Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=lspr20
Separation & Purification Reviews
ISSN: 1542-2119 (Print) 1542-2127 (Online) Journal homepage: http://www.tandfonline.com/loi/lspr20
Organic Solvent Nanofiltration in PharmaceuticalIndustry
M. G. Buonomenna & J. Bae
To cite this article: M. G. Buonomenna & J. Bae (2015) Organic Solvent Nanofiltrationin Pharmaceutical Industry, Separation & Purification Reviews, 44:2, 157-182, DOI:10.1080/15422119.2014.918884
To link to this article: https://doi.org/10.1080/15422119.2014.918884
Accepted author version posted online: 15Jul 2014.Published online: 15 Jul 2014.
Submit your article to this journal
Article views: 1054
View Crossmark data
Citing articles: 34 View citing articles
Separation & Purification Reviews, 44: 157–182, 2015Copyright © Taylor & Francis Group, LLCISSN: 1542-2119 print / 1542-2127 onlineDOI: 10.1080/15422119.2014.918884
Organic Solvent Nanofiltration in PharmaceuticalIndustry
M. G. Buonomenna1 and J. Bae2
1Ordine dei Chimici della Campania, Napoli, Italy2Department of Applied Chemistry, Dongduk Women’s University, Seoul, Republic of Korea
Organic solvent nanofiltration (OSN) is a promising energy and waste efficient unit processto separate mixtures down to a molecular level, which gained attention in the pharmaceuti-cal industry, in particular in the process development of Active Pharmaceutical Ingredients(APIs). This article focuses on all aspects related to OSN (i.e., membrane materials, commer-cial membranes, transport theories, applications) to understand the role of this technology inpharmaceutical industry. The most important results in the last five years on OSN applicationsin the process development of APIs are reviewed extensively.
Keywords: Organic solvent nanofiltration, solvent resistant nanofiltration, active pharmaceu-tical ingredients (APIs), solvent exchange, product purification, genotoxin impurities (GTIs),membrane enhanced peptide synthesis (MEPS), hybrid processes
INTRODUCTION
Separation processes are of utmost importance for the chem-ical and pharmaceutical industry: 50 to 90% of the capitalinvestments in the chemical industry involve separation pro-cesses. Organic syntheses are often carried out in organicsolvents and involve products with high added value thathave to be separated from the organic solvents. All theseprocesses are solvent intensive.
In these last years, production plants increased focus onprocess energy efficiency as well as mass efficiency (1).The overall mass efficiency of a process can be improvedthrough solvent recovery and recycle, and recovery process-ing is currently practiced in various chemical industries.Solvent recovery can offer significant benefits with regardsto reduced purchase, storage and waste costs. Increased com-pliance with environmental legislation and reduced emissionof greenhouse gases are also important. Solvent use has
Received 17 September 2013, Accepted 15 April 2014.Address correspondence to M. G. Buonomenna, Ordine dei Chimici
della Campania, Via A. Tari 22, 80138 Napoli, Italy. E-mail: [email protected]
Color versions of one or more of the figures in the article can be foundonline at www.tandfonline.com/lspr.
been reported to account for approximately 60% of the over-all energy consumption for active pharmaceutical ingredient(API) production indicating that solvent recovery could be ofinterest for improving energy efficiency (2).
In this context, organic solvent nanofiltration (OSN), orsolvent resistant nanofiltration (SRNF) represents an inter-esting membrane technology with enormous potential as itallows separations of organic mixtures down to a molecularlevel by simply applying a pressure gradient over a mem-brane and recovery with possibility of reuse of the organicsolvents.
In fact, one of the significant and recognized benefitsof membrane operations is their low direct energy con-sumption because of the absence of phase transformations.Another feature is the possibility for reducing indirect energyconsumption through the recycling and reuse of raw mate-rials and secondary materials minimizing the formation ofwastes. In addition, the association of membrane opera-tions with other conventional techniques or working withhybrid systems combining or integrating different membraneoperations can lead to more rationale applications (3–5).
Traditional applications for industrial membrane separa-tions in liquid systems have been in a water environment.Membrane materials are now available that can work inorganic solvents. In the range of membrane filtrations where
158 M. G. BUONOMENNA AND J. BAE
discrimination occurs at molecular level, mutual interac-tions between solute and solvent, solvent and membrane,as well as between solute and membrane play a key rolein addition to mere molecular size. It renders selection ofa suitable membrane type for a given separation relativelydifficult and requires a multidisciplinary approach, involvingmaterials science and engineering, chemical synthesis andcharacterization of membrane materials as well as membranemanufacturing, modification and module design. The firstlarge-scale application of OSN technology was in solventrecovery from the dewaxing operation in lubes processing(6, 7).
In a pharmaceutical context, OSN-membranes can beapplied in drug synthesis between reaction steps or in thedownstream processing. Practically, OSN can be used toeither retain a larger target molecule, or allow the targetmolecule to permeate while retaining the impurity. In caseof thermo-labile compounds, OSN has an additional bene-fit compared to conventional thermal unit operations such asdistillation.
In 2008, Vandezande et al. (8) published a critical reviewon all aspects related to OSN and, concerning its appli-cations, the authors reviewed exhaustively literature up to2007 on applications in food chemistry, petrochemistry,catalysis and pharmaceutical manufacturing. In the sameyear, Volkov et al. (9) reviewed the applications of OSN inpetrochemical industry, homogeneous catalysis, separationof ionic liquids, and for the solvent exchange in multisteporganic synthesis and in food industry. In the chapter byPeeva et al. (10), dated 2010, after two sections on typi-cal membranes for OSN and their characterization, the mainapplications of OSN in fine chemical and pharmaceuticalsynthesis, food and beverage, and refining were reported.
In this review, after an updated overview of the mem-brane materials and transport theories of OSN, the role of thistechnology in the specific field of pharmaceutical industry isdiscussed on the basis of recent literature and specificallyin nonthermal solvent exchange, removal of excess reagents,reaction product purification and peptides synthesis.
OSN: BACKGROUND
Membrane Materials
Polymeric Membranes
Nowadays the majority of OSN membranes are based onpolymers. The reasons for that are wide choice of materials,relatively easy processing and good reproducibility. It is alsomuch easier to tailor polymeric membranes to the applica-tion compared to ceramic membranes. In Table 1, polymericmaterials used to manufacture nanofiltration membranes fornonaqueous solvents are reported. In Table 2 a list of somecommercial membranes with their characteristics is given.Commercial OSN membranes include membranes both
especially designed for OSN (such as Starmem, DuraMem,PuraMem, HITK-T1, etc.) and developed for the watertreatment segment market (Desal-5 and Desal-5-DK).
Cross-linked silicon rubbers are used to produce compos-ite membranes in which the selective layer is deposited onthe porous supporting material, generally polyacrylonitrile(PAN) or polyimide porous supports. Commercially avail-able composite polydimethyl siloxane (PDMS)-based mem-branes are: Puramem S380 from Evonik (MET Ltd., UK)developed for filtrations with hydrophobic solvents such asalkanes, consisting of a cross-linked PDMS layer on cross-linked polyimide (PI) (see Table 2, entry 4); Solsep series(some Solsep membranes were proven to have a silicontop layer), Table 2, entries 6-10; GMT-oNF-2 (Table 2,entry 11).
Zeidler et al. (11) reported a systematical experimen-tal investigation of the separation behavior concerning theinteractions between a PDMS-based membrane (GMT-oNF-2) and various solvents and the influence of functional groupsin the solute molecules to identify the parameters which sig-nificantly influence membrane performance. The solventschosen for the study resemble those ones used in the vari-ous synthesis steps in the production of specialty chemicals:ethanol (polar), n-heptane (nonpolar) and tetrahydrofuran(polar aprotic). The solubility parameters of the membrane,the solvent, and the solute (calculated with a group con-tribution method) represent the easily accessible tool usedby the authors to predict the separation behavior of a densePDMS membrane and to estimate the expected rejection ofthe solute molecules investigated.
The strategy used to improve the performance of silicone-based membranes for separations in nonpolar solvents con-sists in filling PDMS top layer with different filler such asnoble metal nanoparticles (and/or treating via photo-thermalheating (12)), zeolites (13, 14), or molecular organic frame-works (MOFs) (15). The role of these fillers in the PDMSmatrix is different. Noble metal nanoparticles release effi-ciently heat under optical excitation (electromagnetic energyis converted into thermal energy). So, the study of the effectof these nanoparticles on membrane performance is, in real-ity, related to the influence of the photothermal heatingon membrane fluxes without altering selectivity: a laser iscoupled to filtration process. Li et al. (12) indicated that pho-tothermal heating improved PDMS membrane fluxes withoutsignificantly lowering selectivity.
Zeolites and MOFs are porous crystalline materials thatact in the polymeric matrix as molecular sieves: in idealcases, they should discriminate between the organic solventand the solute molecules enhancing the selectivity. Basuet al. (15) reported that retention of mixed matrix mem-branes based on PDMS and MOFs in the separation of RoseBengal from isopropanol was significantly better than whenpure PDMS membranes were used. A size exclusion effectof the filler and reduced polymer swelling were responsiblefor the enhanced separation. Considering polymer swelling,
ORGANIC SOLVENT NANOFILTRATION AND PHARMACEUTICALS 159
TABLE 1Polymeric materials used for the preparation of membranes for organic solvent nanofiltration (OSN)
Polymer class Polymer material Formula
Highly permeablerubbers
Silicon rubbers variation ofpolydimethyl siloxanes(PDMS) Si O
R
CH3
R OSi
CH3
CH3
OSi
CH3
CH3x y
Si
CH3
CH3
R
R=OH, –CH=CH2, or another alkyl or aryl group
Low permeablepolymeric glasses
Polyimides (PI)
N
O
O
N
O
O
R
Polysulfones (PS)
S
O
O
O
n
Polyamides (PA)
C
O
N
H
n
Polybenzilimidazole (PBI)
N
N
H
N
H
N
n
Poly(ether ether ketones) (PEEK)
O O C
O
n
Highly permeablepolymeric glasses
Poly(trimethylsilyl) propylene(PTMSP)
(Continued)
160 M. G. BUONOMENNA AND J. BAE
TABLE 1(Continued)
Polymer class Polymer material Formula
Polymethyl pentene (PMP)
polymer of intrinsic microporosity(PIM)
the authors related it to sorption capacity. The experimentalresults showed a reduced sorption capacity with increasingMOF loading, reflecting a cross-linking effect of the fillerson the membrane polymer. Analogous results were observedwhen zeolite was used as filler (13).
The other class of polymers reported in Table 1 is thatone of glassy polymers that can be low or high permeablepolymers, characterized by different free volume degrees.With polyimides (16–18), polysulfones (19–21), poly(etherether ketones) (PEEK) (22–24) (i.e., low permeable glassypolymers) (Table 1), integrally skinned membranes, in whichselective and support layers are made of the same polymericmaterial, are produced via phase inversion. Many of theOSN membranes developed are integrally skinned. The poly-mers used for their preparation may contain various additivessuch as stabilizers and flame retardants that can influencethe preparation process (i.e., phase separation) (25). Withpolyamides, thin film composite (TFC) membranes, in whichselective and support layers are made of different materi-als, can be prepared by interfacial polymerization. Thesemembranes have the potential to achieve higher fluxes thanintegrally asymmetric OSN membranes (26), even thoughfor their use in nonpolar solvents, such as toluene, theexternal surface properties of these membranes need to bemodified (27).
Many more materials could be used to prepare OSN mem-branes. In particular, intensive research is performed withcommercially available glassy polymers that were not explic-itly developed as membrane materials for OSN. Amongthese polymers, polybenzilimidazole (PBI) and PEEK havegained attention for their chemical stability and mechanicalstrength. PBI, possesses thermal, mechanical and chemicalstability in corrosive environments, with excellent stabilitytowards acids and bases (28). Valtcheva et al. (29) report newOSN membranes based on PBI for applications in organicsolvents containing acids or bases. The new OSN mem-branes exhibit superior chemical stability compared to otherwell-known polymeric membranes such as the polyimideones.
PEEK is an interesting material for OSN membranesbecause it shows very low or no solubility in ordinary sol-vents (30). Recently, Peeva et al. (31) reported a one pot,long-term continuous Heck coupling reaction by means ofOSN performed in N,N-dimethylformamide (DMF) at 80◦Cwith a PEEK-based membrane to retain efficiently the Pdcatalyst. The reaction results obtained in the comparison toPBI and aminopropyl triethoxysilane (APTS) cross-linked PImembranes showed that only the PEEK membrane appearedto have no effect on the reaction rate: the use of APTS cross-linked PI membrane reduced the reaction rate, while the PBImembrane seemed to inhibit the reaction (via Pd catalystinhibition). It was hypothesized by the authors that the nitro-gen of the imidazolium ring (see PBI monomer unit structurein Table 1) had chelating properties quenching the Pd atoms(31).
Modified PEEK polymers were synthesized, making itpossible to prepare membranes via the phase inversion pro-cess. PEEK-WC is a PEEK polymer with a cardo-groupin the polymer backbone. This polymer has been reportedas a material for gas separation, nanofiltration (32–36) andseparation of isopropanol solutions of Rose Bengal (40),a dye used for retention screening of membrane materi-als in OSN (37–40). However, PEEK-WC membranes arenot stable in polar aprotic solvents such as acetone, DMFor acetonitrile. To overcome this problem, Hendrix et al.synthesized a PEEK polymer modified with a valeric acidgroup (VA-PEEK) (41) used for membrane cross-linkingwith diamines.
After activation of the carboxylic acid in the polymercasting solution, the cast polymer film is cross-linked bydiamines that were dissolved in the coagulation bath, as pre-viously reported for PI membranes (37, 42). Cross-linking,which implies that the polymer chains are covalently boundtogether, creating an interconnected polymeric network inthe membrane, is a useful strategy not only for OSN mem-branes but also for gas separation membranes to reduceswelling and plasticization. Vanherck et al. (43) reviewedmany cross-linking methods for polyimide (PI) membranes
TAB
LE2
Com
mer
cial
hydr
opho
bic
(PH
OB
)an
dhy
drop
hilic
(PH
IL)
mem
bran
esfo
ror
gani
cso
lven
tnan
ofiltr
atio
n
Ent
ryM
embr
ane
nam
eM
anuf
actu
rer
Mem
bran
ech
arac
ter
Mol
ecul
arw
eigh
tcu
t-of
f(M
WC
O)
Mem
bran
ety
pean
dm
ater
ial
Solv
ents
tabi
lity
aA
ppli
cati
ons
Ref
.
1M
PF-5
0∗K
och
Mem
bran
eSy
stem
s(U
SA)
PHO
B70
0D
a(b
ased
onre
ject
ion
ofSu
dan)
TFC
,com
pris
ing
ade
nse
silic
one
top
laye
ron
apo
rous
cros
s-lin
ked
PAN
-bas
edsu
ppor
t
Alc
ohol
s,ke
tone
s,es
ters
,alk
ylha
lides
,alk
anes
Rec
over
yof
orga
nom
etal
licco
mpl
exes
from
DC
M,T
HF,
ethy
lace
tate
;rec
over
yof
phas
etr
ansf
erca
taly
sts
from
tolu
ene,
sepa
ratio
nof
tigly
ceri
des
from
hexa
ne,s
olve
ntex
chan
gein
phar
mac
eutic
alm
anuf
actu
ring
79–8
2
2M
PF-4
4∗K
och
Mem
bran
eSy
stem
s(U
SA)
PHIL
250
Da
(bas
edon
reje
ctio
nof
gluc
ose
inw
ater
)
TFC
,com
pris
ing
ade
nse
silic
one
top
laye
ron
apo
rous
cros
s-lin
ked
PAN
-bas
edsu
ppor
t
Aqu
eous
mix
ture
sof
low
eral
coho
ls,h
ydro
carb
ons,
chlo
rina
ted
solv
ents
(DC
M,
chlo
rofo
rm),
arom
atic
s(t
olue
ne,x
ylen
e),k
eton
es(M
EK
),di
ethy
leth
er,e
thyl
acet
ate,
cycl
ohex
ane,
prop
ylen
eox
ide,
acet
onitr
ile,T
HF,
1,4-
diox
ane.
Rec
over
yof
clin
dam
ycin
from
ferm
enta
tion
was
tew
ater
.83
3St
arm
em12
0,12
2,22
8,24
0
bW
.R.G
race
-Dav
ison
(USA
),di
stri
bute
dby
Mem
bran
eE
xtra
ctio
nTe
chno
logy
(ME
T)
(UK
)
PHO
B20
0,22
0,28
0,40
0D
acIn
tegr
ally
skin
ned
asym
met
ric
mem
bran
esw
ithac
tive
surf
aces
base
don
poly
imid
e(p
ore
size
<5
nm)
Alc
ohol
s(e
.g.,
buta
nol,
etha
nol,
and
iso-
prop
anol
);al
kane
s(e
.g.,
hexa
nean
dhe
ptan
e);a
rom
atic
s(e
.g.,
tolu
ene
and
xyle
ne);
ethe
rs(e
.g.,
met
hyl-
tert
-but
yl-
ethe
r);k
eton
es(e
.g.,
met
hyl-
ethy
l-ke
tone
and
met
hyl-
isob
utyl
-ket
one)
;an
dot
hers
(e.g
.,bu
tyl
acet
ate
ande
thyl
acet
ate)
.
Solv
entr
ecov
ery
inlu
beoi
lde
wax
ing
and
arom
atic
sen
rich
men
ts,p
harm
aceu
tical
man
ufac
turi
ngfo
rso
lven
tex
chan
gean
dm
icro
fluid
icpu
rific
atio
n,bi
otra
nsfo
rmat
ions
inm
embr
ane
bior
eact
ors
(MB
Rs)
84–8
7
4Pu
raM
emS3
80E
voni
kM
ET
(UK
)PH
OB
600
Da
TFC
with
cros
s-lin
ked
PDM
Sla
yer
onpo
lyim
ide
ultr
afiltr
atio
nsu
ppor
t
Hep
tane
,hex
ane
App
licat
ions
inap
olar
hydr
ocar
bon-
type
solv
ents
88
5D
uraM
em,
Pura
mem
280
Evo
nik
ME
T(U
K)
PHO
B15
0–90
0D
aIn
tegr
ally
skin
ned
asym
met
ric
cros
s-lin
ked
poly
imid
e-ba
sed
mem
bran
e
Ace
tone
,eth
anol
,met
hano
l,te
trah
ydro
fura
n,di
met
hylf
orm
amid
e∗,
dim
ethy
lsul
foxi
de∗ ,
dim
ethy
lace
tam
ide∗
,is
opro
pano
l,ac
eton
itrile
,m
ethy
leth
ylke
tone
,eth
ylac
etat
e.
App
licat
ions
inm
ostc
alle
dag
gres
sive
solv
ents
such
asso
lven
tsof
the
pola
rap
rotic
solv
entf
amily
(with
the
aste
risk
)
88
6So
lSep
d
NF0
1020
6So
lSep
(The
Net
herl
ands
)−
300
Da,
R95
%T
FCA
lcoh
ols,
este
rsU
pgra
ding
ofso
lven
tmix
ture
sfo
rre
use
inin
dust
rial
appl
icat
ions
89,9
0,91
(Con
tinu
ed)
161
TAB
LE2
(Con
tinue
d)
Ent
ryM
embr
ane
nam
eM
anuf
actu
rer
Mem
bran
ech
arac
ter
Mol
ecul
arw
eigh
tcu
t-of
f(M
WC
O)
Mem
bran
ety
pean
dm
ater
ial
Solv
ents
tabi
lity
aA
ppli
cati
ons
Ref
.
7So
lSep
NF0
1030
6So
lSep
(The
Net
herl
ands
)−
500
Da,
R95
%A
lcoh
ols,
este
rs,k
eton
es,
arom
atic
s,ch
lori
nate
dso
lven
ts,
redu
cing
atm
osph
eres
Upg
radi
ngof
solv
entm
ixtu
res
for
reus
ein
indu
stri
alap
plic
atio
ns89
,90,
91
8So
lSep
NF0
3030
6So
lSep
(The
Net
herl
ands
)−
500
Da,
R95
%(A
ceto
ne);
300
Da,
R99
%(a
lcoh
ols)
TFC
Alc
ohol
s,es
ters
,ket
ones
,ar
omat
ics,
chlo
rina
ted
solv
ents
,re
duci
ngat
mos
pher
es
Upg
radi
ngof
solv
entm
ixtu
res
for
reus
ein
indu
stri
alap
plic
atio
ns92
9So
lSep
NF0
3030
6FSo
lSep
(The
Net
herl
ands
)−
300
Da,
R95
%(A
ceto
ne);
300
Da,
R95
%(E
thyl
acet
ate)
TFC
Alc
ohol
s,ke
tone
s,ar
omat
ics,
chlo
rina
ted
solv
ents
Upg
radi
ngof
solv
entm
ixtu
res
for
reus
ein
indu
stri
alap
plic
atio
ns89
,90,
91
10So
lSep
NF0
3010
5So
lSep
(The
Net
herl
ands
)−
300
Da,
R95
%(E
than
ol,
met
hano
l);
750
Da,
R95
%(A
ceto
ne)
TFC
Alc
ohol
s,ar
omat
ics,
keto
nes
Upg
radi
ngof
solv
entm
ixtu
res
for
reus
ein
indu
stri
alap
plic
atio
ns89
,90,
91
11G
MT-
oNF-
2G
MT
Mem
bran
tech
nik
Gm
bH(G
erm
any)
PHO
BT
FCco
mpr
isin
ga
PDM
S-ba
sed
top
laye
ron
aPA
N(P
olya
cryl
onitr
ile)
−−
93
12D
esal
-5G
E/O
smon
ics
PHIL
340
Da,
R96
%T
FCco
mpr
isin
ga
poly
(pip
eraz
ione
amid
e)to
pla
yer
with
anin
term
edia
tesu
lfon
ated
PSf
laye
r
Tolu
ene,
ethy
lace
tate
,met
hano
l,di
chlo
rom
etha
neSe
para
tion
ofol
eic
acid
from
met
hano
l;se
para
tion
ofPd
-BIN
AP
and
Wilk
inso
nca
taly
stfr
omdi
chlo
rom
etha
ne
94–9
6
13D
esal
-5-D
KG
E/O
smon
ics
PHIL
340
Da,
R96
%T
FCco
mpr
isin
ga
PA-b
ased
top
laye
rL
imite
dch
emic
alst
abili
tyin
solv
ents
such
aset
hyla
ceta
tean
dto
luen
e
−96
14H
ITK
-T1
HIT
K(G
erm
any)
PHO
B22
0D
aSi
lyla
ted
TiO
2-b
ased
cera
mic
mem
bran
eM
etha
nol,
acet
one;
apol
arso
lven
tsR
etai
nof
tran
sitio
n-m
etal
cata
lyst
sin
apol
arso
lven
ts95
,98
15In
opor
0.9
nmT
iO2
Inop
orG
mbH
PHIL
450
TiO
2M
etha
nol,
2-pr
opan
ol,
tetr
ahyd
rofu
ran
(TH
F),
N,N
-dim
ethy
lfor
mam
ide.
99
16In
opor
3nm
Zr
PHO
BIn
opor
Gm
bHPH
OB
600
ZrO
2w
itha
sila
neto
pla
yer
Apo
lar
solv
ents
99
a Acc
ordi
ngto
the
man
ufac
ture
r’s
info
rmat
ion;
∗ Not
Ava
ilabl
eon
the
mar
ket(
MPF
-50,
sinc
e20
07).
Onl
yM
PF-3
4m
embr
ane
isst
illav
aila
ble,
insp
iral
-wou
nd(M
PS-3
4)m
odul
eco
nfigu
ratio
n.bA
sof
Mar
ch1,
2010
,Evo
nik
Mem
bran
eE
xtra
ctio
nTe
chno
logy
Lim
ited
oper
ates
asa
subs
idia
ryof
Evo
nik
Indu
stri
esA
G,M
arch
2010
.Evo
nik
ME
Tdi
stri
bute
son
the
mar
ketD
uraM
em(fi
rstg
ener
atio
n,G
1,so
lutio
n)an
dPu
raM
em(G
2so
lutio
n)se
ries
.B
oth
Dur
aMem
and
Pura
Mem
are
mad
efr
omP8
4po
lyim
ide.
Dur
aMem
mem
bran
ese
ries
for
stab
ility
inso
lven
ts,
such
asac
eton
e,et
hano
l,m
etha
nol,
tetr
ahyd
rofu
ran,
dim
ethy
lfor
mam
ide,
dim
ethy
lsul
foxi
de,
dim
ethy
lace
tam
ide,
isop
ropa
nol,
acet
onitr
ile,
met
hyle
thyl
keto
ne,
ethy
lac
etat
e,an
dm
ore;
and
Pura
Mem
mem
bran
ese
ries
for
stab
ility
inso
lven
ts,
incl
udin
gto
luen
e,he
ptan
e,he
xane
,met
hyle
thyl
keto
ne,m
ethy
l-is
obut
ylke
tone
,eth
ylac
etat
e,an
dm
ore.
c Man
ufac
ture
rva
lues
obta
ined
usin
gto
luen
eas
aso
lven
tand
are
quot
edas
the
MW
at90
%so
lute
reje
ctio
n,es
timat
edby
inte
rpol
atio
nfr
oma
plot
ofre
ject
ion
vers
usM
Wfo
ra
seri
esof
n-al
kane
s.dSo
me
Sols
epm
embr
anes
wer
epr
oven
toha
vea
silic
onto
pla
yer
(91)
.
162
ORGANIC SOLVENT NANOFILTRATION AND PHARMACEUTICALS 163
reported in literature and discussed the important, commer-cialized PI types that are now used in membrane technology,their preparation methods and main applications. Most OSNresearch is done on cross-linked PI membranes as cur-rent state-of-the-art (16, 37, 43, 44) such as commercialDuramem-series by Evonik MET (UK) (see Table 2, entry 5).These membranes are integrally skinned membranes. A threestep process is needed to prepare them: (i) phase inversion,(ii) cross-linking with amine and (iii) post-treatment with aconditioning agent (44). The conditioning agent is requiredto avoid membrane pore collapse upon drying, brittleness,and to ensure ease in handling during membrane modulesmanufacturing.
Recently, Siddique et al. (45) proposed a new gen-eration of compaction free PI OSN membranes. Thereis no more any flux reduction in time and these mem-branes do not require addition of a conditioning agent.Aminopropyl trimethoxysilane (APTMS) was used as cross-linker in the post-treatment of PI membranes (prepared viaa phase inversion process) instead of typical amines (e.g.,1,6-hexamethylene diamine or HMDA) used for conven-tional cross-linking. During the post-treatment the aminogroup of APTMS was reacted with polyimide P84, whilethe trimethoxysilane end groups were expected to hydrolyzeand generate a cross-linked network (Figure 1). The asym-metric structure of the membrane is further maintained aftertreatment with organic–inorganic cross-linker, resulting in amembrane with consistent performance and increased ther-mal and mechanical strength. The resulting decrease in fluxof the APTMS membranes was mitigated using pore formingadditives such as maleic acid.
As previously reported for cross-linked silicon rubberscomposite membranes, thin film composite (TFC) are com-mon polymeric OSN-membranes, which consist of a verythin layer that is deposited on an open support membranevia interfacial polymerization, dip-coating, layer-by-layerdeposition or spin-coating (45–47). By applying the selec-tive thin skin-layers films on a cross-linked support mem-brane and by using a cross-linked top-layer, solvent stabilitycan be achieved. Beside cross-linked PI and cross-linkedsilicon rubber-based membranes, polyamide (PA) mem-branes thin films on polysulfone (PSf) supports, such asthe Filmtech membranes by Dow (US) are commerciallyavailable TFCs. They were developed for aqueous separa-tions and are not stable in polar aprotics, since the PSfsupport is not cross-linked. In this specific context, JimenezSolomon et al. developed solvent stable TFCs based oncross-linked PI support for OSN purposes (26, 27, 47). Thenew hydrophobic membranes are stable in DMF and exhibitsignificantly higher permeabilities with comparable or bet-ter selectivity for nonpolar solvents compared to commercialOSN hydrophobic integrally skinned asymmetric and rubbercoated membranes.
Membranes of both composite and asymmetric types aremanufactured upon the base of highly permeable polymeric
O
N
OO
O
N
O
CH2
Polyimide+
H2N Si
OMe
OMe
OMeAminopropyl trimethoxysilane
O
N
OO
N
O
CH2
PolyimideHNO
SiO O
O
SiO
O
NH
O
CH2
N
O O
N
OO
Cross-linked structure
FIGURE 1 Cross-linking of polyimide via aminopropyl trimethoxysilane(APTMES) (modified from (45)).© Elsevier. Reproduced from (45) with permission from Elsevier.Permission to reuse must be obtained from the rightsholder.
glasses, such as poly(1-trimethyl- silylprop-1-yne) (PTMSP)(Table 1) (49–51) and polymer of intrinsic microporosity(PIM) (52–54). The presence of a bulky substituent and dou-ble bond in the main chain of PTMSP ensures a high degreeof rigid free volume (up to 25%). The nanoporous structure(at a level of 1 nm) of the selective layer is spontaneouslyformed as in the case of the preparation of membranes basedon low permeable glassy polymers (49,50). Volkov et al.(50) reported that methanol, ethanol or acetone permeabil-ity through composite membranes based on PTMSP caston PAN porous support exceeds that of some commercialNF membranes (Desal-5-DK, MPF-44 and MPF-50), withretention of 85–90% of a negatively charged dye (molecu-lar weight (MW) 626.5 Da). However, it is known that highfree volume polymers such as PTMSP are subjected to fastphysical aging (reduction of gas permeability in time) (55).Concerning this issue, Volkov et al. (50) reported a slightdecrease of ethanol transport during an operation periodlasting between 80 and 230 h.
In a recent study, the effect of solute nature on theOSN performance of membrane materials based on highlypermeable glassy polymers such as PTMSP, PMP and PIM-1 was investigated in ethanol media (54). PTMSP, PMP and
164 M. G. BUONOMENNA AND J. BAE
PIM-1 show the same level and order of retention for lowmolecular weight dyes (350 Da) regardless of the differencein polymer nature. It was shown that membrane swellingcould enhance the solvent transport across the membrane;for example, PMP and PIM-1 have comparable ethanol fluxwhile the gas permeability for PIM-1 is at least three timeslower than that for PMP. The swollen nanoporous structureof these polymers possesses sufficient mechanical resistancetowards significant applied pressures (10–30 bar).
Ceramic Membranes
In general, solvent-stable polymer membranes are optimizedto resist to a specific class of similar solvents (e.g., via cross-linking, as reported above), while ceramic membranes arestable in all solvents, including the aprotic solvents (e.g.,ethyl acetate, tetrahydrofuran, dimethylformamide, acetoni-trile) that are known to dissolve membrane-forming poly-mers very well (56). With ceramic membranes there isno leaching of chemicals as observed with the progressivedegradation of organic membrane or due to the chemicalconditioning agents used to keep nanopores open in NFmembranes (57). Therefore, ceramic membranes are gen-erally more suitable for use in GMP environments (“GoodManufacturing Practice” as used in the pharmaceuticalindustry) (56). Most of the existing ceramic nanofiltrationmembranes are made of metal oxides such as alumina(Al2O3), zirconia (ZrO2) or titania (TiO2).
The Inopor Company produces a range of ceramic forultrafiltration (UF) and nanofiltration (NF) membranes in theform of monochannel and multichannel tubes with lengthsup to 1.2 m. In Table 2 two membranes with top lay-ers based on ZrO2 and TiO2 are reported (Table 2, entries15 and 16). Due to the presence of hydroxyl groups onthe surface, these membranes are hydrophilic and there-fore in nonpolar organic solvents, show solvent fluxes.They can be prepared to be hydrophobic doing surfacefunctionalization with apolar group (such as alkyl, perflu-oroalkyl, etc.). This proved to enhance the flux of apolarsolvents (56). The most frequently applied functionalizationmethod is an organosilane coupling treatment on the reactivehydroxyl groups of metal oxides. The commercial HITK-T1 (HITK, Germany) is a silylated TiO2-based membrane(Table 2, entry 14). Organosilane reagents have the gen-eral structure SiXX’X”R. with X the reactive group of the
reagent, and R is the functional group grafted onto the metaloxide surface; groups X’ and X” can be on itself reactivegroups or nonreactive groups.
Alkyl groups or perfluoroalkyl groups with varying num-ber of carbon groups are the most frequently used functionalgroups for hydrophobization of the membrane surface, butalso polymers are chemically linked to ceramic membranes.Pinheiro et al. (58) developed a PDMS grafted α-aluminamembrane (5 nm). The first step was the ceramic sur-face modification via formation of an aminosilane layer(APTES) with subsequent reaction of the amine groupwith an epoxy-terminated PDMS proposed (Scheme 1). Thegrafted PDMS imparts the desired membrane selectivitywhile the ceramic support provides the mechanical, thermaland chemical stability.
Meynen and Buekenhoudt (56) reviewed the methods(both post-synthesis modification or in-situ methods) forthe preparation of hybrid organic-inorganic membranes(i.e., ceramic membranes functionalized with organic moi-eties) for OSN membranes. Synthesis methods such asorganosilane and phosphonic acid grafting as well as a morerecently developed method are discussed (59–62). A direct,covalent bonding of the aimed functional group (R) to themetal (M, with M = Al, Zr or Ti) of the metal oxide matrix,M–R was obtained using a Grignard reagent (Scheme 2).
Recently, the application in OSN of Inopor titania (1 nm)membranes functionalized with a series of alkyl groupsin Grignard reagents on the basis of Scheme 2 has beenreported (62). The functionalized ceramic membranes gavean increased retention in acetone, a typical aprotic solvent,compared to the acetone retention by native membranes. Thisconfirmed the hydrophobic character of the functionalizedmembranes. In particular, the retention results of modi-fied membranes are comparable to those obtained with theamphiphilic polymeric Duramem 300 membrane (Table 2,entry 5), a reference membrane often used as a benchmarkfor OSN.
Transport Models
Implementation of membrane processes at industrial scalerequires a good descriptive and predictive model based onreadily accessible physical property data (63). Challenges forprocess design include the selection of a suitable membraneand a suitable solvent (mixture) in the process. Until now, it
OH
O
O
O
OH
Si
NH2
+
O
O
Si
CH3
CH3
O
Si
CH3
CH3
O
Si
CH3
CH3
n
C4H
9
R1
OH
O
O
O
OH
Si
NH
R1
OH
SCHEME 1 Functionalization of ceramic membranes by polymer grafting via organosilanes (58).
ORGANIC SOLVENT NANOFILTRATION AND PHARMACEUTICALS 165
SCHEME 2 Functionalization of ceramic membranes via Grignard’s reagents (59, 60).
has been difficult to identify the best suitable combinationof OSN membrane and solvent (pure or mixture) due totheir nontargeted nature. Therefore, modeling of the under-lying permeation mechanism to predict solvent fluxes/soluterejections and, more important, to generate insight into thetransport, is recommended.
Several transport models have been proposed for OSN,most of them being extension of the existing modelsfrom aqueous to the nonaqueous NF systems. Generally,three models, briefly described below, have been used todescribe mass transport through OSN membranes: 1) theSpiegler-Kedem model; 2) the pore-flow model and 3) thesolution-diffusion model. Both solution-diffusion and pore-flow models take into account membrane properties. TheSpiegler-Kedem model originates from irreversible thermo-dynamics, treating the membrane as a black-box.
Many of the tighter NF membranes (certainly all reverseosmosis (RO)-membranes) are considered having a densetop-layer where the free volume elements between thepolymer chains allow transport. However, for other OSNmembranes, for which the membrane material is typicalfor NF in the spectrum of pressure-driven membrane pro-cesses, solvent transport occurs through the pores whilesolute separations relies on sieving.
Recently, Schmidt and Lutze (64) presented aphenomena-based model for multicomponent perme-ation through polymeric OSN membranes based onsolution-diffusion, pore-flow and mutual coupling terms.Instead of developing a very detailed model, the objectiveof the study is to attain a fixed set of binary interactionparameters for a given OSN membrane, solvents and solutesin analogy to vapor–liquid equilibria (VLE) parametersets. Membrane modelling maps (MMM) were introduced.MMM highlight the shares of the permeation phenomenaand give recommendations for rejection improvement independence of the solute and the applied solvent (mixture).As experimental basis for parameter estimation and modelevaluation, published permeability and rejection data inbinary and ternary mixtures of toluene, n-hexane and2-propanol with Starmem 122 were used (64).
Spiegler-Kedem Model
The membrane is considered to be a “black box.” The trans-port of solvent and solute through a membrane considersonly the driving forces and their resulting fluxes. In thismodel based on irreversible thermodynamics, the drivingforces are differentials across the membrane. The solute fluxis a combination of diffusion and convection [Eq. (1)]:
J2 = P2dc
dz+ Jvc2 (1 − σ ) (1)
The first and second term represent the contributions of dif-fusion and convection, respectively. P2 is the local solutepermeability coefficient, Jv is the volume flux. C2 is the aver-age concentration of the solute in the membrane and ρ is thereflection coefficient, which can be interpreted as the fractionof solute reflected by the membrane in convective flow.
Pore-Flow Model
The pore flow model assumes that the mass transport occursby pressure driven convective flow through the pores of themembranes. Darvishmanesh et al. (65) developed a modelfor nanofiltration membrane units, implemented in a com-mon process simulation software (Aspen Plus). The modelis based on the pore flow mechanism and describes a singlemembrane module. In the case of liquids, the flux behaviorcan be described by the Hagen–Poiseuille equation [Eq. (2)],which contains membrane structural factors, like membranepore size, surface porosity and tortuosity and in whichviscosity (η) is an evident solvent parameter.
J1 = εrp
8ητ· �P
l(2)
εrp
8ητ l is the expression for the membrane permeability (Lp) forcylindrical pores.
However, several experimental OSN results suggestedthat the Hagen-Poiseuille law is not longer valid for organicsolvents due to possible interactions between solvent and
166 M. G. BUONOMENNA AND J. BAE
membrane: pore size (rp) might depend on the type of solventused, due to different swelling of the membrane polymer.Machado et al. (66) characterized transport properties of sol-vents (alcohols, paraffins, ketones, acetates and water forcomparison) permeating through MPF-50 (Table 2, entry1). In this study, the commonly observed linear relation-ship [Eq. (2)] was obtained with paraffins, acetates andketones. However, nonlinear behavior was observed in thecase of alcohols (methanol, ethanol, iso-propanol, n-butanol,n-pentanol). The flux-pressure curve exhibits a falling ratebehavior accentuating as molecular weight increases.
Machado et al. (66) determined for the alcohols the αp,i.e., an empirical pressure coefficient, expressing the fallingrate behavior, defined by Eq. (3):
Lp = Lp0exp (−αpΔP) (3)
where Lp0 is a permeability constant. The calculated αp val-
ues for the alcohols were twice as large as those for water.The data suggested that the flux of either pure or mixedsolvents was mainly affected by the surface tension and vis-cosity of the solvents. For apolar solvents such as paraffins,the dielectric constant also affected the flux. In particularfor the influence of surface properties, the flux level of var-ious solvent families seems to be ordered by their relativehydrophobicity.
In a subsequent work, the same research group describedthe proportionality constant of Eq. (2) as the inverse of aseries of three resistances against mass transport (67). Thefirst two resistances are related to viscous flow in the toplayer of the membrane and in the porous support layer,respectively; the third resistance reflects the influence of ahydrophobic/hydrophilic resistance. The viscous resistancecan be expressed as
R1μ = k1
M
μ
(d1p)2 and R2
μ = k2M
μ
(d2p)2 (4)
and the resistance related to hydrophobicity/hydrophilicityis
R0S = k0
M
(d1p)2 (γC − γL) (5)
where k0M , k1
M , k2M are interaction constants, μ is the sol-
vent viscosity, d1p and d2
p are the membrane pore size in thetop and support layers, respectively and γ c-γ L is the sur-face energy difference between membrane and solvent. Theresulting equation for the solvent flux is
J = �P
φ[(γC − γL) + f1μ] + f2μ(6)
with
f1 = k1M/k0
M , f2 = k2M/(d2
p)2, φ = k0M/(d1
p)2φ (7)
On the basis of this so-called wetting model, membranecharacteristics are into one single parameter (φ). It can beexpected that hydrophobicity/hydrophilicity plays an impor-tant role for the solvent flux: apolar solvents with low surfacetension are expected to have a high flux with hydropho-bic membranes and a low flux with hydrophilic membranes.Polar solvents with a high surface tension have a lowflux with hydrophobic membranes and a high flux withhydrophilic membranes.
The surface force-pore flow (SFPF) model considerssolute-solvent-membrane interactions including the type ofsolvent (68). Solute and solvents physical properties are usedfor calculations making it possible for extension of the modelto different solvent-solute systems relatively easily as com-pared to traditional solution-diffusion models. The modelconsiders a potential function expressing the force exerted onthe solute molecule (φ (r)) by the pore wall or the membranesurface. φ (r) is a function of the distance r between the porewall or membrane surface and the solute molecule. A posi-tive φ (r) value represents a repulsive force, a negative valuean attractive force. The model uses momentum balances toobtain a relation between the observed solute rejection andother parameters. The resultant equation [Eq. (8)] containinga parameter BSFPF can be given by
ϕ(x) = ϕ(r)
RT= − BSFPF/R3
a
((Rb/Ra) − x)3 (8)
where Ra is the effective radius of the membrane pore afterpreferential solvent wetting, Rb the radius of the membranepore and x the dimensionless pore distance. The parameterBSFPF can be used as a measure of the interaction betweenthe solute and the membrane material.
Geens et al. (63) studied the removal of 5 specific activepharmaceutical ingredients (APIs) with different molecularweight (from 189–721 Da) from toluene, methylene chlo-ride, and methanol by using solvent resistant nanofiltration.Although the rejections expected from the size differencebetween solutes and membrane pores were high, the resultslargely depended on the solvent used. The authors analyzedfor the solute retention some of the models based on nano-sieving such as the Steric Hindrance Pore Model (69), themodel of Zeman and Wales (70), the log-normal model (71)and the Verniory model (72). The authors observed thatwhereas the nano-sieving approach seemed to be fundamen-tally different from solution-diffusion, both the nano-sievingand solution-diffusion models lead to similar results, becausethe diffusivity of a given compound is inversely proportionalto its size through the Stokes–Einstein equation:
Ds = kT
6πηrs(9)
where Ds is the diffusion coefficient (m2/s), T is the tem-perature (K), k is the Boltzmann constant, η is the viscosity(Pa.s), and rs is the solute’s radius (63).
ORGANIC SOLVENT NANOFILTRATION AND PHARMACEUTICALS 167
Solution-Diffusion Model
In the classic model developed by Lonsdale et al. (73) andrevisited by Wijmans and Baker (74), the primary assump-tion is that the flux of the solute and solvent are independent.Ji, the flux of a species i through a membrane, is given by
Ji = DiKi
l
[cif − cip exp
(−νi(pf − pp)
RT
)](10)
where Di is the diffusion coefficient of i through the mem-brane; Ki is the partition coefficient; l is the membranethickness; cif, cip are the feed and permeate concentrationsof species i, respectively; υ i is the partial molar volume ofspecies i; pf, pp are the feed and permeate sides pressures,respectively; R is the gas constant and T is the temperature.
White (75) used this model to study the transport proper-ties of toluene solutions of six solutes (from 142 to 311 Da)through integrally skinned membranes based on PI (Table 2,entry 3). However, several experimental findings suggestedthat coupling of the solute and solvent fluxes cannot beneglected and thus the solution-diffusion model cannot beused without modifying it to consider coupling effects(76, 77).
The experimental data with two types of membranes(PDMS-based NF membrane and aromatic polyamide-basedmembrane) obtained by Bhanushali and co-workers (76)allowed them to evaluate two traditional transport theoriesthat consider coupling: the Spiegler–Kedem model and thepore-flow model (see above). The Spiegler–Kedem modelwas used to obtain the convective and diffusive contributions,however the model does not have specific parameters forsolute-membrane interactions; the pore-flow model was usedconsidering convective coupling and interaction parameters.
Paul (78) used the Maxwell–Stefan formulation for mul-ticomponent diffusion and proposed the following equations
for the solvent [Eq. (11)] and solute flux [Eq. (12)],respectively:
n1 = ρD1m
wml(w10 − w1l) (11)
n2 = ρD2m
Wml(
1 + ε2w1wm
) (w20 − w2l)
(ε2
w2wm
)(
1 + ε2w1wm
)n1 (12)
where ρ is the mass density of the membrane; l is the mem-brane thickness; D1m, D2m are multi component diffusioncoefficients of solvent and solute versus membrane, respec-tively; w10 and w1l are mass fraction of solvent in membraneat the upstream and downstream sides, respectively; w20 andw2l are mass fraction of solute in membrane at the upstreamand downstream sides, respectively; ε2 is frictional cou-pling coefficient. The two mechanisms for solute transportthrough polymer films, i.e., diffusion and convection canbe discerned, as suggested by the diffusion and convectionterms of Eq. (12) (78).
PHARMACEUTICAL APPLICATIONS
In the pharmaceutical industry and specifically in APIs man-ufacturing process, OSN can be applied in the synthesisbetween different reaction steps and/or in the downstreamprocessing including separations (extraction, distillation)and particle forming unit operations (crystallization, filtra-tion, drying). OSN can be used to either retain a targetmolecule (retentate stream) or allow the target molecule topermeate while retaining the impurity (permeate stream).During membrane separation, usually one or the two streams(permeate and retentate, Figure 2) are almost always contam-inated with a minor amount of a second component. In some
Retentate
Permeate
Feed
FeedRetentate
Permeate
a)
b)
c)
Feed
Retentate
Permeate
FIGURE 2 a) One stage, b) two-stage stripping cascade, and c) two-stage enriching cascade membrane separations.
168 M. G. BUONOMENNA AND J. BAE
cases the permeate stream can contain significant amount ofmaterials which is supposed to be concentrated in the reten-tate, because the membrane selectivity is not infinite. If theproduct is retained by the membrane and the impurities pass,then pure solvent can be added to the feed tank, so-called dia-filtration operation, to wash out the impurities and to increasepurity. This, however, is at the expense of the product yield.
The extent to which a feed mixture can be separatedis limited. A single membrane module or a number ofsuch modules arranged in parallel or in series without recy-cle constitute a single-stage membrane separation process.To achieve a higher degree of separation, cascades of mem-brane modules with recycle are often used. Assuming thatthe pressure drop on the upstream side of the membrane isnegligible, only the permeate phase must be pumped (for aliquid) or compressed (for a gas).
A 2-stage stripping cascade (Figure 2b) is designed toobtain a purer retentate, whereas the 2-stage enriching cas-cade (Figure 2c) is designed to obtain a purer permeate. Theconcept of membrane cascades has been applied to solvent-based separations as discussed by Vanneste and co-workers(100) and Lin and Livingston (101). In particular, Kim et al.(102) in a recent study with the impressive title “When themembrane is not enough: A simplified membrane cascadeusing Organic Solvent Nanofiltration (OSN)” emphasizedhow the differences in rejection of the solutes by the mem-brane are often in sufficient to separate them in a singlefiltration stage. Membrane cascades can meet this challengeand have potential to implement process intensification (PI)strategy.
Here, the OSN applications in pharmaceutical industryare considered on the basis of relevant literature and patentsin four case-studies: peptides synthesis, removal of excessreagents, reaction product purification and nonthermal sol-vent exchange. In Table 3 some details for each case-studyare reported. For an interesting analysis of the use of OSNin particle formation unit operations such as crystallization,further reading of the review article by Rundquist et al. (103)is suggested. In particular, the authors proposed the integra-tion of APIs crystallization with solvent recovery and recyclevia OSN.
Technology for Peptide and OligonucleotideProduction
The cell membrane is a highly selective barrier, limitingthe cellular uptake of molecules including DNA, oligonu-cleotides, peptides and proteins used as therapeutic agents.Cell Penetrating Peptides (CPPs) represent a new and inno-vative concept, which bypasses the problem of bioavail-ability of drugs. A large number of different therapeuticagents have been efficiently delivered by CPPs. These rangefrom small molecules to proteins and even liposomes andmagnetic particles (104). Typical applications of CPPs inthe delivery of biopharmaceutical agents include: i) genedelivery; ii) siRNA delivery; iii) antisense oligonucleotide
delivery; iv) protein delivery; v) delivery of drug carriers;vi) CPP as APIs.
Munyendo et al. (105) reviewed the development of CPPsfor cell-specific delivery strategies involving biomoleculesand discussed conjugations of therapeutic agents to CPPsfor enhanced intracellular delivery. Produced by almost allliving organisms as a component of their innate nonspe-cific immune system, antimicrobial peptides are consideredas lead compounds for the discovery of human therapeuticsto tackle the development of antibiotic resistance. As drugs,peptides show unique features, high biological activity andspecificity, and low toxicity, thereby making them attractiveas therapeutic agents (106).
CPPs have been reported to act as APIs when used alone.A transactivator of transcription, called TAT peptide, con-taining a cysteine residue, was shown to be able to inhibitinfection by irreversibly inactivating virions exposed to thespecific TAT-C peptide prior to cell infection, blocking entryof cell-adsorbed viruses, or inducing a state of resistance toinfection in cells pretreated with TAT-C (107).
The technology for manufacturing many of these mate-rials is based on solid-phase synthesis. Solid-phase peptidesynthesis (SPPS) is the most widely used technology, sinceit neatly solves the critical purification problems encounteredat each stage in solution-phase synthesis. However it facesserious challenges including mass transfer, steric hindrance,and resin handling. The concept of membrane separationcoupled to solution-phase synthesis offers major advantagesover SPPS by combining the advantages of ‘‘classical’’solution-phase synthesis with the ease of purification of thesolid-phase method (108, 109).
In prior application of membrane separation in peptidesynthesis (108, 109), peptides built on poly(ethylene gly-col) (PEG) were separated from impurities by ultrafiltra-tion. After each coupling and each deprotection step ofthe peptide synthesis, three consecutive steps of solventevaporation, neutralization after deprotection, and uptake inwater prior to ultrafiltration were necessary. Water was thenremoved by evaporation and/or azeotropic distillation beforere-dissolving the PEG–peptide into organic solvent for thenext coupling step. All these steps were required becauseno solvent resistant membranes were available thirty yearsago. OSN represents an ideal separation method for in-cycle purification during peptide synthesis and MembraneEnhanced Peptide Synthesis (MEPS) is a new technologyplatform that advantageously combines OSN with solution-phase peptide synthesis (Table 3, entry 1).
So et al. (110, 111) illustrate clearly that MEPS benefitsfrom the advantages of SPPS, while avoiding the purificationsteps that have until now made this synthesis path practi-cally difficult. MEPS was patented by the same researchgroup (112). Peptide chain assembly occurs via: (1) amidecoupling; (2) a washing step for removal of excess reagentsvia constant volume dia-filtration; (3) deprotection; and (4)a washing step for removal of deprotection by-products andexcess reagents again via diafiltration. The cycle is repeated
TAB
LE3
OS
Nin
reac
tions
and
sepa
ratio
nsin
phar
mac
eutic
alpr
oces
sing
Ent
ryP
harm
aceu
tica
lpr
oces
sing
step
OSN
role
OSN
mem
bran
eA
dvan
tage
sD
isad
vant
ages
Ref
.
1R
eact
ion
Sepa
ratio
nm
etho
dfo
rin
-cyc
lepu
rific
atio
ndu
ring
pept
ide
synt
hesi
s−
Zir
coni
umox
ide
coat
edm
embr
ane
(3nm
pore
size
),hy
drop
hobi
csu
rfac
em
odifi
ed(I
nopo
rG
mbH
,G
erm
any)
;−
Che
mic
ally
cros
s-lin
ked
poly
imid
em
embr
ane
(Dur
aMem
,ME
TL
td,U
K)
All
the
adva
ntag
esof
solu
tion
phas
esy
nthe
sis
avoi
ding
the
puri
ficat
ion
step
s
−R
equi
red
high
mem
bran
ese
lect
ivity
betw
een
the
pept
ide
and
side
reac
tion
prod
ucta
ndex
cess
reag
ent
isre
ques
ted
−R
equi
red
long
term
mem
bran
est
abili
tyin
the
reac
tion
solv
ent
(DM
F)
108–
112
2R
eact
ion
Rem
oval
ofex
cess
ofre
agen
ts(b
enzy
lal
coho
l)St
arm
em-1
20St
arm
em-1
22D
irec
tand
effic
ient
sepa
ratio
nal
tern
ativ
eto
chro
mat
ogra
phy
whi
chre
quir
esa
larg
equ
antit
yof
silic
a
An
acyl
atio
nst
epto
incr
ease
the
mol
ecul
arw
eigh
tof
benz
ylal
coho
lis
nece
ssar
yfo
rco
mpl
ete
rete
ntio
nth
roug
hth
em
embr
anes
115
3Po
st-r
eact
ion
puri
ficat
ion
Sepa
ratio
nof
1-(5
-bro
mo-
fur-
2-il)
-2-
brom
o-2-
nitr
oeth
ane
(API
)fr
omim
puri
ties
(pyr
idin
e,ac
etic
anhy
drid
ean
dbr
omin
e)
Dur
amem
150
Dir
ect,
scal
able
sepa
ratio
nal
tern
ativ
eto
crys
talli
zatio
nan
dpr
epar
ativ
eco
lum
n
Mor
eco
nsec
utiv
efil
trat
ion
stag
esar
ere
quir
ed98
4Po
st-r
eact
ion
puri
ficat
ion
Sepa
ratio
nof
arom
atic
smin
e(3
91g
mol
−1)
(API
)fr
oma
dim
eric
hydr
azo
impu
rity
(781
gm
ol−1
)
Dur
amem
200
Effi
cien
tsep
arat
ion
alte
rnat
ive
toch
rom
atog
raph
y;in
this
latte
rlo
wso
lubi
lity
ofth
eso
lute
inth
eso
lven
tsus
edre
sulte
din
anex
cess
ivel
yla
rge
volu
me
ofso
lven
tsan
dlo
ngpr
oces
sing
times
−Slig
htre
duct
ion
offlu
xdu
eto
mem
bran
eco
mpa
ctio
n99
5Po
st-r
eact
ion
puri
ficat
ion
(rem
oval
ofG
TIs
)
Sepa
ratio
nof
Mom
etas
one
furo
ate
gluc
ocor
ticoi
d(A
PI)
from
Met
hyl
mes
ylat
e(M
eMS)
and
4-d
imet
hyla
min
opyr
idin
e(D
MA
P),
chos
enas
mod
elG
TIs
GM
T-oN
F-2
Low
estA
PIlo
ssco
mpa
red
tofla
shch
rom
atog
raph
yan
dre
crys
talli
zatio
n;ea
sysc
ale-
upw
ithlo
wla
bor
inte
nsity
Hig
hso
lven
tcon
sum
ptio
n(7
diav
olum
es)
(fro
man
inve
stm
ent
pers
pect
ive,
the
optio
nof
solv
ent
recy
cle
beco
mes
econ
omic
ally
feas
ible
for
OSN
)
122
6Po
st-r
eact
ion
puri
ficat
ion
App
licat
ion
ofdu
alm
embr
ane
diafi
ltrat
ion
(DM
D);
API
-IN
T(M
W=
675
g·m
ol−1
),its
isom
er(I
som
erB
),an
da
seri
esof
olig
omer
icim
puri
ties
base
don
API
-IN
Tw
ithM
W>
1000
g·m
ol−1
Dur
amem
300
Dur
amem
500
Rem
ovin
g99
%of
the
part
icul
arly
chal
leng
ing
high
erol
igom
eric
impu
ritie
s(i
.e.,
tetr
amer
and
high
erof
API
-IN
T)
Hig
hfr
esh
solv
entc
onsu
mpt
ion
redu
ced
byin
tegr
atin
ga
dow
nstr
eam
OSN
-bas
edso
lven
tre
cove
rysy
stem
into
the
diafi
ltrat
ion
puri
ficat
ion
proc
ess
via
Dua
lM
embr
ane
Dia
filtr
atio
n(D
MD
)
123
7Po
st-r
eact
ion
puri
ficat
ion
Mar
tius
Yel
low
(MY
,2,
4-D
initr
o-1-
naph
thol
sodi
umsa
lt)(M
.W.2
74.1
6g·m
ol−1
)(A
PIm
odel
)an
dB
rilli
antB
lue
R(B
BR
,M.W
.82
6g·m
ol−1
)(c
olor
impu
rity
inm
anuf
actu
ring
proc
ess)
STA
RM
EM
228
Effi
cien
tim
puri
tyre
mov
alM
ore
cons
ecut
ive
filtr
atio
nst
ages
are
requ
ired
126
8Po
st-r
eact
ion
puri
ficat
ion
4-ph
enyl
azop
heno
l(So
lven
tYel
low
7,SY
7)(A
PIm
odel
)an
dB
rilli
antB
lue
R(B
B)
(mod
elim
puri
ty)
Dur
aMem
300
Hig
hpr
oduc
tpur
ity(9
9.7%
)w
itha
prod
ucty
ield
of90
%M
ore
cons
ecut
ive
filtr
atio
nst
ages
are
requ
ired
123
9So
lven
texc
hang
eTo
luen
e/m
etha
nol,
exch
ange
with
quat
erna
ryam
mon
ium
salt,
tetr
aoct
ylam
mon
ium
brom
ide
(TO
AB
r,m
olec
ular
wei
ght5
47D
a),
chos
enas
am
arke
rso
lute
STA
RM
EM
122
75.3
%so
lven
texc
hang
eis
obta
ined
inth
ree-
stag
esca
scad
esT
heco
nfigu
ratio
nus
edis
nots
uita
ble
for
com
plet
eex
chan
geof
two
solv
ents
asth
isw
ould
requ
ire
anin
finite
num
bers
ofst
ages
102
169
170 M. G. BUONOMENNA AND J. BAE
FIGURE 3 Schematic of membrane enhanced peptide synthesis (MEPS) (112).
as many times as necessary, adding a further amino acideach cycle, until the desired peptide sequence is obtained(Figure 3).
For successful realization of MEPS the membrane mustpossess excellent long term stability in the reaction solvent(in the case of patent, by Vasconceles et al. (112), the solventis DMF) and high selectivity between MeO–PEG–peptide,and side reaction products and excess reagents, includingunreacted amino acids, activators and deprotection reagents.In a recent article by Valtcheva et al. (29) it is reported thatcross-linked polybenzilimidazole membranes (see Table 1)have great potential to be used as OSN membranes as alter-native to polyimide membranes in various pharmaceuticalprocesses and among them in typical reaction conditionsused in peptide synthesis (20 wt.% piperidine in DMF) (113).The membranes reported in the patent (112) are two commer-cial membranes: zirconium oxide coated membrane (3 nmpore size), hydrophobic surface modified (Inopor GmbH,Germany) and a chemically cross-linked polyimide mem-brane (DuraMem, MET Ltd, UK) (see Table 2, entries 5 and16, respectively). Tubular membrane modules were used.
Both coupling and deprotection reactions were performedin the reaction vessel where mixing was provided via the cir-culation pump. The reaction solution is recirculated throughthe membrane cartridge and ensures good liquid mixingthroughout. Upon completion of each reaction, the system ispressurized. The resulting solvent flow permeating throughthe membrane is balanced by a constant flow of fresh solvent(DMF) supplied to the feed tank from the solvent reser-voir via an HPLC pump. The same procedure is applied ateach reaction/washing cycle. The peptide is assembled ona soluble polymeric support, methoxy–amino–PEG (MeO–PEG–NH2) with MW 5000 g mol−1, to increase retentionby the membrane (114).
Removal of Excess Reagents
An interesting example of OSN application in pharmaceu-tical industry for removal of excess of reagents has been
proposed by Ormerod (115) (Table 3, entry 2). The reactioninvestigated was the transesterification from a methyl esterto a benzyl ester in a molecule whose functionality includesa secondary alcohol (Scheme 3).
This reaction requires a large excess of benzyl alco-hol, but this reagent must be removed prior to the fol-lowing synthetic reactions. Removal of the excess benzylalcohol via distillation is not an option; chromatographyworks well, but requires a large quantity of silica. Thedirect OSN of the reaction mixture in Scheme 3 by usingStarmen-120 membrane yields insufficient separation ofbenzyl alcohol from product. Therefore OSN has been car-ried out on the reaction mixture after alcohols acylation(Scheme 4).
The acylated product and reagent have enough differ-ent molecular weights compared to the benzyl ester toallow their separation. In Table 4, the OSN results by usingStarmem-120 and Starmem-122 are reported. The impuritiesremain in the retentate stream and can be removed with asecond filtration over a membrane with MWCO of 400 Da.After the selective oxidation reaction step, the product isobtained in the permeate, whilst the impurities are retainedin the retentate (Table 5).
Purification
Removal of Impurities
The removal of impurities formed during the synthesis oforganic molecules used as drug substances is a major con-cern for all processes in the pharmaceutical industry (116).Crystallization and chromatography are mainly used forpurification. Crystallization can be a rather complex processthat is difficult to control and scale-up and requires sig-nificant optimization to generate acceptable process yields(117). Preparative column chromatography is widely usedin process development and is regarded as a reliable purifi-cation technology. However, it consumes large quantitiesof solvents, which require further downstream processingnot only to recover the solvents but also to concentrate
ORGANIC SOLVENT NANOFILTRATION AND PHARMACEUTICALS 171
R'
OH
R
CO2Me
benzyl alcoholR'
OH
R
O O Ph
+ impurity A + impurity B
278 g/mol 372 g/mol 354 g/mol
SCHEME 3 Transesterification reaction with excess of benzyl alcohol.
O
R
R1
R'
O O Ph
M.W. = (R1 = CH3) 320 g/mol
(R1 = CH3CH2) 334 g/mol
+ impurity A + impurity B OR1
372 g/mol 354 g/mol (R1 = CH3) 150g/mol
(R1 = CH3CH2) 164 g/mol
O
O
+
SCHEME 4 Reaction mixture reported in Scheme 3 after acylation (115).
TABLE 4OSN of reaction mixture of Scheme 4 (115)
R1 in Scheme 4 Membrane Pressure (bar) Flux (Lm−2h−1)Rejection %
(acylated product) Result
CH3 Starmem-120(MWCO 200)
20 24 85 Benzyl ester separated fromacylated product, impuritiesremain
CH3CH2 Starmem-122(MWCO 220)
20 48 90
TABLE 5OSN of reaction mixture after selective oxidation (115)
OH
RR'
O O Ph
+ impurity A + impurity B
278 g/mol 372 g/mol 354 g/mol
MembranePressure
(bar)Flux
(Lm−2h−1)Rejection %
(product)Rejection %(impurity A)
Rejection %(impurity B) Result
Starmem-122(MWCO 220)
20 7 67 92 100 Product obtained in thepermeate not 100% pure
the diluted products. In addition, when processing solutionscontaining some oligomeric impurities, the active sites of thestationary phase can become blocked by these oligomericimpurities, thus making the chromatography difficult andtedious (118).
In Table 3 (entries 3-8) some example of OSN appli-cations in APIs purification are listed. Martinez et al. (98)
studied the application of OSN for the recovery of thepharmaceutical compound 1-(5-bromo-fur-2-il)-2-bromo-2-nitroethane, referred to as G-1 (Table 3, entry 3). G-1 is themain active ingredient for the preparation of pharmaceuticalproducts such as keratofural, which is used in veterinarianapplications as ophthalmic ointment to treat bacterial dis-eases and fungi; vitrofural, which sterilizes chemicals for
172 M. G. BUONOMENNA AND J. BAE
Ethanol Activated carbon
Pure G-1
B
A
NFRecovered
G-1
Ethanol
Raw G-1(G-1+ impurities)
Proposed scheme for G-1 recovery
FIGURE 4 Schematic of 1-(5-bromo-fur-2-il)-2-bromo-2-nitroethane (G-1).© Elsevier. Reproduced from (98) with permission from Elsevier. Permission to reuse must be obtained from the rightsholder.
vitro plantlets production; and dermofural, which is an oint-ment for treatment of fungal skin diseases. G-1 is wasted inlarge amounts during the purification step with ethanol asthe washing agent. Figure 4 shows the proposed approachbased on OSN for G-1 recovery. In particular, the feasibilityof OSN process in dead-end configuration for G-1 recov-ery in realistic conditions, i.e., in presence of the impuritiesthat are typically present in the mixture ethanol/G-1, suchas pyridine, acetic anhydride and bromine, is evaluated. Fourmembranes were used: three based on polyamide (NF 90, NF270, and BW30 XLE) and the last on polyimide (Duramem150).
The experimental results showed that purification of G-1is technically viable via OSN by using Duramem 150 ifconsecutive filtration stages are applied. In Figure 5, theaccumulative recovery of G-1 respect to the total amount ofG-1 present in the initial solution is shown. After ten filtra-tion stages, a recovery of about 80% of G-1 is obtained. Theanalysis of the costs and profits indicated that this recoverypercentage is economically feasible with a payback period of0.72 years (98).
Another case-study concerning post-reaction purifica-tion via OSN is that one investigated by Ormerod et al.(99) to purify an API intermediate from a dimeric impu-rity (Table 3, entry 4). The API intermediate formationoccurs via the reduction of an aromatic nitro group to anamine (MW 390.5 Da), but a small quantity of a dimerichydrazo impurity (MW 781 Da) is formed. The specifica-tion limit of the hydrazo impurity was 0.05 wt%, whilst alarge scale batch contained ten times more. Considering theMW difference between the amine and the hydrazo impurity,OSN was a particularly suitable technique to attempt thispurification.
FIGURE 5 Performance of OSN process (10 stages) using Duramem150 for the recovery of 1-(5-bromo-fur-2-il)-2-bromo-2-nitroethane (G-1).© Elsevier. Reproduced from (98) with permission from Elsevier.Permission to reuse must be obtained from the rightsholder.
Cross-linked polyimide-based membranes of Duramemseries (Table 2, entry 5), i.e., Duramem 150 and 200 showedbetter performance in THF compared to other tested mem-branes such as Starmem 122, Starmem 240 (based onuncross-linked polyimide) and Solsep. The observed rejec-tion with the two Duramem membranes for the dimerichydrazo impurity was 98% in the solvent of the syntheticprocess, i.e., THF. However, in this solvent the desiredAPI was also rejected. It was reported that the structureof membranes based on polyimide such as the Duramemmembranes changed after treatment with polar solvents(methanol, acetone, acetic acid), leading to an increase influx and a reduction in rejection (119).
ORGANIC SOLVENT NANOFILTRATION AND PHARMACEUTICALS 173
In the specific case of Duramem 150 and 200 membranes(based on cross-linked polyimide), the addition of 20% v/vof water to THF caused an increase of flux without rejec-tion modification. Adding an acid in order to fully protonatethe amine moieties of both the API intermediate and thedimer, caused the rejection of the API-intermediate to dropby approximately 20% with both Duramem 150 and 200,while the rejection of the dimer impurity remained suffi-ciently high to ensure its concentration is below the detectionlimit of the analysis technique used. However, the additionof acid restored the initial flux values without the addition ofwater in the solvent, maintaining the same relation betweenthe two membranes, i.e., Duramem 200 more permeable thanDuramem 150. Therefore, Duramem 200 in acidic (2.6 molequivalents) water/THF (20:80) mixture was the membraneselected to purify the dimeric API. After 5 diafiltration vol-umes, a yield of 82% of the purified API intermediate aminewas obtained.
The third case-study of post-reaction purification via OSNis the removal of genotoxic impurities (GTIs) (Table 3,entry 5). Among the various impurities to be removed dur-ing the API purification, GTIs represent an actual unsolvedproblem of utmost concern for pharmaceutical industry.
Several processes have been proposed to address this issue,including preparative column chromatography (117), use ofresins (118), and fractional distillations (120). The mostcommon process for API purification from GTIs is asequence of stages of solvent exchanges followed by recrys-tallization. Such process includes several steps, repeated APIrecrystallizations that result in losses in the mother liquor andhigh solvent consumption, hence increase of the total cost ofthe final API. Székely et al. (121) discussed the use of OSNas a general platform for the removal of GTIs to replace mul-tistep purification processes, as schematically illustrated inFigure 6.
The chemical class and molecular weights of the GTIsand API selected by Szekely et al. (121) such as model com-pounds to study their separation by OSN are resumed inFigure 7. The authors selected methyl ethyl ketone (MEK)and THF as solvents for the study for two reasons: the sol-ubility of the model compounds and the challenge offeredby the use of polymeric membranes in OSN with these sol-vents which dissolve many polymers. Two OSN membraneswith MWCO range of 250–350 Da were selected, namelySolSep NF010206 and GMT-oNF-2 (Table 2, entries 6 and11, respectively).
FIGURE 6 Schematic comparison of active pharmaceutical ingredient (API) purification by a conventional (left) and organic solvent nanofiltration (OSN)based process (right).© Elsevier. Reproduced from (121) with permission from Elsevier. Permission to reuse must be obtained from the rightsholder.
174 M. G. BUONOMENNA AND J. BAE
FIGURE 7 Compounds classes and molecular weights (MWs) of themodel genotoxin impurities (GTIs) and active pharmaceutical ingredients(APIs) used in the organic solvent nanofiltration (OSN) experiments bySzékely et al. (121). The dotted lines indicates an MW of 300 Da.© Elsevier. Reproduced from (121) with permission from Elsevier.Permission to reuse must be obtained from the rightsholder.
Aiming at a widespread use of OSN, the report by Székelyet al. (121) describes in simple terms the operating princi-ples of diafiltration and highlighted the steps for operationdesign. The recommendation for this process is to select amembrane with high rejection for the API (> 95%) and toextend the number of diavolumes according to the require-ments for GTI removals. However, from an environmentalpoint of view, the diafiltration operations require a consider-able amount of fresh solvent to achieve a high yield of theproduct. Thus, the authors associated OSN with a solventrecovery step (distillation or organic solvent reverse osmosis)(121).
The necessity of a solvent recovery stage in OSNdiafiltration was recognized in two interesting studies: oneis a subsequent article by Székely et al. (122) and theother is by Sereewatthanawut and coworkers (123). In theSzékely study, two examples were selected for the eval-uation of OSN (122). The API was mometasone furoate(Meta) glucocorticoid, and methyl mesylate (MeMS) and4-dimethylaminopyridine (DMAP) were chosen as modelGTIs (Table 3, entry 5). The performance efficiency andsustainable impact of OSN for degenotoxification of APIshas been compared to recrystallization and flash chromatog-raphy. Successful degenotoxification was achieved in allcases with DMAP and MeMS reaching final GTI levelsbelow the regulatory thresholds with the exception of DMAPusing recrystallization. API losses were 5% and 6.4% forOSN, 6.4% and 11.9% for flash chromatography and 14.9%and 16.4% for recrystallization during the removal of DMAPand MeMS, respectively.
The API loss occurring during the purification processeshas a significant impact on the outcome of cost analysis.Mass and carbon intensity values are highest for OSN andlowest for recrystallization, while flash chromatography hasintermediate values. It is clearly possible to dramaticallyreduce mass losses by recycling the solvent but at the cost ofan increase in energy consumption. This issue is of particularimportance for the OSN process, which is the process withthe highest solvent requirements, and therefore with high-est energy requirements, when the option of recycling thesolvent is chosen.
In this context, Sereewatthanawut and coworkers (123)proposed the dual membrane diafiltration (DMD) processto reduce the use of solvent to achieve a high yield of theproduct (i.e., the diafiltration limitation). The DMD process(Figure 8) combines two membrane stages, i.e., a purifica-tion stage combined with a solvent recovery stage. Insteadof adding fresh solvent to the process, recovered solventfrom the solvent recovery stage is returned to the purifica-tion stage to provide further purification. The feasibility ofthe DMD process is demonstrated through the purification ofan intermediate compound involved in the synthesis of a drugcandidate in the Janssen Pharmaceutica portfolio (API-INT).This is the fourth case-study selected in the present reviewas example of post-reaction purification via OSN (Table 3,entry 6).
During the synthesis of API-INT (MW 675 Da), its iso-mer B, and a series of oligomeric impurities based on API-INT with MW >1000 Da (i.e., dimers, trimers, tetramers,pentamers, etc.) are also formed. The solvent used is THF.Permeate containing the purified product from the primarymembrane filtration (Membrane 1) is fed into the solventrecycle stage (Membrane 2). API-INT, isomer B, and theoligomeric impurities are retained by Membrane 2 in the sol-vent recovery stage, whilst the solvent freely passes throughthe membrane (Figure 8).
The membranes used in stage 1 and stage 2 areDuraMem500 and DuraMem300, respectively. The integra-tion of solvent recovery in the DMD process can massivelyreduce the fresh solvent requirement for purifying a fixedmass of feed solution I by up to 90%, in an ideal pro-cess where 0% of solvent is lost from the system viaevaporation and/or no dead volume is present in the sys-tem. Furthermore, DMD advantageously avoids generatinga diluted product which would require further downstreamprocessing. Sereewatthanawut et al. (123) compared the effi-ciency of OSN with that of crystallisation, and charcoaltreatment to remove oligomeric impurities from API-INT.The obtained results show that OSN is a very promis-ing technology as it successfully removed the problematicoligomeric impurities, with a yield of 99.2%.
The last two examples of post-reaction purification viaOSN discussed in this review concern removal of dye impu-rities such as Brilliant Blue R (BBR, MW 826 Da) (Table 3,entries 7, 8). One of the common process-related impurities
ORGANIC SOLVENT NANOFILTRATION AND PHARMACEUTICALS 175
FIGURE 8 Dual membrane diafiltration (DMD) process: separation of compound X (in retentate stream) from compound Y (in permeate stream) during firststage; separation of compound Y (in the retentate stream) from the solvent, which recovered, is recycled back into the primary stage.© (2010) American Chemical Society. Reprinted with permission from (123). Permission to reuse must be obtained from the rightsholder.
in pharmaceutical manufacturing process is colored by-products. Because they are normally present at a trace level(�0.1%) and have structural similarities with the API, col-ored impurities are not easy to separate by conventionalprocesses and visibly affect the product quality. Muller et al.(124) pointed out that a major hurdle in APIs process devel-opment is to separate an organic synthesis intermediate froma mixture comprising multiple components, including inor-ganic salts, polymers, isomers, and colored by-products. Thelast two items were found to be the most challenging task inpurification when the first two could be easily removed byquenching and charcoal treatment, respectively.
Lin et al. (125) compared different purification schemesof OSN membrane processes for dyes impurities removal,i.e., a three-stage configuration and diafiltration. MartiusYellow (MY, 2,4-dinitro-1-naphthol sodium salt, MW274.16 Da) and Brilliant Blue R (BBR, MW 826 Da) werechosen respectively as API and color impurity in manufac-turing process. STARMEM 228, was chosen as membrane inthis study (Table 3, entry 7). Average rejections of the BBRmodel impurity in the first, second and third stage were 99.6,
99.1 and 100%, respectively, confirming the efficiency of theOSN process to separate the product and impurity into twostreams, i.e. the first stage permeate (Permeate 1) and thefinal retentate.
BBR was also the model impurity in another OSN study(Table 3, entry 8) for 4-phenyl-azophenol (Solvent Yellow7, SY7) purification (123). The objective of the investi-gation was to demonstrate the use of OSN technologyfor diafiltration purification of the model product in harshorganic solvent such as DMF. A schematic overview of thecase-study is shown in Figure 9. To separate BBR from themodel product (SY7), the ideal membrane should completelyretain BBR whilst allowing SY7 to pass. A high permeateflux is also desirable to minimize the required membranearea. A screening of four flat-sheet DuraMem membranes,DuraMem150, 200, 300, and 500 was carried out at differentDMF operating pressures. DuraMem300 showed the highestretention for BBR (99.6 % at 20 bar) with a high permeateflux (29 L m−2 h−1 at 20 bar).
The feasibility of the OSN process was demonstratedby using a spiral-wound membrane module in diafiltration
FIGURE 9 Objectives of the separation in N,N-dimethylformamide (DMF) of organic dye Solvent Yellow 7 (SY7) from organic dye Brilliant Blue (BB).© (2010) American Chemical Society. Reprinted with permission from (123). Permission to reuse must be obtained from the rightsholder.
176 M. G. BUONOMENNA AND J. BAE
100 100.0
99.9
99.8
99.7
99.6
99.5
99.4
99.3
99.2
99.1
99.0
0.0
80
60
40
20
00 5 10
Yield of product (Experimental)Yield of product (Calculated)Purity of product (Experimental)Purity of product (Calculated)
15 20Diafiltration volume. N (-)
Yie
ld, Y
SY
7 (%
)
Pur
ity, P
SY
7 (%
)
FIGURE 10 Yield and purity profiles for the purification of Solvent Yellow 7 (SY7) from Brilliant Blue R (BBR) with N,N-dimethylformamide (DMF) usinga 1.8-in. × 12-in. DuraMem300 membrane module at 30◦C and 30 bar.© (2010) American Chemical Society. Reprinted with permission from (123). Permission to reuse must be obtained from the rightsholder.
mode using a kilo scale filtration unit. The system con-sisted of a 5 L capacity feed vessel and module housing(s)designed to hold the 1.8-in. × 12-in. spiral-wound module.Figure 10 shows the high product purity of 99.7% and prod-uct yield of 90% obtained after 10 diafiltration volumes. Thecalculated yield and purity were in good agreement withexperimental values, suggesting that the model is valid forprocess prediction and scale-up. The results obtained in thepurification of SY7 from BB demonstrated the feasibilityof OSN technology for the purification of organic solventsolutions containing products and impurities in the molec-ular weight range of 200–1000 Da. The concept can bereadily transferred to various applications in pharmaceuti-cal and natural products production (e.g., separation of APIfrom lower/higher MW by-products in the pharmaceuticalindustry and removal of free fatty acids (MW 200–300 Da)from glycerides (MW 600–800 Da) present in natural oils.Depending on the application target, the process can beadapted to deliver the target yield and purity (123).
Hybrid Processes
The utilization of membrane operations as hybrid systems,in combination with other conventional techniques or inte-grated with different membrane operations, is considered theway forward more efficient and rational processes. Threerepresentative examples of hybrid processes based on OSNfor purification are reported in the present review. The firstone focuses on the energy comparison among three purifi-cation systems: OSN, distillation, and a hybrid combinationof both techniques (126). The second example concerns the
synergic combination of OSN with a new emerging selectivepurification strategy, the molecularly imprinted polymers(MIP) (127). The third example describes the improvementof countercurrent chromatography (CCC) when combinedwith OSN (128).
Rundquist et al. (126) calculated that even with optimizedconditions for distillation, its overall energy consumption is32 times higher than that of OSN. However, for OSN, themaximum amount of solvent recovered is limited by solu-bility: the concentration should remain above the solubilitylimit to prevent solute precipitation that would damage themembrane module. Therefore, the maximum amount of sol-vent recovered by OSN is limited to 80% whereas distillationcan be continued until 90% of the original volume has beenrecovered. Rundquist also compared OSN, distillation, anda hybrid OSN/distillation combination on the basis of thesustainability metrics by Curzon et al. (129). The sustain-ability metric criterion is obtained as the ratio of the energyused for solvent recovery over the total mass of purifiedproduct.
This criterion showed that OSN was the more favorableprocess. However, as the volume of solvent recovered withOSN is lower compared to distillation, a relatively largeramount of waste will remain after OSN recovery process-ing. An equivalent volume recovery of 90% could be reachedby using OSN to recover 80% of the original solvent added,and distillation to recovery an additional 10% up to a 90%total. Equivalent volume recovery can be obtained throughuse of a combined approach of OSN and distillation. Energyassessment shows that the hybrid approach is nine timesmore efficient than distillation used alone (126).
ORGANIC SOLVENT NANOFILTRATION AND PHARMACEUTICALS 177
Székely et al. (127) proposed a hybrid approach for thepurification of mometasone furoate (Meta), an API, fromGTIs: a synergic combination of OSN and MIP where OSNcan remove genotoxic 1,3-diisopropylurea (IPU) at highloads and MIP can remove IPU effectively at low concen-trations. IPU (MW 144 Da) is partially retained by OSNmembranes. A significant IPU removal requires diafiltra-tions at high dilution ratios, which leads to high API losses.Considering rejections of about 25% and 99%, for GTI andAPI, respectively, one can calculate about 90% GTI removal(with an API loss of 3%) for a dilution ratio of 3. However,applying a dilution ratio of 5 would allow for a GTI removalof 97% at the cost of 5% API loss.
In the hybrid approach OSN/MIP, diafiltration is firstused to remove the greater part of the IPU. On the basis ofthe first membrane screening to select the best OSN mem-brane, GMT-oNF-2 (see Table 2, entry 11) resulted to be themost suitable for separating Meta from IPU when dissolvedin dichloromethane, with a lower IPU rejection and a highMeta rejection. The transition between the OSN and MIPstages takes place with an IPU remaining concentration ofapproximately 100 ppm, a value particular suitable for MIPoperations. MIP is then used to polish the Meta rich-retentatestream with in ultra-low IPU levels.
Rundquist et al. (128) reported a combined CCC/OSNapproach for separating an API from a heavily contami-nated waste stream. The approach uses OSN to improvethe application of counter-current chromatography (CCC)in an industrial process. In this case, the OSN impact ison the mass intensity metric of the process, related to theorganic solvent recovery. OSN provides an efficient routefor exchanging solutes from the process solvent into thedesired mobile phase for CCC, generating a CCC feed con-taining less than 0.01% (area % by GC) of the originalprocess solvents. The recovered solvent was then success-fully recycled into a subsequent CCC run with no indicationof impurity build-up. Coupling OSN with CCC improvedthe mass-intensity of the CCC process, reducing the solvent
use by 56%. OSN can be a useful tool in facilitating theapplication of CCC to pharmaceutical process streams.
Additional examples about isolation and concentration ofpharmaceuticals by means of OSN are reported in the reviewby Vandezande et al. (8), who analyzed literature publishedup to 2007.
Solvent Exchange
In the pharmaceutical industry, a significant part of APIs isproduced in a multistep chemical synthesis, each of whichmust be performed in another solvent, while the isolationof the product itself occurs in a specific solvent. Thereforea problem of solvent exchange exists on transition fromone step to the other. Usually distillation is used for sol-vent exchange: the major part of the first solvent is distilledoff, and then the second solvent is repeatedly added to, anddistillated from the residue. Geens et al. (63) made an eval-uation for the energy consumption of nanofiltration insteadof distillation for a pharmaceutical company in the FlemishBelgium region. OSN is a nonthermal technique and themajority of power required for operation is consumed bya pump generating the required back-pressure. Therefore,in the study by Geens et al. (63), the energy used for theOSN pump was compared to that needed for the distilla-tion boiler. The comparison was in favor of the OSN processthat might become a profitable alternative to traditional dis-tillation. This conclusion was confirmed by Rundquist et al.(103), who investigated the feasibility of using OSN as analternative to distillation for solvent recovery showing thatOSN used 25 times less energy per L of recovered sol-vent than distillation. Membrane-based solvent exchangescan take place at room temperature or at the operationtemperature, regardless of the boiling point of the solventsinvolved.
Livingston (130) patented a membrane-based solventexchange process to alter the mixture of solvents in whichone or more solutes is dissolved. The invention provides
a) b)
SCHEME 5 Membrane-based solvent exchange process reported in Ref. (130): a) dead-end and b) cross-flow configurations.
178 M. G. BUONOMENNA AND J. BAE
1st
stage TOABr rejection: 96.7 %
2nd
stage TOABr rejection: 99.2 %
TOABr mass balance: 91.7 wt% (after1 hr)
Flowrate ratio: 0.92
1 st
stage
Permeate 1
TOABr: 0.0049 wt%
Methanol: 31.8 wt%
Flowrate: 5.80 ml min–1
Initial solvent stream
TOABr: 0.139 wt%
Toluene: 99.8 wt%
Flowrate: 6.24 ml min–1
2nd
stage
Permeate 2
TOABr: 0.0012 wt%
Methanol: 55.7 wt%
Flowrate: 7.44 ml min–1
Retentate
TOABr: 0.147 wt%
Methanol: 59.2 wt%
Flowrate: 5.46 ml min–1
Replacing solvent stream
TOABr: 0 wt%
Methanol: 100 wt%
Flowrate: 5.74 ml min–1
Retentate
TOABr: 0.172 wt%
Methanol: 47.8wt%
Flowrate: 3.60 ml min–1
Replacing solvent stream
TOABr: 0 wt%
Methanol: 100 wt%
Flowrate: 5.02 ml min–1
1 st
stage
Permeate
TOABr: 0.0106 wt%
Methanol: 48.6 wt%
Flowrate: 6.10 ml min–1
Initial solvent stream
TOABr: 0.139 wt%
Toluene: 99.8 wt%
Flowrate: 5.00 ml min–1
TOABr rejection: 93.8 %
TOABr mass balance: 103.53 % (after1 hr)
Flowrate ratio: 1.00
1st
stage TOABr rejection: 74.0%
2nd
stage TOABr rejection: 90.8%
3rd
stage TOABr rejection: 98.4%
TOABr mass balance: 101.2 wt% (after1 hr)
Flowrate ratio: 0.95
2nd
stage
Permeate 2
TOABr: 0.0138 wt%
Methanol: 55.0 wt%
Flowrate: 6.28 ml min–1
Initial solvent stream
TOABr: 0.15 wt%
Toluene: 99.8 wt%
Flowrate: 4.33 ml min–1
3rd
stage
Permeate 3
TOABr: 0.0024 wt%
Methanol: 74.4 wt%
Flowrate: 7.37 ml min–1
Retentate
TOABr: 0.135 wt%
Methanol: 75.3 wt%
Flowrate: 3.30 ml min–1
Replacing solvent stream
TOABr: 0 wt%
Methanol: 100 wt%
Flowrate: 4.09 ml min–1
1st
stage
Permeate 1
TOABr: 0.0392 wt%
Methanol: 29.3 wt%
Flowrate: 5.36 ml min–1
FIGURE 11 Comparison of the technical data for the single, two- and three-stage solvent exchange processes studied by Kim et al. (102).© Elsevier. Reproduced from (102) with permission from Elsevier. Permission to reuse must be obtained from the rightsholder.
a process for carrying out a solvent exchange by alter-ing the composition of an organic liquid containing atleast one first solvent, at least one second solvent and atleast one solute. The obtained final organic liquid mix-ture has a reduced concentration of the first solvent andan increased concentration of the second solvent. Preferredmembranes reported in this patent are integrally skinnedPI membranes. Two general membrane operation modes,the dead-end (Scheme 5a) and cross-flow filtration pro-cesses (Scheme 5b), are proposed for the OSN-basedexchange process.
In a dead-end process (Scheme 5a), pressure is appliedto the organic liquid through an inert gas. The organic liq-uid above the membrane is kept stirred using a stirrer toreduce fouling of the membrane surface. A portion of theorganic liquid permeates through the membrane. The sec-ond solvent is fed to the cell via a pump and a second pumpremoves organic liquid from the cell. This style of opera-tion may lead to lower fluxes than the same configurationoperating continuously, as in the batch process the soluteduring each filtration stage may tend to build up as a layeron the surface of the membrane as it becomes concentrated
ORGANIC SOLVENT NANOFILTRATION AND PHARMACEUTICALS 179
in the retentate. However, it will provide a solvent exchangeto the same degree as the continuous process while using lessamount of the second solvent. In the cross-flow configuration(Scheme 5b), it is possible to use the cross-flow velocity toavoid build-up of solute layers on the membrane surface.
Kim et al. (102) proposed a continuous process for sol-vent exchange (Table 3, entry 8): a counter-current mem-brane cascade. The effect of the process parameters, suchas number of stages and flow rate ratio of replacing solventto initial solvent, on the solvent exchange performance istested through simulations and experiments. Two solventscommonly employed in organic synthesis such as tolueneand methanol were chosen for use in this study. In thisstudy, a quaternary ammonium salt, tetraoctylammoniumbromide (TOABr, MW 547 Da), was chosen as markersolute to represent an API intermediate. A STARMEM122 membrane (Table 2, entry 3) was used. Figure 11 sum-marizes key operational parameters and compares detailedresults of the single-, two- and three-stage solvent exchangeprocesses. From these experiments, a general conclusionwas that increasing numbers of stages in the cascadegave higher concentrations of the replacing solvent inthe product stream. The experimental results are in goodagreement with the simulations and indicated that with athree-stage cascade a 75.3% solvent exchange is obtained.However, this configuration will not be suitable for com-plete exchange of two solvents as this would require aninfinite numbers of stages (102). Therefore, it was suggestedto integrate it into other solvent exchange processes or tomodify it.
CONCLUSIONS
The success of the application of OSN for the removal ofAPIs from organic solvents depends on an in-depth analysisof the system, in particular of the solute-solvent combi-nation. The selected membrane should be resistant againstthe solvent(s) used in the application, which might be dif-ficult when aggressive solvents are considered. The intenseresearch of this last decade in this direction gave interestingnew polymeric materials, which can be used with harsh sol-vents such as DMF or THF. On the other side, the membraneis not efficient enough by itself. An accurate study of theexperimental process in terms of the optimal stage numberfor the membrane cascade is mandatory.
The advantages of OSN in terms of energy consumption,operational feasibility and scalability compared to other sep-aration techniques such as flash chromatography, distillationor crystallization, prevail significantly on the disadvantageof using considerable amount of fresh solvent to achievea high yield of the product, if OSN is integrated with asolvent recovery stage. The interest of OSN is confirmedby the promising results obtained with membrane enhancedpeptide synthesis (MEPS) in a new technology platform thatadvantageously combines OSN with solution-phase peptidesynthesis.
REFERENCES
1. Pink, C. and Rundquist, E. (2012) Organic solvent nanofiltration(OSN) in pharmaceutical processing. The Network Young Membrains14 Conference (NYM14), Imperial College London, London, UK,September 20–22, 2012.
2. Jiménez-González, C., Curzons, A.D., Constable, D.J.C., andCunningham, V.L. (2004) Expanding GSK’s solvent selection guide— Application of life cycle assessment to enhance solvent selections.Clean Techn. Environ. Pol., 7: 42–50.
3. Charpentier, J.C. (2007) Modern chemical engineering in the frame-work of globalization, sustainability, and technical innovation. Ind.Eng. Chem. Res., 46: 3465–3485.
4. Baker, R.W. and Lokhandwala, K. (2008) Natural gas processing withmembranes: An overview. Ind. Eng. Chem. Res., 47(7): 2109–2121.
5. Buonomenna, M.G. (2013) Membrane processes for a sustainableindustrial growth. RSC Advan., 3: 5694–5740.
6. Bhore, N.A., Gould, R.M., Jacob, S.M., Staffeld, P.O., McNally, D.,Smiley, P.H., and Wildemuth, C.R. (1999) New membrane processdebottlenecks solvent dewaxing unit. Oil Gas J., 97(46): 67–74.
7. Gould, R.M., White, L.S., and Wildemuth, C.R. (2001) Membraneseparation in solvent lube dewaxing. Environ. Progr., 20 (1): 12–16.
8. Vandezande, P., Gevers, L.E.M., and Vankelecom, I. F. J. (2008)Solvent resistant nanofiltration: separating on a molecular level.Chem. Soc. Rev., 37: 365–405.
9. Volkov, A.V., Korneeva, G.A., and Tereshchenko, G.F. (2008) Organicsolvent nano-filtration: Prospects and application. Russian Chem.Rev., 77 (11): 983–993.
10. Peeva, L.G., Malladi, S., and Livingston A.G. (2010) Nanofiltrationoperations in nonaqueous systems. In Comprehensive MembraneScience and Engineering, Volume 1; Drioli, E., Giorno, L., eds.;Elsevier: London.
11. Zeidler, S., Kätzel, U., and Kreis, P. (2013) Systematic investiga-tion on the influence of solutes on the separation behavior of aPDMS membrane in organic solvent nanofiltration. J. Membr. Sci.,429:295–303.
12. Li, Y., Verbiest, T., and Vankelecom, I. (2013) Improving the flux ofPDMS membranes via localized heating through incorporation of goldnanoparticles. J. Membr. Sci., 428: 63–69.
13. Gevers, L.E.M., Vankelecom, I.F.J., and Jacobs, P.A. (2006) Solvent-resistant nanofiltration with filled polydimethylsiloxane (PDMS)membranes. J. Membr. Sci., 278: 199–204.
14. Gevers, L.E.M., Aldea, S., Vankelecom, I. F.J., and Jacobs, P.A.(2006) Optimisation of a lab-scale method for preparation of com-posite membranes with a filled dense top-layer. J. Membr. Sci., 281:741–746.
15. Basu, S., Maes, M., Cano-Odena, A., Alaerts, L., De Vos, D.E.,and Vankelecom, I.F.J. (2009) Solvent resistant nanofiltration (SRNF)membranes based on metal-organic frameworks. J. Membr. Sci., 344:190–198.
16. Soroko, I., Bhole, Y., and Livingston, A.G. (2011) Environmentallyfriendly route for the preparation of solvent resistant polyimidenanofiltration membranes. GreenChemistry, 13: 162–168.
17. Soroko, I., Lopes, M.P., and Livingston, A. (2011) The effect ofmembrane formation parameters on performance of polyimide mem-branes for organic solvent nanofiltration (OSN): Part A. Effect ofpolymer/solvent/non-solvent system choice. J. Membr. Sci., 381:152–162.
18. Soroko, I., Sairam, M., and Livingston, A.G. (2011) The effectof membrane formation parameters on performance of poly-imide membranes for organic solvent nanofiltration (OSN). PartC. Effect of polyimide characteristics. J. Membr. Sci., 381:172–182.
19. Holda, A.K., Aernouts, B., Saeys, W., and Vankelecom, I.F.J. (2013)Study of phase inversion parameters for polysulfone-based SRNFmembranes. J. Membr. Sci., 442: 196–205.
180 M. G. BUONOMENNA AND J. BAE
20. Hołda, A.K. and Vankelecom, I.F.J. (2014) Integrally skinned PSf-based SRNF-membranes prepared via phase inversion—Part A:Influence of high molecular weight additives. J. Membr. Sci., 450:499–511.
21. Hołda, A.K. and Vankelecom, I.F.J. (2014) Integrally skinned PSf-based SRNF-membranes prepared via phase inversion—Part B:Influence of low molecular weight additives. J. Membr. Sci., 450:512–521.
22. Hendrix, K., Koeckelberghs, G., and Vankelecom, I.F.J. (2014) Studyof phase inversion parameters for PEEK-based nanofiltration mem-branes. J. Membr. Sci., 452: 241–252.
23. Buonomenna, M.G., Golemme, G., Jansen, J.C., and Choi, S.-H.(2011) Asymmetric PEEKWC membranes for treatment of organicsolvent solutions. J. Membr. Sci., 368: 144–149.
24. Hendrix, K., Vaneynde, M., Koeckelberghs, G., and Vankelecom, I.(2013) Synthesis of modified poly(ether ether ketone) polymer for thepreparation of ultrafiltration and nanofiltration membranes via phaseinversion. J. Membr.Sci., 447: 96–106.
25. Hołda, A.K., De Roeck, M., Hendrix, K., and Vankelecom, I.F.J.(2013) The influence of polymer purity and molecular weight on thesynthesis of integrally skinned polysulfone membranes. J. Membr.Sci.,446: 113–120.
26. Jimenez Solomon, M.F., Bhole Y., and Livingston, A.G. (2012)High flux membranes for organic solvent nanofiltration (OSN)-Interfacial polymerization with solvent activation. J. Membr. Sci.,423–424:371–382.
27. Jimenez Solomon, M.F., Bhole, Y., and Livingston, A.G. (2013)High flux hydrophobic membranes for organic solvent nanofiltration(OSN)—Interfacial polymerization,surface modification and solventactivation. J. Membr. Sci., 434: 193–203.
28. Woishnis, W. and Ebnesajjad, S. (Eds.) (2012) Chemical Resistanceof Specialty Thermoplastics; Elsevier Inc.: New York, 863–870,1399–1417.
29. Valtcheva, I., Kumbharkar, S.C., Kim, J.F., Bhole,Y., Livingston, A.G. (2014) Beyond polyimide: Cross-linked polybenzimidazole mem-branes for organic solvent nanofiltration (OSN) in harsh environmentsJ. Membr. Sci., http://dx.doi.org/10.1016/j.memsci.2013.12.069.
30. Mark, J.E. (Ed.) (1999) Polymer Data Handbook; Oxford UniversityPress: London, 466–470.
31. Peeva, L, Arbour, J., and Livingston, A. (2013) On the potential oforganic solvent nanofiltration in continuous heck coupling reactions.Org. Process Res. Dev., 17:967–975.
32. Buonomenna, M.G., Figoli, A, Jansen, J. C. and Drioli, E. (2004)Preparation of asymmetric PEEKWC flat membranes with differ-ent microstructures by wet phase inversion. J. Appl. Polym. Sci., 92:576–591.
33. Jansen, J.C., Buonomenna, M.G., Figoli, A., and Drioli, E. (2006)Asymmetric membranes of modified poly(ether ether ketone) withan ultra-thin skin for gas and vapor separations. J. Membr. Sci., 272:188–197.
34. Buonomenna, M.G., Gordano, A., and Drioli, E. (2008)Characteristics and performance of new nanoporous PEEKWCfilms. Eur. Polym. J., 44: 2051–2059.
35. Buonomenna, M.G., Gordano, A., Golemme, G., and Drioli, E. (2009)Preparation, characterization and use of PEEKWC nanofiltrationmembranes for removal of Azur B dye from aqueous media. React.Funct. Polym., 69: 259–263.
36. Buonomenna, M.G., Lopez, L.C., Davoli, M., Favia, P., D’AGOSTINO, R., and Drioli, E. (2009) Polymeric membranesmodified via plasma for nanofiltration of aqueous solution containingorganic compounds. Micropor. Mesopor. Mat., 120:147–153.
37. Hendrix, K., Vanherck, K., and Vankelecom, I.F.J. (2012)Optimization of solvent resistant nanofiltration membranes pre-pared by the in-situ diamine cross-linking method. J. Membr. Sci.,421–422:15–24.
38. Ahmadiannamini, P., Li, X., Goyens, W., Joseph, N., Meesschaert, B.,and Vankelecom, I.F.J. (2012) Multilayered polyelectrolyte complex
based solvent resistant nanofiltration membranes prepared from weakpolyacids. J. Membr. Sci., 394– 395: 98– 106.
39. Hendrix, K., Vaneynde, M., Koeckelberghs, G., and Vankelecom,I.F.J. (2013) Cross-linking of modified poly (ether ether ketone) mem-branes for use in solvent resistant nanofiltration. J. Membr. Sci., 447:212–221.
40. Buonomenna, M.G., Golemme, G., Jansen, J.C., and Choi, S.-H.(2011) Asymmetric PEEKWC membranes for treatment of organicsolvent solutions. J. Membr. Sci., 368: 144–149.
41. Hendrix, K., Vaneynde, M., Koeckelberghs, G., and Vankelecom,I. J. (2013) Cross-linking of modified poly(ether ether ketone) mem-branes for use in solvent resistant nanofiltration. J.Membr. Sci., 447:212–221.
42. Vanherck, K., Cano-Odena, A., Koeckelberghs, G., Dedroog, T.,and Vankelecom, I.F.J. (2010) A simplified diamine cross-linkingmethod for PI nanofiltration membranes. J. Membr. Sci., 353:135–143.
43. Vanherck, K., Koeckelberghs, G., Vankelecom. I. F.J. (2013) Cross-linking polyimides for membrane applications: A review. Progr.Polym. Sci., 38: 874–896.
44. See-Toh, Y. H, Silva, M., and Livingston, A. G. (2008) Controllingmolecular weight cut-off curves for highly solvent stable organic sol-vent nanofiltration (OSN) membranes. J. Membr. Sci., 324: 220–232.
45. Siddique, H., Rundquist, E., Bhole, Y., Peeva, L.G., and Livingston,A.G. (2014) Mixed matrix membranes for organic solventnanofiltration. J. Membr. Sci., 452: 354–366.
46. Aerts, S., Vanhulsel, A., Buekenhoudt, A., Weyten, H., Kuypers, S.,Chen, H., Bryjak, M.,Gevers, L.E.M., Vankelecom, I.F.J., and Jacobs,P.A. (2006) Plasma-treated PDMS-membranes in solvent resistantnanofiltration: Characterization and study of transport mechanism. J.Membr. Sci., 275: 212–219.
47. Jimenez, M.S., Bhole, Y.S., and Livingston, A. G. (2011) Solventresistant polyamide nanofiltration membranes. Great Britain PatentPCT/GB2011/051364, July 12, 2011.
48. Li, X., Goyens,W., Ahmadiannamini, P., Vanderlinden, W., De Feyter,S., and Vankelecom, I.F.J. (2010) Morphology and performanceof solvent-resistant nanofiltration membranes based on multilayeredpolyelectrolytes: Study of preparation conditions. J. Membr. Sci., 358:150–157.
49. Volkov, A.V., Stamatialis, D.F., Khotimsky, V.S., Volkov, V.V.,Wessling, M., and Plate, N.A. (2006) Poly[1-(trimethylsilyl)-1-propyne] as a solvent resistance nanofiltration membrane material. J.Membr. Sci., 281: 351–357.
50. Volkov, A.V., Parashchuk, V.V., Stamatialis, D.F., Khotimsky, V.S.,Volkov, V.V., and Wessling, M. (2009) High permeable PTMSP/PANcomposite membranes for solvent nanofiltration. J. Membr. Sci.,333:88–93.
51. Volkov, A., Yushkin, A., Grekhov, A., Shutova, A., Bazhenov, S.,Tsarkov, S., Khotimsky, V., Vlugt, T.J.H., and Volkov, V. (2013)Liquid permeation through PTMSP: One polymer for two differentmembrane applications. J. Membr. Sci., 440: 98–107.
52. McKeown, N.B., Budd, P.M., and Fritsch, D. (2005) Thin layer com-posite membranes with microporous layers and supporting layers.U.S. Patent WO 2005113121, December 2005.
53. Fritsch, D, Merten, P., Heinrich, K., Lazar, M., and Priske, M.(2012) High performance organic solvent nanofiltration membranes:Development and thorough testing of thin film composite membranesmade of polymers of intrinsic microporosity (PIMs). J. Membr. Sci.401–402:222–231.
54. Tsarkov, S., Khotimskiy, V., Budd, P. M., Volkov, V., Kukushkina, J.,and Volkov, A. (2012) Solvent nanofiltration through high permeabil-ity glassy polymers: Effect of polymer and solute nature. J. Membr.Sci. 423–424:65–72.
55. Nagai, K., Masuda, T., Nakagawa, T., Freeman, B.D., andPinnau, I. (2001) Poly[1-(trimethylsilyl) 1-propyne] and relatedpolymers: synthesis, properties and functions. Prog. Polym. Sci.,26:721−798.
ORGANIC SOLVENT NANOFILTRATION AND PHARMACEUTICALS 181
56. Meynen, V. and Buekenhoudt, A. (2012) Hybrid organic-inorganicmembranes for solvent filtration in Advanced materials for membranepreparation, Eds. M.G. Buonomenna, G. Golemme. BenthamPublishers.
57. Aran, H.C., Salamon, D., Groote, R., Jani, J.M., Chinthaginjala, J.K.,Lammertink, R.G.H., and Wessling, M. (2009) Ceramic membranemicroreactors. Proceedings of Euromembrane, Montpellier, France,6–10 September, 2009.
58. Pinheiro, A.F.M., Winnubst, L., and Nijmeijer, A. (2013)Functionalization of ceramic membranes by polymer grafting forsolvent nanofiltration, Proceedings of 13th International Conferenceof the European Ceramic Society 2013, Limoges, France, 23–27 June2013.
59. Buekenhoudt, A., Wyns, K., Meynen, V., Maes, B., and Cool, P.(2010) Surface-modified inorganic matrix and method for preparationthereof. U.S. Patent WO 2010/106167, March 2010.
60. Wyns, K., Buekenhoudt, A., Van Heetvelde, P., Cool, P., and Meynen,V. International Conference on Inorganic Membranes (ICIM12),University of Twente, Twente, Netherlands, July 9–13, 2012.
61. Van Heetvelde, P., Beyers, E., Wyns, K., Adriaensens, P., Maes,B.U.W., Mullens, S., Buekenhoudt, A., and Meynen, V. (2013) A newmethod to graft titania using Grignard reagents. Chem. Commun., 49:6998–7000.
62. Hosseinabadi, R. S., Wyns, K., Meynen, V., Carleer, R. Adriaensens,P., Buekenhoudt, A., and Van der Bruggen, B. (2014) Organic solventnanofiltration with Grignard functionalized ceramic nanofiltrationmembranes. J. Membr. Sci., 454: 496–504.
63. Geens, J., De Witte, B., and Van der Bruggen, B. (2007) Removal ofAPI’s (active pharmaceutical ingredients) from organic solvents bynanofiltration. Sep. Sci. Technol., 42: 2435–2449.
64. Schmidt, P. and Lutze, P. (2013) Characterisation of organic solventnanofiltration membranes in multi-component mixtures: Phenomena-based modelling and membrane modelling maps. J. Membr. Sci., 445:183–199.
65. Darvishmanesh, S., Firoozpour, L., Vanneste, J., Luis, P., Degr`eve,J., and Van der Bruggen, B. (2011) Performance of solvent resistantnanofiltration membranes for purification of residual solvent in thepharmaceutical industry: experiments and simulation. Green Chem.,13: 3476–3483.
66. Machado, D.R., Hasson, D., and Semiat, R. (1999) Effect of solventproperties on permeate flow through nanofiltration membranes. Part I:investigation of parameters affecting solvent flux. J. Membr. Sci., 163:93–102.
67. Machado, D.R., Hasson, D., and Semiat, R. (2000) Effect of solventproperties on permeate flow through nanofiltration membranes: PartII. Transport model. J. Membr. Sci., 166: 63–69.
68. Matsuura, T. and Sourirajan, S. (1981) Reverse osmosis transportthrough capillary pores under the influence of surface forces. Ind. Eng.Chem. Proc. Des. Dev., 20: 273–282.
69. Nakao, S.I. and Kimura, S. (1982) Models of membrane transport phe-nomena and their applications for ultrafiltration data. J. Chem. Eng.Jap., 15(3): 200–205.
70. Ferry, J.D. (1936) Ultrafilter membranes and ultrafiltration. Chem.Rev. 18: 373–455.
71. Van der Bruggen, B., Schaep, J., Vandecasteele, C., and Wilms, D.(2000). A comparison of models to describe the maximal retentionof organic molecules in nanofiltration. Separ.Sci. Technol., 35(2):169–182.
72. Verniory, A., Dubois, R., Decoodt, P., Gassee, J.P., and Lambert,P.P. (1973) Measurement of the permeability of biological mem-branes. Application to the glomerular wall. J. Gen. Physiol., 62(4),489–507.
73. Lonsdale, H., Merten, U., and Riley, R. (1965) Transport proper-ties of cellulose acetate osmotic membranes. J. Appl. Polym. Sci. 9:1341–1362.
74. Wijmans, J.G. and Baker, R.W. (1995) The solution-diffusion model:a review. J. Membr. Sci., 107:1–21.
75. White, L.S. (2002) Transport properties of a polyimide solvent resis-tant nanofiltration membrane. J. Membr. Sci., 205: 191–202.
76. Bhanushali, D., Kloos, S., and Bhattacharyya, D. (2002) Solute trans-port in solvent-resistant nanofiltration membranes for non-aqueoussystems: experimental results and the role of solute–solvent coupling.J. Membr. Sci., 208: 343–359.
77. Stafie, N., Stamatialis, D.F., and Wessling, M. (2004) Insight intothe transport of hexane–solute systems through tailor-made compositemembranes. J. Membr. Sci., 228: 103–116.
78. Paul, D.R. (2004) Reformulation of the solution-diffusion theory ofreverse osmosis. J. Membr. Sci., 241: 371–386.
79. Vankelecom, I.F.J., De Smet, K., Gevers, L.E.M., Livingston, A.G.,Aerts, S., Kuypers, S., and Jacobs, P.A. (2004) Physico-chemicalinterpretation of the SRNF transport mechanism for solvents throughdense silicone membranes. J. Membr. Sci., 231: 99–108.
80. Gibbins, E., D’Antonio, M., Nair, D., White, L.S., Freitasdos Santos, L.M., Vankelecom, I.F.J., and Livingston, A.G.(2002) Observations on solvent flux and solute rejection acrosssolvent resistant nanofiltration membranes. Desalination, 147:307–313.
81. Van der Bruggen, B., Geens, J., and Vandecasteele, C. (2002)Influence of organic solvents on the performance of polymericnanofiltration membranes. Sep. Sci. Technol., 37: 783–797.
82. Schmidt, M., Mirza, S., Schubert, R., Rödicker, H., Kattanek, S.,and Malisz, J. (1999) Nanofiltrationsmembranen für Trennproblemein organischen Lösungen. Chem.-Ing.-Tech., 71: 199–206.
83. Hu, A, Zhu, W., Wu, Z., and Jing, Y. (2003) Recovery of clindamycinfrom fermentation wastewater with nanofiltration membranes. WaterRes., 37(15): 3718–3732.
84. White, L.S. and Wildemuth, C.R. (2006) Aromatics enrichment inrefinery streams using hyperfiltration. Ind. Eng. Chem. Res., 45:9136–9143 b) US Patent 5,624,166 and WO 00/06293.
85. Lin, J.C. and Livingston, A.G. (2007) Nanofiltration membranecascade for continuous solvent exchange. Chem. Eng. Sci., 62:2728–2736.
86. Rundel, J.T., Paul, B.K., and Remcho, V.T. (2007) Organic sol-vent nanofiltration for on-chip purification of PAMAM dendrimer. J.Chromatogr. A, 1162: 167–174.
87. Valadez-Blanco, R., Ferreira, F.C., Jorge, R.F., and Livingston, A.G.(2008) A membrane bioreactor for biotransformations of hydrophobicmolecules using organic solvent nanofiltration (OSN) membranes. J.Membr. Sci., 317: 50–64.
88. EVONIK Industries. Fluid processing in a spiral wound membrane.http://duramem.evonik.com/product/duramem-puramem/en/Pages/default.aspx (accessed September 16, 2013).
89. Cuperus, F.P. (2005) Recovery of organic solvents and valuable com-ponents by membrane separation. Chem. Ing. Tech., 77: 1000–1001.
90. Cuperus, F.P. Final report DSTI-GUTS Techno project. Evaluationof the use of SolSep membranes in process industries. http://ispt.eu/cusimages/Projects/CS-01-01%20Final%20report%20Evaluation%20of%20use%20of%20Solsep%20membranes.pdf (accessed Sept-ember 16, 2013)
91. Bargeman, G., Albers, J., Westerlink, J.B., Manahutu, C.F.H., and tenKate, A. (2006) International Workshop on Membranes in SolventFiltration, Campus Library Arenberg, Leuven, Belgium, March 23–24, 2006.
92. Wessely, L. and Samhaber, W.M. Organophilic nanofiltration to sepa-rate solvent mixtures. http://www.ivt.jku.at/Forschung/Publikationen/pdf/Forschungsbericht%20LWe%20en%2018-03-2011.pdf (accessedSeptember 16, 2013).
93. Scarpello, J.T., Nair, D., Freitas dos Santos, L.M., White, L.S.,and Livingston, A.G. (2002) The separation of homogeneousorganometallic catalysts using solvent resistant nanofiltration. J.Membr. Sci., 203: 71–85.
94. Raman, L.P., Cheryan, M., and Rajagopolan, N. (1996)Deacidification of soybean oil by membrane technology. J. Am.Oil Chem. Soc., 73: 219–224.
182 M. G. BUONOMENNA AND J. BAE
95. Yang, X.J., Livingston, A.G., and Freitas dos Santos, L. (2001)Experimental observations of nanofiltration with organic solvents. J.Membr. Sci., 190: 45–55.
96. Geens, J., Boussu, K., Vandecasteele, C., and Van der Bruggen, B.(2006) Modelling of solute transport in non-aqueous nanofiltration. J.Membr. Sci., 281: 139–148.
97. Mutter, M. (2006) International Workshop on Membranes in SolventFiltration, Campus Library Arenberg, Leuven, Belgium, March 23–24, 2006.
98. Martínez, M.B., Jullok, N., Rodriguez Negrin, Z., Van der Bruggen,B., and Luis, P. (2013) Effect of impurities in the recoveryof 1-(5-bromo-fur-2-il)-2-bromo-2-nitroethane using nanofiltration.Chem. Eng. Proc., 70: 241–249.
99. Ormerod, D., Sledsens, B., Vercammen, G., Van Gool, D., Linsen,T., Buekenhoudt, A., and Bongers, B. (2013) Demonstration ofpurification of a pharmaceutical intermediate via organic solventnanofiltration in the presence of acid. Sep. Purif. Technol., 115:158–162.
100. Vanneste, J., Ormerod, D., Theys, G., Van Gool, D., Van Camp, B.,Darvishmanesh, S. and Van der Bruggen, B. (2013) Towards highresolution membrane-based pharmaceutical separations. J. Chem.Technol. Biotechnol., 88: 98–108.
101. Lin, J.C.-T. and Livingston, A.G. (2007) Nanofiltration membranecascade for continuous solvent exchange. Chem. Eng. Sci., 62:2728–2736.
102. Kim, J.F., Freitas da Silva, A.M., Valtcheva, I.B., and Livingston,A.G. (2013) When the membrane is not enough: A simplified mem-brane cascade using Organic Solvent Nanofiltration (OSN). Sep. Purif.Technol., 116: 277–286.
103. Rundquist, E.M., Pink, C.J., and Livingston, A.G. (2012) Organicsolvent nanofiltration: a potential alternative to distillation for sol-vent recovery from crystallisation mother liquors. Green Chem., 14:2197–2205.
104. Murriel, C.L. and Dowdy, S.F. (2006) Influence of protein trans-duction domains on intracellular delivery of macromolecules. ExpertOpin. Drug Deliv., 3: 739–746.
105. Munyendo, W.L.L., Benza-Ingoula, H.L.H., Baraza, L.D., and Zhou,J. (2012) Cell penetrating peptides in the delivery of biopharmaceuti-cals. Biomolecules 2: 187–202.
106. Iannucci, N.B., González, R., Cascone, O., and Albericio, F. (2011)Novel strategy for designing antimicrobial peptides: an answer to thedevelopment of drug resistance? In Science Against Microbial Patho-gens: Communicating Current Research and Technological Advances;Méndez-Vilas, A. ed., Formatex Research Center, 961–967.
107. de Coupade, C., Fittipaldi, A., Chagnas, V., Michel, M., Carlier, S.,Tasciotti, E., Darmon, A., Ravel, D., Kearsey, J., Giacca, M., andCailler, F. (2005) Novel human-derived cell-penetrating peptides forspecific subcellular delivery of therapeutic biomolecules. Biochem. J.,390: 407–418.
108. Bayer, E. and Mutter, M. (1972) Liquid phase synthesis of peptides.Nature, 237: 512–513.
109. Bayer, E., Gillessen, D., Kuenzi, H., Mutter, M., and Studer, R.Hoffmann, F. Homogenous peptide synthesis. U.S. Patent 3772264A, November 1973.
110. So, S., Peeva, L.G., Tate, E.W., Leatherbarrow, R.J., and Livingston,A.G. (2010) Membrane enhanced peptide synthesis. Chem. Commun.,46: 2808–2810.
111. So, S., Peeva, L.G., Tate, E.W., Leatherbarrow, R.J., and Livingston,A. G. (2010) Organic solvent nanofiltration: A new paradigm inpeptide synthesis. Org. Proc. Res. Dev., 14: 1313–1325.
112. Vasconceles, R.C., Gaffney, P.R. J., Leatherbarrow, R.J., Livingston,A.G., Peeva, L.G., So, S. W.J., and Tate, E.W. (2011) Solvent resis-tant diafiltration of peptides, pna or oligonucleotides. U.S. Patent20110245460 A1, October 2011.
113. Zompra, A.A., Galanis, A.S., Werbitzky, O., and Albericio, F. (2009)Manufacturing peptides as active pharmaceutical ingredients. FutureMed. Chem., 1(2): 361–377.
114. Bayer, E., Mutter, M., Uhmann, R., Polster, J., and Mauser, H. (1974)Kinetic studies of liquid-phase peptide synthesis. J. Am. Chem. Soc.,96: 7333–7336.
115. Ormerod, D. (2008) Proceedings of Innovation for SustainableProduction, Bruges, Belgium.
116. Anonymous (2006) Impurities in new drug substances: (adoptedguideline), in: International Conference on Harmonization of techni-cal requirements of Pharmaceuticals for Human Use.
117. Maddula, S.R., Kharkar, M., Manudhane, K., Kale, S., Bhori, A., Lali,A., Dubey, P.K., Sarma, J.K.R., Bhattacharya, A., and Bandichhor,R. (2009) Preparative chromatography technique in the removal ofisostructural genotoxic impurity in rizatriptan: Use of physicochem-ical descriptors of solute and adsorbent. Org. Proc. Res. Dev., 13:683–689.
118. Lee, C., Helmy, R., Strulson, C., Plewa, J., Kolodziej, E.,Antonucci, V., Mao, B., Welch, C.J., Ge, Z., and Al-Sayah,M.A. (2010) Removal of electrophilic potential genotoxic impuri-ties using nucleophilic reactive resins. Org. Process Res. Dev., 14:1021–1026.
119. Darvishmanesh, S., Degrève, J., and Van der Bruggen, B. (2010)Performance of solvent-pretreated polyimide nanofiltration mem-branes for separation of dissolved dyes from toluene. Ind. Eng. Chem.Res., 49: 9330–9338.
120. Yang, Y. and Tjia, R. (2010) Process modeling and optimization ofbatch fractional distillation to increase throughput and yield in man-ufacture of active pharmaceutical ingredient (API). Comput. Chem.Eng., 34: 1030–1035.
121. Székely, G, Bandarra, J., Heggie, W., Sellergren, B., and Ferreira,F.C. (2011) Organic solvent nano-filtration: a platform for removalof genotoxins from active pharmaceutical ingredients. J. Membr. Sci.,381: 21– 33.
122. Székely, G., Gil, M., Sellergren, B., Heggie, W., and Ferreira, F.C.(2013) Environmental and economic analysis for selection and engi-neering sustainable API degenotoxification processes. Green Chem.,15: 210–225.
123. Sereewatthanawut, I., Lim, F.W., Bhole, Y.S., Ormerod,D., Horvath, A., Boam, A.T., and Livingston, A.G. (2010)Demonstration of molecular purification in polar aprotic sol-vents by organic solvent nanofiltration. Org. Process Res. Dev., 14:600–611.
124. Muller, M., Schneeberger, R., Wieckhusen, D., and Cooper, M. (2004)Example of finishing technologies as key elements for successfulactive pharmaceutical ingredient process development. Org. Proc.Res. Dev., 8(3): 376–380.
125. Lin, J.C.-T., Peeva, L.G., and Livingston, A.G. (2006) Separationof pharmaceutical process-related impurities via an organic solventnanofiltration membrane cascade. AIChE 2006 Annual Meeting, SanFrancisco, CA, November 12–17, 2006.
126. Rundquist, E.M., Pink, C.J., and Livingston, A.G. (2012) Organicsolvent nanofiltration: a potential alternative to distillation for sol-vent recovery from crystallization mother liquors. Green Chem. ,14:2197–2205.
127. Székely, G., Bandarra, J., Heggie, W., Sellergren, B., and CasteloFerreira, F. (2012) A hybrid approach to reach stringent low genotoxicimpurity contents in active pharmaceutical ingredients: Combiningmolecularly imprinted polymers and organic solvent nanofiltration forremoval of 1,3-diisopropylurea. Sep. Purif. Technol., 86: 79–87.
128. Rundquist, E., Pink, C., Vilminot, E., and Livingston, A. (2012)Facilitating the use of counter-current chromatography in pharma-ceutical purification through use of organic solvent nanofiltration. J.Chromatogr A, 1229: 156–163.
129. Curzons, A.D., Constable, D.J.C., Mortimer, D.N., and Cunningham,V.L. (2001) So you think your process is green, how do you know?—Using principles of sustainability to determine what is green–acorporate perspective. Green Chem., 3: 1–6.
130. Livingston, A.G. (2008) Membrane for solvent exchange process.European Patent EP 1 372 829 B1, December 3, 2008