ELSEVI ER

Desalination 131 (2000) 299-305

DESALINATION
New composite membrane for water softening
S . B e q u e t a*, T . A b e n o z a a, P . A p t e l a, J . - M . E s p e n a n b, J . - C . R e m i g y a, A . Ri c a r d a
"Universit~ Paul Sabatier, Laboratoire de G~nie Chimique, CNRS UMR 5503, 118, route de Narbonne,
31062 Toulouse Cedex 4, France
Tel . +33 (5) 61 55 88 77; Fax + 33 (5) 6I 55 61 39; e-mail: bequet @l gc. ups-t l se. f r
bPolymem, route de Revel, 31450 Fourquevaux, France
Received 26 June 2000; accepted 10 July 2000
Abs t ract
A novel route to the preparation of nanofiltration membranes has been proposed. The method consists of the
photografting of polymer on to the surface of porous membranes. The method was tested with polysulfone
ultrafiltration membranes and poly(acrylic acid) in water solution. It has been shown that the surface modification is
stabilised by crosslinking with N, N'-methylene bis acrylamide and that the grafting does not require any
photoinitiator and any purification of the reactants. Moreover, by the proper choice of the wavelengths, the UV
irradiation does not induce any significant copolymerisation of the reactants inside the aqueous solution. The
irradiation conditions, the monomer concentration and the permeability of the support membrane were examined.
The water softening properties of the novel membranes were measured in dead-end mode and at a 95% recovery. A
large range of performances has been obtained which demonstrates the feasibility of the UV irradiation as an
environmentally friendly and versatile (flexible) technology to prepare nanofiltration membranes.
Ke y wo r d s : Nanofiltration; Softening; UV photografting; Surface modification; Membrane preparation
1. I nt r oduc t i on
Conventional softening technologies are
bei ng replaced by nanofiitration as the
membrane operation does not add by-products to
finished wat er or by-product disposal cost.
Specifications for an ideal softening membrane
include l ow cost, high fluxes, high rejection of
*Corresponding author.
bivalent ions, low rejection of monoval ent ion
and high rejection of small organic mol ecul e
(pesticides, NOM) , no or little pre- or post-
treatment, low fouling, chemical and biological
resistance. For domestic needs, the ability to
operate at very low pressures - - typically
400 kPa, with a high recovery - - typically 95%,
is an important requirement. As for reverse
osmosis, the usual approach to nanofiltration
membranes involves thin-film composi t e techno-
logy generally based on interracial polymeri-
Presented at the Conference on Membranes in Drinking and Industrial Water Production, Pads, France, 3-6 October 2000
International Water Association, European Desalination Society. American Water Works Association, Japan Water Works Association
0011-9164/00/$- See front matter 2000 Elsevier Science All rights reserved
PII: S0011- 9164( 00) 00123- 5
300 S. Bequet et al. / Desalination 131 ( 2000) 299- 305
sat i on. Thi s t e c hnol ogy needs t he use o f sol vent s
and r eact ant s l eadi ng t o wast e di sposal costs.
In t hi s paper , we r epor t about anot her r out e t o
t he f or mat i on o f nanof i l t r at i on me mbr a ne by
modi f i cat i on o f ul t rafi l t rat i on me mbr a ne s wi t h
gr af t ed pol ymer . Thi s was accompl i s hed by UV
i rradi at i on.
Modi f i c a t i on o f me mbr a ne s by phot o-
gr af t i ng has al r eady been des cr i bed by several
aut hor s wi t h di f f er ent obj ect i ves s uch as f oul i ng
r educt i on [1,2], bi omol e c ul e i mmobi l i zat i on [3]
or pr ovi di ng s wi t chabl e per meabi l i t y [4]. In our
knowl e dge , phot ogr af t i ng as a r out e t o el aborat e
nanof i l t r at i on by modi f i cat i on o f por ous me m-
br anes has not yet been suggest ed.
As one obj ect i ve was t o devel op an envi r on-
me nt a l l y f r i endl y t echnol ogy, an ul t rafi l t rat i on
pol ys ul f one me mb r a n e was chos en si nce t hi s
p o l y me r is i nt r i nsi cal l y phot oact i ve and and t hat
no phot oi ni t i at or s are r equi r ed [5]. Acr yl i c aci d
has been sel ect ed as a wel l st udi ed model for
phot opol yme r i s a bl e mo n o me r , pol y( acr yl i c aci d)
is b o u n d t o sur f ace me mb r a n e wi t h coval ent
bond i n or de r t o i ncr ease life t i me o f surface
modi f i cat i on; it c oul d al so be expect ed t hat a
negat i vel y char ged modi f i ed me mbr a ne wi l l
pr ovi de r epul s i ve i nt er act i on wi t h ani ons and
i mpr ove t he r ej ect i on o f salts [6,7]. Si nce t he
p Ka o f pol y( acr yl i c aci d) is about 4.8, t he pH
wi l l s t r ongl y i nf l uence t he salts rej ect i on; mor e-
ove r t he p H- d e p e n d e n t swel l i ng o f t he pol yme r
[4] c oul d al so i nf l uence t he per meabi l i t y o f t he
nanof i l t r at i on me mbr a ne .
Thi s wor k pr esent s t he first resul t s obt ai ned
on t he gr af t par amet er s ( wavel engt hs, mo n o me r
concent r at i on, i rradi at i on t i me) on t he me mbr a ne
per f or mance.
2. Materi al s and method
The me mb r a n e us ed is a fl at sheet pol y-
s ul f one me mbr a ne . It has been obt ai ned by phase
i nve r s i on ( c ol l odi on was ki ndl y pr ovi ded by
N2
Cooling system
(glass or quartz)
Monomer solution
UVL ~ p
Membrane
Stirring rod
Bubble
Fig. 1. Photoreactor schema.
Po l y me m SA, Four quevaux, France). Prepara-
t i on o f t hi s ul t rafi l t rat i on me mbr a ne is repro-
duci bl e and fl ux obt ai ned was 400 l . h-l . m -2
(+ 10%) at 400 kPa. Acr yl i c aci d (Caa), N, N' -
met hyl ene bis acr yl ami de (Cr), 1-met hyl
2- pyr r ol i di none pr oduct s wer e pur chas ed f r om
Al dr i ch and used as recei ved. Acr yl i c aci d con-
cent rat i on is gi ven in mass per cent age; N, N' -
met hyl ene bis acr yl ami de ( cr ossl i nker ) con-
cent rat i on (Cr) is cal cul at ed in mol ar per cent age
referri ng t o acryl i c aci d mol e number . Wat er
used for nanofi l t rat i on exper i ences was ei t her
reverse os mos i s wat er f or synt het i c sol ut i ons, or
di r ect l y t ap water.
The phot or eact or (Fig. 1) is c o mp o s e d by a
cyl i ndri cal c ha mbe r in whi ch me mbr a ne has
been i nt r oduced. A UV l amp is pl aced in t he
cent re o f reactor. It is a pol ychr omat i c l amp,
model UV Hanau Her aeus TQ 150 ( Hg me d i u m
pressure). The cool i ng syst em can be in gl ass
(Duran 50) or in quartz, so t hat wavel engt hs
r ecei ved by t he me mbr a ne can be cont r ol l ed
(glass ~ >300 nm, quart z ~, >220 nm) . Di f f er ent
S. Bequet et aL / Desalination 131 (2000) 299-305 301
light intensities are received by the membrane
when the cooling system is made in glass
(470 mW/ cm 2) or in quartz (248 mW/cm2).
Experimental met hod was as follows: as-
casted membrane was washed with reverse
osmosis water, introduced in the reactor, then
laid on the wall (skin exposed to UV light).
Solutions of acrylic acid and crosslinker, in
adequate concentration, were bubbled with
nitrogen. The UV lamp was dipped in the reactor
when its permanent regime had been established
(t = 60 s). The membrane was irradiated during a
given time. Then it was put out of the reactor and
washed with reverse osmosis water. Membranes
were kept in aqueous formal dehyde solution (1%
weight).
Wat er flux was measured at 400 kPa using
reverse osmosis wat er with an unstirred cell
(Ami con 8050) having a 50ml volume and
13.2 cm 2 active membrane area. The rejection of
ions was measured for a volumic reduction
factor, VRF, equal to 15 (94% recovery). Values
wer e obtained, either for a synthetic solution of
calcium chloride at 50 mg. F 1 in calcium ion, or
wi t h tap wat er (calcium concentration bet ween
20 and 50 mg.l-~).
Calcium concentrations in initial feed Co and
in total permeat e C o were measured by a sequen-
tial plasma torch ICP (Jobin Yvon JY 24). The
mean rejection of calcium was then given by
Eq. (1):
- - Cp
R = 1 ( 1)
Co
Solutions for UV spectophotometry were
prepared in 1-methyl 2-pyrrolidinone for poly-
sulfone and reverse osmosis wat er for mono-
mers. They were analyzed using a UV spectro-
phot omet er Ant hel i e (Secoman).
3. Resul t s and di scussi on
First works carried out in the laboratory have
shown that t he photografting technology can be
used to elaborate nanofiltration membrane from
an ultrafiltration membrane. Agei ng tests (7 d in
reverse osmosis water at 60C) have shown that
surface modification obtained was not stable. An
addition of a low quantity of crosslinker has t hen
been proved efficient to stabilise t he
modification in this ageing test (Table 1). The
uncrosslinking surface modification is not stable:
as indicated by the flux and calcium rejection
values whi ch are the same, than for the initial
ultrafiltration membrane. In opposite, cross-
linking surface modification is permanent: flux
and calcium rejection are maintained after the
ageing test.
3.1. Acrylic acid concentration o f grafting solu-
tion influence onto membrane properties
Flux and rejection values vs. acrylic acid
concentration are represented in Fig. 2. This
study has been done with a quartz reactor and a
0.1% crosslinker concentration.
At constant irradiation time, we can observe a
flux decrease and a rejection increase with an
acrylic acid concentration increase. However, in
these conditions the flux and rejection values are
still relatively low.
Flux and rejection tend respectively to a
constant value when acrylic acid concentration
increases. From Fig. 2 we can say that the
grafting is no more efficient when acrylic acid
concentration is higher than 5%: flux and
calcium rejection tends to a constant value. This
efficiency decrease could be explained by UV
absorption. With an increase of acrylic acid
concentration, UV absorption by the solution
increases (Beer-Lambert law) and light intensity
received by membrane decreases. Therefore, the
number of radicals created on the membrane
surface decreases.
Consequently to this important absorption,
UV irradiation will also induce copolymerization
of acid acrylic and crosslinker in solution. We
have observed that solution viscosity increases
3 0 2 S. Bequet et aL / Desalination 131 (2000) 299- 305
Table 1
Crosslinker effect on ageing test (7 d in reverse osmosis water at 60C); Grafting parameters: CAA = 2.5%, [Cr] = 0.5%
Irradiation Without crosslinker With crosslinker
time, min Calcium rejection, % Flux (400 kPa), I.h -~.m 2. Calcium rejection, % Flux (400 kPa), l.h -~.m -2.
at 25C at 25C
Before After 7 d Before After 7 d Before After 7 d Before After 7 d
2 14 0 55.6 360 7 7 6.8 6.8
3 39 0 4.4 380 18 17 3.6 3.6
5 51 0 2.8 368 57 57 1.6 1.6
. . . . . . . . . . . . . . T 2 5
6 ~ L
" ~ " . . . . . . . . i g
~ " i " x . ' " + F l u x i " ~
" " ~ . " - - - R e j e c t i o n ] 1 o
I o i
0 ~.0
0 2 4 6 8
Acrylic a c i d concentration ( we i ght % )
Fig. 2. Acrylic acid concentration influence on membrane
properties; Quartz reactor [Cr] = 0.1%, t~ = 5 rain; Pressure
= 400 kPa, [Ca 2+] = 50 mg.F I.
3.5 I
3 " r k ' X . . . . . . . . . . . " " ' P o l y s u l f o n e
i - - A c r y l i c a c i d + c r o s s l i n k e r
oa
2 ~ ,~ . . . . . ~% . . . . . . . . . . . . . . . . . .
: ' ~ ' . .
~ ; ~ ;
240 290 340 390 440
W a v e l e n g t h s ( n m )
Fig. 3. Polysulfone and monomers solutions UV spectra.
up t o gel f or mat i on and a l ayer is deposi t ed on
t he react or wal l s and on the UV lamp. This
sol ut i on copol ymer i sat i on is a dr awback for an
i ndust ri al scal e-up for several reasons. First, this
sol ut i on copol ymer i sat i on consumes monomer s,
whi ch are not used for graft i ng, and produces by-
product s. Secondl y, t he deposi t ed layer on the
wal l of t he UV l amp woul d cause a progressive
decrease o f the light i nt ensi t y i nduci ng a varia-
t i on in pr oduct quality.
Fr om this point, there are t wo probl ems to
solve. How to i mprove fl uxes and cal ci um
rej ect i on and reduce sol ut i on copol ymer i sat i on?
First, we can adj ust graft i ng paramet ers as such
as irradiation time, wavel engt h, light i nt ensi t y or
crosslinker concent rat i on. Secondl y, we can
choose di fferent pol ysul f one membr anes wi t h
varyi ng pore size, porosi t y or permeabi l i t y. In
this paper, first dat a on the i nfl uence of
irradiation time, wavel engt h and support
membr anes are presented. Nevert hel ess, the first
poi nt to st udy is how to limit sol ut i on copol y-
merisation.
3.2. Wavelengths and irradiation time influence
onto membrane properties
Trials conduct ed wi t h a wavel engt h
irradiation above 3 0 0 n m (by usi ng a glass
cool i ng syst em) s how a l ower copol ymer i sat i on
S. Bequet et al. / Desalination 131 (2000) 299-305 303
35
~,~o
E
r. "25
e-
=~o
" 5
i - ' G l a s s reactor _
, , " ~- " Quar l z reactor
l l
-d
1.5 2 2.5
I rradi ati on t i me (mn)
16
A14
o
0
' E a
2
- - e - Q u a r t z reactor
- * - Gl ass reactor / / - ' ~
1.5 2 2.5
Irradiation t i me (ran)
Fig. 4. Wavelengths and irradiation time effects on flux
values; CAA = 2.5%, [Cr] = 0.5%; Pressure = 400 kPa,
[Ca 2+] = 50 mg.i -1.
Fig. 5. Wavelengths and irradiation time effects on
calcium rejection values; CAA=2.5%, [Cr]=0.5%;
Pressure = 400 kPa, [Ca 2+] = 50 mg/l.
(limpid solution, no deposit and no viscosity
increase). This is explained by examining the
absorption spectra of monomers and polysulfone
solutions. For wavelengths above 300 nm, Fig. 3
shows that photosensibilisation of polysulfone is
still possible without photoinitiator (i.e: poly-
sulfone absorbs UV light). In the same con-
ditions, monomers solution does not absorb and
then the copolymerisation of acrylic acid and
crosslinker is certainly negligible.
Agei ng tests equally show that nanofiltration
membranes in these conditions are able to resist
to ageing test for 7 d at 60C.
Figs. 4 and 5 show the influence of wave-
lengths range and irradiation time on flux and
rejection. These figures show that an increase in
irradiation time gives lower fluxes and higher
calcium rejection.
These results also indicate that, for an
irradiation with a glass reactor, a significant
i mprovement of performance is possible in
comparison with an irradiation with a quartz
reactor: flux values are 6 times higher while
rejection values remain the same. Wavelengths
used are consequently a very important
parameter since experiments with ~, >300nm
rather t han ~,>220 nm can limit copol ymer
formation in solution and increase the flux at
constant rejection value.
Explanation of this phenomenon is not easy.
On the one hand, light intensity received by
membrane surface is not t he same for a glass and
a quartz reactor, this can probably induce a
difference in the number of radicals created onto
the polysulfone surface and in depth into the skin
layer grafting. On the other hand, the grafted
layer composition obtained with glass reactor is
different from the grafted layer formed using
quartz reactor. In the latter, the grafted copoly-
mer chains are physically associated with copo-
lymers formed in solution.
3.3. Membr ane support i nf l uence
Anot her idea to improve the flux of the
modified membrane is to use a porous support of
the membrane having a higher initial wat er flux.
In adjusting the parameters of the inverse phase
process, different ultrafiltration membranes can
be prepared with a wat er flux in the range 400 to
2400 l.h -l.m -~ (at 400 kPa). Aft er irradiation with
the quartz reactor, the results given in Fig. 6
show that when the support membrane
permeability increases, t he fl ux for t he modi fi ed
304 S. Bequet et al. /Desalination 131 (2000)299-305
8 ! . . . . . : I i 1
7 i "*'Flux + ~ -t- !o9
" ~ [ " ~ - R e i e c t i o n ] ! I ~ 0 . 8
t ~ 5 i ~ - i : ! - ~ ! . , :
'~2 ! ~!=k. ~ - ? . ~ i 03 "~
= 1 { - : i - r ~ - : 0 1
0 5 0 0 1 0 0 0 1500 2 0 0 0 2 5 0 0 3 0 0 0
F l u x ( i n i t i a l membrane) at 25C ( l . h q . m "2)
Table 2
Calcium rejection and flux vs. water type (CAA = 2.5%,
[Cr] = 0.5%, glass reactor)
Irradiation Calcium rejection, % Flux (400 kPa),
time, rain l.hq.m -2 at 25C
Synthetic Tap water Synthetic Tap water
solution solution
1. 5 ] 24 32 24
2 5 35 18 12
2.5 7 68 6 4
Fig. 6. Ultrafiltmtion membrane porosity effect on flux and
rejection values; Quartz reactor, [Cr] = 0.5%, t~= 5 min;
Pressure = 400 kPa, [Ca 2+] = 50 mg/l.
membrane increases and rejection does not drop
significantly. This unexpected important
observat i on woul d deserve analyses, more
particularly ultrafiltration membrane needs to be
characterised in terms of pore size, porosity.
3. 4. Tap wat er sof t eni ng
In order to prove the efficiency of the
modified membranes for tap water softening,
f ew tests were performed. The results in Table 2
show that calcium rejection is higher (and flux
l ower) with tap water than with synthetic
solution. This difference can be attributed to
several parameters. In particular, tap water
contains bivalent anions (as SO42-) and repulsion
is more important for these bivalent ions than for
chloride ions, and calcium rejection increase can
be expected. Anot her effect is the pH, which has
an influence on poly(acrylic acid) properties
( pKa=4. 8) . So, with a synthetic solution
( p H=5 . 5 ) , few carboxylic groups are still
protonated, inducing a decrease o f the charge
density. For tap wat er (pH = 7.5), all acid groups
are ionised ( CO0 - form) and the membrane has
a maxi mum charge density. This charge density
difference combined to grafted copol ymer chains
swelling might also be an explanation for the
performance obtained with tap water. The higher
rejection woul d also induce a stronger concent-
ration polarisation effect and woul d also explain
the lower flux.
4 . C o n c l u s i o n s
A novel route to the preparation of nano-
filtration membrane has been demonstrated. The
method consists in photografting a pol ymer on
the surface of porous membranes. The method
was tested with polysulfone ultrafiltration mem-
branes and poly(acrylic acid) in water solution. It
has been shown that the surface modification is
stabilised by crosslinking with N, N' -methylene
bis acrylamide and that the grafting does not
require any photoinitiator and any purification of
the reactants. The water softening properties of
the novel membranes were measured in dead end
mode and at a 95% recovery. A large range of
performances has been obtained which demon-
strated the feasibility of the UV irradiation as an
environmentally friendly and versatile techno-
logy to prepare nanofiltration membranes.
Several ways to improve the membrane
performance have been discussed. Flux was ira-
S. Bequet et al. / Desalination 131 (2000) 299- 305 305
proved by a factor 6 by choosing a wavelength
range above 300 nm without any significant drop
in calcium rejection; moreover in this irradiation
condition, no significant copolymerisation inside
the monomer s solution occurred. Flux was also
improved by a factor 3 by using a more per-
meable support membrane without any signi-
ficant change in calcium rejection. Irradiation
time and monomers concentration could be
optimised in order to adopt the membrane
performance to a defi ned application.
This preliminary study will be continued to
clarify multiple points that have been raised
during this work. To achieve the objective it is
necessary to determine ultrafiltration membrane
characteristics grafting yield and the grafted
layer structure (inside the membrane and on the
surface).
The influence of grafting parameters (light
intensity, wavelengths, irradiation time and
monomer concentration) on membrane charac-
teristics and properties will be also further
examined.
Acknowledgements
Financial support for this wor k was provided
by the R6gion Midi Pyr~n~ees and by the Centre
National de la Recherche Scientifique (Pro-
gramm ECODEV-PROSETIA). The authors
want to thank Miss Sandrine Descl aux for her
excellent technical assistance.
References
[1] M. NystrOm and P. Jarvinen, J. Membr. Sci., 60
(1991)275.
[2] Y.M. Lee, S.Y. lhm, J.K. Shim, JM. Kim, C.S. Cho,
Y.K. Sung, Polymer, 36 (1995) 81.
[3] M. Ulbricht and M. Riedel, Biomatefials, 19 (1998)
1229.
[4] M. Ulbdcht, React. and Funct. Poly., 31 (1996) 165.
[5] H. Yamagishi, J.V. Crivello and G. Belfort, J. Membr.
Sci., 105 (1995) 237.
[6] W.R. Bowen and H. Muktar, J. Membr. Sci., 122
(1996) 263.
[7] C. Combe, C. Guizard, P. Aimar and V. Sanchez,
J. Membr. Sci., 129 (1997) 147.