PSF20k PSF35k PSF150k
ABCDE
12．6 11．1 9．2 8．8 8．6
810842
0．65 0．65 0．66 0．65 0．63
0．98 0．98 1．01 1．00 0．98
CHAPTER 3
0ptima1 dope viscosiリノブ〜）r nucleatゴon and gro・wth
Table 3．4 shows residual PEG content， tensile elongation at break， and tensile strength
at break for the five hollow−fiber species produced． Both tensile elongation at break and
tensile strength at break were much higher with P SF 150k than with the other species．
Residual PEG content was lowest in P SF150k， indicating that residual content is affected by
the concentration of PEG additive in the dope． No apparent relationship between residual
PEG content and the tensile test results can be ascertained．
Ta b夏e 3．4
Residual amount of PEG tensile elom ation at break and tensiRe stress at break
Fiber
Residual amount of PEG（wt°／・）
Tensi且e elongation at break （％）
Tensile strength at break （MPa）
PSF600 PSF6k PSF20k PSF35k PSF150k
2．73 2．74 3．23 3．58 2．25
80 58 56 60
138
3．3．3、Pure「meatθ7プZux
hFigure 3．3， pure water permeate flux ofthe obtained hollow一飾er membranes is
plotted against Mw o f additive PE G． The membrane made with PEG600 had a flux of 160
1・m−2・h−1，and that made with PEG6k had an only slightly higher flux of 1851・m−2・h−1． Flux
ofthe membrane made with PEG20kjumped sharP ly to 7，7001・m−2・h−1． With still higher
CHAPTE1〜3
0pti〃7al dope visco5「ゴζyプわr nucleation a〃d growlh
（」
10000
10eo
100
102 ¶03 104 鱒5 106 Molecular weight of aClditive PEG
Figure 3．3． Effect of mo且ecular weig血t of additive PEG on pure water permeate flux．
PEG Mw， the rate of increase in flux became quite moderate；且uxes of membranes made with
PEG35k and PEGl50k were 7，8001・m−2・h−l and 8，2761m−2・h−1， respectively． As with the
dope viscosity relationship， there was a clear inflection between PEG6k and PEG20k， and this
inflection was much more pronounced in the case of flux．・
Residual PEG content is not considered to have significantly affected flux results：The
pattem of residual PEG content， increasing moderately from P SF600 to P SF35k and then
decreas血g with PSF150k・was quite inc・ngrU・us with the pattern seen in flux results・
3．3．4SE7レfα刀αるソs is
SEM images of cross sections， outer surfaces， and inner surfaces ofhollow−fiber
membranes obtained with each dope are shown in Figures 3．4，3．5， and 3．6， respectivel）乙
Porosity and mean pore size of the outer surfaces and inner surfaces analyzed from Figures
CHAPTER 3
ρρ伽01ゐPθ幡oo吻力ア灘1θα ∫o刀and grow th
3．5and 3．6 are shown in Tables 3．5 and 3．6． Only in PSF600 finger−1ike voids can be seen in
the cross section． The other membranes all display a sponge−like intemal structure．
咀able 3．5
Porosit an⊂1 mean ore size of outer surface．
Fiber
Porosity（％）
Mean pore siZe
（m）
SEM photo no．
（Fi．6）
PSF600 PSF6k PSF20k PSF35k PSF150k
18．2 19．9 35．5 35．0 43．9
0．195 0．167 0．214 0．319 0．985
abCde
Table 3．6
Porosity and mean Ore SiZe Of inner SUrfaCe．
Fiber
Porosity（％）
Mean pore siZe
（μm）
SEM photo no・
（Fig．7）
PSF600 PSF6k PSF20k PSF35k PSF150k
10．2 6．9
（not analyzable）
（not analyzable）
（not anal zable）
0．125 0．141
（not ana且yzable）
（not ana艮yzable）
（not analyzable）
ah彫Cαe
The outer surfaces of all meml）rane species showed a fine porous structure， with pore
size nearly the same in PSF600 and PSF6k， and increasingly larger in PSF20k， PSF35k， and
PSF150k． The outer−surface porosity increased significantly between PSF6k and P SF20K．
Regarding the㎞er surface， both P SF600 and PSF6k had very dens ely structured
surface・m・rPh・1・gies・with ave・age p・re sizes・fO・125 pm and O・141μm， respectively・ aPd
both had low porosity． The pore sizes on the inner surface of PSF20k， P SF35k， and PSF150k
were apParently larger than those of PSF600 and PSF6k， but the software we used could not
CHAPTER 3
の伽α1ゆe・visco吻for nucleation and grorvth
analyze the porosity and average pore size fbr these species apParently due to the irregular，
lacerated appearance ofthe structure． It is nevertheless clear in visual observation ofthe SEM
images that the size of the inner−surface pores were increasingly larger in P SF20k， P SF35k
and P SF150k， and that for each ofthese species the inner−surface pores were larger than the
outer−surface pores． We find these observations ofthe outer−surface and inner−surface
structures to be consistent with the water permeability results obtained．
The SEM images also show that P SFI50k alone featured a smooth lattice pore
structure on the outer surface． We believe that thi s outer−surface structure imparted
considerable mechanical strength， explaining why this fiber s tensile elongation and tensile
strength at break were so much higher than those of the other species． This would be
consistent with Ohishi and Ohya［2刀， who disclosed a fabrication method for P SF hollow
fiber having a smooth lattice pore structure， claiming that this PSF hollow fiber had higher
mechanical strength than hollow fibers without such a structure．
3．4Discussion
3．4．1」協〃zbrane／eor〃1ing〃iechanis〃2ρ／outer surface
ln addition to higher outer−surface pore size with increasing Mw of additive PEq
SEM images show a clear change in pore shape． From PSF600 to PSF35k， pores have a
highly irregular structure． This changes dramatically with P SF150k， which has a smooth
lattice pore structure． W巳consider the mechanism in light of Tsai et al．， who carefUlly
investigated the effect ofAG relative humidity on the structure of P SF hollow−fiber
CHAPTER 3
0ρtima1 dope viscosity／for nucleation and gr・wth
（a） （b）
（c） （の （e）
Figure 3．4． SEM photogmphs of cross section of PSF hollow fibers．（a）hollow血ber PSF−600 x70；（b）ho置且ow fiber PSF−6k x70；（c）hollow fiber PSF−20k x70；（d）hollow fiber PSF−35K x70；
（e）hol且ow血ber PSF−150k x70．
CH41）TER 3
の伽α1ゆθv漉・吻力7刀〃・lea伽and gr傭乃
（a）
（c） （d）
（b）
（e）
Figure 3．5． SEM photograp血s of outer surface of PSF血ollow血bers．（a）holEow茄er PSF600 x5，000；（b）hollow血ber PSF6k x5，000；（c）hollow fiber PSF20k x5，000；（d）hollow 血ber PSF35K x5，000；（e）hollow fiber PSF150k x5，000．
CHAPTER 3
0ptimal d・pe vis・・吻for nu・leatio〃and gr・wth
（a） （b）
（c） （d） （e）
Figure 3．6． SEM phbtographs of inner sur血ce of PSF血01且ow舳ers．（a）hollow fiber PSF600 x5，000；（b）ho皿ow fiber PSF6k x5，000；（c）血oIEow fiber PSF20k x5，000；（d）ho夏bw 血ber PSF35K x5，000；（e）血ollow fiber PSF150k x5，000．
CH41）TER 3
の 〃ial dope viscosiりノノわ7 n〃cleation and gアrow 乃
membrane， particularly in the vicinity of the outer surface［21］． They concluded that
penetration of water vapor into the membrane dominated the effect of solvent evaporation
du血g the short period oftime when nascent fiber passes through the AG Especially in case
of high boiling point solvent such as NMP（boiling point，202°C）， they fbund evaporation
induced phase separation to be negligible compared to vapor induced phase separation．
They also studied mass variation of cast film using polysulfbne casting solution under
various relative humidities， and fbund that the mass of cast film increased and that greater
increases corresponded to higher relative humidity． In this study， with the AG at 80°C and
relative humidity・f 100％， abs・lute humidity was determined t・be 2929／m3 using Equati・ns
land 2． in accordance with Tasi et al．， we believe that N正PS occurred in our study due to
penetration of water vapor in the AG
Further valuable insight is fbund in Lee et al．， who studied the water uptake of
PSF∠NMP binary cast film with different polymer concentrations in an atmosphere of 90％
relative humidity［29］． They reported that binary s olutions of different polymer
concentrations showed similar water uptake 1）ehavior， even though the viscosity increased and
the diffUsion coefficient decreased with higher polymer concentration． In Iight of Lee et al．，
we therefore believe that the different dope compositions in this study all had similar water
uptake behavior in the Aq despite having different PEG Mws and different viscosities． On
the other hand， the interdiffUsion coefficient of each dope must decrease as PEG Mw is
increased． As mentioned in Theory， Kim and Lee reported that the precipitation rate
decreased with increased PEG Mw and with increased PEG additive loading， though the
precipitation type remains instantaneous demixing in each case． They attril）uted these
CU 4PTEI〜3
0pti脚1 d・pe visc・吻力7鷹Zθ㈱刀and gn傭乃
phenomena to dope viscosity・ We thus believe that our re sults can be explaine d as follows：a）
phase separation ofthe outer surface ofnascent hollow fiber was i lduced in the A G siエnilarly
for each species， although b）as dope viscosity was raised with higher PE G MW， there was a
greater amount oftime betWeen when phase separation was induced and when coagulation
composition was reached， i．e．， nucleation and growth of the polymer−lean phase was able to
progress fUrther． Outer−surface pore size therefore became larger when dope vi scosity was
higher．
With regard to the皿e chanism of formation of the smooth latti ce structure on the outef
surface of PSF1 50k， we believe that demixing betWeen PEG and polysulfone in accordance
with the long time scale proposed by Boom et al．［14】changed dramatically between PSF35k
and P S F 150k． ln the case of low Mw additive PEq dope viscosity is low， implying a short
time to reach coagulation composition and a low degree nucleation and groWth．㎞case of
high Mw additive PEq dope viscosity is high， implying a long time to reach coagulation
composition and high degree ofnucleation and groWth．
Apparently， in the range of additive PEG600 to PEG35k， the long time scale
polymeric additive and polysulf（）ne demixing begin in mid−course of short time scale
nucleation and growth ofthe polymer−lean phase． Instantaneously， the virtual binodal state
shifts toward the real bino da1 state， causing a j ump to the coagul ation composition， and
structure is fixed． On the other hand， in case of PEG150k， because the PEG chain length is s o
long， the nucleation and groWth of the polymer−lean phase saturates bef（）re the begiming
CHAPTER 3
0ptimal dope viscosめノノb7 nucleation and growth
of PEG and polysulfbne demixing． High Mw PEG then solves out， entering the saturated
polymer−lean phase；this serves to expand portions ofpolymer−lean phase as well as
interconnections between them． As a result， a smooth lattice structure as shown in Figure 3．6
（e）is飴rmed． We thus believe that a smooth lattice structure is formed only when gro舳of
polymer−lean phase has been saturated befbre polymer demixing begins．
3。42、娩〃7braneプbア〃7ing〃zechan is〃70finner surface
imer−surface pores of PSF20k， PSF35k， and PSF150k were clearly larger than those
of PSF600 and PSF6k， though porosity and average pore size fbr the three fbrmer species
could not be analyzed because of the irregular， lacerated apPearance of their inner−surfage
po「es・
The fact that nascent hollow fiber would break apart血the 400 mm AG when we attempted to
use NMP concentration of 75 wt％indicates that a support layer which can withstand the
tensile stress of fiber drawing was formed in the vicinity ofthe i皿er sur魚ce a負er dope passed
through the spinneret only when the water concentration of bore liquid was raised丘om 25
wt％to 30 wt％． The coagulating power of 70 wt％NMP solution as bore liquid being
stronger than water vapor in the Aq non−solvent concentration at the inner surface arrived at
coagulation composition sooner than at the outer surface． Thus phase separation at the inner
surface stopped progre s sing and the structure 1）ecame丘xed more quickly than at the outer
surface． The average pore sizes for P SF600 and P SF6k on the imer surface were therefore
smaller than on the outer surface．
CHAPTER 3
ρρ伽α1ゆe・vis・・吻for nu・1θα伽αnd gr傭h
The reason why the average size became larger with higher Mw of additive PEG was
as discussed with regard to the outer surface． Dope viscosity increased with highe士 PEG Mw，
so the mutual diffUsion rate between non−solvent water and solvent NMP become slower． As
aresult， the time to reach coagulation composition become longer， allowing greater
development ofthe polymer−lean phase．
The irregular shapes and lacerated app earance of inner−surface p ores with P SF20k，
PSF35k， ahd PSF150k is believed t・be due t・a partial ripPing and pulling apart・fthe㎞er
surface under drawing tension． This tension， which was enough to completely break nascent
〔一．∫・
hollow fiber when bore liquid NMP concentration was 75 wt％， apparently caused the血ner
surface of nascent hollow fiber to p artially rip and extend， though the nascent fiber had j ust
enough strength to prevent c omplete breakage， when bore liquid NMP concentration was 70
wt％． Ih other words， the structure of the polymer−rich phase in the midst of coagulation was
extended by drawing tension． This structural deformation was greater with higher PE G Mw
because the formation ofthe support layer is slower， as higher dope viscosity corresponds to
slower mutual diffUsion rate betWeen non−solvent water and solvent NMP and a longer time
before coagulation composition is reached．
As such irregular㎞er−surface structures would generally be imdesirable in practical
MF application， we believe fUrther study regarding control of inner−surface structure i s
required．
CHAPTER 3
0pti〃zal d∂pe viscosめノプわr nucleation and g7¶oWth
3．4．3／lfacrovoid／Cor〃zation
Among the membrane specimens， only P SF600 had fnger−like macrovoids in the
membrane wall． The other membrane specimens showed a sponge−like structure． The finger−
like macrovoids of PSF600 were only observed on the inner−layer side；none were observed
on the outer・layer side． It is interesting that no macrovoids formed in P SF6k， despite only a
slight difference in average i lner−surface pore size，0．141 pm versus O．125 pm with P SF600，
and only moderately higher dope viscosity，9．13 Pa・s versus 5．70 Pa・s with the dope fbr
PSF600． This suggests that a small difference in pore size and／or dope viscosity can affect
whether macrovoids form or not．
Figure 3．8 shows a segment of the PSF600 membrane wall cross−section at greater
magnification． It is clear that the walls of the macrovoids are quite dense in the region of the
membrane inner surface， with more and more circular pores fbrming as they extend toward
the middle of the membrane wal1． Based on this observation， we believe the macrovoids were
formed not through progression ofphase separation， but rather through a process whereby， a
dense skin having formed on the inner surface ofthe nascent membrane， discrete masses of
bore liquid are drawn into the nascent membrane wall by osmotic pressure． The process of
phase separation and coagulation then initiates at the interface between these masses of bore
liquid and the su皿ounding dope solution， the se interfaces eventually forming the Walls of the
macrovoids． The relatively high water concentration in the region ofthe inner surface of the
membrane results in faster coagulation and macrovoid walls forming as dense skin−like
structures． Ih the region toward the middle ofthe membrane wall， in contrast， NMP
CHAPTER 3
（〜pti〃ZOI dope viscos iりノノわ7 nucleatio刀α〃∂「gア「owth
concentration is relatively high， slower coagulation allows the growth of polymer−lean phase
to progress， and circular pores form ih the macrovoid walls．
Stratimann［30］expla血s that macrovoids are the result ofrapid penetration of non−
solvent at certain weak spots血the top of a flat−sheet membrane． Regarding macrovoid
growth， Smolders et al．［31］assumed that this occurred due to diffUsion between the polymer
solution and the macrovoid， with the macrovoid acting as a local coagulation bath． Our
results correspond well with their models．
However， if macrovoids are fbrmed by discrete masses of bore liquid， macrovoids may
be expected to exist in PSF6k， because the血ner−surface pore size differential between
PSF600 and PSF6K is only slight． Pore size difference alone， therefbre， seems an insuf丘ciellt
explanation． We therefore believe dope viscosity has a critical influence on macrovoid
l
formation． The reason macrovoids fbrmed in P S F600 but not in P SF6k， in spite oftheir
similar inner−surface pore sizes， is believed to be that there is an upper limit to dope viscosity
which allows such masses of bore liquid to enter the nascent membrane wall by osmotic
pressure， and that the viscosity ofDope B exceeded this limit while that ofDope A did not．
3．5Conclusion
This study examined the formation of P SF hollow−fiber MF membranes by NIP S
using a dry／wet spin血lg process． A 3−component dope composition of PSF∠NMP／PEG was
used with different PEG Mws． The effect of different PEG Mw on dope viscosity， membrane
morphology， and membrane performance were examined based on the theory ofnucleation
．CHAPTER 3
（〜ρ ゴ〃7a1∂Ope visco・∫ノζyプわr nucleatioア2 and growth
Figure 3．7． SEM photographs of cross section of PSF hol且ow fiber PSF600 x250．
CH4PTER 3
ρρ 加2al d（）pe viscosiz＞ソbr nucleation and growth
Dope viscosity rose sharPly when PEG Mw was changed丘om 6，000 Da to 20，000 Da：
From 9・13Pa°s to 26・8 Pa・s The same change ofPEG Mw also resulted in a sharP rise i l
water permeate flux o f the membranes obtained， from 185 1・m 2・h｝1 to 77，0001・m−2・h−1．
SEM analysis revealed that outer−surface pore size increased with higher PEG Mw and
higher dope viscosity． Outer−surface pore structure was irregular with PEG Mw丘om 600 Da
to 35 kDa， but changed to a smooth lattice structure when PEG Mw was raised to 150 kDa．
㎞er−sur飴ce pore sizes ofPSF20k， PSF35k， and PSF150k were apparently larger than
those of PSF600 and P SF6k， though the porosity and mean pore size ofthe former could not
be analyzed because oftheir irregular， lacerated appearance． The irmer−surface pores of
PSF20k， P SF35k， and P SF 150k were血creasingly larger among the three species in this order
Only P SF600 had fngeri−like macrovoids hl the membrane wall． The other membralle
specimens showed a sponge−1ike structure． The macrovoids in P SF600 were only fbund on
the inner−layer side．
The macrovoids are believed to be formed through a process whereby， a dense skin
having formed on the inner surface of the nascent membrane， discrete mas ses of bore liquid
are drawn into the nascent membrane wall by osmotic pressure． Our results correspond well
with the models of Strathmarm［30］and Smolders［31］． Additionally， our results suggest that
there is an叩per limit to dope viscosity with which such masses of bore liquid are al）le to thus
enter the nascent membrane wall．
We found that it is possible to fabricate a PSF hollow−fiber membrane us血g PEG Mw
of 20 kDa or higher， with dope viscosity of 26．8 Pa・s or greateL However， more precise
＼I
CHAPTER 3
の伽al d・pe vis・・sityf・r nu・leati・刀and gr・働
practical MF membrane having a fiIter layer in the region ofthe outer surface and gradualIy
increasing pore size across the thickness ofthe membrane wall toward imer surface．
CHAPTER 3
0ptimα1 d・pe vis・・吻for nueleation and gr・wth
3．6 References
［1］S．Loeb， S S ourir勾an， Sea water demineralization by means of an osmotic membrane，
Adv．Chem．Ser．，ACS 38（1963）117．
［2］M．W．V Mulder， Basic Principle of Membrane Technology， Kluwer Academic Publishers，，
1996．
［3］S．Jain， J． B ellare， A Review of TIP S and Dll｝S Technique s for Membrane ManufactUre，
Indian Chem．， Engr．，46（2004）155．
［4］R．E． Kesting， Synthetic Polymeric Membranes， John Wiley＆Sons，1985， pp．238−251，
［5］1．Pinnau and W． J． Koros， Structures and gal separation properties of asymmetric
polysulfone membranes made by dr）x， wet， and dry／wet phase inversion， J． App1， p olym S ci，
43（1991）1491．
［6］H．Caquineau， P． Menut and C． Dupuy， lnfluence ofthe Relative Humidity on Fil皿
Formation by Vapor lnduced Phase S eparation， Polym Eng． Sci．，43（2003）798．
［7］D．R． Lloyd， J．W． B arlow， Microporous Membrane Formation via Thermally−induc ed
Phase S eparation， AIChE． Symp． S er．，84（1988）28．
［8］B．J． Cha， K． Char， J．−J． Kim， S．S． Kim and C．K． Kim， The effects of diluents molecuユar
weight on the structure of thermally−induced phase separation membrane． J． Membrane，
Sci．，108（1995）219．
［9］H．Matsuyama， H． OkafUj i， T． Maki， M． Teremoto and N． Kubota， Preparation of
polyethylene hollow fiber membrane via thermally induced phase separation， J． Memblanc．
Sci．223（2003）119．