Casting and performance of polyvinylidene fluoride based membranes

Journal of Membrane Science, 16 (1983) 181-193 Elsevier Science Publishers B . V ., Amsterdam - Printed in The Netherlands

181

CASTING AND PERFORMANCE OF POLYVINYLIDENE FLUORIDE BASED MEMBRANES*

STELIO MUNARI, ALDO BOTTINO and GUSTAVO CAPANNELLI Institute of Industrial Chemistry, University of Genoa, Corso Europa 30, 16132 Genoa (Italy) (Received July 27, 1982 ; accepted in revised form February 15, 1983)

Summary The investigation of factors governing membrane preparation is of paramount importance in order to understand the formation of the membrane and its behaviour under test conditions . A large amount of fundamental work on membrane formation has been carried out on cellulose acetate membranes, but less information is available for other polymers . The purpose of the present work is to provide a better understanding of the parameters governing the formation of polyvinylidene fluoride membranes . The parameters controlling the casting process have been investigated in some detail .

Introduction Most of the synthetic membranes used in ultrafiltration (U .F .) and reverse osmosis (R .O .) processes are prepared using the so called "phase inversion" technique . The preparation involves several steps : (1) choice of a correct polymer ; (2) choice of a proper solvent and/or solvent + cosolvent and/or additives ; (3) casting of the dope prepared ; (4) partial evaporation ; (5) coagulation : choice of the bath and its operating condition ; (6) eventual post-treatment of the membranes . Membranes morphologies and performances are strictly correlated to all the above steps . Some years ago in our laboratory we prepared membranes with interesting properties for U .F . purposes from polyvinylidene fluoride (PVDF) modified by sulfonation [1] . In this paper some results obtained in a more systematic study of membrane preparation from this polymer are reported . In order to evaluate how operating conditions affect the membrane properties, membranes have also been cast from unmodified PVDF .

*Paper presented at the Symposium on Membranes and Membrane Processes, May 19-22, 1982, Perugia, Italy .

0376-7388/83/$03 .00 © 1983 Elsevier Science Publishers B .V .



182

Experimental Materials The PVDF was a commercial product (Foraflon 1000 HD, Ugine Kuhlmann) with an average viscometric MW of 80,000 . The sulfonated polymer (PVDFS) was obtained by heterogeneous reaction with fuming sulfuric acid to yield a sulfonic group content of 2% [1, 21 . During the treatment some other reactions occur leading to partial chain degradation . The average viscometric MW was 21,000 . From the practical point of view the sulfonated polymer is more hydrophilic than the unmodified version . Only a few solvents can be used to prepare casting solutions from these polymers ; among these are NN-dimethylformamide (DMF) and N-methyl-2pyrrolidone (NMP) . These solvents have good solubility properties towards both polymers up to reasonable concentrations for casting purposes ; this study was limited to these two solvents . In all casting solutions, a fixed amount (3 g per 100 ml) of LiCl was added . In several casting dopes, acetone (Ac) and tetrahydrofuran (THF) were also used in mixture with the above solvents . Both Ac and THF swell the PVDF and PVDFS . All products were reagent grade . Viscosity measurements The absolute viscosities, 77abs, of the pure solvents, solvent mixtures (without polymer) and casting solutions were calculated by the equation

n abs

1 8

where 77 is the kinematic viscosity and S the density . The kinematic viscosities were determined at 30'C with a Cannon-Fenske glass viscometer. The densities were determined, at the same temperature, with a picnometer . The reduced viscosities, rl Ye d, of the casting solutions were calculated using the well known equation 77 sp

1l red c

where c is the concentration of the polymer and rtsp the specific cosity . The specific viscosity is the viscosity increase (due to the presence of the polymer) divided by the viscosity of the pure solvent . Thus f7 sol

r1 soly

17 SP _ 77 soly

Membrane preparation Membranes were prepared from 16 to 32% polymer solutions (16-32 g polymer per 100 ml solution) . The filtered solutions were cast on a glass

183 plate with a knife, preset for the desired thickness . The evaporation period was varied between 5 and 480 sec . The coagulation bath was deionized water kept at 5'C . All the operations were carried out in a room controlled at a constant temperature of 20 ° C and at 60% humidity .

Membrane morphologies Optical microscopy (O .M .) was used to evaluate the morphology of the membranes. More accurate investigations were carried out using scanning electron microscopy (S .E .M.) and transmission electron microscopy (T .E .M .) techniques . Membrane specimens were embedded in an epoxy resin . The specimens were cross sectioned with an ultramicrotome to a thickness of 5 p m and then observed with an optical microscope . Cross-sections of 500 A were also prepared ; they were shadowed with carbon and then observed in an electron microscope . Scanning electron photomicrographs were taken of crosssections of membranes prepared by fracturing in liquid nitrogen and shadowing with gold .

Membrane properties Burst pressure resistance of the wet membrane was used as the criterion for mechanical properties . Flat-sheet membranes were tested in a U .F . pilot unit with an aqueous solution of dextran and an emulsion of cutting oil under the operating conditions listed in Table 1 . All the membranes were tested as cast, without post treatment . Fluxes and rejections were measured after 1 hr . Dextran in feed and permeate streams was determined by colorimetric method [3] . Oils were analyzed according to the standard methods of APIIA--AWWA-WPCF [4] . TABLE 1

Operating conditions for U .F . tests Feed

Pressure (kPa)

Temperature (°C)

Recirculation rate (rn/sec)

Dextran, Mw 40,000 (1000 ppm) Cutting oil (20%)

200 220

40 40

5

Results

Viscosity of the casting solutions The viscosities of the pure solvents, solvent mixtures (without polvul and casting solutions are reported in Table 2 . Obviously an increase in polymer concentration causes an increase in viscosity . The addition of Ac or always causes a decrease in viscosity . When the reduced viscosity is considered in order to eliminate the effects of solvent and polymer concentration, the

184

TABLE 2 Viscosities of pure solvents, solvent mixtures and casting solutions at 30°C ; all casting solutions contain a fixed amount of LiCI (3 g per 100 ml of solvent or solvent mixture) Polymer

Polymer concentration

Solvent

Cosolvent

Volumetric ratio solvent : cosolvent

Absolute viscosity

-

-

Ac THF -

1 :1 1 :1 -

3,609 0,962 1 .232

Ac THF -

1 :1 1 :1 -

Ac THF THF -

1 :1 1 :1 1 :1 -

Ac THF -

1 :1 1 :1 -

Ac Ac THF THF -

1 :1 1 :1 1 :1 1 :1 -

Ac Ac THF THF

1 1 1 1

(g1100 ml)

PVDF PVDF PVDF PVDF PVDF PVDF PVDF PVDF PVDF PVDF PVDFS PVDFS PVDFS PVDFS PVDFS PVDFS PVDFS PVDFS PVDFS PVDFS PVDFS PVDFS PVDFS

-

NMP NMP NMP DMF DMF DMF

12 16

NMP NMP NMP NMP NMP NMP DMF DMF DMF DMF NMP NMP NMP NMP NMP NMP NMP DMF DMF DMF DMF DMF DMF

24 16 16 24 16 24 16 16 16 24 32 16 24 16 24 16 30 16 24 16 24

:1 :1 :1 :1

Reduced viscosity

(cP)

0 .822 0 .71 0 .776 1110 3076 22227 1001 930 3590 1250 9263 453 643 1115 4939 21107 235 887 348 1570 495 3100 131 442 176 651

25 .55 53 .22 256 .64 64 .97 47 .12 121 .37 94 .98 469 .5 39 .81 51 .72 19 .25 56 .98 182 .73 15 .20 38 .38 17 .59 53 .06 37 .57 125 .67 11 .47 25 .90 14 .11 34 .41

behaviour can be different as the additives vary the thermodynamic quality of the solvent . From the table, it appears that the overall viscosities of the sulfonated polymer are lower than the viscosities of solutions of the unmodified polymer, as can be expected on the basis of MW data . The presence of sulfonic groups introduces further effects, due to the changes in the polymer chain conformation .

Fig . 2 . Effect of evaporation time on fluxes and morphologies of membranes prepared from solutions of PVDF and PVDFS in NMP .



185 DM F

14000-

o 16°J0 PVDF o 24°k FVDF • 15°!o PVDFS • 24 0 1° °VDFS

Fig . 1 . Effect of evaporation time on fluxes and morphologies of membranes prepared from solutions of PVDF and PVDFS in DMF .

NMP

t

OIL

o 16°i° PVDF a 24 01° PVDF • 16°!0 PV -FS * 24 PVDFS



18 6

Morphologies Optical photomicrographs of the cross-sections of the membranes as a function of evaporation time are reported for different dope formulations in Figs . 1-6 . Although optical microscopy and low magnification S .E .M . do not provide sufficient information to allow speculation about the fine membrane morphology, nevertheless the pictures are extremely useful in giving an indication of the whole membrane structure, as can be seen in Fig . 7 . Binary solutions of polymer and solvent yield membranes with finger-like cavities . All these membranes have a dense skin-layer on their upper surface and a sponge-like structure on their bottom surface . The thickness of the skin layer can be clearly seen in the T .E .M . photomicrograph in Fig . 7 . The upper surface of the membrane corresponds to the air--solution interface of the cast solution, the bottom surface corresponds to the glass- -solution interface . The size of finger-like cavities decreases when the concentration of the polymer and the reduced viscosity of the casting solutions are increased (Figs . 1---6) . A membrane with either finger-like or sponge-like structure can be obtained for

0hA -AC

14000

_,'0000

d

000

JF:

0' T'Ji

Fig . 3 . Effect of evaporation time on fluxes and morphologies f from solutions of PVDF and PVDFS in DMF--_Ac .

hrane ; prepared



18 7

DM F-THF

o 16°I° PVDF • 16 0 /. PVDFS

14000

10000 w E

x

3

6000

2000

Fig. 4 . Effect of evaporation time on fluxes and morphologies of membranes prepared from solutions of PVDF and PVDFS in DMF--THF, ternary solutions (polymer + solvent + cosolvent) by selecting the cosolvent, the concentration of the polymer and especially the evaporation period before the phase inversion . Typical examples of structures from different experimental preparations are shown in Figs, 8-10,

Performances Flux performances in the U .F, application as a function of evaporation time are reported for different dope formulations in Figs . 1---fi, In all membranes, the sulfonated polymer gives higher fluxes than the unmodified polymer . This obviously means that the specific properties of the polymer are extremely important . When a high boiling solvent is used, the evaporation period, at least in the range explored, has very little effect on the properties of the membranes (Figs . 1 and 2) . However, when a highly volatile cosolvent is present, the



188

",J MP -Ac

ivDFS 24 'VDFS o 16°I° PVOF

I 14000

10000

6000 L

2000

2

3

OF G( jTTIhiG OIL

4 L voscration

t :-7e

( C-

:r )

Fig . 5 .

Effect of evaporation time on fluxes and morphologies of membranes prepared from solutions of PVDF and PVDFS in NMP-Ac .

M

-THF

o 1 6 ° /, PVD 24°? ° P~IDFS

10000L N

E

X

6 i',00

l1

2000

3

4

G

Evaporc'. ioo

Fig. 6 . Effect of evaporation time on fluxes and morphologies of membranes prepared from solutions of PVDF and PVDFS in NMP-THF .

S

lag : : n»«xe ae :

»yE

: Sb .

mV w :iS, . > « ««

2»3

* 200 Fig . 7 Results obtained with different microscopy techniques used for morphological studies .

16% PVDFS

16% 9VDF

24% PVDFS

30% PVDFS

24% 2VDF

mr y Sections (S £ : .-o M) of membrane prepared from solutions o PVDF and PVDF m DNm

190

24% PVDFS/DMF

16% PVDFS/DMF

24% PVDFS/DMF-Ac

16% PVDFS/DMF-Ac

24% PVDFS/DMF-THF

16% PVDFS/DMF-THF

Fig. 9 . Sections (S .E .M .-O .M .) of membranes prepared from solutions of PVDFS in DMF and DMF-cosolvent mixtures .

16% PVDFS/NMP

16° PVDF/NMP

16% PVDFS/NMP-Ac

16% PVDFS/Nrr4P-THF

16% PVDF/NMP-THF

Fig. 10 . Sections (S .E .M .-O .M .) of membranes prepared from solutions of PVDS and PVDFS in NMP and NMP-cosolvent mixtures .



191 properties of the membranes are significantly affected by the evaporation period (Figs . 3--l) . Among the factors that may contribute to this effect in the latter case are the noticeable evaporation, with consequent increase in polymer concentration before the phase inversion ; formation of a concentration gradient from the surface to the bottom ; changes in viscosity and in solvent thermodynamic quality ; etc . Under some conditions, when the evaporation time is increased, the solubility limit of the polymer can be reached and some polymer precipitation can occur before the cast film is immersed in the coagulation bath . In these latter cases, the membranes resulting after immersion in water are extremely brittle and cannot be used for practical purposes . This fact is evident, for example, in the data in Table 3, where the burst strength is reported as a function of evaporation period . Oil rejection has not been plotted in Figs . 1-6 because it was higher than 99 .9% for all the membranes, with total rejection of all oil droplets in the emulsion . Because the droplets in the oil emulsion were relatively large compared to the size of U .F . membrane pores, the test does not supply detailed information about the cut-off of the membranes .

TABLE 3 Effect of evaporation time on burst pressure of PVDFS membranes Evaporation time (min) 0 .5

1

2

Casting solution 16% PVDFS/DMF ; burst pressure (kPa) 6 .7 6 .7 6 .7 Casting solution 16% PVDFS/DMF-Ac ; burst pressure (kPa) 11 .5 15 .7 13 .1 Casting solution 16% PVDFS/DMF THF ; burst pressure (kPa) 22 .7 23 .1 23 .3

4

6

8

7 .9

6 .2

6 .8

1 .3

0 .6

1 .4

0 .4

TABLE 4 Comparison of membrane performances in ultrafiltration of cutting oil and dextran Casting solution composition

16% PVDFS/NMP 16%PVDFS/NMP-THF 24% PVDF/NMP-THF 16% PVDFS/DMF 16% PVDFS/DMF-THF 16% PVDFS/DMF-THF

Evaporation time (sec)

Cutting oil Flux (1/m2 -d)

Permeate

Flux (1/m 2 -d)

Rejection (%)

30 30 30 30 30 30

12,900 8,600 180 10,400 6,000 3,500

clear clear clear clear clear clear

10,500 7,500 100 7,800 5,500 3,200

39 .4 95 90 .7 45 .2 96 .3 98 .1

Dextran, Mw

40,000

1 92

The performance of the membrane was investigated using a dextran solution as U .F . feed . In Table 4 the results are reported for membranes obtained from different casting solutions with an evaporation period of 30 sec . All membranes offer a clear permeate in cutting oil ultrafiltration while a certain difference in rejection of dextran is obtained . This indicates that membrane quality is dependent on preparation conditions and is in agreement with the general statement that membrane structure, as is presently known, tells us very little about membrane performance [51 . This fact is evident by the comparison of the fluxes of the membranes in Figs . 1-6 with the optical photomicrographs reported in the same figures and the electron scanning photomicrographs reported in Figs . 7-10 . As can be seen, membranes with an apparently similar finger-like or sponge-like structure offer quite different performances . Conclusions A large variety of membranes with different structures and properties can be obtained from PVDF and PVDFS casting solutions by varying the conditions of preparation . Binary solutions of polymer and solvent yield membranes with finger-like cavities . The size of finger-like cavities decreases when the concentration of the polymer and the viscosity of the casting solutions are increased . For equiconcentrated solutions with different solvents (DMF or NMP), an increase in reduced viscosity also causes a decrease in membrane flux . Morphological structures and properties of the membrane prepared from these solutions containing a high boiling point solvent are not significantly affected by the evaporation period. Ternary solutions, where a low boiling point cosolvent (Ac or THF) is present, yield membranes with either finger-like or sponge-like structure . No clear relation can be found between the reduced viscosity of these solutions and the ultimate characteristics of the membranes, but an important factor is the noticeable evaporation of the cosolvent before phase inversion . In these cases the mechanical properties and performance of the membranes are seriously damaged by allowing the evaporation period to extend too long, due to resulting precipitation phenomena . The results obtained and the latter statements allow some points to be made about the best conditions for membrane preparation . The sulfonated polymer is an attractive material for preparation of asymmetric ultrafiltration membranes because of the noticeably high fluxes obtained . In the treatment of cutting oil emulsions, the best results are obtained with membranes prepared from a 16% polymer solution in NMP . In the ultrafiltration of aqueous solutions containing dextran with an MW of 40,000, a choice can be made in a wider range of preparation conditions .

193

Acknowledgement This work was financially supported by the Consiglio Nazionale delle Ricerche (CNR), Programmi Finalizzati Chimica Fine e Secondaria - Sottoprogetto Membrane, Rome, Italy .

References 1 2

3 4 5

S . Munari, F . Vigo, G . Capannelli, A . Bottino and C . Uliana, Methods for the preparation of asymmetrical membranes, U .S . Patent 4, 188, 354, Feb . 2, 1980 . F . Vigo, G . Capannelli, C . Uliana and S . Munari, Modified polyvinylidene fluoride (PVDF) membranes suitable for ultrafiltration purpose application, Chim . Ind . (Italy), 64 (1982) 74 . M . Dubois, K .A . Gilles, J .K . Hamilton, P .A. Rebers and F . Smith, Colorimetric method for determination of sugars and related substances, Anal . Chem ., 28 (1956) 350 . APHA, AWWA and WPCF, Standard Methods for the Examination of Water and Wastewater, 14th edn . 1976 . A.J. Staverman, Structure and function of membranes, J . Membrane Sci .,' 16 (1983) 7 .