was analysed using the Stokes-Einstein equation,
Shear rate at the membrane surface was calculated from the simple relatIOnship between surface shear and rotational stirrer speed, as discussed In Appendix C,
Yw
Y·w
= 1.50
(0.18)
= 1.5 x 400 = 600 S·I
0-8
APPENDIX D. Supporting Calculations for the Particle Deposition Model Effective back transport due to particle diffusion,
Vd,
was calculated according to the
following relationship:
(0.19)
Where the mass transfer coefficient, kd, is descnbed by Equation 0.20.
(0.20)
Substitution of Table 0.3 values mto Equation 0.20 gives the mass transfer coefficient, kd,
(S.7XIO·14 )067 X (600)°55 X (2.05 X10.2 )° 1 )
_
kd -0.069
X
6 022
(l.OxIO- )
Finally, the back transport related to particle diffusion,
Vd,
for this particular system is
6.0 x 10.7 m S·l.
Vd
= 5.9xlO
-8
X
In ( 5 2 52 ) 2.0xlO-
D-9
APPENDIX D. Supporting Calculations for the Particle Deposition Model
D.4 SHEAR-ENHANCED HYDRODYNAMIC DIFFUSION (vs) An example calculation of the effective back transport velocity due to shear-enhanced hydrodynamic diffusion,
Vs,
is presented for 5 J.Un diameter latex particles. Table D 4
below, lIsts the values of various physical properties and operational settings for use m calculation of the back transport velocity due to shear-enhanced diffusion.
Table D.4 Shear-enhanced hydrodynamic dIffusion mechanism calculatIOn data. Particle radius (rp) StIrrer rotational speed (fJ) Membrane radIUS (b) Kinematic viscosity (v) Concentration at membrane surface (Cw) Concentration of particles in bulk (Cb)
2.5 x 10-6 400 2.05 x 10.2 1.0 x 10.0 525 2.0 x 10-2
m revs min·' m rn's·' kgm- j kgm- j
Eckstein et al. (1977) estimated the effective shear-enhanced hydrodynamic dIffusivity as,
D, = 0.03r;
Yw
(D.21)
Shear-enhanced diffusiVIty, D" was calculated as a function of particle radius and shear rate usmg values reported in Table DA.
D,
= 0.03 X
(2.5xl0-6)2 x 600
Shear rate was estimated from a simple relationship with the stirrer speed (Equation
D.IS).
Yw = 1.5 x
400
= 600 S-I 0-10
APPENDIX D. Supporting Calculations for the Particle Deposition Model The effective back transport due to shear-enhanced diffusion,
Vs.
was calculated from
the simplified relationship:
W
=ks In(C C
Vs
)
(D.22)
b
Where the mass transfer coefficient, kg, is given by Equation D.23
_
ks - 0.069
(D,067 rw 055 bO!) v
022
(D.23)
For the conditions specified the mass transfer coefficient is 7.1 x 10-6 ms-I.
_ (1.1XlO·!0)067 x (600)°55 X (2.1XlO·2)0!) k, - 0.069 x 6 022 (1.0 X 10· )
ks = 7.1xlO-6 ms·!
The back transport velocity caused by shear-enhanced hydrodynamic diffusion across a concentration gradient,
Vs,
was determined for these specific expenmental
conditions as 7.3 x 10-5 m S·I .
Vs
= 7.1xlO •6 X In
(525) 2.0xlO
2
D-11
APPENDIX D. Supportmg Calculations for the Particle Deposition Model
D.S INERTIAL UFf (VI) An example calculation of the effective back transport velocity due to inertial lift,
VI,
is presented for S J.Un diameter latex particles. Selected system data are highhghted in Table D.5 below. Table D.S Inertial lift mechanism calculation data. Fluid density
kgm- j m revs min-' kg m·' s·'
The back transport velocity due to inertial lift, VI, was evaluated usmg Equation D.24,
(D.24)
Using Equation D.18, the shear rate at the membrane surface was resolved easily as,
i'w = 1.5 x
400
= 600 s·'
Once Equation D.24 was fully specified, the back transport due to inertial lift,
VI,
was
calculated as 2.0 x 10.7 m S·l.
IVI
= 2.0xlO·7
ms·'1
0-12
APPENDIX D. Supportmg Calculations for the Particle Deposition Model D.6 INTERACTION ENHANCED MIGRATION (VI) Electrostatic double layer (usually repulsive) and London-van der Waals (u sually attractive) interactions form the basis of the Dejagun-Landau-Verwey-Overbeek (DLVO) theory of colloidal stability. The interaction potentials depend
0n
the
distance between surfaces and electrostatic double layer mteractions and are strongly mfluenced by surface charge and ionic strength. van der Waals mteractions predominate at small and large inter-particle distances, whereas double layer repulsion dominates at intermediate distances. The summation of the mdl vidual interaction potentials between particles gives the total Interaction energy, wh ich if repulsive, can result In release of the fine particle from the surface.
D.6.1 Electrostatic Double Layer Repulsion (VDLR) An example calculation for 1 J.Lm diameter latex particles and particle separation distance, h = 1 nm is presented in this Section. Reerink and Overbeek reported an approximate expression for double layer repulsive energy, VDLR , as
VDU
= 641l"cw Tp (kB TJ (tanh zelf/o ze
4kB T
J
exp(-KD
h)
(D25)
Table D.6 Electrostatic double layer repulsion calculation data. Permittivity of free space (in a vacuum) (to) Relative permittivity of water (cr) (Dielectric constant) Particle radIUS (Tp) Boltzmann constant (kB) Average feed suspension temperature
885 x 80.5
10,12
5.0 x 10'7 1.381 x 10'23 297.3 I 1.60 x 10,19 -0.055 3.1 x 1022 1.0 x 10'9
Fm" dimenslOnles s m J K"' K dlmensionles s C V numberm,j m
0-13
APPENDIX D. Supporting Calculations for the Particle Deposition Model The individual tenns of Equation D.25 have been introduced elsewhere in Chapters Five and Six of this theSIS. Additional reference is made in the Nomenclature Section. Selected constants and measured values are hsted in Table D.6, on the previous page. The permittivity of the dispersing medium (in this case ultra pure water), e,., was calculated from the permittivity of free space and relative permittivity of water using Equation D.26,
(D.26) The permittivity of water was calculated In tenns of er and eo (values from Table D.6).
low
= 80.5 X 8.85xIO-'2
D.6.1.1 Molarity of Solution as if NaCI Electrolyte
The particle-particle interactions descnbed by the electrostatic double layer model occur in an ionic dispersing medium. It was desirable therefore to determine the equivalent concentration of the suspensIOn as If NaCI strong electrolyte rather than ultra pure water as used in experiments. The ionic (electrolytic) conductivity,
7(,
of a
0.01 % w/w latex feed suspension was measured using a Phihps conductivity meter as
Ir
= 6.59 f.l S cm-'
(at 21.5 0C).
[Note: The SI unit of conductivity is siemens per metre, S m·l • The older c.g.s metric units are ohm· 1 cm-I]
The molar conductivity, Am, of a solution is defined as,
2
Am (S cmmo
I·')
= 1000 x
Ir(Scm·') c(gmoldm·3 )
(D.27)
D-14
APPENDIX D. Supporting Calculations for the Particle Deposition Model This is a practical relationship for the most commonly available units, where Am is the molar conductivity expressed in S cm2 g mOri,
IC
is the conductivity measured in
S cm· l , and c represents the molarity of the added electrolyte with units of mol dm·).
[NOTE: Am is commonly referred to as the 'equivalent conductance' and expressed in the older c.g.s metric units ohm· 1 cm2 equiv·1• In SI units these become S cm2 mor 1 for a univalent ion.]
The molar conductivity is dependent on the concentration of electrolyte, and, in the case of strong electrolytes, decreases sbghtly as the concentration is mcreased. At dilute concentrations Kohlrausch showed experimentally that the molar conductivity of strong electrolytes varies lmearly with the square root of concentration, Atkins (1998), and Maron and Prutton (1959, pp.443-451):
A.. Where,
A:
IS
.
= A" -
B.Jc
(0.28)
the molar conductivity at infinite dtlution (ions can have no interactton
with one another), and B is an empirical coefficient.
Tabulated data from Table 5 "Equivalent conductances of electrolytes in aqueous solution at 25°C", in Maron and Prutton (1959, pp.443-451), is plotted in Figure 0.1, as Am versus.Jc for NaCI electrolyte. The molar conductiVity of NaCI was considered lmear up to .Jc = 0 032.
A best-fit line through these initial data points describes the relationship:
y = mx
..
A =
+c
-B.Jc + AO. 0-15
APPENDIX O. Supporting Calculations for the Particle Deposition Model
130,--------------------------------------,
--+-
Tabulated data Best fit line
120 U
IV
Z
11 ; -79.1 -le + 126.3
110
i:
<
100
90+-------.-------.-------.-------r-----~
0.0
0.1
0.2
Figure D.l Am vs
,Jc
0.3
0.4
0.5
for NaCl strong electrolyte,
Maron and Prutton (1959, pp.443-451).
As displayed in Figure 0.1 the equation of best fit is:
Am = - 79.1,Jc + 126.3
The molar conductivity at infinite dilution A"m, and constant B are 126.3 and -79.1 respectively.
Combining both Equations 0.27 and 0 .28 gives,
1000 K
c
= A"m _B ,Jc
(0.29)
0-16
APPENDIX D. Supporting Calculations for the Particle Deposition Model Equation D.29a is born from substituting measured values and those determined graphically into equation D.29 and then simplifying.
1000
6
X
6.59 x lO· = l26.3-79.1.../c c
6.59 x I0·3 = 126.3c-79.1c J5
(D.29a)
The Solver function in Microsoft ExceJTM was used to solve Equation D.29a for the molar concentration c. The target cell,
(l26.3xAl)-(79.l2 x (Al~0.5))
was set equal
to 6.59 x 10.3 by changing the contents of cell Al. The equivalent molar concentration c for the aqueous latex system was 5.240 x 10-5 g mol dm-]
c
= 5.240xlO·'
gmoldm·3
c
= 5.240x 10"
x 1000
= 0.052 g mol m·3
The Particle number concentration, no, is simply the product of molar concentration and Avogadro's number of the number of atoms comprising one mole of substance, l.e:
(D.30)
Where, NA = 6.02xlO·23 g mor l
no
= 0.052
g mol m·3 x 6.02x10 23 g mor l
D-17
APPENDIX D. Supporting Calculations for the Particle Deposition Model The inverse Debye length,
KD '
was calculated from the well-known relationship,
(D.31)
Replacing parameters in this Equation for the appropriate values leads to the following expression for
= (2
K D
KD'
19 X (1.60 x I0· )' x 3.1 x lO" x (1),)0.5 7.12 x lO·1O X 1.381 x lO·21 x 297.3
For ease of calculation, the tanh (a) function was determined independently of the fmal calculation.
= 647r E r III
( :--------:)' ( ) k. T
P
Z
'
e
tanh: ze lfo: exp(-K h) : 4 k8 T 1 D ~-
_____ _ I
a
In this case,
a
__ _z_e-'.I-f,o,-
4kB T
=
I x 1.60 x 10. 19 X - 0.055 4 X 1.381 x lO·'3 x 297.3
= -0.54
The tanh (a) function is described by,
tanh(a)
=
sinh(a) cosh(a)
0-18
APPENDIX D. Supporting Calculations for the PartIcle Deposition Model Where, sinh(a) = (ea _e-IX)/2
= (eO S4 +e-O s4 )/2
sinh(a) = (eO S4 _e-O S4 )/2
cosh(a)
sinh(a) = - 0.56
cosh(a) = US
The tanh (a) function was therefore evaluated as
tanh(a) = sinh(a) = _-0_.5_6 cosh(a) US
tanh(a) = - 0.49
The electrostatic double layer repulsion, particles of radius O.S
VDIR = 64 x
7r:
~,
VDLR,
was estimated to be 1.1 x 10- 17 J for
at 1.0 nm separation distance, h, usmg Equation D.2S.
x (7.12xlO-'o ) x (S.OX10·7 ) x
1.381XlO-23 x 297.3 ( 1 x 1.60xlO-'9
)2
x (-0.49)' exp(-2.3x10 7 x 1.0xlO-9 )
D.6.2 London-van der Waals Attraction (VLVA) An example calculation of London-van der WaaIs attractive interaction energy,
is presented for 1.0
~
VLVA,
diameter latex particles at a separation distance of 1.0 nm.
VLVA is a function of the Hamaker constant Am, particle radius rp, and separation
distance between particles h, as shown in Equation D.32.
D-19
APPENDIX D. Supporting Calculations for the Particle Deposition Model
v LVA
Where, H
Am [2(I+H) +In(~)~ H (2+H) 2+H U
=_
6
(D.32)
= hI rp
D.6.2.1 Hamaker Constant The effective Hamaker constant between two identical particles (I) in a dispersing medium (3) is given by:
(D.33)
Vlsser (1972) reviewed many different Hamaker constants including those of water An and polystyrene All. The underlined values presented in Tables 1 and 2 of Visser
(1972) were considered the most accurate, i.e.
An
= 6.6 X 10.20
J
The effective Hamaker constant, Am, was determined from a completely specified Equation D.33, such that
The ratio of separation distance to particle radius, H, was calculated by
0-20
APPENDIX D. Supporting Calculations for the Particle Deposition Model
9 H = IOxlO· = 0.002 5.0xlO·7
Evaluation of these individual tenns enabled the calculatIon of interactIon energy due to London-van der WaaIs attraction,
21
= _2.27xlO· 6
Iv
LVA
= - 1.9xl0·'9
[
VLVA,
as -1.9 x 10. 19 J.
2 x (1 + 0.002) I ( 0.002 )~ 0.002 x (2 + 0.002) + n 2 + 0.002 U
JI
_
D.6.3 Total Interaction Energy (Vrotal) The total interaction energy, Vrotal, between charged particles is realIsed by summation of the mdlvidual repulSIve and attractive interaction energIes.
(D.34)
Acid-Base (VAB) and Born repUlsive (VBR) interaction energy were considered negligible. The total interaction potential describes the variation with the distance between particles. The total interaction energy, Vrotal, for I !lrn spherical latex particles was found to be approximately J.J x 10- 17 J.
Vro,al = 1.1xl0-17
Iv:
Total
= J.JxlO- 17
-
1.9xlO-19
JI
.
0-21
APPENDIX D. Supportmg Calculations for the Particle Deposition Model An expression for dimensionless total interaction energy (Equation D.35) is commonly used in literature when descnbing the potential energy curve.
Dimensionless total interaction energy =
(D.35)
VTo,o{
kB T
D.7 TOTAL BACK TRANSPORT (Vtotal) Back-transport velocities described by particle diffusion, shear-enhanced diffusion, and inertial lift mechanisms are presented in Table D.7 for the stirred-cell system and 5 llII1 (mean diameter) latex particles.
Table D.7 Particle transport velocIties for 5 llII1 diameter particles. Brownian diffusion (Vd) 6.0 x 10-7 m sShear-enhanced dIffusIOn (vs) 7.3 x 10-5 m sInertial lift (VI) 2.0 x 10-7 m s-
The total back transport velocity away from the membrane surface,
Vtotab
was
calculated from the summation of the individual mechanisms. Where the, back transport velocity is larger than the forward-acting permeation velocity then the particle will not deposit. The effect of particle-surface interactions on the total velocity opposing permeate flux was evaluated as far as possIble, and ultlmately onutted from the total back transport term. The reader is referred to Section 6.16 (Interaction enhanced migration) on page 6-10 for a full and considered account of this decision. Equation D.36 descnbes the total transport velocity m a direction normal to the membrane surface for individual mechanistic terms of particle diffusion (Vd), shear-enhanced dIffusion (vs) and inertial lift (VI) terms:
(D.36)
D-22
APPENDIX D_ Supportmg Calculations for the Particle Deposition Model The total back transport velocity,
V'o'al,
was evaluated from the individual velocity
terms for 5!11Il particles (summarised in Table D_7) as 7.3 x 10-5 m S-l_
V,o,a'
= 6.0xlO-7
+ 7.3xlO-s + 2.0xlO-7
0-23
APPENDIX D. Supporting CalculatIOns for the Particle Deposition Model
D.S EXPERIMENTAL DATA Table D.8 is a comprehensive account of all generic experimental data used in the supporting calculations including equipment specifications and operatIonal settings, measured physical properties, and universal constants etc., and selected calculated parameters.
Table D.8 Experimental data and calculated parameters. Units
Parameter All Particle-partIcle interaction constant All DIspersion medium-dispersion medium interaction constant Am Effective Hamaker constant between identical particles and dispersing medium a Stirrer (cone) radIUs b Active membrane radIUs Cb Concentration of particles in bulk suspension Cw ConcentratIOn of partIcles at membrane surface c Molar concentratIOn of electrolyte d, Column internal dIameter dm Membrane dIameter dp Particle dIameter e Elementary charge of electron Acceleration due to gravIty g H Ratio of separation distance to partIcle radius h Separation dIstance between partIcles k8 Boltzmann constant m Stirrer (cone) heIght NA Avogadro's constant no Particle number concentration Q Feed suspension volume to overflow level Membrane (full) radIUS rm Particle radius rp Axial gap between stirrer (cone) tip and s membrane Absolute temperature T w Stiffer shaft diameter y Column heIght to overflow z Charge number of electrolyte Yw Shear rate at membrane surface e VOId fractIOn (porosity)
6.60 x 10.20 4.38 x 10.20
dimensionless dimensionless
2.27 x 10.21
dimensionless ,
0.0185 0.0205 0.02 525 5.240 x lO.s 0040 0047 5.0 x lOo{) 1.60 x 10.19 9.81 0.002 1 x 10.9 1.381 x 10.23 1.29 x 10.3 6.02 x 1023 3.13 x 1022 1.95 x 10-4 0.0235 2.5 x 10-6 1.0 x 10.3
m m , kg m·> kg m·' gmol r' m m m C m s·· dlmenslOnless m J K' m gmor' m·' m>
297.3 0021 0.219 1 600 0.5
K m m dimensionless s·' dlmensionless
m m m
0-24
APPENDIX D. Supporting Calculations for the Particle Deposition Model
Table D.S (continuetl) Experimental data and calculated parameters. Parameter
er
Permittivity of water Eo Permittivity of free space (in a vacuum) (J Stirrer (cone) angle 1( Conductivity of solution (at 21.5 0c) tl FlUid viscosity v Kmematic viscosity Pr Fluid density J!p_ Particle denSity ( Zeta potential of particle Stirrer rotational speed t;.
n
Units
+-:::R_e_Ia_tl_v_e-,:p_erm_I:-·tt_iv_i~tY,-of_w_at_e_r..:.(d_i_el_e_ctn_·c_c_o_n_st_an---'.t)+-8_0_.5_ _--;-;;-+::d:-lm--,.en_s_io_n_l_es_s-J
7.12 x 10. 10 8.85 x 10. 12 0.07 (4) 6.59 x 104 1.0 x 10.3 1.0 x 10.6 998 1050 -0.055
400
Fm' Fm' rads (degrees) S m' kg m' s· m skg mkg mV revs min-
0-25
APPENDIX E. PDESOL Model for Interaction Enhanced Migration
APPENDIXE PDESOLTM MODEL FOR INTERACTION ENHANCED MIGRATION A complete text version of the Pdesol™ (v2.0) model for interaction-enhanced migration applied to the cross-flow microfiltration of multi-dispersed iron oxide particles described by Y oon et al. (1999) is presented here. The Pdesol™ solutIOn output Table E.l for this model is reported on page E-4.
E.1 TEXT VERSION OF PDESOLTM MODEL 'Notes on use: '************************************************************************
To solve the equatIon the lower boundary condition IS required. 'We do know that VB=O when the repulsion energy (or force) 'is equal to the attractIOn energy: VA=VR. So, one method of 'solutIOn IS to solve from the lower boundary value of t=O '(assuming that VB=O at t=O - whIch is wrong), and to check the 'answer for where Vtot=O. Then It is possible to use the 'value of t (height) as the correct lower boundary value 'for the integration and VB=O which IS now correct. Iteration 'could be used to detennine the value of t (at VB=O) more precisely. 'However, the overall mfluence on the value of effective back 'transport velocity of this more rigorous procedure, compared 'to simply taking VB=O at t=O, is marginal. '***********************************************************************
'CONSTANTS 'Pennittivity of water (F mA-I): etaw=6.96E-1O 'Particle diameter (m) varDiameter= lE-7 rp=varDiameter/2 'Boltzman constant (J KA-l): kB=1.38IE-23 Temperature (K): T=298.15 'Charge number of electrolyte z:l
E-l
APPENDIX E. PDESOL Model for Interaction Enhanced Migration 'Charge of electron (C): e=1.60E-19 'Zeta potential of particle (V): phl=O.05 'Molecular density (molecules m"-3) n=602E+23 'Hamaker constant (J): A=3.4E-20 'A=IE-19 'Hydraulic mean diameter (m): dh=5.36E-03 'viscosity (kg m"-I s"-I): mu=O.OOI 'fluid density (kg m"-3): rho=IOOO 'mean velocity (m s"-I): Umean=O.24 'membrane length (m):
L--o.ll 'CONSTANTS CALCULATED kappa=(2*e"2*n*z"2)/(etaw*kB*T) kappa=kappa"O 5 'Diffusivity (m"2 s"-I) Db=(kB*T)/(6*3.I42*mu*rp) 'ReynoJds number Re=rho*Umean*dhl(mu) 'Schmidt number Sc=(mulrho)/(Db) 'Boundary layer thickness (m): delta=(dhl( 1.62*(Re*Sc*(dhIL»"O.33»
'VARIABLES WIm HEIGHT 'ELECTROSTATIC DOUBLE LAYER REPULSIVE INTERACTION ENERGY Vdlr=64*3.142*etaw*rp*«kB*T/(z*e»"2)*«tanh(z*e*ph1l(4*kB*T»)"2)*exp(-kappa*t) 'LONDON-VAN DER WAAL ATTRACTION ENERGY Vlva=-Al6*(rp/(t>lE-20)+rp/(t+2*rp)+log«t>1E-20)/(t+2*rp»)
E-2
APPENDIX E. PDESOL Model for Interaction Enhanced Migration 'NET INTERACTION VtotaI=Vdlr+Vlva 'WHAT DO WE CALL THIS? - THE ONLY DIFFERENTIAL EQUATION TO SOLVE VB_t=(exp(Vtot/(kB*T))-l) 'VB_t=(exp«(Vtot)/(kB*T))-l) 'Lower boundary condition 'VALID ONLY WHEN VR=VA BUT APPEARS TO MAKE LITTLE DIFFERENCE TO RESULT VB@tO=O 'EFFECTIVE BACK TRANSPORT VELOCITY vl=Db/delta*log«(VB> le-20)/delta) 'vi=Db/delta*log«(VB)/delta)
B-3
APPENDIX E. PDESOL Model for Interaction Enhanced Migration "
Table E.! Pdesol™ output table for interaction enhanced migration model.
E-4
APPENDIX E. PDESOL Model for Interaction Enhanced Migration Table E.! (cont••• ) Pdesol™ output table for interaction enhanced migration model.
t _(nm) 5.1 5.2 53 5.4 5.5 5.6 5.7 5.8 5.9 6.0 6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9 7.0 7.1 7.2 7.3 7.4 7.5 76 7.7 7.8 79 80 81 82 8.3 8.4 8.5 8.7 8.8 89 90 91 9.2 9.3 94 9.5 96 97 98 99 10.0
Yjt
~)
~j}'
V,,-t
-4.11 E-20 -4.01E-20 544E-19 5.38E-19 5.32E-19 527E-19 522E-19 5.16E-19 511E-19 506E-19 500E-19 495E-19 4.90E-19 4.85E-19 4.80E-19 475E-19 4.70E-19 4.65E-19 4.60E-19 4.56E-19 4.51E-19 4.46E-19 4.42E-19 4.37E-19 4.33E-19 4.28E-19 4.24E-19 4.19E-19 4.15E-19 4.11E-19 4.07E-19 4.02E-19 3.98E-19 3.94E-19 3.90E-19 3.82E-19 3.78E-19 3.74E-19 3.70E-19 367E-19 3.63E-19 359E-19 355E-19 3.52E-19 348E-19 3.44E-19 3.41 E-19 3.37E-19 3.34E-19
5.14E-19 5.09E-19 -392E-20 -383E-20 -375E-20 -3.66E-20 -3.58E-20 -351 E-20 -3.43E-20 -3.36E-20 -329E-20 -3.23E-20 -3.16E-20 -3.10E-20 -304E-20 -298E-20 -2.93E-20 -2.87E-20 -2.82E-20 -2.77E-20 -2.72E-20 -2.67E-20 -2.62E-20 -2.58E-20 -2.53E-20 -249E-20 -245E-20 -241 E-20 -237E-20 -233E-20 -2.29E-20 -2.26E-20 -2.22E-20 -2.19E-20 -2.15E-20 -209E-20 -206E-20 -202E-20 -1 99E-20 -1.97E-20 -1.94E-20 -1.91E-20 -1.88E-20 -1.86E-20 -1.83E-20 -1.80E-20 -1.78E-20 -1.76E-20 -1.73E-20
1.60E+54 5.04E+53 5.04E-19 5.00E-19 4.95E-19 4.90E-19 4.86E-19 481E-19 4.76E-19 4.72E-19 4.67E-19 463E-19 458E-19 4.54E-19 4.50E-19 4.45E-19 4.41E-19 4.37E-19 4.32E-19 4.28E-19 4.24E-19 420E-19 416E-19 411E-19 407E-19 403E-19 3.99E-19 3.95E-19 3.91 E-19 3.88E-19 3.84E-19 3.80E-19 3.76E-19 3.72E-19 3.69E-19 3.61 E-19 3.58E-19 3.54E-19 350E-19 347E-19 343E-19 340E-19 3.37E-19 333E-19 3.30E-19 326E-19 323E-19 320E-19 3.17E-19
4.44E-05 4.44E-05 1.59E+53 5.07E+52 1.62E+52 5.22E+51 1.69E+51 5.51E+50 1.81E+50 5.98E+49 1.99E+49 6.68E+48 2.26E+48 7.70E+47 2.64E+47 9.14E+46 3.19E+46 1.12E+46 3.97E+45 1.42E+45 5.11 E+44 1.85E+44 6.79E+43 2.51E+43 9.33E+42 3.50E+42 1.33E+42 5.06E+41 1.95E+41 7.55E+40 2.95E+40 1.17E+40 4.64E+39 1.86E+39 7.51E+38 1.26E+38 5.21E+37 2.18E+37 9.16E+36 3.89E+36 1.66E+36 7.16E+35 3.11E+35 1.36E+35 6.01E+34 267E+34 120E+34 540E+33 246E+33
(~i 555E-19 549E-19 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 511E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 511E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11 E+58 5.11E+58 5.11E+58 5.11 E+58 5.11 E+58 5.11 E+58 5.11 E+58 5.11 E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58 5.11E+58
V,
(ms·\ 5.11E+58 5.11E+58 4.44E-05 444E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 4.44E-05 444E-05 4.44E-05 4.44E-05 4.44E-05 444E-05 444E-05 4.44E-05 444E-05
E-5
APPENDIX F. Stirrer Calibration
APPENDIXF STIRRER CALIBRATION The rotational speed of the plastic stirrer rod, 12, was measured using a Radio Spares TM-3011 Hand Tachometer. Black tape was wrapped around the stirrer motor chuck and thin vertical strips of reflective tape (supplied) were fixed to the chuck and the stirrer rod respectively. Normal operation of the stirrer was maintained during the calibration procedure, with the stirrer located in the acrylic column filled to the mark with ultra pure water. The stirrer was operated at each set speed (gears I and II) for five minutes before any measurements. Rotational speeds of the stirrer motor chuck and stirrer rod were measured using the hand tachometer. A distance at least 50 mm was maintained between the reflective tape and the photoelectric probe. The presence of the clear acrylic column wall between the reflective strip and probe did not seem to affect the measurements. Ten tachometer measurements were recorded for each stirrer speed.
1000 , -- ---------- ---------- ---------- ---------, --- Linear regression "C QI QI
800
Measured speed (rpm) ; 0.99 x test speed (rpm)
•
E
Q.Q.
Ul ...
Cii.....::. ......
-
600
... QI
.-
QI Ul
E "Co l!!.c ::J
Ul
400
u
ns
ns-
QI-
::i!:
200
O~------~-------.------_.--------~----~
o
200
400
600
800
1000
Test stirrer speed (LeO display), rpm.
Figure F.l Heidolph RZR 2102 electronic stirrer calibration. F- l
APPENDIX F. Stirrer Calibration Care was taken to shade the tachometer from external light sources in the laboratory, i.e. sunlight and fluorescent light strips. Figure FI is a calibration graph for the Heidolph stirrer motor. The measured rotational speed of the stirrer motor was very similar to the set speed indicated on the LED display and all tachometer measurements were only accurate to
± 0.01
%. No 'slip ' was experienced between the stirrer motor and the stirrer rod.
F-2
APPENDIX G. Pressure Transducer Calibration
APPENDIXG PRESSURE TRANSDUCER CALIBRATION The Radio Spares pressure transducer (l bar g, 0 - 100 m V) was calibrated using a Druck DPI 603 portable pressure calibrator. A satisfactory pressure hold test confinned that the pressure transducer was securely attached to the calibrator. The transducer was connected to a Pico Technology 16 bit analogue-digital converter and PC. A test pressure was set in the range 0 - 1 bar g. Once the system had stabilised, the transducer output voltage was logged every five seconds for one minute, and the average voltage output (mV) was recorded. The calibrator was purged and the. pressure reset between test pressures. Figure G.l is the calibration graph.
120
:>
E oi
- - Linear regression
-Cl CO
"0
100
Pressure (mbar) = 9.4 x (voltage (mV) + 0.85)
80
>
::I Co ::I
60
0 l-
40
"t:I Q)
20
::I tIl CO
0
ll.
... Q)
::i!: -20 0.0
0.2
0.4
0.6
0.8
1.0
Test pressure, bar.
Figure G.t Calibration graph for Radio Spares pressure transducer (PT).
Although the pressure transducer was tested across the entire pressure range, particular attention was paid to the normal operating envelope. A pressure parameter
G- l
APPENDIX G. Pressure Transducer Calibration was created in the PicoLog for windows software (release 5.02.5) from the Linear Regression Equation G.l.
Pressure (mbar)
= 9.402 (voltage (mV)
+ 0.85)
(G.!)
The pressure transducer was calibrated regularly to maintain accurate pressure readings.
G- 2
APPENDIX H. Pore Size Distributions of Nickel SPO and CPO Membranes
APPENDlXH PORE SIZE DISTRIBUTIONS OF NICKEL MEMBRANES
The pore sizes of nickel ci rcu lar pore and slotted pore membranes were analysed using several different techniques, which are described in Section 3.5.2. Figures H.I to H.4 are the pore size distributions of 10.0 Ilm, 4.9 Ilm, and 4.1 Ilm nicke l SPO membranes, and 4.8 Ilm nickel CPO membrane determined by Image Anal ysis of high-resolution electron micrographs. Gaussian di stribution of the pore sizes is al so indicated.
H- I
APPENDIX H. Pore Size Distributions of Nickel SPO and CPO Membranes
10
=
10.0 ~m nickel SPO membrane
- - Gaussian distribution
8
...
Q)
Ir
I 1\
6
JJ
E
:l
Z
4
j
2
0
6
8
10
12
14
Pore width, !lm. Figure H.1 Pore size distribution of 10.0 !lm nickel SPO membrane.
10.-------- - - - - - - - - - - - - -- - - -- - - -- - - - - -- ,
=
4.9 ~m nickel SPO membr~e
- - Gaussian distribution
8
... Cl)
6
JJ
E
~1(
:l
Z
4
1\
~~
1/
2
r-
1\
_/ f / 2
3
4
5
~ 6
7
8
Pore width, !lm Figure H.2 Pore size distribution of 4.9 !lm nickel SPO membrane.
H- 2
- --
------------------------------------------------------------------------------APPENDIX H. Pore Size Distributions of Nickel SPO and
epo Membranes
15,------ -- -- -- -- -- -- -- -- --------------,
=
4.1 ~m nickel
spa membrane r-
- - - Gaussian distribution
('
...
10
Q)
.Q
E ::s
Z
5
Pore width , Ilm.
Figure H.3 Pore size distribution of 4.1 Ilm nickel SPO membrane.
20
=
4.8 ~m nickel
epa membrane
- - - Gaussian distribution
15
... Q)
.Q
E ::s
10
Z
1\
5
0 0
2
d
I~ 4
6
8
10
Pore width , Ilm.
Figure H.4 Pore size distribution of 4.8 Ilm nickel
epo membrane. H-3
APPENDIX 1. Constant-rate Filtration: Experimental Details and Results
APPENDIX I CONSTANT·RATE FILTRATION: EXPERIMENTAL DETAILS AND RESULTS Table 1.1 Clean water backflush (BF) and membrane resistance tests (MRT) prIor to filtration experiments with ID 0 Ilm nickel SPO membrane.
-
c
c
0 ;:
Il)
E
a. .;:
.;: Il)
a.. ><-
CD
1ii
wl!!
c
311Ulx1 31julx2 31julx3 31julx4 311Ulx6 31julx7 01augx1 01augx2 01augx4 01augx5 01augx7 01augx8 02augx1 02augx2 02augx3 02auQx3 06augx1 06augx2 06auQx4 06augx5 07augx1 07augx2 07auQx4 07augx5 08augx1 08auQx2 08augx4 08augx5 09augx1 09augx2 11augx1 11augx2 11augx4
31/07/01 31/07/01 31/07/01 31/07/01 31/07/01 31/07/01 01/08/01 01/08/01 01/08/01 01/08/01 01/08/01 01/08/01 02108/01 02108/01 02108/01 02108/01 06/08/01 06/08/01 06/08/01 06/08/01 07/08/01 07/08/01 07/08/01 07/08/01 08/08/01 08/08/01 08/08/01 08/08/01 09/08/01 09/08/01 11/08/01 11/08/01 11/08/01
u
III Il)
c BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF
Cl
.5~ _Ill c.s
-
(I).!.
a::~
0.
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2
Il)
...... as
"tU
Il)
c
~III
Il)
Il)
~
.co;"
~
E
a. >a.
o.ll)
ECI ~ c
E ~
PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC
o.f! (3) (3) 5 5 3 (3) 3 3 3 3 3 3 (3) 3 5 5 (3 3 3 3 3 3 3 3 (3) (3) (3) (3) (3) (3) 3 3 3
60 05-80 40 05 -80 60 05 -80 60 05 -80 60 05 -80 60 05 -80 60 05 -80 60 05 -80 60 05 -80 60 05-80 60 05 -80 60 05-80 60 05 -80 60 05 -80 60 05 -80 60 05 -80 60
><
~~
;::~
... .c ~':'
o
E
...I:::'
>< ~~
;:"i
... .c Il),:,
~E
::I:::.
518
861.8
186.6
24036
52.1
870.0
53.4
865.5
50.7
865.5
52.7
873.6
53.4
882.3 "
152.7
2383.6
52.7
860.5 1
54.8
874.1
55.5
876.4
53.4
8736
51.4
8636
52.7
865.5
54.6
863.6
53.5
860
1-1
Table 1.1 (continued) Clean water backflush and membrane resistance tests (MRT) pnor to filtration expenments wIth 10.0!lm nickel SPO membrane.
-
c
c CD E .;:
CD CL.
)(-
0 :;:: CL
.;: CD
"la
0 III
CD
wl!!
c
c
11auQx5 12augx1 12augx2 12auQx4 12auQx5 17augx2 17augx3 17auQx4 17augx5 18augx1 18augx2 20auQx3 20augx4 21augx1 21augx2 21auQx4 21augx5 22augx1 22auQx2 22augx4 22augx5 24augx1 24auQx2 24augx4 24augx5 25augx1 27auQx1 27augx2 27augx4 27auQx5 27augx7 27augx8 28auQx1 28augx2 25augx2
11/08/01 12108/01 12108/01 12108/01 12108/01 17/08/01 17/08/01 17/08/01 17/08/01 18/08/01 18/08/01 20/08/01 20/08/01 21/08/01 21/08/01 21/08/01 21/08/01 22108/01 22108/01 22108/01 22108/01 24/08/01 24/08/01 24/08/01 24/08/01 25/08/01 27/08/01 27/08/01 27/08/01 27/08/01 27/08/01 27/08/01 28/08/01 28/08/01 25/08/02
MRT
SF MRT
SF MRT
SF MRT
SF MRT
SF MRT
SF MRT
SF MRT
SF MRT
SF MRT
SF MRT
SF MRT
SF MRT
SF SF MRT
SF MRT
SF MRT
SF MRT MRT
-...... CD
01 01
Cl
CD
.5_ -Ill
en.!!!.
c.: :l E a:~
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
3 2 3 2 3 2 3 2 3 5 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 2 3 2 3 2 3 2 3 3
CD.co;"
c
-
=
:l
c..
c..~
PLC (3) PLC (3) PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 5 PLC 5 PLC (5) PLC 5 PHC 6 PHC 6 PHC (6) PHC 6 PHC 6 PHC 6 PHC 6 PHC 6 PHC 6 PHC(6 PHC 6 PHC 6 PHC 6 PHC 6 PHC 6 PHC 6 PHC 6 PHC 5 PLC 5 PLC 5 PLC 5 PHC 6 PHC 6 PHC (6)
05 -80 60 05 -80 60 05 -80 60 05 -80 40 05 -80 40 05-80 15 05-80 15 05 -80 15 05 -80 15 05 -80 15 05 -80 15 05 -80 15 05 -80 15 15 05 -80 40 05 -80 40 05 -80 15 05 -80 05 -80
CD CL
=-CL
E
CD III CL CD
ECI
:l C
_.
)(
)(
:l_
:l_
;:~
-~
... .c
... .c
~'l' o E
CD'l'
8:E :J:::.
..J:::'
54.1
876.4
54.3
861.4
534
8636
530
870.0
141.1
2396.8
149.1
2356.4
374.3
106236
367.5
109146
3668
113400
3268
11130.0
4146
11252.7
.
330.7
10846.4
5175
108382
10870.9
403.6 ~
1498
2370.7
1552
24150
410.0 521.6
108546 11150.0
Table 1.2 Clean water backflush and membrane resistance tests (MRT) prior to filtration experiments with 4.9 Ilm nickel SPO membrane.
-
c
cCl) E .;:: Cl)
c...
x-
0
;: c.. .;:: Cl)
"tU
u III CI)
wl!!
c
c
23aprx3 23aprx4 24aprx1 26aprx2 03mayr1 03mayr2 04mayx1 04mayx2 04mayx3 18iunx1 18Junx2 19junx1 19junx2 191unx4 19Junx5 20lunx1 201unx2 20Junx4 20junx5 21iunx1 21iunx2 21junx5 21junx6 221unx1 22Junx2 26junx1 26junx2 261unx4 26junx5 27junx1 271unx2 27Junx4 27junx5 021Ulx1 02Julx2 02julx4 021Ulx5 03Julx1 03julx2 03julx4
23/04/01 23/04/01 24/04/01 26/04/01 03/05/01 03/05/01 04/05/01 04/05/01 04/05/01 18/06/01 18/06/01 19/06/01 19/06/01 19/06/01 19/06/01 20/06/01 20/06/01 20/06/01 20/06/01 21/06/01 21/06/01 21/06/01 21/06/01 22106/01 22106/01 26/06/01 26/06/01 26/06/01 26/06/01 27/06/01 27/06/01 27/06/01 27/06/01 02107/01 02107/01 02107/01 02107/01 03/07/01 03/07/01 03/07/01
MRT MRT MRT MRT
BF MRT MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF
Cl
._- Cl)
Cl)
"tU ...
c..
Cl)
...ca
>-
E
_Ill
en.!!!..
c.!: :l E a:~
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
3 3 3 3 3 3 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2
Cl) -
.t:'
c..
c
c..CI)
x _ --. --.
:::I C
o E
= Cl)
III
x
:l_
:l_
... .t:
... .t:
CI)'l'
E
ECI
a.
a.E
...I:::;'
::J:::;.
PLC 3 PLC 5 PLC 3 PLC (3) PLC (3) PLC (3) PLC (3) PLC 3 PLC 3 PLC 3 PLC (3) PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC (3) PLC 3 PLC 3 PLC 3 PLC (3) PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC (3) PLG(3t PLC (3) PLC (3) PLC (3)
05-80 05-80 05-80 05-80 60-80 05-80 05-80 60-80 05 -80 60 05-80 60 05-80 60 05-80 60 05-80 60 05-80 60 05-80 60 05-80 60 05-80 60 05-80 60 05-80 60 05-80 60 05-80 60 05-80 60 05 -80 60 05 -80 60
54.8 136.6 53.9 53.0
867.7 2225.5 871.8 856.4
54.3 53.8
857.3 856.4
50.7
856.4
51.9
862.7
53.1
863.2
52.8
864.1
55.5
870.5
53.6
848.2
53.0
860.5
49.1
790.5
54.2
854.6
51.0
850.0
:l
;'l'
' ,
g;E
,
51.5
854.6
53.2
861.4
.0
.~
51.3 -'-
-,
-
860.0 ~,
,
53.8
864.6
52.5
854.1
53.1
861.2
Table 1.2 (continuet!) Clean water backflush and membrane resistance tests (MRT) prior to filtration experiments with 4.9 J.lI11 nickel SPO membrane.
..
c
e
CD
:ll!!
:l_ :l_ -; = ---. .. -. ECI
..
0
.5_ .. 0
.c' (/).!!!.
a:~
:lE
c.
a..~
..J::::'
:::I::::.
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
3 2 2 3 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2
PLC (3) PLC (3) PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC (3) PLC 3 PLC 3 PLC 3 PLC (3) PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC (3) PLC 3 PLC 3 PLC 3 PLC (3) PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3) PLC (3) PLC (3) PLC (3) PLC (3) PLC(3) PLC (3) PLC 3 PLC 3 PLC 3
05 -80 53.0 60 60 05 -80 49.8 05-80 49.8 60 05 -80 53.9 60 05 -80 51.8 60 05 -80 52.7 60 05 -80 53.2 60 05 -80 50.5 60 05 -80 53.9 60 05 -80 50.7 60 05 -80 54.3 60 05 -80 50.7 60 . 05 -80 53.2 , 60 05 -80 53.0 60 05 -80 54.8 60 05 -80 53.0 60 05 -80 54.3 60 05-80 54.3 60 05 -80 52.1 60 05 -80 52.7 60
856.4
CD
:;::
-;
CD
1\1
1\1
..
e
....
0
CD
E "i:
Cl
c. .;:
c.. xw l!!
c
c
031ulx5 04julx1 04julx1 041UIx2 041ulx2 06julx1 06julx2 061Ulx4 06julx5 06julx7 06julx8 071Ulxl 07julx2 09julxl 09julx2 09iulx4 09julx5 091Ulx7 091Ulx8 10julxl 10julx2 10julx4 10lUlx5 11julxl 11julx2 14julx1 14julx2 14julx4 14julx5 14iulx7 14julx8 16julxl 16iulx2 16julx4 16julx5 16iulx7 16julx8 17julx1 171UIx2 17julx4
03/07101 04/07101 04/07101 04/07101 04/07101 06/07101 06/07101 06/07101 06/07101 06/07101 06/07101 07107101 07107101 09/07101 09/07101 09/07101 09/07101 09/07101 09/07101 10107101 10107101 10107101 10107101 11/07101 11/07101 14/07101 14/07101 14/07101 14/07101 14/07101 14/07101 16/07101 16/07101 16/07101 16/07101 16/07101 16/07101 17107101 17107101 17107101
MRT
0 CD
BF BF MRT MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF
CD-
CD
c.!:
CD C.
>C.
:lE
CD 0 C.CD
:l
e
X
X
.. .c
.. .c
~~E
CD~
g;E
CD
c.
ECDO
I-~
i
:
855 855
26.0 26.0
866.4
25.5
868.2
26.0
875.9
26.5
866.4
25.0
865.9
.
,
872.3
23.5
642.3
,
870.9 "
872.3 865.0 "
858.6 878.6
22.5
869.6
22.0
870.9
21.5
855.9
22.0
859.1
22.0
868.6
21.5
Table 1.2 (continued) Clean water backflush and membrane resistance tests (MRT) prior to filtration experiments with 4.9 Ilm nickel SPO membrane.
..
eI
C
C
0
CD
~
E .;:
CL .;:
Cl)
()
CL.
)C-
Cl)
«i
III
CD
w!!!
c
c
171Ulx5 271ulx3 27Julx4 230ctxl 230ctx2 230ctx4 230ctx5 240ctxl 24octx2 24octx4 240ctxS 250ctxl 250ctx2 15novxl 15novx2
17107101 27107101 27107101 23/10101 23110/01 23/10101 23/10101 24/10101 24/10101 24/10101 24/10101 25110101 25/10/01 15/11/01 15/11/01
MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
CD
«i ...
._CD
E
...ca
.. Ill
CD-
.c':'
CC
en'!!'
='E 11: ......
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
3 2 3 2 3 2 3 2 3 2 3 2 3 2 3
C
>-
..
E CD
CL
E =' n.
CL CD
Eel =,C
CD CL
PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC
III
(3) (3) 3 3 3 3) 3 3 3 (3) 3) 3) 3) (3) (3)
n.E 05-80 60 05 -80 60 OS -80 60 05 -80 60 05- 80 60 05 -80 60 05- 80 60 05 - 80
--. --_='-. )C
)C
... .c
... .c
..... :::.
;:):::.
53.6
872.3
52.3
864.1
SO.9
85S.S
49.1
863.2
49.1
865.0
50.0
861.8
S3.2
864.1
50.7
854.6
='-
CDN :;:. o E
CD,:,
~E
Table 1.3 Clean water backflush and membrane resIstance tests (MRT) prior to filtration experiments with 4.1
...c
c
.;:
a. 'l:
1U
Cl
.c' IJ)!!!..
0
C1> E
C1> Q,.
-
i
...as C1>
>C-
we!
c
19novx1 19novx2 19novx3 19novx4 19novx6 19novx7 20novx1 20novx2 20novx4 20novx5 21novx1 21novx2 21novx4 21novx5 22novx1 22novx2 22novx4 22novx5 22novx7 22novx8 23novx1 23novx2 23novx4 23novx5 25novx1 25novx2 25novx4 25novx5 27novx1 27novx2
19/11/01 19/11/01 19/11/01 19/11/01 19/11/01 19/11/01 20/11/01 20/11/01 20/11/01 20/11/01 21/11/01 21/11/01 21/11/01 21/11/01 22111/01 22111/01 22111/01 22111/01 22111/01 22111/01 23/11/01 23/11/01 23/11/01 23/11/01 25/11/01 25/11/01 25/11/01 25/11/01 27/11/01 27/11/01
., 0
c BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
Ilm nickel SPO membrane. Cl
.... C1>
C1>
as C1>~ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
C1> Q,
>-
~.,
Q, E
Q,C1> ECI
c.
c.l!
PLG(3) PLC (3) PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC (3) PLC 3 PLC 3 PLC 3 PLC (3) PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC (3) PLC 3 PLC 3 PLC 3 PLC (3) PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3
60 05-80 60 05 -80 60 05 -80 60 05 -80 60 05 -80 60 05 -80 60 05 -80 60 05-80 60 05 -80 60 05 -80 60 05 -80 60 05 -80 60 05 -80 60 05 -80 60 05 -80
., ...
.5~ ... c .5 ~ E
a:~
2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3
c
~
~
-,-- ---, >C
>C
.. .c
.. .c
~~
c
~~
o
~'l' E ...1::::-
C1>N Q,' Q,E ::1::::-
53.9
864.6
53.4
865.9 ,
53.0
872.3
55.7
870.5
51.8
854.1
50.9
871.4 ' ,
49.6
860.5
50.5
862.3
50.9
865.9
52.7
868.2
, 51.8
864.6
532
866.4
550
874.1
53.2
872.3
52.7
872.3
-----
-----------
Table 1.4 Clean water backflush and membrane resistance tests (MRT) prior to filtration experiments with 5.0 Ilm PC isopore CPO membrane.
..
Cl
c
c III E .;:
:;::: Co .;:
III
III
U
....
>ewe
III
.co;"
Q
0
Co.
07junx1 071unx3 11Junx1 11junx3 11junx5 121unx1 12junx4 12junx6 131unx1 131unx3 14junx1 14junx3 15junx1 15iunx2 15junx4 151unx7 181Ulx1 18julx3 20julx1 20iulx3 24aprx1 24julx1 24julx3 24iulx5 25julx1 25julx3 25julx5 26iulx1 26julx3 27junx7 271ulx1 31mayx1 31mayx3
1;
I/)
1;
ra III
E~ :;:::1/) ~
BF BF BF BF BF BF BF BF BF BF BF BF BF BF BF BF BF BF BF BF MRT
BF BF BF BF BF BF BF BF BF BF BF BF
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
c
..
=
E
E
a.
a.
III Co
III
(/).!!!.
Q
07/06/01 07/06/01 11/06/01 11/06/01 11/06/01 12106/01 12106/01 12106/01 13/06/01 13/06/01 14/06/01 14/06/01 15/06/01 15/06/01 15/06/01 15/06/01 18/07/01 18/07/01 20/07/01 20/07/01 24/04/01 24/07/01 24/07/01 24/07/01 25/07101 25/07/01 25/07101 26/07101 26/07101 27106/01 27/07101 31/05/01 31/05/01
III
c.!: :::lE a:~ 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 2 2 2 2 2 2 2 2 2 2 2 2
III
>-
I/)
Co
Co
:::I
PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC
:::I
3) 3) 3) 3 3 3 (3 3 3 3 3 3 3 3 3 3 3 (3) 3) 3) 3) 3 3 3 3) 3 3 3 (3) 3) 3) 3) (3)
60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 05-80 60 60 60 60 60 60 60 60 60 60 60 60
Table 1.5 Clean water backflush and membrane resistance tests (MRT) prior to filtration experiments with 4.8 Ilm nickel CPO membrane.
C III
r:t..
><-
"0 III III
r:t. .;:
...
III
III ::E .- r:t.
0 ;:
E .;: III
C
III
'la
wE!
c
11octx1 11octx2 11octx3 11octx3 11octx5 11octx6 11octx7 12octx1 12octx2 12octx3 150ctx1 150ctx2 150ctx3 150ctx5 150ctx6 16octx1 16octx2 16octx4 16octx5 16octx7 16octx8 17octx1 17octx2 18octx1 18octx2 18octx4 18octx5 18octx7 18octx8 190ctx1 190ctx2 190ctx4 190ctx5
11/10101 11/10101 11/10101 11/10101 11/10101 11/10101 11/10101 11/10101 11/10101 11/10101 15/10101 15/10101 15/10101 15/10101 15/10101 16/10101 16/10101 16/10101 16/10101 16/10101 16/10101 17/10101 17/10101 18/10101 18/10101 18/10101 18/10101 18/10101 18/10101 19/10101 19/10101 19/10101 19/10101
...1/1
c BF MRT
BF MRT
BF MRT MRT
BF MRT MRT
BF MRT MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
BF MRT
r:t.
1/1
Cl C
~1/1
III
US.=.
E ;;0 c.S :::l E 11:_
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2 3 2 3 2 3 3 2 3 3 2 3 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3
r:t.
:::l_
... ... .c
r:t.1II
-~
~':'
Ill,:,
E
Eel
D.
D.~
PLC (5) PLC (5) PLC 3 PLC 3 PLC 3 PLC (3) PLC (1.6) PLC (3) PLC (3) PLC (1.6) PLC (3) PLC (3) PLC (1.6) PLC 1.6 PLC 1.6 PLC 1.6 PLC 1.6 PLC 1.6 PLC 1.6 PLC (1.6) PLC 1.6 PLC 1.6 PLC 1.6 PLC (1.6) PLC 1.6 PLC 1.6 PLC 1.6 PLC 1.6 PLC 1.6 PLC 1.6 PLC 1.6 PLC (1.6) PLC (1.6)
' 60 05 -80 147.8 60 540 05 - 80 60 05 -80 52.7 6.1, 02 - 80 ,', . ' 60 05 - 80 869.6 05 - 80 16.0 , 60 05 -80 51.4 05 -80 16.1 80 05 -80 160 , ' 80 05 -80 164 ,: ',.' 80 05 -80 15.7 , , 80 05 -80 15.7 I~~';~~ ~ 80 05 -80 15.9 80 ' 05 -80 16.1 " ",.'; 80 05 -80 150 80 -' 05 -80 14.6 ,,' . ~ 80 05 -80 15.2 o' . " 80 05 -80 14.8
:::l
>< _ . ... .c
><
:::l_
;
§:E
o E
:::l C
::J:::::-
....1:::::-
,
~
2348.5
..
,
.. ,
.
'
.'
,
.
,
873.6 ..,
"
870.5 256.8 ,
,
'
"
,
264.1 c" •
869.1 266.4 ,
,
266.8
'. '. : ' 268.1 .
'.
,~'"
.
M';;" ,
f!
2636 .
"
259.6 ~,'
.
'"~,
263.2
.
"
2623 ,
"":,
'
2600 .':" "
."
260.9
.,
,
257.3 ,. ,
256.4
Table 1.6 CntIcal flux evaluated for 10 0 ).lm nickel SPO membrane across a range of shear rates (experimental parameters: 0 - 1200 S-1 surface shear, 0020 kg m-3 latex feed suspension)_
Si
~
'0
c
Cl) Cl)
~
)(
W
.. i..
..
Cl)
Cl) "C
1::: ,. Q
0
1'0
I?,
7
I
27augx9
27/'08/'01
11
I
?"~lInY~
JI
I
?n~lInY"
?nmR/D1
1
I 21augx3 21/'08/'01 I 27augx3 '01
2 9
5 8
6
?R.'n~
'01
<:"IUO" JI
12
4
CI)~
C
t: a. E .-
US.:. '0 67 10'0 133 20'0 267 40'0 533 60'0 667 733 80'0
'OE
t: .-
I I I I I Ii I! I
'0 1'0'0.5 15'0 199.5 30'0 4'0'05 60'0 7995 90'0 100'05 1099.5 12'00
2.1575 21565 2.1586 2.1565 21546 21504 2.1551 21585
2~
2.1507
Cl
~o
~0
E :::I
E :::I
a.
C.: :::I E
Cl) Cl
en
nla
.§~ ... 0
8.r
Cl)
Cl
C
c.i
o
~
127augx6
I ?R~lInY~
C
a.
.~
!!!
Cl
tfc.
0:: .....
'0 '020 (D) '0 '020 Dl '0 '02'0 D) '0 '02'0
11-2
» »
'0.'02'0 '0 '02'0 '0 '02'0 '0 '02'0 '0 '02'0 D '0'02'0 D '0.'02'0 '002'0 :D
!!!
C
:::I
~Cl)
a.
a.
a.. a.. 30 32 913.0 95Q.5 9317
1'0_'0 1'0_1 1'0.7 10.'0 1'00 1'0 '0 10.5 10'0
595 9 9282
118_2 318-9
1'086_1 26'08 118'0.2 2112.3 3124.1 3279 6 3990'0
16'0_2 390.3 5782 323.9 1'07.'0 250.1 31'09
PL(; PHCD8 PLC 85 PHC (6) 23 PHC (6) 1'0 PHC (6) 20 PHC (6) 25 PHC (6) 32 PHC (6) 3D
10 '0 10.1
4821-8 6148 6
286 8 364.1
PHC PHC
38 46
1'0 9'0
928,2
1247_1 255'0_'0 12687.7 25 '>00"-;; 1-32982 14 118'02 1758.4
E~ Cl) 0 I-~
24])
', 24.'0 3145.726.0 1469.3 26'0 22 2487.3 -V60() 2623 6 24.5 26 3225'0 133382 3281.6 33 4167.3 r42800 42236' ,230 32 399'0 '0 1430'09 41455 225 4'0 48
5'052 3
5395 5
1'087_6
5223 9
6'i486 65127 633D.7
23 5 23 5
Table 1.7 CritIcal flux evaluated for 4.9 f.1m nickel SPO membrane across a range of shear rates
(experimental parameters: 0 - 1400 S·1 surface shear; 0.020 kg m·3 latex feed suspension).
S
.
...: CD
-
"tI CD CD
C
CD
.;:E
-..,
"tI
C
0
07julx3 061ulx9 02julx3 02julx6 06julx6 03julx3 091Ulx3 11julx3 031Ulx6 10julx6 091Ulx6 04julx3 091Ulx9 17julx6 141Ulx6 16julx3 061Ulx3 10julx3
07/07/01 06/07/01 02107/01 02107/01 06/07/01 03/07/01 09/07/01 11/07/01 03/07/01 10/07/01 09/07/01 04/07/01 09/07/01 17/07/01 14/07/01 16/07/01 06/07/01 10/07/01
9 8 1 2 7 3 10 15 4 14 11 5 12 22 17 19 6 13
CD
..,
Cl
c
~ . ... .... . ... en.;. ~
a. >C w
CD
"tI CD "tI "tI
CD
CD
a. (/)
(/)
CD-
CD
cE .a. 0 67 100 100 133 200 267 267 400 400 533 600 667 667 733 733 800 800
1ij
co
CD .co;"
nla I I I
)(
CD
S (/) (/)
en.!!!.
co ::!is
0 100.5 150 150 199.5 300 400.5 400.5 600 600 799.5 900 1000.5 10005 1099.5 1099.5 1200 1200
2.1605 2.1476 2.1483 21612 2.1588 2.1518 2.1556 2.1546 2.1600 2.1541 2.1598 2.1616 2.1597 2.1588 2.1457 2.1455 2.1531 2.1590
NOF refers to no obvIOUS sign of fouling
- ~
CD
N
::::.
'iij
u c
0_
U"I
"tiE
CD Cl
d!c 0.020 (0 0.020 0 0020 0 0020 0 0020 0 0020 0 0020 0 0.020 0 0.020 0 0020 0 0.020 (D) 0020 (D) 0020 (D) 0.020 (D) 0.020 (D) 0.020 (D) 0020 (D) 0020 (D)
-
Cl
a. CD
CD
)( ;,~
.5_ ;::::c -(/)
c.: ;, E
~~E
(/))(0;"
;,.c
;;:::~
:> E
CD
c
a. >a.
=
;,
;,
E
c:~
c_
-~
et::::.
a.
108 10.5 208 10.6 104 11.0 10.5 107 10.5 12.1 106 11 0 104 126 9.7 126 104 106
21.7 219 160.2 21.00 625 852 173.2 170.9 169.6 240.9 281.4 2805 3036 351.4 400.9 3580 3321 3252
21.1 210 nla 21.7 217 303 18.6 209 327 19.1 215 21.1 208 29.6 15.7 136 27.7 22.6
PLC (3 PLC (3 PLC (3 PLC (3 PLC (3 PLC (3 PLC 3 PLC 3 PLC 3 PLC 3 PLC (3 PLC 3 PLC (3) PLC (3) PLC (3) PLC (3) PLC (3) PLC (3)
CD
(/)
a.
E c..
06 08 15 08 10 14 25 18 24 22 32 30 36 46 36 32 34 NOF
.c
Cl
:c
~
)(
Cl
c
~
(/)
a.
E c.. ;,
08 10 10 12 16 26 20 28 24 34 32 38 48 38 34 36
--. )(
;,-
.. .c
)(
;,;";'
.. .c
-.
. .
;:
;,
CD ;,
c-
1ij
·o:C a.",
CD
a.
'1' !iE
...I::::'
:J::::.
' :EE ::!i:::.
I-~
648 848 1602 833 1064 148.4 2736 192.1 258.2 240.9 3482 327.5 392.7 504.6 4009 358.0 371.6
849 105.9
748 95.3
250 26.0
~'1'E
0
CD
ECD()
107.7 127.7 1732 2846 212.7 300.5 2600 367.5 3436 407.7 529.1 4166 371.6 387.5
95.5 117.0 160.8 279.1 202.4 279.3 250.5 357.8 335.6 400.2 516.8 408.8 364.8 3795
27.0 24.5 240
24.5 28 23.0 23.5 22.5 265
Table 1.7 (continued) Critical flux evaluated for 4.9 !lm nickel SPO membrane across a range of shear rates (expenmentaI parameters: 0 - 1400 S·1 surface shear; 0 020 kg m-3 latex feed suspension).
Si
...: CII
...
"C CII CII
<:
a.
CII
Cl
<:
==1/1CII
W
C
-
... ... ... ... t: E ... "C ...0 ;e:0:=
14]ulx3 16]ulx6 141Ulx9 17]ulx3 15novx3
14/07/01 16/07/01 14/07/01 17/07/01 15/11/01
16 20 18 21 23
E .;: CII
a. ><
1/1
CII
01
CII
CII -
CII
(f)~
(f)
800 867 933 933
11 11 11 11
CII N
0;
......
>< CII
0;
d
<:
8.;-
CII-
"CE
(f)o!!.
:E~
Ifc
1200 13005 13995 1399.5
2.1553 2.1475 2.1497 2.1507
s:':"
01-
:=-
-
Cl
a.
1/1 1/1
01
0
'iij
01
CII
-:: -:c s:
"C CII "C "C
CIICl
0020 (0 0020 (0 0020 (0 0020 (0
CII
E
><
-
::::1::-;:c
E
.-<:.5 -1/1
ca~
CII
1/1-
><':"
::::IS:
;:~
:> E
CII
<:
a. >a. E
i
1/1
a.
E
Cl
~
><
Cl
<:
==1/1CII
a.
E
-. ><
::::1-
_. ><
::::l-
-~
-~
~'1'
CII'1'
... s:
... s:
a:~
<:-
0..
0..
::I
-~
c:(:=-
0..
oE ..J:=-
g;E :::I:=-
104 108 9.9 10.4
3246 501.1 545.5 5066
246 303 618 450
PLC (3) PLC (3) PLC (3) PLC (3)
50 58 50
54 60 55 70
547.5 6327 5455 734.3
580.2 652.5 607.3
::::I
EE
::I
::::I
66
776.6
...::::ICII
::::I
;;::
c- O;...
·01: a.
.. :E!E N
CII
a. ECIIO
:E:=-
I-e...
5639 6426 5764 7555
225 24.0 22.5 230
Table I.S Repeatability of critical flux analysis for 4.9 J.1m nickel SPO membrane across a range of shear rates (experimental parameters: 0 - 1200 S·I surface shear; 0.020 kg m·3 1atex feed suspension). ~
.. ~
Cl)
Cl)
Cl.
E .;: w
c
... ... t: E "C ...0 .-en.:. Cl.
24octx3 230ctx3 230ctx6 24octx6
24/10/01 23/10/01 23/10/01 24/10/01
3 1 2 4
Cl)
Cl
"CCl)
C
Cl.
Cl)
)( Cl)
S
..:
)(
C
"CCl) "C "C lIS
1/1
Cl)
"tU
Cl)
C1)~
0 267 533 800
Cl)
......
"tU lIS
Cl)
.c
en
= .. ... 1/1
...~
;:
en
.! 1/1 1/1
IIS~
:i!S 0 n/a 2.1586 400.5 I 2.1532 799.5 11 2.1534 1200 11 2.1548
]'
Cl)
N
cl C
)( ::::s~
Cl)
Se?
;Ci)
Cl
C.:
E
"CE Cl)
E
If::. 0.020 0.020 0.020 0.020
::::s a:~
D D
D D
10.2 10.0 100 10.1
;::~
'iiiC1
.iij
:::.
Cl.
Cl
.. .. Cl)
Cl)
I/I~
Cl.
)('0
::::s.c
;;::~
c
.c
Cl
:c
~
~1/1
~1/1
Cl.
Cl.
Cl.
:> E
-~
et:::.
E c..
20.7 156.4 2309 265.5
17.0 25.0 28.6 27.0
PLC PLC PLC PLC
::::s
E c.. ::::s
(3) 06 (3) 21 (3) 25 (3) 34
;:'" ... .c
.. _ . -- 'o:C
o E
liE
::::s
C
>-
EE c_
)(
Cl
E c..
)( ::::s~
~'1'
;:
)(
::::s~
... .c Cl)
'1'
c~
9-~
...
.. Cl)
::::s
f!
Cl)
Cl.
:i!:::.
E~ C1)t) I-~
07 60.9 71.8 66.4 22 220.0 231.4 225.7 27 265.9 288.2 277.0 36 3591 3736 366.4
24.0 23.0 235 23.0
::::s
..J:::'
::J:::.
:EE
Table 1.9 Critical flux evaluated for 4.1 ).lm nickel SPO membrane across a range of shear rates (experimental parameters: 0 - 1200 S·I surface shear; 0.020 kg m·3 latex feed suspension). ~
...GI
"Cl
Cl) Cl)
C
c..
Cl)
... ...
E .;:
III
C1)~
Cl)
Cl)
><
"C
w
1ii
c
... 0
t::Ec.. .-
20novx3 27novx3 20novx6 22novx6 23novx6 19novx5 22novx9 23novx3 19novx8 21novx3 22novx3 25novx3 21novx6 25novx6
20/11/01 27/11/01 20/11/01 22111/01 23/11/01 19/11/01 22111/01 23/11/01 19/11/01 21/11/01 22111/01 25/11/01 21/11/01 25/11/01
3 14 4 8 11 1 9 10 2 5 7 12 6 13
0 67 133 133 200 267 267 267 400 533 667 667 800 800
c..
"Cl
S
...:
Cl)
en.:.
"Cl "Cl
Cl
~III
Cl)
1ii 01
Cl)
~
III
Ill!!.
nla
0 100.5 199.5 199.5 300 400.5 400.5 400.5 600 799.5 10005 10005 1200 1200
NOF refers to no obvious sign of fC:lUIing
>< Cl)
.!l! III III 01
:::.
Ui
~
:5.2! 2.1530 2.1552 2.1547 2.1550 2.1533 2.1521 2.1548 21539 2.1549 2.1567 21541 2.1542 2.1564 21538
cl
C
8<1
"CIE
GI
E~
;:111
c.:
><
::lr ;::c 1ii~
af~
::lE a:~
EE
0.020 D 0020 D 0020 D 0020 D 0020 D 0020 D 0020 D 0020 D 0020 D 0.020 D 0020 D 0020 (D) 0.020 (D 0020 (D)
10.0 10.1 103 10.1 10.0 10.0 60 10.0 10.0 10.1 10.0 100 10.0 10.1
21.7 85.0 605 81.8 156.8 41.4 183.4 1889 103.6 2332 2966 2700 315.7 400.9
Cl)
- -. c.. Cl)
01
... ...... ...... ..c':" ;: Cl)
GI N
GI
C
'i 0
Cl
c-~
III~
><':"
::l..c :: E
-N
Cl)
c.. >c..
E ::l
C:(:::.
ll.
212 314 244 318 284 36.1 37.7 375 244 30.0 34.1 34.8 31.4 35.3
PLC (3) PLC (3) PLC (3 PLC (3 PLC (3 PLC (3 PLC (3) PLC (3) PLC 3 PLC 3 PLC 3 PLC 3 PLC 3 PLC 3
Cl
c
= Cl)
III
c..
E ::l
ll.
10 08 27 17 29 12
..c
Cl
:c
~
Cl
c
~III
c..
E ::l
>< ::l
-. f1! ::l
-.-- --_.
·01: c..
c..
oE
:EE
C1)(.)
><
::l~
... ..c ~'l'
><
::l~
... ..c C1)'l' g;E
;:
c~
'
N
~
Cl)
E~
127.7 116.4 3048 209.1 3273 149.6
:5:::. 1161 1007 2963 3905 320.1 1386
... ~ 21.0 21.0 22.5 23.0 21.5 22.0
263.9 299.1 3518 330.7 443.9 472.5 542.1
2482 2903 343.3 3136 4285 4568 5274
21.0 22.0 24.0 22.5 220 240
ll.
.J:::'
~:::.
12 11 29 20 31 14
1046 850 287.7 181.4 313.0 127.7
25 28 33 31 41 44 50
2325 281.6 3348 296.6 4132 441.1 512.7
NOF
22 26 31 28 38 41 47
Table 1.10 Critical flux evaluated for 5.0 J.1m PC isopore CPO membrane across a range of shear rates (experimental parameters: 0 - 1400 s'\ surface shear; 0.020 kg m·3 latex feed suspension). ~
...
-
"CGI
C
GI
GI
Do 0
E .;: GI
Do
><
w 24julx4 241Ulx2 20Julx4 20Julx2 181Ulx4 18julx2 24juJx6 25iulx2 25Julx4 25julx6 26Julx2
"C GI "C "C ca
S
...: GI
Cl C
i
c
...0 ... ... ...... I: E "C ...0 ;E;:;
24/07/01 24/07/01 20/07/01 20/07/01 18/07/01 18/07/01 24/07/01 25/07/01 25/07/01 25/07/01 26/07/01
6 5 4 3 2 1 7 8 9 10 11
GI
16
GI
GI-
GI
GI
><
GI
0 0
"CE
ca-
(/)
(/)~
~S
0 67 133 200 267 400 533 667 800 867 933
nla
0 100.5 199.5 300 400.5 600 799.5 1000.5 1200 1300.5 1399.5
2.1474 2.1589 2.1628 2.1668 2.1474 2.1434 2.1567 2.1481 2.1522 2.1560 2.1552
11 11 11 11 11
cl
GI-
.co;"
- Do
.ll!
...ca
~ :cCl
~ Ui
16 ...
(/)~
I I I 11
.c
C
><
GI
:::J~
1.)'1
.5~ _0
;::c
GlCI
C.:
u..~
a:~
EE c_
o~
GI..>=
0.020 D 0.020 D 0.020 D 0.020 (D) 0.020 (D) 0.020 D 0.020 D 0.020 D 0.020 D 0.020 D 0.020 D
:::J
E
11.1 10.7 10.6 10.4 10.8 10.4 11.4 10.6 10.7 10.6 105
ca"
GI
0~
GI
Do
><0;"
>-
;:'1
E :::J
:::J.c
;:;E
Do
::::.
~
Cl C
Cl C
== GI
0 Do
E :::J
GI
0 Do
E :::J
><
><
:::J~
;:;::::"i
... .c ;'1'
o
E
_. ><
:::J~
-~
... .c
- -
:::J ;:
c~
.-"i
o.c DoN
~E
•• :2E
GI'1'
-~
~::::.
Do
Do
D..
.J::::.
:J::::.
~::::.
21.9 65.1 86.4 85.5 43.4 65.1 105.2 127.7 127.7 126.4 150.0
21.5 21.5 20.6 22.0 20.8 21.7 22.3 23.1 22.2 22.0 21.1
PLC (3) PLC (3) PLC (3) PLC (3) PLC (3) PLC (3) PLC (3) PLC (3) PLC (3) PLC (3) PLC (3)
12 12 12 12 12 14 14 14 16 16 18
14 14 14 14 14 16 16 16 18 18 20
128.9 130.7 128.6 128.6 128.9 1489 150.5 149.8 173.2 169.3 192.1
150.7 151.1 148.2 151.4 147.5 173.6 172.1 173.9 194.3 192.5 213.2
139.8 140.9 138.4 140.0 138.2 161.3 161.3 161.8 183.8 180.9 202.6
S!:::J
...GIca
Do
Eu ~~ 23.5 23.0
22.5 22.5 24.5 240 245 24.5
Table 1.11 Critical flux evaluated for 5.0 ~m PC isopore cpa membrane across a range of shear rates (experimental parameters: 0 - 1200 S-I surface shear; 0 093 kg m-3 latex feed suspension). ~
S
..-
...: Cl)
'0
Cl)
Cl.
)(
'lO
C1).:.: u.._
Cl)
'0
w
c
0
12aprr3 13aprr3 12aprr2 14aprr2 21aprr3 12aprr1 13aprr2 21aprr1 11aprr2 14aprr1 21aprr2 11aprr1
12104/01 13/04/01 12104/01 14/04/01 21/04/01 12104/01 13/04/01 21/04/01 11/04/01 14/04/01 21/04/01 11/04/01
4 6 3 8 11 2 5 9 1 7 10 12
Cl)
(/)
C1)~
:: E .- Cl. 0 0 200 200 200 400 400 400 800 800 800 800
:;::
'lO
as Cl) .s:: ":'(1)
)( Cl)
'lO (/) (/)
!fI
!fl-
::as
nla nla I I I 11 11 11 11 11 11 11
0 0 300 300 300 600 600 600 1200 1200 1200 1200
10.0182 10.0007 10.0622 10.0076 10.0232 100426 10.0254 100230 100230 10.0090 10.0141 10.0134
..2
Cl
Cl)
0 c
o~
Cl)
Cl
0.093 (C) 0.093 C 0.093 C 0.093 C 0.093 Cl 0.093 Cl 0.093 Cl 0093 C 0.093 C 0093 C 0.093 (Cl 0.093 (Cl
Cl)
)( :::I~
.5_ ;::c -(/)
c.s :::I E Il:_ 10.0 10.3 10.2 10.1 11.2 10 10.1 10.4 10.0 10.2 10.7 10.0
"i~
--
-c
Cl
:c
(/)~
)(0;'
:::I-C
;:~
:>
Cl)
Cl.
c
>-
Cl) (/)
Cl.
Cl.
E
E
)(
Cl
- - == E Cl.
~"IE
(/)
Cl)
as
as _
Cl.
Cl)
-
Cl)
.!::! (/)
= .. .. ...I!! .... .. US.;. Cl) Cl)
cCl) E .;:
Cl
c
'0 Cl) '0 '0
'i
c
Cl) (/)
Cl.
E
)( :::I~
;:~
.. -C
;'l'
_ . .. -c )(
:::I~
-~
C1)'l'
:::I
-
:;:
c~
·o:C Cl. N
:E 'E
EE c_
Il.
Il.
:::I
:::I l1.
...I:::;'
;:):::;.
::a:::;.
47.2 20.1 47.6 20.1 22.6 58.3 20.9 23.1 57.1 21.4 22.4 76.9
9.2 9.8 9.9 10.1 10.6 9.4 98 10.1 9.5 9.8 10.2 88
PLC (3) PLC 3 PLC 3 PLC 3 PLC (3) PLG(3) PLC 3 PLC 3 PLC 3 PLC 3 PLC (3 PLC (3
06 07 06 04 05 07 06 05 08 06 06 08
07 08 07 05 06 08 07 06 09 07 07 09
57_1 70.5 58.4 41.5 54.0 67.6 59.2 53.9 76.7 60.9 63.3 76.9
65.6 79.1 67.4 51.1 64.9 77.1 69.8 63.3 85.7 70.6 73.2 85.7
61.3 74.8 629 46.3 59.4 72.3 645 58.6 81.2 65.7 682 81.3
:::I
o E
8:E
Table 1.12 Critical flux evaluated for 4.8 Ilm nickel CPO membrane across a range of shear rates (experimental parameters: 0 - 1400 S·I surface shear; 0.020 kg m·3 latex feed suspension). ~
"C
S
...: Q)
......c
"C Q) Q)
a.
Q)
"C "C
Cl
C
~1/1
Q)
1;j
......
Q)
a.
Q)
><
1;j
W
Q
... ... ... I:: E "C '" ...0 ;e- ;:;...!1! .c"'" (/)~
170ctx3 18octx9 16octx6 190ctx3 160ctx9 11octx8 120ctx4 18octx3 150ctx4 190ctx6 150ctx7 16octx3 18octx6
17/10/01 18/10/01 16/10/01 19/10/01 16/10/01 11/10/01 12110/01 18/10/01 15/10/01 19/10/01 15/10/01 16/10/01 18/10/01
8 11 6 12 7 1 2 9 3 13 4 5 10
.E;:
1/1
Q)
Q)~
Q)~
(/)~
(/)
0 67 133 200 267 400 400 533 667 667 800 800 933
nla I I I I 11 11 11 11 11 11 11 11
0 1005 1995 300 4005 600 600 7995 1000.5 10005 1200 1200 1399.5
NOF refers to no obvious sign of fouling
.....!!!'><" Q)
1/1 1/1
"'~ :!is
2.1534 2.1554 2.1577 2.1511 2.1538 2.1532 2.1562 2.1548 2.1499 2.1523 2.1534 2.1539 2.1550
~ ::=.
I!l 'iij
Q)
($
c
o~
U'2
"CE
Q)CI Q)~
u.~
0.020 (D) 0020 (D) 0.020 D 0.020 D 0.020 D 0.020 D 0.020 D 0.020 D 0.020 D 0.020 D 0.020 D 0.020 D 0.020 D
Q)
....5~ 1/1
a. Q)
...
Cl
c
~
Cl
~
:c
(.)
~
><
Cl
c
= = ---.
Q)
><
::I
><
...
:;::
2.,..
... 1;j ... Q)
::I
"i~
;;::f1
...>-a.a. E
E
E
;'l'
Q.
o E
..I::='
~E :::l::=.
:EE
Q.
:!i::=.
{!
10 05 05 05 11 10 15 15 15 26 NOF 29 34
20 08 08 08 14 30 20 17 25 28
66 157 159 152 341 318 490 47.3 50 81.8
13.3 255 255 25.7 44.1 90.9 63.2 51.8 81.4 882
99 20.6 20.7 205 391 614 561 495 65.7 850
22.0 235 24.0 23.5 235
250 255 225
32 37
955 1105
102.7 1196
991 1150
255 260
>< .a.r -~
I/I~
><"' " ::I~
c.= ::I E
EE
a:~
c-~
C(::=.
Q.
100 100 10.1 100 10.1 100 100 100 100 100 100 10.0 10.0
6.6 5.9 5.8 152 159 6.3 160 318 500 36.8 496 49.8 64.6
100 9.8 9.8 10.5 9.4 28.2 15.7 100 31.4 10.3 10.5 10.6 11.0
PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC PLC
:> E
Q)
Q)
1/1
1/1
a.
::I
::I
16 1.6 1.6 1.6 1.6) 1.6) 16 16 16 1.6 16 1.6 1.6
a.
::I
::I~
...
~
::I~
...
;;::'"i
~ Q)'l'
c~
·o~ Iil-'l'
Q)
a.
E
Table 1.13 Critical flux evaluated for 10.0 Jlm nickel SPO membrane across a range of latex feed concentrations (experimental parameters: 600 g"1 surface shear; 0.01 - 0.40 kg m·3 latex feed concentration).
"tI
...
Gl
Gi C Gl
Gl
(I)
"tU
.. ..l::E .... .. ;eCl.
E .;: Gl
Gl
>< w
"tU
c
0
02augx5 18augx3 17auQx6 01augx9 09augx3 01augx6 12augx3 12augx6 llaugx6 llaugx3 08augx6 01augx3 31julx5 311Ulx8 08augx3 06augx3 07augx6 06auQx6 07augx3
02108/01 18/08/01 17/08/01 01/08/01 09/08/01 01/08/01 12108/01 12108/01 11/08/01 11/08/01 08/08/01 01/08/01 31/07/01 31/07/01 08/08/01 06/08/01 07/08/01 06/08/01 07/08/01
6 19 18 5 13 4 16 17 15 14 12 3 1 2 11 7 10 8 9
Cl.
"Cl Gl Gl
Gl
Gl--
'C
IV
Gl - -
'C 'C
'iij
>< Gl
Gl
(I) (I)
IV--
IJ)~
"c"'cn IJ)~
:i!S
400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400
600 600 600 600 600 600 600 600 600 600 600 600 600 600 600 600 600 600 600
1.0894 2.1518 2.1535 2.1563 3.2331 32357 32359 32378 32388 32390 53933 53953 10.7706 10.7774 10.7804 21.5495 21.5564 323281 430973
;:: 0
.. .. Cl.
IV
"tU
-- :c--
Gl N
cl C
8cr
"tiE
Glm Gl..:.=
u...~
0.01 0 002 0 002 0 0.02 0 0.03 0 003 (0 003 (0 003 (D) 003 (D) 003 (D) 0.05 (D) 0.05 (D) 0.10 (D) 0.10 0 0.10 0 0.20 0 0.20 (D) 0.30 (D) 040 (D)
Gl
E
.-
..
c
-.= (I)
E
::::I a:~
10.4 10.2 102 10.3 106 12.4 104 104 104 107 104 10.3 10.7 103 105 104 105 105 10.5
><
::::I::""
;;:::c ca~
EE
(1) __
><0;" ::::I.c
;:~
:> E
Gl
Cl.
>-
Cl.
E ::::I
-~
c-
C:(::::.
Il.
856.4 898.0 739.8 280.0 1968 220.5 2155 2132 215.91 2141 1936 108.0 636 634 435 434 226 21.3 23.9
173.6 129.6 536 40.5 26.2 30.7 45.1 44.9 448 445 236 21.4 119.5 328 21.5 208 21.6 15.8 216
PLC 5 PLC 5 PLC 5 PLC 3 PLC 3 PLC (3 PLC (3 PLC (3) PLC (3) PLC (3) PLC (3) PLC (3 PLC (5 PLC 3 PLC 3 PLC 3 PLC (3) PLC (3) PLC (3)
.c
Cl
::::.
~
Cl
Cl
C
i
(I)
Cl.
E ::::I
Il.
30
><
.. _. 'o'b = -.-- -C
Gl (I)
Cl.
E ::::I
::::I
><
><
::::1--
::::1--
.. .c
.. .r:
~'1'
o E
;:
c--
R:E
' . :2E
Gl'1'
Cl.N
... Gl
::::I
~
Gl
Cl.
EO'
Il.
..J::::'
;:)::::.
:i!::::.
~~
34
2727.3
3112.7
2920.0
250
78 46 26 56 60 64 44 20 22 05 12 22 06 22 12 14
8400 471.1 260.5 5639 611.1 666.1 434.6 193.6 220.0 63.6 106.1 2152 43.4 219.6 110.5 1307
8877 5107 281.8 621.1 6621 7084 481.4 217.3 236.4 183.2 129.1 2368 64.2 239.1 131.8 153.6
8639 4909 271.1 592.5 6366 687.3 4580 2055 2282 123.4 117 6 2260 538 229.3 121 1 1422
270 225 27.5 230 23.0 23.0 22.5 23.0 26.5 260 25.0 23.0 23.5 24.0 24.5 23.5
NOF NOF
74 42 24 52 56 60 40 18 20 02 10 20 04 20 10 12
Table 1.14 Critical flux evaluated for 4.9 Ilm nickel spa membrane across a range of latex feed concentrations (experimental parameters: 600 S,l surface shear; 0.0052 - 0 40 kg m,3latex feed concentration).
"C
GI
.,.:
.. ...GI
"C "C 111
"C
GI GI
C
GI
Q,
GI
W
C
~ ...111 ... ... :: E ... ;e- .c" 0
21junx3 261unx6 20Junx6 271unx3 26junx3 20junx3 19junx3 03mayr3 18junx3 21junx7 19iunx6 22junx3 27junx6
21/06/01 26/06/01 20/06/01 27/06/01 26/06/01 20/06/01 19/06/01 03/05/01 18/06/01 21/06/01 19/06/01 22106/01 27/06/01
7 11 6 12 10 5 3 1 2 8 4 9 13
C .;: GI
Q,
><
1/1
GI
10
GI
GI ..-
"C
IJ)~
GI - IJ)!J,
400 400 400 400 400 400 400 400 400 400 400 400 400
600 600 600 600 600 600 600 600 600 600 600 600 600
,
.. ><
GI
..!l!
1/1 1/1
111--
::!is
0.5591 0.8556 1.0837 2.1522 2.1595 32456 5.3914 10.0038 10.7584 10.7785 21.5545 32.3184 43.1078
'i 0
GI N
'0 Q,
c.i
C
8cr "Cc GlCI GI"II:
u..~
0.0052 C) 0.0079 Cl 0010 C 0.020 C 0.020 C 0.030 C 0.050 C 0.093 C 0.10 (C) 0.10 C 0.20 C 0.30 C 0.40 (C)
GI
.5... . 1/1 c.: ::I C
a:~
109 10.4 10.6 10.7 10.5 10.8 10.4 10 10.5 10.5 10.5 10.4 11.6
>< ::I~ =i::S:
ca '1 EE c_
-~
319.6 377.1 191.8 214.6 221.4 86.1 22.5 42.8 12.2 20.6 10.8 22.4 21.9
21/1..><" ::I.c ;;:::~
:: E
Cl C
Cl C
E ::I
E ::I
>-
Q,
Cl
:c
..
Q,
.c
::;:.
EGI
GI
..-
1/1 Q,
~
~1/1
Q,
E ::I
><
.. _. ·o:C -::I
><
::1..-
;:';'
....c
~'l'
o E
><
::1..-
... .c GI'l'
;;:
C"Q,N
••
8:E
:2c
eI:::;:.
Q.
Q.
Q.
...I::;:'
~::;:.
::!i::;:.
29.1 39.9 15.5 43.2 20.9 15.7 10.5
PLC (3) PLC (3) PLG(3) PLC 3 PLC 3 PLC 3 PLC 3) PLC 3) PLC 3) PLC 3 PLC 3 PLC 3 PLC (3)
58 40 34 28 36 19 10
59 42 35 29 37 20 11
604.8 436.4 368.2 300.9 388.2 206.4 108.9
610.2 456.8 377.3 303.2 402.3 211.4 117.3
607.5 446.6 372.7 302.0 395.2 208.9 113.1
03 10 07 08 04
04 11 08 09 06
326 102.7 743 85.5 42.6
43.4 110.9 84.8 94.8 63.8
38.0 106.8 79.5 90.1 532
10.4 12.9 10.6 18.1 20.9
Table 1.15 Critical flux evaluated for 5.0 Ilm PC isopore CPO membrane across a range of latex feed concentratIOns (experimental parameters: 600 S·I surface shear; 0.0035 - 0040 kg m·3 latex feed suspension).
...:
...
-. GI
'C
E ;: GI
e.. ><
GI
e..
GI
GI
1;
w
c
lljunx2 07junx2 07junx4 13junx4 31mayx4 14runx2 lljunx4 l1junx6 13runx2 23aprxl 22aprx2 14runx4 12junx7 31mayx2 24aprx2 12junx5 12junx3 14aprr3
11/06/01 07/06/01 07/06/01 13/06/01 31/05/01 14/06/01 11/06/01 11/06/01 13/06/01 23/04/01 22104/01 14/06/01 12106/01 31/05/01 24/04/01 12106/01 12106/01 14/04/01
... 'C ...
I:
GI
0
10 8 9 17 7 18 11 12 16 4 3 19 15 6 5 14 13 1
...... 111
1;
en
...
:!I 'iij
><
GI
GI-
E
GI-
GI
1;
U C
8.r
en en
'CE
.f~ 0.0035 (C) 0.0035 (C) 0.0036 (C) 00051 (C) 0.0069 C) 0.0079 C) 0.0094 C) 0.0094 C) 0.010 C 0.019 C 0.019 C 0.020 C 0.030 C 0.035 (C) 0.047jC) 0.050 (C) 0.10 (C) 0.19 (C)
(/)-
=e-
.c' (/).!!!.
:!is
400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400
600 600 600 600 600 600 600 600 600 600 600 600 600 600 600 600 600 600
0.37315 0.3758 0.3862 0.5463 0.7466 0.8523 1.0086 1.0135 1.0781 1.9934 2.0035 2.1573 3.2203 3.7315 5.0422 5.3868 10.7773 200190
111-
Glm
GI
E "0 c.s j E 11:_ 11.6 7.7 11.7 9.8 704
16.0 6.6 60 11.5 8.4 9.8 9.5 9.9 10.4 11.9 9.6 10.3 10
>< j::", :;:J:: "i~
--
!2E c-
377.7 419.6 329.1 163.0 162.7 174.3 387.1 310.9 43.8 218.2 24.0 85.6 22.7 44.6 43.1 225 23.8 21.5
j.c ;::~
:> E
.c
Cl
.9
- en_ ><,
-:c - 'i
e..
111
GI GI
C
'C Cl 'C 'C
GI
e..
>-
e..
E j
Cl
c
GI
GI
==en
==en
E j
E j
D.
D.
e..
C(::.
e..
27.0 12.7 11.6 16.2 55.3 15.3 96 nla 10.8 10.7 10.6 18.2 10.6 10.8 11.0 10.5 10.3 9.8
PLC (3) NOF PLC (3) 38 NOF PLC 3 34 PLC 3 40 PLC 3 19 PLC 3 35 PLC 3 PLC 3 PLC 3 16 21 PLC 3 20 PLC 3 3 12 PLC PLC 3 07 PLC (3) 07 PLC (3) 07 PLC (3) 05 PLC (3) 04 PLC (3) 04
>< j
Cl
c
e..
39
-
>< ....><... .c .... ... .c ~~ ~E
;:
::J::.
:!i::.
419.6
432.3
4259
370.5 4389 206.1 384.1
3900 494.3 220.2 3936
380.2 466.6 213.2 388.9
173.2 224.1 218.6 129.6 75.9 76.5 76.4 54.7 44.9 40.1
184.6 239.6 432.7 140.2 86.3 87.7 87.1 65.3 54.7 51.0
178.9 231.8 325.7 134.9 81.1 82.1 81.7 60.0 49.8 45.5
j-
-~
j-
-~
Cl')'
co.c e..OI :E 'E .-":'
"
36 45 20 36 "
17 22 40 13 08 08 08 06 05 05
..............-----------------------------------------------------------------------------
~
Table 1.15 (continued) Critical flux evaluated for 5.0 I1m PC isopore CPO membrane across a range of latex feed concentrations (experimental parameters: 600 S·l surface shear; 0.0035 - 0040 kg m· 3 latex feed suspensIOn).
...... Gl
a. )( w
c
" a. 1ii ... ...... ... :: E .c':" "0... ==een'!!'
15junx3 15junx8 22aprx1 27junx8
15/06/01 15/06/01 22104/01 27/06/01
20 21 2 22
Gl Gl
C Gl
E
Gl
III
.;: Gl
" ""
Gl
...:
Gl
1ii
Gl
Gl-
IU Gl-
en~
400 400 400 400
600 600 600 600
Gl
1ii III III IU-
:5~
21.5932 323248 40.0061 43.0866
:;:
0
N
'iii
a. Gl
IU )(
- -:c
Gl
U C
0_ U'1
"E Glen
.f~ 0.20 (C) 0.30 C 0.37 C 0040 C
Gl
.S_ ... 111 c.!:
E a:~ :l
12.0 12.9 10.6 1004
)(
:l:;"'
:;::j:
me:BE c-
...
ili-
-.
)(':"
:l.c
-N
;:;E
Gl
a. >a.
...
E :l
.c
Cl
:::.
~
Cl C
Cl C
::Gl ::Gl III
III
E :l
E :l
a.
a.
)(
--.;':' )(
:l_ ... .c
o
E
:l
...c _. ·o:C -)(
;::
:l_ ... .c Gl,:,
8:E
-~
C:(:::.
Il.
Il.
Il.
...I:::'
::J:::.
22.7 29.1 250 20.7
11.0 7.3 10.5 10.7
PLC (3) PLC (3) PLCI3l PLC (3)
03 03 03 03
04 04 04 04
33.4 34.2 35.8 32.2
44.7 43.8 46.0 42.1
..
a.N
:5!E
:5:::. 39.0 39.0 40.9 37.1
Table 1.16 Extended filtration of 4.9 !lm nickel SPO membrane at various concentrations of feed suspension (experimental parameters: 600 s'\ surface shear; 0.01 - 0.10 kg m·3 latex feed concentration).
"Cl
GI
...:
.....
"Cl
E .;:
...GI
GI
GI
W
C
16]ulx9 27iulx2 27julx5
1;j
...GI "Cl ... 0
01
>< GI
GI 01
a.
GI
a. ><
.....
GI GI
C GI
"Cl "Cl
III
~
t: E .a.
en':'
~
.c ";'u)
UI~
16/07/01
400
600
27/07/01
400
600
III III
Her
01
"CIE ~
::sS 1.0947 10.7775 10.7703
.. GI
C
1;j
... 01 GI
U
GlC)
Jf!~
0.Q1 (D) 0.10 (D)
GI
.5~ .. Ill
c.s
:s E
a:~
159.5 1400 152.0
a. >a.
E :s
Co
C) C
= GI
III
a.
E :s
Co
PLC (3)
28
PLC (3)
05
... 1;j ... GI
><
:s~
;::c
"i'1 EE c_
-~
304.1 24 51.2
:s GI
a.
E~ GlO ~
...
21.5 26.5
Table 1.17 Extended filtration of 5.0 Jlm PC isopore CPO membrane at vanous concentrations of feed suspension (experimental parameters: 600 S-I surface shear; 0.01 - 0 093 kg m-3 latex feed concentration).
.. I/)
...: Q)
.....
"
Q) Q)
cQ) E .;: Q)
Co
><
w 26julx4 15iunx5 23aprx2 12aprr4 13aprr1 26Julx2
" ""as
cQ) E E
Q)
Co
... ...... I/)
Q)
iii
c
26/07/01 15/06/01 23/04/01 12104/01 13/04/01 27/07/01
Q)
~
iii ...
iii
Q)
I/) I/)
~
.c "i'U)
I/)~
400 400 400 400 400 400
NOF refers to no obvious sign offouling
cl c
>< Q)
Q)
...as
Cl
600 600 600 600 600 600
o~
U"I
as ==,9 ~
1.0805 1.0849 100061 10.0184 10.0416 10.7775
c
..
>-
~
Co
E
Co
E
D.
D.
Q)
"E Q)CI
Ifc 0.Q1 (D 0.Q1 (C 0093 (C 0093 C 0093 C 0093 D
Co
Q)
.5~ c.s
. . I/)
::I
0:
E .....
1460 126.3 125 192 60.0 137.8
I/)
::I
PLC PLC PLC PLC PLC PLC
::I
(3) (3) (3) (3 (3 (3
11 13 06 06
>< .=r -.c ca~
EE c_
-~
114.8 138.0 71.3 56.7 38.2 24.1
0 U Cl
c
:;
.f
... iii ... Q)
::I Q)
Co
E~
Q)O
I-~
26.0
NOF NOF
27.0
APPENDIX 1. Relative importance of membrane open area in relation to critical flux
APPENDIXJ RELATIVE IMPORTANCE OF MEMBRANE OPEN AREA IN RELATION TO CRITICAL FLUX The open area of a porous medium can be defined as the ratio of area through which flow can pass compared to the total filtration surface area. Open area is generally specified as a percentage. Comparattve critical flux experiments of membranes with different pore sizes and pore geometries are reported in Chapter EIght Results and Discussion (Surface Microfiltration). There was considerable variation in the open area of 4
~m
and 5
~m
slotted (SPO) and CIrcular (CPO) pore membranes evaluated,
ranging from 8 per cent (5
~m
isopore CPO membrane) down to 0.3 per cent (4.8 per
cent nickel CPO membrane). In a study to investigate the potential influence of membrane open area on cntical flux PVA glue was carefully applied to the nonfiltration side of the 4.9 ~ nickel SPO membrane to block off selected pores. Minute
Figure J.l Non-filtration side of 4.9 ~m nickel SPO membrane following PVA glue addition to block off pores and consistently reduce flow area.
L-___________________________ _ _ __
1-1
APPENDIX J. Relative Importance of Membrane Open Area to Critical Flux drops of PVA glue, which is waterproof, were bonded uniformly to sites across the surface using a fine metal point. This time consuming operation was necessary to ensure a uniform reduction of flow area throughout the membrane. The glue was allowed to dry overnight before use. Figure J.l is a digital image of the non-filtration side of the modified membrane, taken usmg an Olympus digital camera. The many glue sites are clearly visible in the Figure. What seem like areas of significant glue coverage are actually aggregates of many smaller glue sites. Even after considerable preparation time It is estimated that the open area of the 4.9 !lm nickel SPO membrane was reduced marginally by 10 per cent (approximately) from the origmal 2.1 per cent measured open area.
Five constant-rate filtration experiments were performed using this modified membrane. The experimental method was similar to that described in Section 8.3.3 on page 8-29. A flux-stepping technique was utilised whereby the permeate flux is increased step by step from an imtlally low flux until the pressure drop across the membrane increases 'significantly' with time. At each step the permeate flux was maintained for 10 minutes with permeate recycled to the column top. There was no backflushing during cntical flux evaluation. Following each IO-minute flux run; the recycle stream was temporanly sampled to allow flux measurement by stopwatch and electronic balance. Pressure readmgs were recorded at 2-second intervals. As the experiments were performed at a constant rate, any increase in the pressure drop across the membrane was indicative of membrane fouling. The effect of shear rate at the membrane surface on critical flux was investigated at stirrer speeds of 267 to 800 rpm. This corresponds approximately to a shear rate range of 400 - 1200 S·I. A dilute latex feed suspensIOn of 0.020 kg m'3 was used in all experiments.
Filtration performance of the open area adjusted 4 9 !lm membrane is compared in Figure J.2 against that of 4.9 !lm (original open area) and 4.1 !lm nickel SPO membranes, 5.0 !lm isopore CPO membrane, and 4.8 JlID nickel CPO membrane, The particle deposition model for 5 0 !lm spherical latex particles is overlaid. As seen in the Figure, critical fluxes for the 4.9 !lm nickel SPO membrane are very similar regardless of the open area, that is however, with the exception of 400 S'I surface
J-2
APPENDIX J. Relative Importance of Membrane Open Area to Critical Flux shear. The limited pore size reduction of around 10 per cent is insufficient to seriously evaluate a relationship between critical flux and flow area. A greater reduction of perhaps 50 per cent would be required before a satisfactory study can be completed and for this a new technique to deliver uniform pore blocking must be sought.
250
••
"';"
4.9 ~ m nickel SPO membrane (2.1 % open area) 4.1 ~m nickel SPO membrane (1 .8 % open area) 5.0 ~m isopore CPO membrane (8 % open area) 0 4.8 ~m nickel CPO membrane (0.3 % open area) 4.9 ~ m nickel SPO membrane (open area reduced to 1.9 %) - - Model (shear-enhanced effects)
VI
E :::1.. .:
•
200
ra
"
Q)
J:
VI
150
0
•
Q)
::J "tl
><
100
::J
I;:
.. ra
u
.;:
•~
•
50
•• • "0
0 0 0
200
400
•
•
•
• • • "t •• ~ i •
~
~
600
800
8
•
1000
0
•• 1200
1400
Shear rate at the membrane surface (r) , S-1_
Figure J .2 A study of critical flux behaviour in relation to membrane open area
for assorted surface filters.
J-3
--
--
---
---
---------------------------------------------------------------
APPENDIXK PUBLICA TIONS
Kl Holdich, R.G., Cumming, LW., Awang, A.R.B. and Bromley, A.I. , (2000) , "Seeded Microfiltration for Base Metal Removal from Effluents", Proceedings of 8th World Filtration Congress, The Filtration Society, Brighton, pp.355-358.
K2 Bromley AJ., Holdich, R.G. and Cumming, LW., (2002), "Particulate fouling of surface micro filters with slotted and circular pore geometry", 1. Membr. Sci., 196, pp.27-37.
K3 Holdich, R.G., Cumming, I.W., Kosvintsev, S., Bromley, AJ. and Stefanini , G., (2003), "Clarification by slotted surface microfilters", Miner. Eng., 16, pp.121128.
SEEDED MICROFlLTRATION FOR BASE METAL REMOVAL FROM EFFLUENTS R.G. Holdich, LW. Cumming, A.R.B. Awang and A.J. Bromley, Department of Chemical Engineering, Loughborough University, Leicestershire, LE!! 3TU.
Copper in solution at low concentrations has been removed by contacting the water with modified activated carbon. A crossflow micro filter retained the carbon particles. Extremely high and stable flux rates were measured, approximately 6000 I m- l h- I , whilst employing a transmembrane pressure of less than 0.1 bar. These conditions resulted from the use of a slotted true surface micro filter, instead of a conventional microfilter medium. Slots of as little as 2 f.U!1 in diameter and 400 !lm in length have been used instead of conventional filter pores, and the micro filters were formed into tubes 14 mm in diameter and 350 mm long. The mass transfer of the copper into the carbon particles was reliably modelled using a model that took account of both transfer across an aqueous film around the particles and diffusion within the particle. KEYWORDS: ion exchange, modelling, filter media INTRODUCTION The crossflow membrane filtration of sorption media (activated carbon, ion exchange beads) is a method for the removal of solutes from process streams and is called seeded rnicrofiltration [1]. The solid sorption medium is retained on the crossflow filter and complexes the dissolved species in the liquid. The application of crossflow micro filtration in preference to conventional contacting equipment, such as packed columns, has the following advantages: inert suspended solids in the feed may be tolerated, much finer sorption species may be used, rapid mass transfer onto the smaller particles, simple contacting equipment and straightforward process control. However, the successful application of the process is very dependent on high permeate flux rates. Conventional process scale microfiltration membranes rely on Fig 1(a) Conventional MF medium (b) Slotted membrane - 2 !lm depth filtration mechanisms to achieve their pore size rating [2]; thus they require a tortuous flow pore channel which precludes the seeded micro filtration of effluents containing suspended solids that may become lodged within the rnicrofilter membrane. An example of a conventional MF is illustrated in Fig lea). However, a true surface rnicrofilter may be successfully employed as it
355
does not have a tortuous pore flow channel. These types of filters are currently under development for process scale use and an example of one with a slotted pore is provided in FIg 1(b). The slot shown in the figure is approximately 2 IJ.lll wide and 400 !lm long. A slot geometry is preferable to a hole as it is less likely to suffer from pore plugging on the surface. For the mass transfer of a dissolved species (such as copper ions) transferring from the bulk solution, through an aqueous film surrounding a sorption particle and then diffusing within the particle, a mathematical model based on two resistances to mass transfer acting in senes can be developed: 1. aqueous film mass transfer resistance around the suspended particle, and 2. dlffusion of copper within the carbon particle. The mass transfer can be described by the followmg two equations respectively dC =_Ak (C-C) and 8q(r) =!l...E..(r2 oq(r») (1) and (2) dt V, ' ot r2 or Or where C is concentration of copper in the solution, Cs is concentration at the surface of the carbon particle, A is the total surface area of carbon present, V is system volume, k is the film mass transfer coefficient, D is the copper diffusivity within the carbon particle, r is the local radial position within the carbon particle, t is time and q(r) is the local concentration of copper within the carbon on a mass per unit mass basis. The copper concentration on the surface and within the carbon are linked via a Langmuir type isotherm
aC, q, = l+bC,
(3)
where a and b are constants, and material balances can be used to deduce the average mass on the carbon
q = -R33 lR0 q(r)r2 dr
(4)
where R is the radius of the carbon particle. The above equations can be combined with a material balance on the copper during extraction which will take two separate forms, depending on the method of operation. In a simple batch tank the copper concentration in solution can be described by
C=C_Wqo V
(5)
where W is the mass of carbon adsorbent present and Co is the initial copper concentration. However, in a system continuously fed by liquid containing the solute but with the sorption particles retained by the micro filter - the seeded micro filtration process, the material balance is dC dV-=FC -FC-w--i (6) dt 0 dt where F is the permeate rate and is matched by the rate of addition of copper bearing effluent into the process. In all cases equations (1) to (6) can be solved by first rendering dimensionless and then solving numerically using the proprietary mathematical package PDESOLTM[3]. EXPERIMENTAL AND DISCUSSION The seeded micro filtration of copper from dilute aqueous solutions using carbon modified by the addition of ion exchange groups onto the particle surface has been studied, using the experimental rig illustrated in Fig 2. The carbon particles were oxidised by contacting with concentrated nitric acid, resulting in acidic ion exchange groups distributed throughout the carbon particles. The median particle size of the resulting adsorbent was 50 !lID. Two separate batches of modified carbon were produced with capacities for copper of 0.023 and 0.079 grams
356
of copper per gram of carbon. !bis is a capacity close to a commercially available ion exchange resin. Experiments were conducted using the filtration equipment illustrated in Fig. 2; where the feed rate balanced the permeate filtration rate. The feed contained the copper bearing solution at a pH greater than 4 and the permeate was depleted in copper by the carbon maintained within the crossflow filtration circuit When the carbon became loaded with copper it was regenerated by contacting with acid to give a solution pH of approximately 1.2. !bis was performed in the regeneration circuit pH control which also contained Feed a crossflow filter. An tanks 400 litres example of the flux rate and pressure drop achieved during the loading of the Ceramic membrane copper onto the Process carbon is illustrated Permeate lank in Fig 3. The 20L '---r---' DesorptJon extremely high flux 4!><~~4+....J Tank 1 L rate achievable from a slotted surface filter is evident from this Fig 2 Schematic of experimental rig
T;;~="
10000 , . - - - - - - - - . . 0 . 1
..:
"
.0
figure, and the low operating pressure.
"~ .<:
Experiments were also performed using very low concentrations of carbon in suspension to deliberately exhaust the adsorption species, providing extraction kinetic data suitable for full scale process modelling. These were performed using a similar crossflow experimental rig to that shown in Fig. 2, but recycling the permeate back to the feed tank. The concentration of copper in the permeate with respect to tune was monitored. 1~---------------r
8000
"E 6000 x
"
0:
~CD
§ a."
...1
0.09
~
" UI
•••••
~
c-
"c
4000 0.06
i
~
c
0 1 - - - - - - -........-10.05 .~
Fig 3 Flux and transmembrane pressure
Figure 4 illustrates one example of the analysis, solved by a finite difference solution to the above equations, 0.8 together with experimental measurements when o 06 filtering using 0.2 g r1 of carbon in a solution -uu containing 478 ppb of copper. The mass transfer 04 coefficient and diffusivities have been found to be 02 consistent with all the operating conditions of the many • expenments and, therefore, permit modelling of the o 1000 2000 3000 4000 process at higher loadings of carbon to copper as well as in different modes of filtration, such as feed and lime. s. bleed operation. An example of the application of the Fig 4 Measured and predicted copper numerical model and a duplicate experiment using the seeded micro filtratIOn arrangement illustrated in Fig. 2 is given in Fig. 5. The process tank contained initially 20 litres of water and copper bearing water, at 0.527 ppm, was fed into the tank at a rate that matched the permeate flowing through 1200 rpm
357
020 E c. c.
C
0.16
o
0
;;
-"
.a slotted tubular microfilter of 14 mm internal diameter and 350 mm long, operating in crossflow as illustrated in Fig. 2. The measured permeate rate from this small filter was approximately one litre per minute.
~
c
0.12
u
c
0
u
~
"c.c.
0.08
•
Exp data -first readIng
o
Exp data-second readIng
0
u
~
" E "
~
0.04
-
Model. k=0 014. D=7E-11
000
o
1800
3600
5400
7200
9000
10800
The carbon employed had the higher capacity Time,s. for copper and the Fig 5 Measured and predIcted copper concentration ID permeate diffusivity for copper within the carbon particle was calculated to be 7x10-1l cm2 S-I. In the seeded micro filtration system the agitation is not as great as that for the batch system, illustrated in Fig. 4, hence the aqueous film mass transfer coefficient was ouly 0.014 cm S-I. During seeded microfiltration of copper, this transfer coefficient was found to be constant for both types of carbon used but was a function of the hydrodynamic conditions. Whereas the copper diffusivity within the carbon depended on the carbon type used but was independent of the prevailing hydrodynamic conditions [3]. However, both diffusivities for copper within carbon are low. A BET investigation of the carbon showed that the intemal porosity was only 4%, and these diffusivities are consistent with ones recorded for highly cross-linked (low porosity) ion exchange media [4]. The result illustrated in Fig. 5 suggests that the mass transfer is primarily influenced by aqueous film control until about 600 seconds and then chffusion within the particle becomes increasingly important. When the latter condition applies the rate of increase of solute concentration in the process tank, and hence the permeate from the system, reduces to a low value until saturation of the sorption medium is approached. Other solutes to be removed from water may be treated in a seeded microfiltration process, such as pesticides and dissolved organic species sorbed onto dispersed carbon particles. CONCLUSIONS True surface microfilters provide much higher permeate flux rates during microfiltration than conventional macroporous media. These rates are sufficient for the process of seeded microfiltration to be considered. The process may be reliably mathematically modelled by a two resistance in series model due to an aqueous film surrounding the sorption particle and diffusion of the transferring species within the particle. Seeded microfiltration with surface microfilters could be considered as a contacting procedure for the removal of organic species from water and the treatment of contaminated land. References 1. Cumming, LW., Turner, A.D., 1989, Future Industrial Prospects of Membrane Processes, 76-96, Ed. Cecille,L., Toussaint,J.C., Elsevier Applied Science. 2. Rushton, A., Ward, A.S. and Holdich, R.G., 1996, Solid-Liquid Filtration and Separation Technology, VCH, Weinheim, Germany. 3. Awang, A. R. B., 1999, PhD Thesis, Loughborough University. 4. Helfferich, F., 1995, Ion Exchange, Dover Pub. Inc., Mineola, NY, USA.
358
joumalof MEMBRANE SCIENCE
ELSEVIER
Journal of Membrane SCIence 196 (2002) 27-37
wwwelsevlercomllocatelmemscl
Particulate fouling of surface microfilters with slotted and circular pore geometry A.J. Bromley, R.G. Holdich*, I.W. Cumming Department of ChemICal Engmeenng. Loughborough UnwersJly, Loughborough, Le,ceslersh'Te LEII 3TU. UK Received 1 Marcb 2001, received ID revised form 2 July 2001; accepted 3 July 2001
Abstract During Jrucrofiltrauon It IS possIble to obtain a permeate rate equal to that of the clean lIquid permeation rate under certam cnucal condmons. ThIs occurs when a depoSIt does not foul the surface of the membrane and internal deposluon of matenal WIthin the filler does not occur. Surface filters do not possess an internal structure, therefore, the only partlculate foulIng possIble IS that on the surface Expenments in a surred cell compared the surface foulIng of two types of true surface filters. a commercIally available track-etched filter WIth circular pores 10 I1m m dIameter and a filter WIth slotted pores 10 I1m x 420 I1m m sIze. When uSlOg a challenge suspension containing dIlute latex parl1c1es of diameters 15 I1m, and less, the circular pore membrane exhibIted a cnhcal flux at 260 I m-2 h- I. Underthe same condluons, the cnucal flux for the slotted membrane was m excess of 1500 I m- 2 h- I. Surface foulIng was removed by back-flushing for both filters These results IOdlcate that slotted pores are less lIkely to suffer from parncles bndgmg the pores leading to cake depoSItIon, or secondary membrane formahon, and that further development of filters WIth slotted pores of smaller slot WIdths would be worthwhIle. © 2002 Elsevler SCIence B.V. All rights reserved. Keywords Mlcrofiltraton, Pore, Geometry. Slot, Cntlcal flux
I, Introduction
In Dllcrofiltration, the problem of irreversIble membrane fouhng by internal penetration of material withm the filter medium IS well documented. It arises WIth most convenhonal micro filtration membranes because the pore rating of the filter is determmed by the tortuous flow channel through the membrane rather than by the true surface pore opening [I] Hence, a Dllcrofilter rated at 111m may have pore opemngs many microns wider a1lowmg fine matenal to enter, and IrreversIbly foul, the membrane There are several commercIally available rnicrofilters deSIgned to over• CorrespondIng author E·mQl/ address' r g holdlch@lboroacuk (R G. Holdlch)
come this problem, where the pore rating is equal to the pore opemng size Most of them are manufactured by a track-etching process and the nucleopore filter is one such example. Other examples have recently been reported [2] and they all filter by retaining the particles on the surface of the filter, rather than allowing internal filtration. Expenmental evidence suggests that these true surface filters provide high, long-term flux rates compared to the convenuonal "macroporous" type of DllcrofiltratlOn membrane [2] However, foulmg may stdl take place on a true surface filter, due to cake or gel layer formation and because of pore plugging. If the depoSIt is retamed m a cake layer, the filter operatmg condItIons may be altered to Dllmmlse the fouling resistance to permeate flow, e g. a hIgh crossflow rate may reduce the cake layer thickness, or even
0376-73881021$ - see front matter © 2002 Elsevler SCience B V All nghts reserved PIl 50376·7388(01)00573·7
28
A J Bromley et aUJournal of Membrane SCience 196 (2002) 27-37
prevent deposition of a cake altogether. 1111S effect has become known as "critical flux"; which represents a sensitive balance between the permeate rate, operating conditions and the material to be filtered [3-5]. For flux rates below a cntlcal value, there is little fouling because the deposit can be removed by the shear, or other hydrodynamic condlttons. Hence, the permeate rate during filtration is sinular to that found when permeating clean liquid. These conditions represent the "ideal" operating regime for a microfilter. However, critical flux is very sensitive to operating parameters and It is often found that a shght change in the condlttons can lead to a considerable reduction in permeate rate. It is hkely that cntlcal flux is easier to maintain with a true surface filter because, in the absence of pore plugging. the membrane medIUm resistance remains constant since particles finer than the pore Size will pass through the filter. Whereas, a macroporous membrane will experience internal depoSition of the finer particles and pore plugging, leading to a greater hqUld flow through the rema.ming pore channels, which Will cause further fouling. It is possible, therefore, to argue that the opttmum microfilter design is for a true surface filter With a pore geometry that does not easily surface plug. ThiS paper reports work companng a true surface mlcrofilter havmg circular pore openings (CPO) with another true surface nucrofilter having slotted pore openings (SPO). The filters compared m this study have large pore openings for membranes: the CPO membrane was a Mtlhpore 10 ILm filter with an open area of pores equal to 8% of the total filter area. The SPO membrane was obtained from Stork Veco and had a slot geometry of 10 I1m x 420l1rn, With an open area of 5% Nominally, the maximum diameter of a solid particle that should pass through either membrane is 10 I1m. The greater open area of the CPO membrane suggests that, for the same permeate rate, the velOCity of the liquid towards the pore openings Will be greater for the SPO membrane. Thus, the convective velocity of depositing particles WIll be greater for the SPO and It would be expected that the SPO membrane would be more susceptible to surface fouhng, and more dtfficult to maintain a flux rate below the critical value It IS weB known that when the pore and panicle size distnbutions overlap the tendency for membrane fouling is greatest [6], hence, the challenge suspension used in this study was selected to provide these
conditions. However, one of the advantages of a true surface filter is that the finer particles should not deposit within the membrane, unhke a macroporous filter. So, It should be possible to select a surface filter to retam a species of speCified particle size, e.g. bacteria, CrytpospondlUm parvum oocyts, precipitates, and to aVOId the retention of finer material provided a secondary membrane, or cake, does not form. This investigation considers the formation of a secondary membrane on the surface filters with different pore geometry and its influence on the critICal flux. Clearly, if it can be shown that the slotted geometry has advantages over a circular pore, WIth the size of particles used in this study, then there is an mcentive for further research and development to produce finer pore size slotted microfilters that are capable of removing the matenals mentioned earlier. As a surface filter does not exhibit any depth filtration mechanisms it will pass parttcles smaBer than the pore size if a filter cake, or pluggmg, does not occur. Hence, an important parameter m filter performance is the "filter grade efficiency", which IS defined as the fraction of particles (usually by mass) that IS retamed on the filter. Clearly, If a cake forms the grade efficiency Will Improve during filtration, as particles become filtered m the cake or secondary membrane layer. However, if a deposit does not form on a surface filter then the grade efficiency should not alter significantly.
2. Experimental techniques Stirred ceB expenments were performed The design ofthe sttrrer used in the sttrred ceB was based on a cone-and-plate constant shear rheometer. The stirred ceB is IBustrated In Fig. 1. In a constant shear rheometer, the tangenttal shear rate throughout the flUId IS mdependent of radial position [7]. Hence, it IS possible to calculate the umform shear field used ID these experimental studies. The ceB consisted of a clear acrylic column, with a volume of 195 ml. Towards the top of the column were two inlets which could be used to introduce fresh ultra-pure water, or to recycle the permeate stream back to the column. A removable membrane holder was also manufactured from acrylic and polished to a transparent fimsh. The membrane to be tested was located on a perforated
A J Bromley et al /Journal of Membrane SCience 196 (2002) 27-37
29
Adjusting screw
With ---..,1-- Stirrer shallow cone Fig 2 Picture of the Mllbpore track-etched membrane clfCular pore opemngs
oo:::.-r-- Membrane
Permeate
Recycled to column, or collected m receiver
Penstalne pump Fig 1 Schematic representation of the surred cell eqUipment
brass support plate. A tIuck, soft, rubber O-ring was fitted between the filter and the column. The flanges of the column and base-section were bolted together tightly and the compressed O-nng provided a tight seal around the filter. The acrylic stirred cell was finnly supported by a specIally designed clamp stand. Polycarbonate isopore membranes wIth a nom1Oal pore size of 10 ",rn, supphed by MIlhpore, were used. The physical propertIes quoted for this track-etched filter are given ID Table 1. Each 47 mm diameter mem-
With
brane was used as supplied. The data sheet states that the nonnally hydrophobIc polycarbonate surface has been treated with a wettmg agent, polyv1Oylpyrrohdone (PVP) to make the surface hydrophlhc. Slotted filters were fonned from "Veconic Plus" metal screen, supphed by Stork Veco, WIth rectangular slot dimensIons of 10 ",m x 420 ",m. Manufacturer's physical property data IS also gIven m Table 1 Images of both types of filter are provided in Figs. 2 and 3. Polystyrene latex particles produced by the balance swell method (BSM) were used as the challenge rnatenal [S]. A 500 mllatex stock suspensIOn was prepared using filtered (at 0.1 ",m) ultra-pure water (1S.2 Mrl cm conductivity), and stored at room temperature. The addItion of Coulter dIspersant (Coulter Electronics Ltd.) was found to have no sIgnIficant effect on measured parttcle sIze dlstnbutlOn by Coulter Mulllsizer 11 and it was, therefore, assumed that the stock suspensIOn was fully dIspersed. Table 2 proVIdes a parttcle sIze dlstnbutlOn, by mass, of the stock suspension analysed us10g a Coulter MulllSlzer 11. As can be seen, the majority of partIcles were less than both membranes nominal pore size of 10 ",m.
Table 1
PhYSical properties of CPO and SPO membranes
Membrane
Matenal
Isopore track-etched Metal screen
Pore geometry
Pore size
Polycarbonate film
Circular
10
100% D1ckel
Slotted
10 x 420
(~m)
Open area (%)
Thickness (I'm)
8 5
300
10
30
A J Bromley et al /Journal of Membrane SCience 196 (2002) 27-37
Fig. 3 Scanntng electron micrograph picture of slotted nucrofilter screen
A radio spares pressure transducer (1 barg, 0-1 ()() mV) was connected to the permeate line between the filter and peristaltlc pump. The transducer was connected to a Pico Technology 16 bit analogue-
back to the column or collected in a receiver on an electronic balance (OHAUS precision plus). The permeate lme was 6 mm bore slhcone rubber tubing. Each expenment Involved a new test membrane and freshly diluted latex suspensIOn from an eXlsnng stock. A 47 mm test membrane was carefully placed on the perforated support plate. In the case of the SPO membrane, the two sides of the filter were shghtly dIfferent and a microscope was used to ensure that the same side was consistently facing upwards. The column was lowered onto a rubber O-nng and bolted nghtly to the base-section. The compressed O-nng made a seal around the filter, whIch waS clearly viSIble as a thm black lIne. A MI1lipore M111i.Q 185 PLUS water punficatlOn system proVIded ultra-pure water WIth very low levels of ions, organics and parllculates (18.2MS"lcm conducnviry) for all expenments. Ultra-pure water was initially pumped around the system to remove rur. The polystyrene latex stock suspension was sonicated for 2 min in an ultrasonic bath and a further lTIInute usmg a sonic probe dehvenng 70 W power. Then 10 ml
Table 2
PartIcle Size dlstnbutton of challenge matenal Particle diameter (JLm) Cumulative mass undersize (%)
18 12
20 25
31 44
37 50
S4 70
7S 87
10S
130
96
98
A J Bromley et al/JouTMl of Membrane SCience 196 (2002) 27-37
of stock suspension was pipetted into the column, halffilled With ultra-pure water. The stirrer was located m position and the column level made up to the 195 ml mark with ultra-pure water The feed suspension was well-mixed before commencmg an expenment. Most experiments were performed under conditIOns of constant permeate rate, rather than constant pressure. Hence, pressure measurements were used to assess the occurrence of foulmg and critical flux. In most experiments, the permeate was recycled back to the column top to maintain the feed concentration, m the absence of slgmficant cake formation. It also served to maintain the hqUld level. A constant stirrer speed of 800 rpm was employed in all the experiments descnbed. This corresponds to a shear rate at the filter surface of approximately 1200 s-I.1t is roughly equivalent to a cross-flow velocity of 1 ms-I in a 5 mm flow channel filter. Each constant-rate filtration expenment consisted of a series of runs at different flux rates. The penstaltic pump was operated for 10 nun during each run. After a short break, and with the pump direction reversed, the filter was gently back· flushed for 2 mm. The pernieate flow rate was measured between runs using an electronic balance and stopwatch. All the permeate collected was returned to the column Pressure readmgs were logged at 2 s intervals. Samples were collected from the column for particle size analysIs usmg a 10 ml syringe With a 300 mm metal sample tube attachment (1.5 mm bore). Typically, 4-lOml sample volumes were collected from the middle of the column. Ultra-pure water was added to remstate the column level. Constant·rate filtration expenments were also performed without the permeate recycle to the column. In these "once-through" filtration experiments, the permeate was pumped and collected m a receiver, situated on an electronic balance. The permeate flow rate was measured at 1 mm mtervals. The hqUld level in the column was maintained by manually addmg ultra-pure water. The grade efficiency of each membrane was investigated. In this case, gravity drainage was used instead of the pump so that permeate samples, for size analysis, were obtained as close to the filter as possible. The short permeate Ime discharged into a collectIOn beaker, and two valves in the permeate line were used: the lower valve was set to control the flow, whtlst the upper valve openedlclosed the permeate
31
line. There was no pressure measurement, but the control valve provided an mitlal flux rate simtlar to that obtained dunng pumping. The column was allowed to drain and the experiment was stopped when the hquid level reached the stirrer cone. Permeate samples were collected for particle concentration and size analysis. Samples from the column were also gathered at the begmning and end of the experiment for concentration and size analysIs. The size distnbution oflatex partlcles in the samples was measured using a Coulter Multlslzer 11. A sample tube With a 70l1m onfice was used for all analyses. Aliquots of each well-mixed sample were transferred to the electrolyte solution, using a Volac high precIsion nucropipette, enablmg calculation of the particle concentration present in the stirred cell for each size grade.
3. Results and discussion 3.1. Membrane pore sIze
The isopore track-etched (CPO) membrane was examined using optical microscopy. As seen in Fig. 2, the manufacturer's quoted nominal pore sizes are similar to the sizes observed. The manufacturer states that the pore size distnbution is ±20%, giving a pore size range from 8 to 1211m. The scanning electron microscope (SEM) picture of the nickel (SPO) membrane, Fig. 3, shows the consistent width of the slot. The manufacturer's stated slot dimensions are close to those measured. The slot width, however, was not entirely uniform as there is a mimmum value at each end, where the slot tapered shghtly. 3.2. Constant·rate filtration experiments permeate recycled
To compare the filtration performance of the Circular and slotted pores, each membrane was tested in the stirred cell under Identical expenmental conditions. The penstaltlc pump was operated for 10min during each run at a set flow rate, With the permeate recycled to the column top, followed by back-flushing for 2mm at 970Im- 2 h- l • Pressure readmgs were recorded and, since the expenments were performed at a constant-rate, any increase m the pressure drop across the membrane mdlcated membrane fouling.
A J Bromley et al /Journal of Membrane SCience 196 (2002) 27-37
32
250r------------------------------------------------,
a.:
.aE
200
+
OPO membrane
:I':
SPO membrane
CPO membrane
CPO membrane
i ~
j
+
•*
• o
20
40
80
60
100
120
140
Time, mlns. Fig 4 Companson of CPO and SPO pressure dIfferences
Fig. 4 shows the pressure dIfference across the membrane (in mbar) for the CPO and SPO membranes, at vanous permeate flux rates. Flux rate values are dIsplayed next to the relevant data points. The oscIllatory nature of the peristaltic pump cycle is eVIdent m the pressure measurements, which vary between a maximum, and minimum value. Inillally, the CPO membrane showed no sign of fouling at a flux rate of 125 I m-2 h- I . DoublIng the flux rate resulted m a rise m the pressure drop during the 10 mm run, which suggests that the membrane fouled under these conrullons. Movmg from a non-foulIng to a foulIng regIme, after mcreasmg the flux rate, may indicate the existence of a entlcal flux between 125 and 269 I m-2 h- I . In a repeat expenment, there was further evidence of a critical flux between 130 and 262Im-2 h- 1 for the CPO membrane. Followmg a back-flush, and for the experIments at permeate rates <805 I m-2 h- I , the pressure drop returned to the clean water value of 5 mbar, which suggests that the fouling was reversible. This behavIOur IS consIstent WIth foulIng by cake layer formation rather than pore plugging, where back-flushing IS usually less successful m returning the membrane to Its origmal state.
With
penneate recycle
Studymg the feed suspensIOn m the column provIded a further insight into the fouling of the membrane. Since any partIcles passing through the membrane were returned back to the column, It is to be expected that the challenge size rustnbution would change sigmficantIy dunng the expenments when a surface cake formed. FIve samples were taken from the middle of the column at vanous tImes and analysed us10g the Coulter MulllSlzer IT. Particles m the recycle pIpe at the tIme of samplIng, were not conSIdered SIgnificant because the recycle volume was small and the residence lime low. Table 3 shows the change in the number of 2, 5 and 8 J1m partIcles m suspension during the set of Table 3 Particles remammg concentration Sample
In
Descnptlon
suspension as a funcnon of
irubal number of parbcles (%) 2 fUll
2 3 4 5
Run Run Run Run
2 4 6 8
(end) (end) (end) (end)
initial
5 fUll
8 fUll 867 764 722 532
676
706
405
585
169 122
353 257
feed
A J Bromley et aLlJournal of Membrane SCience 196 (2002) 27-37
experiments. The rate at whIch partIcles were deposited on the membrane does appear to be related to particle sIze. At the end of the experiment at a flux rate of 2080 I m- 2 h- I, there were only 12% of 2!lm partIcles stIlI m suspensIon compared to 53% of 8!lm particles. Superficially, it is surprising that there should be less reduction in larger particles than smaller ones. However, segregation of particles WIthin a deposited cake is well known [9] and the size dlstnbutlon data show that stirring was more effectIve at removmg the larger particles than the smaller ones from the tlnn cake layer deposIted on the membrane. It is noticeable that the percentage of all partIcles reduced as the flux was increased, indicatmg that the cake thIckness increased WIth increasing flux rate. Using this membrane, the pressure drop mcreased whenever the stirred cell was operated at a flux rate of 2691 m-2 h- I , or hIgher. The pressure drop across a surface filter will increase whenever pores become obstructed. It is clear from the particle size analyses that large numbers of particles were deposited on the membrane surface during the filtration runs. As the challenge matenal was substantially smaller than the pore sIze it IS lIkely that pore bridging occurred, leadmg to the formation of a cake layer
33
In contrast, the SPO membrane showed no appreCIable SIgn of foulIng during the eight runs; this is evident from the data illustrated m FIg. 4. The transmembrane pressure increased only shghtly between the operating fluxrates of 125 and 17241 m- 2 h- I and there was no SIgn of a critical flux for the SPO membrane. On completion of the expenments, the number of particles in the feed suspension of sIzes 2, 5 and 8 !lm were withm 75-80% of the Imtial values. The SPO membrane showed no sign of fouhng at a flux rate of 17241 m-2 h- I• In a further experiment, the flux rate was gradually mcreased to 44,000 I rn-2 h- I, over 14 stages, to investigate the eXIstence of a critical flux for this membrane. All other expenmental conrutlons were as stated previously. The performance of the membrane across the flux range 130-44,000 I m-2 h- I IS presented in FIg. 5. The pressure plot has been spht mto four zones A, B, C and D. Zone A was essentially a repeat of the experiment featured in Fig. 4. There was no suggestion of foulmg dunng these eIght runs. When the flux rate was increased from 1500 to 87oolm- 2 h- 1 over three runs (zone B), there was a 2 mbar increase in the pressure drop. In an attempt to elucidate the cntlcal flux, the flux rate was mcreased consecutIvely
Fig 5 Pressure differences across the SPO membrane at elevated flux rates
A J Bromley et ai/Journal of Membrane SCience 196 (2002) 27-37
34
to 18,300, 28,000, and 43,800lm-2 h- l • The results from these Iugh flux runs can be seen in zone C. There was eVidence of fouhng as the pressure drop increased to 55 mbar at the highest flux rate, and rose during each run. The pressure drop consistently returned to the clean water value after back-flushing, which shows that the fouling was not permanent. On reducmg the flux rate to 960Im-2 h- 1 (zone D) the clean water pressure drop of 5-7 mbar, was once agam maintamed, suggesting that the previous fouling was completely removed.
3.3. Constant· rate filtration experiments permeate collected The filtration performance of the CPO and spa membranes was also assessed usmg the "oncethrough" experimental set:up. In this case, the column volume was mamtamed by the addition of ultra·pure water. The removal of the recycle meant that as the experiment progressed there was less challenge matenal avrulable to foul the membrane. The other operating condittons were as stated prevIOusly. The performance of both membrane types can be seen m Fig. 6. The CPO membrane first showed signs of fouhng at a flux rate of 530 I m- 2 h- I ; when the pressure drop
increased steadily as the membrane fouled. This result differs from the expenments descnbed m Fig. 4, where there was significant fouhng at 269Im-2 h- l . However, cntICaI flux is not unique to the membrane itself, but depends upon filtratton condtttons and, during these expenments, the concentratton of fine pamcles WlthlO the challenge would be less than m the recycled experiments. It is, therefore, difficult to define a single cnttcal flux for a pamcular membrane when studying different experiments. The success of the back-flush (at 120Im-2 h- 1 flux rate) at clearing the membrane pores again suggests reversible fouling. The spa membrane showed no sign of fouling at a flux as Iugh as 17721 m-2 h- I , as the average pressure drop during a filtratton remained constant. This supports the existence of a much higher cntical flux for the slotted membrane. 3.4. Grade effiCIency
The results from the prevIous experiments demonstrate that the slotted pore geometry IS less prone to fouling than the circular pore, for similar shear and other filtration conditions. However, good permeate flux perfonnance would also be provided by a filter that did not retain any particles. Hence, It was
250~-----------------------------------------------,
• CPO membrane
..:
~
200
E
8'c I!!
x SPO membrane
o
Flux rate. I m4 hO'
150
:! '6
!
:s
100
'"'" I!!
11.
50
20
40
60
80
Time, m1ns. Fig 6 Companson of CPO and SPO pressure differences Without penneate recycle
100
A 1. Bromley et aLl Journal of Membrane SCIence 196 (2002) 27-37
35
. ..
100~~__~~~~~~~~~-------------------------,
~
o
U
..
+
~
"c:..
+ +
eo 60
+ +++ + +",,++ .. +
+ ++.......... +~
i
§
40
8
20
~
o+---~--~----~--~--~----~--~--~----~--~ 18 6 10 12 14 20 o 2 4 6 16
Particle diameter, I1m. + permeate sample 1
x permeate samples 2 - 5 (average)
Fig 7 Grade efficiency of CPO membrane
important to investigate the particle removal efficiency of the filters. Clearly, a filter providing cake filtratIon wIll show high removal effiCiency because of the formatIon of a secondary filtenng membrane; so care has to be exercised m both the experimentalmvestigatIon and in the interpretation of the data. For these experiments, a single pass was used; the column dramed under gravity and samples of penneate and challenge suspensions were taken for concentration and particle size measurement as described earher. Five penneate samples were collected and analysed using the Coulter MultiSlZer 11. The grade efficiency ofthe CPO membrane can be seen m FIg. 7. There are two sets of data presented: the imtIal penneate sample (taken over the filtration of the first 43 ml) and an average of the subsequent four samples (43-184 ml filtered). Consldenng the first penneate sample, it is surprising that at least 60% of partIcles 1.4-8 I1m were removed by the 10 I1m membrane. Complete cut-off appears to take place at approximately 8 I1m. After the initial 43 ml had passed through the membrane, particles of all sizes were removed With an efficiency of 99% or higher. This result supports the conclusion that a secondary membrane (or cake) quickly fonned.
The grade efficiency data for the SPO membrane IS shown in Fig. 8. Once agam, the mltIal collection efficiency (40ml filtered) and average collectIon efficiency for the remainder of the expenment (40-179 ml filtered) are presented. Collection efficiency increases WIth partIcle Size, but the data is scattered WIth no obvIOUS trend WIth respect to volume filtered. This result IS consistent WIth a filter on wluch a secondary membrane dId not fonn. The SPO membrane does not SIgnificantly remove smaller particles, which were 10 the majonty in the challenge matenal. Less than 30% of particles below 5 I1m were filtered. An approxImately fitted curve to the data mdlcates that the cut-off size for this membrane type IS around 1211m which IS consistent WIth the maximum slot width measured usmg the SEM images (Fig. 3). Column samples collected at the end of each grade efficiency expenment were analysed to study the concentration of challenge suspension particles during the filtration. When a membrane is successfully filtering, the concentratIon of the challenge should be greater than that of the initial feed concentration as particles are retained and, poSSibly, redIstnbuted by the shear conditions. The initial particle concentration in both grade effiCiency expenments was approximately
A J Bromley et aUJournal of Membrane SCience 196 (2002) 27-37
36 100
oi
BO
,::
x
. ;: ..
u c: 60 U c: 0
;:
+
• *'
40
u .!!
'0 0
+
+
+
• x
x x + x
20
0 0
2
4
6
S
10
12
14
16
IS
20
Particle diameter, I'm. + permeate sample 1
x permeate samples 2 - 5 (average)
Fig 8 Grade efficiency of SPO membrane
8.4 X 106 particles per ml, the final concentrations in the retained suspensions for the SPO and CPO expenments were 10 X 106 and 61 x 106 , respectively. Hence, it is evident that filtrallon occurred for both the SPO and SPO membranes, but was more efficient in the case of the CPO because of the formation of the secondary membrane.
4. Conclusions Mlcrofiltratlon membranes With two different types of pore shape have heen 1Ovesllgated: one had Circular pores 10 I1m m diameter wlulst the other had slots 10l1m x 420l1m m size. Both filters had s10ular areas open for filtration and can be descnbed as surface filters in that their pores passed directly through each filter; neither filter exlublted depth filtration charactensllcs. The circular pores formed a surface cake at fluxes above 260 I m- 2 h- I and tlus cake could be removed by back-flushing. In contrast, the slotted pores showed no sign of suffering from cake deposition at fluxes as high as 1500 I m-2 h- I • Hence, under these operating conditions, the critical flux of the slotted microfilter is much greater than that of the Circular pore filter. The slotted pores were effectively cleared
by back-flushing when cake depoSition was observed at fluxes greater than 8000 I m-2 h- I • This expenmental study shows that, compared to circular pores, slotted pores are less likely to promote particle bridg10g over the pores lead10g to cake deposition or secondary membrane formation. Slotted filters are, therefore, more likely to allow the passage of partJcles smaller than the pore size. The present slot size of 10l1m is too large for many useful applications so work is now in progress to develop smaller Width slots, where these separation advantages can be applied to 10dustnal processes of commercial 10terest
Acknowledgements The authors wish to acknowledge the financial support of the Eng10eenng and PhYSical Sciences Research Councd (EPSRC) for the provIsion of a studentshlp for AlB.
References [11 RG Holdlch, IW Cummmg, ID Smllh, Cross-flow 1lI1ClOfiltration of 011 ID water dispersIOns usmg surface filtration with Imposed flUId TOlalIon, J Membr Sel 143 (1998) 263-274.
A I Bromley e/ aUlournal of Membrane SCience 196 (2002) 27-37
[2] S KUlper. CJ M van RIJn. W NIJdam. M C Elwenspoek, Development and applications of very high flux tnlcrofiltratton membranes. J Membr Sel 150 (1998) 1-8 [3] R W FIeld. D Wu. J A Howell. B B Oupta. CntIeal flux concept for mlcrofiitrabon foulIng. J Membr. Sel 100 (1995) 259-272 [4] H L1. A 0 Fane. HO L. Coster. S Vigneswaran. An assessment of depolansatton models of cross-flow mtcrofiltratlon by direct observatIon through the membrane. J. Membr Sci 162 (2000) 135-147 [5] D Y Kwon. S VIgneswaran. Influence of particle SIZe and surface charge on cntlcal flux of cross-flow rrucrofiltrauon, Water SCI Tech 38 (1998) 481-488
37
[6] E S Tarleton. RJ Wakeman. UnderstandlDg flux dechne ID cross-flow ffilcrofiltratton Part I Effects of partIcle and pore SIze. Trans IChemE Part A 71 (1993) 399410
[7] R W. Whorlow. Rheologlcal Tetbraques. 2nd Edition. Elhs Horwood, ChIchester. UK. 1992 [8] J W OoodWID. J Hearn. C C Ho. R Ottewdl. StudIes on the preparation and charactensauon on monocbsperse
polystyrene lattIces, Coli Polym SCI 252 (1974) 464471 (9) NJ. Blake, I W Cummmg, M. Streat. Predlctton of steady state cross-flow filtration usmg a force balance model, J Membr SCI 68 (1992) 205-216
Available online at www.sciencedirect.com
MINERALS ENGINEERING
8C •• [email protected]
PERGAMON
Mmerals Engmeenng 16 (2003) 121-128 11ns arttcle
IS
also available onbne at
www elseVler com/Jocateirmneng
Clarification by slotted surface microfilters "* R.G. Holdich ., I.w. Cumming,
s. Kosvintsev, A.I. Bromley, G. Stefanini
Department of Chemical Engmeermg, Loughborough Umverslty. Leicestershire LEl1 3TU, UK
Recetved 10 July 2002; accepted 22 August 2002
Abstract Metal mlCTofilters, WIth a slotted pore geometry, have been used to filter conoIdal suspensIOns USlOg crossflow filtratIOn to provide shear at the filter surface In order to reduce the deposition of solIds on the filter The slot widths tested varIed from 10 to 2 J.1m and the slot length was about 400 J1Ill To assess the degree of pore blockage, challenge suspensIOns With particle diameters slImiar to the pore dIameters were used For companson, a 10 J.1D1 pore Width filter with CIrcular pores was also tested Under Identical conditions, the slotted pore geometry dId not foul as badly as the CIrcular pore filter. FIltration fluxes of up to 9000 I m-2 h- I were poSSIble WIth mmlIDum eVidence of filter blockage or fouhng However, WIth a 21lJIl slot width foulIng was eVIdent at filtration fluxes as low as 200 Im-2 h- l • An average filtration flux of 3000 Im-2 h- 1 could be mamtamed WIth the 2 Ilm pore WIdth filter by mcludmg a backflush, but It was Important to exclude air from the filter, otherWIse the capillary pressure to be overcome m order to remove the air from the filter became excessive This new mlcrofilter membrane deSIgn does not suffer from mternal pluggmg of the filter matnx, because It does not have an mternal structure, and can be made m to filters contamIng flat sheets or selfsupportmg tubes. The filter has many possibilIties m fine particle processmg and may be used to recycle clean bquld back to an industnal process © 2002 Elsevler SCIence Lld All nghts reserved Keywords Fdtrabon. Fme particle processmg, Hydrometa11urgy, Leachmg, Recychng
1. Introduction In many mmeral processmg and hydrometallurgIcal operatIons there IS a reqwrement for clanficatlOn of a hquor, where the intention IS to obtain a clean liquid stream, either for mscharge to the environment, or for recycle within the process Oarifying filtratIOn can be achIeved usmg sand filters, or other depth media, candle filters, filter leaves, conventional cake filters WIth filter aIds and filtenng screens However, most of these operations are batch, where the filter must be taken off-line for clearung, or a duplex system IS required, and two sets of eqwpment are needed to ensure contmuous operation Contmuous clarifying filtration can be achIeved usmg crossflow microfiltratlon, in whIch the suspension to be filtered flows across the filter membrane, rather than perpendIcular to it. Thus, the crossflow shear at the surface prevents cake deposition and high clanfying filtration rates are maintained. ThIS type of filtration IS ~ Presented at So/lt/-LIqUId SeparatIon '02, Falmouth, UK, June
2002 • Correspondmg author E-mall address r g holdlCb@lboro ac uk (R G Holdtch)
appropnate for finely dIvided matenals at low concentration Typically, particle dIameters m the collOIdal range; less than 10 J1ffi are microfiltered. However, a mlcrofiltratlOn process usually needs to be' part of a f10wsheet with at least one other fonn of sohd hquid separation reqwred for the final removal of a cake. An example of how crossflow 11l1crofiltratlOn can fit mto a metallurgIcal f10wsheet IS gIven m FIg I (Metal Fmishers ASSOCiatIOn, 1992). The example is for metal effluent treatment, where the condItioning chemicals cause preCIp,tates to fonn and these are thickened m a conventional thickemng tank. The overflow from the thickener can contain finely dIvided preCIp,tates and 011 covered partIcles, which renders It challengIng for disposal or reuse. Hence, the overflow IS treated m a crossflow 11l1crofiltratlOn Clrewt where the filtrate can be dIScharged, or recycled, and the retained suspension by the microfilter is passed back to the tluckener. The underflow from the thickener is treated in a conventIonal filter and the cake dISposed of, or recycled. A S111l1lar scheme could be enVIsaged for recyclmg water from mineral trubngs ponds; usmg a 11l1crofilter to obtain clean liqwd to recycle back to the mmeral processing operation
0892-68751031S - see front matter © 2002 ElseYler Science Ltd All nghts reserved
dOl 10 IOI6IS0892·6875(02)00176·0
R.G HoldlCh et all Mmerals Engmeermg 16 (2003) 121-128
122
''''''''''
10 to 1000 ppm heavy metals
width rather than a true pore opening size. In filtratIon tests, at low suspended solids content, this membrane dId provIde greater than 99"10 retention of partIcles that were 3 I'm in dIameter (Rushton et al , 2000) but the method by which it achIeves thIS retention is not by slevmg at the surface-instead It relies upon deposItion of fine partIcles within the matrix of the membrane ill much the same way as a depth filter. At higher conoentratIons of suspended material a depOSIt, or secondary membrane, can form on the filter in which case the filtratIon performance is dependent upon the suspended partIcles and not the membrane (Holdlch et aI, 1996). However, at low concentratIon the depOSItIOn of partIcles illSlde the filter leads to long-term decay m the filtratIOn flux: a clean microfilter wIll prOVIde filtration flux rates of several 1000's Im-'h- I (htres of filtrate per square metre of membrane area per hour), whereas an mtemally fouled filter often has a flux rate of less than 100 I m-' h- I. Clearly, the hIgher the flux rate the more economically VIable a nucrofiltratIon process WIll be Hence, research into filters that do not mtemally foul is underway at several locatIons (KUlper et aI, 1998) Surface nucrofilters do not rely on depth filtration mechanisms The pore rating of these filters represents the pore width; typIcally 1-10 f1ID Such filters are commercially avaIlable for laboratory use, often manufactured by a track·etch technique. However, surface filters WIth circular pores can stIlI plug WIth solIds lodged m the throat of the pore. Slotted surface nucrofilters, where the slot WIdth IS stIlI 1-10 I'm, but the slot length can be 400 f1ID, or more, excludes all non-deformable particles bIgger than the slot WIdth from the filtrate and, under carefully controlled CIrcumstances, can also exclude deformable drops. The mam advantage of the slot geometry is that the pore IS less prone to blocking than a CIrcular opemng: a smgle, roughly spherical, partIcle SIttIng in a CIrcular pore opening would block off all the flow from that charmel. However, It is most unlIkely that
Preclprtatnv
or concllllOnll'l9 agents, pH. ate.
many % wlw solids
waste water
equalisation feed tank
"'-
<01 ""'" heavy metal.
todran
FIg. 1. Flowsbeet for heavy metal recovery mcludlOg crossftow nucrofiltration
Microfiltration by conventional membranes has had only Iinuted success for clanficatIon in nunerals processmg because the flow path for the filtrate, and the pore opening of a flow channel, may well be an order of magnitude bIgger than the filter ratmg. Hence, particles become lodged Wlthm the nucrofilter matnx, causmg long-term filtrate flux declIne that IS often irreverSIble. This pnnClple is Illustrated ill Fig 2, where the pore sIZe dlstnbutIon and a photograph taken under a scanning electron mIcroscope (SEM) are Illustrated. The pore sIZe distnbutIon was taken by a standard technique, ASTM F-316-86 (1986), wInch saturates the membrane WIth a lIqUId and then mcrementally mcreases aIr pressure m order to blow the pores free of lIqUId from the filter. The increase in the flow of aIr through the membrane, can be related to the increase m the number of pores open to air flow and, hence, the pore size distnbution of the membrane Thus, the filter Illustrated has a modal pore SIze of 3 5 f1ID and no pores bIgger than 5 5 f1ID. However, the SEM clearly shows pore opemngs at the surface of the filter that are 50 f1ID, or bIgger. Hence, the ASTM test provides an "eqUIvalent" pore
~ 100 C co .c 80
-
30 ~
25
0
~
200.
*. f1!
'0
60
C>
15
.2l
f1! 40
10
c~
0
a. !D
~
"E
"
()
20 0 2
5
25
5
Pore size,llm.
Fig 2 Photograph and pore sIZe cbstnbutton of commerctal nucrofiltratlon membrane
a; c
.2 0 '0 ~ 55 LL
123
R.G Holdtch et all Mmerals Engmeermg 16 (2003) 121-128
Membrane V4
V3
Filtrate
Hiac Royco 051020
Feed tank Fig 3 Surface IDlCfofilter WIth a 3
~
slot Width
a slotted pore could be blocked in a sImIlar way. An example of a true slotted surface pore is Illustrated m FIg 3. The ava!lab!llty of robust lugh flux slotted mlcrofilters, that provide long-term stable flux rates, has led to the posSlb!llty of combinmg microfiltratlOn together wIth other umt operatIOns appropnate to nuneral processing, hydrometallurgy and effluent treatment. For example, It IS possIble to retam finely dlVlded Ion exchange, or adsorphon, mewa Wlthm a flow CirCUIt (Holdlch et al., 2000) 2. Experimental The slotted filter mewa used was m the form of a regular array and IS Illustrated m FIg 4 In the expenments described in this work, typIcal slot length was 400 IIl1l and slot WIdth vaned between I and 10 IIl1l The substrate slotted metal med,a was suppbed by Stork Veco, of the Netherlands, and the slot wIdth was reduced to sizes between I and 10 IIm by a process developed Wlthm the laboratory at The Department of ChemIcal Engmeenng, Loughborough University
Fig 4 Array of slots used
lR
V2
the filters
Fig 5 Crossftow expenmental eqUlpment
(Bromley et al ,2002) In all of the experimental tests, the challenge suspensIOn was a dIspersion of latex parncles with a mean size very close to the pore SIZe, or WIdth, of the nucrofilter. It IS well known that these conwtlOns proVIde the greatest foulmg of the membrane (Tarleton and Wakeman, 1993) and these condItions were selected as a severe test for the membrane A standard techmque was used to make the partIcles withm the laboratory (Goodwin et ai, 1974) A crossflow filtratIOn flow loop was used for the tests, WIth onIme partIcle sIze momtonng by a Hlac Royco model 346BCL. The experimental eqmpment IS Illustrated schemahcally m FIg. 5 The membrane configuranon used was a flat sheet, WIth membrane dimensions 25 by 120 mm, WIthin a flow channel 44 mm WIde and 8 mm lugh. The crossflow velOCIty was 0 25 m S·I, whIch corresponds to a flow Reynolds number of about 3600 In all expenments, the pressure dlfferennal between the feed and filtrate side of the membrane was less than 0.1 bar but, for the purposes of companson, all the filtrate flux data at constant pressure was normabsed to a flux rate at 0.1 bar. For the backflushmg expenments the filter area was apprOJumately 4 cm' and I bar was reqmred to backflush the 2 IIl1l filter, but only 0.2 bar was needed for the 10 IIl1l filters BackflusIung was effected by reversmg the flow through the membrane every 30 s for a duratIon of 0 6 s. A set of expenments were designed to test the stable flux rate that could be achieved WIthout foulmg on the membrane and WIthOut backflusIung Tlus IS called 'cntical flux' operation, whereby control of the flux rate below this threshold value w!ll result m nunlfnal foubng on the surface of the membrane because the crossflow shear WIll prevent deposItion. The tests also conSIdered the filtratIOn grade effiCIency, or r".)ectlOn, of the membrane and how tlus compared to other methods to
124
R.G Holdtch et all Mmerals Engmeermg 16 (2003) 121-128
flux rate of 125 I m-l h- I. However, on increasing the flow to 269 Im-lh- I, for the cIrcular pore membrane, the pressure difference reqUIred to maintam this flux increased WIth filtratIOn lime. Hence, the membrane suffered foulmg. At all higher fluxes the cucular pore membrane fouled and the rate of fouhng mcreased wIth flux rate. However, the slotted pore filter did not show any sigruficant mcrease in pressure, or fouhng, even at flux rates of 1724Im-l h- l . For the slotted pore memo brane only, further tests at much hIgher filtrate flux rates were performed and these are reported m FIg 7. The pressure plot has been split into four zones A, B, C and D In Zone A, there was no indIcatIOn of foulmg during these eight runs When the flux rate was in· creased from 1500 to 8700 I m-l h- I over three runs (Zone B), there was a 2 mbar mcrease in the pressure drop In an attempt to find a crillcal flux, the flux rate was increased consecullvely to 18200,28000 and 43 800 I m-l h- I. The results from these high flux runs can be seen in Zone C. There was eVIdence of fouhng as the pressure drop increased to 55 mbar at the highest flux rate, and rose during each run The pressure drop consistently returned to the clean water value after backfluslung, whIch shows that the fouling was not permanent On reducmg the flux rate to 960 Im-lh- I (Zone D) the clean water pressure drop of 5-7 mbar was once agam maintamed, suggestmg that the prevIous foulIng was completely removed. Clearly, high flUId flux rates would be expected if no actual partIcle retenllon occurred. Hence, It IS unportant to consIder the filtratIon grade effiCIency curve to check for filtration For the 10 IUD slotted pore filter the grade effiCIency curve is shown m FIg. 8 The grade efficiency IS deduced by companng the concentratIOn of particles challenging and passmg the filter and care must be exercised when it IS measured
charactense the filter slot WIdth Finally, If foulmg could not be aVOIded, tests to deduce the filtration rate as a function of tune, or pressure, were assessed and back· flushing to remove the depOSIt from the surface of the filter.
3. Results and discussion The range of partIcle and pore sizes consIdered, 10-2 IUD, cover the region where increasmg collOIdal mter· aCllon forces would be expected to occur Therefore, as the partIcle and pore diameters decrease the resistance to filtratIOn, or fouling condItions, would be expected to mcrease. The results of filtration WIth a slot WIdth of 10 IUD are illustrated m Fig. 6. In thIS instance, and because the flux rates were hIgh, the filtratlOns were conducted at constant filtrate flux rate and the pressure drop required to mamtam the flux was momtored Control of the flux rate was achieved by suclang the permeate through the filter using a poslllve dIsplacement pump This pump was used to regulate the flow of filtrate In these tests a companson of the filtration of the slotted filter geometry WIth a Circular pore geometry was achieved; the CIrcular pore filters tested were commerCIally available track etched Isopore membranes supphed by Mllhpore The pore width of the Isopore filters was the same as the slot width: 10 IUD and the challenge suspension had a mean diameter close to thIS value In FIg 6 the abbreViatIOns CPO and SPO are cIrcular pore and slotted pore membranes respectIvely and the flux rates (m Im-lh- I) are marked on the graph, thus flux rates from 125 to 2080 were tested. FIg 6 Illustrates the concept of "cntical flux": for both the CIrcular and slotted pore membranes, there was no measurable increase in pressure chfference WIth lime when filtenng at a
250r----------------=-=-·
1!E 200
!
• CPO membrane , SPO membrane
~
CPO
150
533
."
l'! 100
iil ~ 0..
i•
,•i
•
50 125
o
-
864
20
40
60
80
100
-1724
120
140
Time, mlflutes Fig 6 Filtrate ftux from 10 JUl1 pore
SiZe
filters slotted and Circular pore geometry
RG Holdlch et of I Mmerals Engmeermg /6 (2003) 121-128
125
80
B
A
~
~'" 60
43824
_I
i~ ~ 40 '0
~
~ 20
0.
0 ...
C
~2071
Flux rates (I m 2 h")
~/\~--.
................ ,... ~ "" "'I! 134
397
o
768
40
8658.W
1352 1574 3608
80
160 120 lime, minutes
:
. ,
...
963 240
200
Fig 7 The pressure ddTerence reqUIred for high filtratton rates With slotted pores
100
'" 80 0
'"m60 0
•
0
lE
5'" 40
ti
~
•
820 0 0
2
4
6
8 10 12 14 PartICle diameter. microns
16
18
20
Fig 8 ParbcJe rejectJ.on curve for 10 JUTl slotted filter
The formation of a secondary membrane, or foulmg layer, wIillead to hIgher grade effiClencles than are truly provided by the filter; hence, it IS important to measure the grade effiCIency at a low concentratIOn of solids, quickly and at the start of a filtration. Under the test condItions, particle retention at the nommal pore WIdth of 10 J1Dl was orIly 60% and an extrapolated 100"10 retention occurred at about 12 I'm Slotted m1crofilters with modal pore sIZes of 5 and 2 J1Dl were tested: both by Image analysIs of the slot wIdths and in the crossflow filtration flow rig illustrated ID FIg. 5. The resrIltlng pore sIZe dlstnbutlons for both filters, from the Image analysis, are dlustrated in Fig. 9. The correspondmg grade effiCIency curves for these filters are shown ID FIg. 10 Agam, the 100% retentIon sIZe is m excess of the apparent maxImum pore size of the filter ThIs could be attnbuted to the nature of the filter test by Image analysIs: a two rumenslonallmage is
140 120 2 micron nominal 1;' 100
5 micron nominal
/
'\..
1660 :J
I
•
5f60 ~
u.
I
40 20 0 0
1
2
3 4 5 6 Slot diameter, I'm
7
8
Fig 9 Slot WIdth dlstnbullOR for 5 and 2 J.1m slotted filters
taken under the microscope and analysed for the number of slot WIdths of a given Imear drmenslOn. However,
R.G Hold
126
10 ~80
E 800
~60
\
water backflush
of E .2 600
~
AJr backflush
g
'" 40
Q)
i 20
1!!
~
(!)
400 200
o 2
3
4
5
6
7
8
9
10
Particle diameter, J.lm Fig 10 Grade effiCiency of the S and 2 IJIII slotted filters
In-8oo0 IIm! I O~~~~~~==~ o
1000
2000
3000
4000
5000
Filtration bme, s Fig 12 Filtrate mass dunng backftushtng with filtrate and then air
dunng filtratIOn the flow field through the slot is three dimensIOnal, and not two as IS the unage analysis, hence a slight bendmg of the slot would permit the passage of a particle bigger than the two dunenslOnallmage. The filtrate flux performance of the filters is Illustrated m Fig. 11, which shows the flux decay, at constant pressure dIfference across the filter of 0.1 bar. The rate of decay is a strong functIOn of the partlcle concentratlon, wluch IS marked on the figure m terms of particles per m!. Clearly, cntIcal flux operatIOn is not apparent in this figure In fact, the critical flux value for the 2 J.lffi filter was found to be 180 1m-z h- I , which compares poorly with a flux rate of about 8000 Im-'h- I for the 10 J.lffi filter (Fig. 7) However, It IS expected that the collOIdal forces, and fouling, would become much greater with the finer particle SIZe. The critIcal flux appears to be strongly dependent on the collOIdal sIZe of the particles. For practlcal use, If critical flux is not viable for mamtaining an acceptable flux rate, an alternatIve means of cleanmg or mimmismg the deposition of partIcles IS reqwred. Backflushmg a fluid through the pores 8000r-----;:::=~~:::;=:::;-]
o n-200 l/nIl, • n- 580 !1mI '" 0-1400 IIrn!
7000
;f
6000
E
-5000 $ !!! 4000
.ll 3000 <:
g ~ u:
2000 1000
o .t-..-..,.......-:::::;::::~~........J o 500 1000 1500 2000 2500 3000 3500 Filtration time,
S
Fig H. Filtrate flux decay at three concentrations of sobds on 5 J.UD fiJter
of the filter IS fairly straightforward and is often found to be suffiCIent for this purpose. FIg 12 Illustrates the success in mamtammg a constant filtratlon flux using filtrate backflushed through the membrane and air, when filtenng With a 2 llm slotted filter. It is clearly eVident on Fig. 12 that backflushing with air does maintam a constant flux, but conSiderably less than that obtained when backfluslung With filtrate (water) This can be explamed by the need to remove the air from the membrane before filtratIOn can recommence. The capIllary pressure reqwred to displace air from a 2 J.lffi pore IS Significant, for two mfimtely long rods with a slot between them the capIllary pressure IS as follows: d= 2y M' where d is the slot Width, y is the surface tension and M' IS the applIed pressure The equatIon IS stnctly valid for mfinitely long rods, but the end effects should be small. The usual surface tensIOn of water is 72 x 10-' N m-I, but m these tests the colloids were stabIlised by surfactant solution and this lowered the tension to 40 x 10-' N m-I. This implies that the pressure reqwred to diSplace air from a 2 J.lffi slot is 04 bar. Hence, it IS lIkely that when backfluslung With air the pores at the finer end of the dlstnbulIon did not become free of rur, thus preventing the resumption of filtrate through them It IS noticeable that backflushing With filtrate, aVOIdmg the mgress of air mto the membrane, did not suffer from thiS effect. The average filtrate flux rate when backftushing With filtrate was 3000 I m-Z h- I . When a depOSit is permitted on the slotted membranes the filtrate rate decays m the fashIOn demonstrated by Fig. 11. At some stage in the decay process the depOSit Will become sufficiently great such that a cake filtratIon mechanism fits the data. This is Illustrated m Fig 13, where penneate volume is loganthmically plotted agrunst tune. Usmg the standard cake filtration
Cake filtration
penod
... '
05+-----.-__- - r -__- - r -__~~ 15
20
25
30
127
~
~
.
5
2
R.G HoldJch et all Mrnerals Engmeermg 16 (2003) 121-128
______________________
35
Loganthm of filtrahon bme Fig 13 llIustratton of decay penods dunng filtration process.
equatIOns, the gradient on thIs loganthnuc plot should be 0 5 during the cake filtralton penod. It IS evident that there IS a period in which pore flow restnctlOn occurs first, followed by cake filtration Dunng the stage before cake filtration it is possible to mathemaltcally model the flow restrictIOn stage usmg a model based on slevmg (Kosvmtsev et al., 2002). The modellmg, and data i1. lustrated in Fig. 13, suggest that c1anfymg filtratIOn takes place at a concentration too low to Immediately fonn a cake hke deposit on the filter. Hence, the filter design IS Important m order to nnnurnse membrane fouhng. Dunng foulmg, particles must be depositmg on the surface of the filter slots, but tlus does not completely plug the pore and prevent any further filtrate flow However, It does prOVIde a surface on wluch other parltcles can attach; leading eventually to a cake deposit when the enltre filter surface IS covered with particles and a relatively poor filtrate rate results. Th,s sltuaIton can be aVOided by operatmg below cntlcal flux condlltons, If the pore and particle size is suffiCiently large, or by backflushmg the particles away from the slotted pore
robust filter WIth high flux rates and resIstance to fouling is reqwred. Novel mlcrofilter membranes WIth pores m the fonn of slots are now avadable to provide high flux rate filtraltons of finely suspended matenals. Slot WIdths are WIthin the range of 2-10 IIm and these WIll exclude particles with diameters greater than thiS from entenng the filtrate. Thus, particles wlthm the colloidal range are filtered. It is pOSSible to maintain a very high flux rate when filtenng WIth the 10 !lID slots, a cntical flux rate of several thousand I m-' h- I was observed, before appreCIable depOSition of matenal on the filter occurred. However, the cnltcal flux rate of the 2 IIm filter was below 200 I m-' h- I. Hence, the collOidal and convectIVe drag forces towards the membrane at 2 IIm appear to be much stronger than the force from the crossflow shear at the membrane surface Backflushing the 2 !lID filter did mamtain a good average filtrate flux rate' approXllIlately 3000 Im-' h- I In all cases, the slot width observed by Image analYSIS was found to be less than the maXllI\um particle size passed by the filter. The challenge particles were not defonnable, so the slot must posses a hnear dimensIOn greater than the two dunensional representalton of it obtamed by Image analysIs Work is contmwng to reduce the slot Width further, to enable filtralton of submicron SIZed particles WIth high retentIOn efficlencles. The stable and robust nature of the metal filters opens up new pOSSlbdlltes for combmmg nucrofiltralton and other operaltons, such as ion exchange and adsorplton, Wlthm a smgle process vessel as a means for process mtenslficatlOn The filters providmg a filtratIOn stage and actmg as flow dlstnbutors wlthm the Ion exchange, or adsorplton, media The mam advantage of thiS arrangement bemg the very high superfiCIal area for f1wd flow compared to conventional packed column operaIton. Hence, overall flow rates could be much greater and the ion exchange, or adsorptIOn, media would be protected from particle blockage
Acknowledgements 4. Conclusions Conventional mlcrofilter crossflow membranes rely upon internal tortuous flow channels to achieve their pore rating. Hence, particles can become lodged Wlthm the structure of the filter leadmg to mcreasing pressure reqwred to maintain a given flow rate, or decreasmg filtrate flow for a given applied pressure. These filters have not, therefore, found use in high throughput mdustnes such as mmeral processmg However, there is a requirement for recycle of clean flwd within many mmeral processes, such as: metal removal from effluents and the return of clean water from a talhngs dam. A
The authors would like to express thelf gralttude for the award of a grant from the Engineering and PhYSical SCiences Research CounCil (EPSRC) to support this work, reference number: GRfN05697.
References ASTM F·316--86, 1986 Test Method for Pore SlZC Charactensttcs of Membrane Filters for Use WIth Aerospace FlUids, West Cons-
hohocken. PA 19428·2959. USA. Bromley, A J. Holdlch, R G. Cummmg, lW, 2002 Partlculate
fouhng of surface rrucrofilters WIth slotted and cU'Clllar pore geometry J Mem. Sa 196,27-37
128
RG Holdfch et all Minerals Engmeermg 16 (2003) 121-128
GoodWIn. J W • Heam,. J • Ho, CC. Ottewdl, R • 1974 Studies on the
preparation and charactensatton on monodtsperse polystyrene latbces CoU Polym Sa 252,464-471 Holdlcb, R G • Cummmg, I W • Ismad. B. 1996 Cross8ow mlcrofil.
trabon (or nuneral suspensions thtckemng and washmg. Mmer Eng 9 (2), 243-257 HoldlCh, R G. Cummmg. I W • Awang. A R B • 8romley. AJ • 2000 Seeded MICfofiltratlon for Base Metal Removal, Proc 8th World Fdtn Cong. FIItratton SOCIety. Bnghton pp 355-358
Kosvmtsev, S. Holdlch, R G. Cumnung, I W • Starov, V M • 2002 Modelhng of dead-end mJcrofiitrahon WIth pore blockmg and cake formation J Mem Set 208,81-192
KUlper. S. van RIJn. C J M. NIJdam. W. Elwenspoek, M C, 1998 Development and apphcatlons of very high flux tnIcrofiltratlon membranes J Mem SCI 150,1-8 Metal Finishers AssOClauon. 1992. Effluent treatment 10 electrotreat· IDg plants 10 the USA, 10 Vyse Street, Hockley, Bmnmgham, BI8 6LT, UK
Rushton, A, Ward, AS. Holdlch, R 0 • 2000 Sobd-LIqwd Fdtra· bon and Separabon Technology VCH·WlIey, Wemhenn. Ocr· many. Tarleton, E S. Wakeman, R J. 1993 Understandmg flux dechne m crossflow mIcrofiltratI0n Part I-Effects of partIcle and pore sIZe Tram ICbemE, Part A, 399-410
--------------------------------------------------------------------------------------------,