Nanofiltration-Enabled In Situ Solvent and Reagent Recycle for Sustainable Continuous-Flow Synthesis

Article abstract of DOI:10.1002/cssc.201701120

Solvent usage in the pharmaceutical sector accounts for as much as 90 % of the overall mass during manufacturing processes. Consequently, solvent consumption poses significant costs and environmental burdens. Continuous processing, in particular continuou

Full text of DOI:10.1002/cssc.201701120

Accepted Article
Title: Nanofiltration-enabled in situ solvent and reagent recycle for
sustainable continuous-flow synthesis
Authors: Tamas Fodi, Christos Didaskalou, Jozsef Kupai, Gyorgy T
Balogh, Peter Huszthy, and Gyorgy Szekely
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: ChemSusChem 10.1002/cssc.201701120
Link to VoR: http://dx.doi.org/10.1002/cssc.201701120
A Journal of
www.chemsuschem.org

ChemSusChem
10.1002/cssc.201701120
FULL PAPER
Nanofiltration-enabled in situ solvent and reagent recycle for
sustainable continuous-flow synthesis
Tamas Fodi,†[a-c] Christos Didaskalou,†[a] Jozsef Kupai, Gyorgy T. Balogh, Peter Huszthy, and
[b]
[c]
[b]
Gyorgy Szekely*[a]
Abstract: The solvent usage in the pharmaceutical sector accounts
for as much as 90% of the overall mass during manufacturing
processes. Consequently, solvent consumption poses significant
costs and environmental burden. Continuous processing, in
particular continuous-flow reactors have a great potential in the
sustainable production of pharmaceuticals but subsequent
downstream processing remains challenging. Separation processes
for the concentration and purification of chemicals can account for as
much as 80% of the total manufacturing costs. In this work, a
nanofiltration unit was coupled to a continuous-flow rector for in situ
solvent and reagent recycle. The nanofiltration unit is straightforward
to implement and control in a continuous operation. The hybrid
process was continuously operated over 6 weeks recycling about
energy usage.[3a-b] The solvent usage in the pharmaceutical
industry accounts for 60% of the total energy consumption
required for the production of active pharmaceutical
ingredients.[ In another perspective, solvents account for
80-90% of the overall mass during the manufacturing processes
in the pharmaceutical industries.[5] Continuous-flow reactors can
4]
only be truly sustainable if solvent
recovery is
employed.[6a-b] Solvent waste has a large carbon footprint and
subsequently it is detrimental to the sustainability of the process.
[
7]
Organic solvent nanofiltration (OSN) is a green technology,
that has been used for solvent recovery.[
8a-g]
OSN can be
coupled to continuous processes for in situ separations.[
This approach is in line with the European Commission’s
"Roadmap to a resource efficient Europe" initiative seeking to
eliminate waste through deliberate process design with resource
efficiency and recycling in mind.[10]
8g,9a-i]
90% of the solvent and the reagent. Consequently, the E-factor and
the carbon footprint were reduced by 91% and 19%, respectively.
Moreover, the nanofiltration unit concentrated the product 11 times
and simultaneously increased the purity from 52.4% to 91.5%. The
boundaries for process conditions were investigated to facilitate
implementation of the methodology by the pharmaceutical sector.
With profit margins growing thin, there is an imperative drive
for minimising both the cost and environmental impact via
process intensification. As effective tools, continuous
processing[11a-b] and membrane separations[3b,7,12] have been
recognized as core technologies that can provide green process
engineering. The newest generation of nanofiltration membranes
have stable performance in aggressive media including extreme
pH and organic solvents,[7,13] which opens new possibilities for
continuous solvent recovery, product purification and
concentration. Figure 1 shows the completive relationship for
continuous-flow synthesis and nanofiltration. These technologies
offer the opportunity towards smooth integration with further unit
operations. However, attempts for the synergistic coupling of
flow reactors with nanofiltration are scarce (Table 1). In this
context nanofiltration was used for the recovery of
homogeneous catalysts as well as for solvent exchange. In this
work, the authors present the development of a nanofiltration-
enabled in situ solvent and reagent recycle process in order to
improve the sustainability of flow reactors.
Introduction
The continuous-flow reactors can provide cost savings, facile
automation, higher level of consistence and certainty, improved
overall safety, and reduced carbon footprint.[1a-c] The application
of polymer-supported catalysts in flow reactors have been widely
researched in recent years because i) the use of a polymer
support eliminates the need for the removal of the catalyst from
the desired product, and ii) readily enables its reuse. However,
the downstream processing (i.e. concentration, purification and
isolation) in continuous production remains challenging. Recent
efforts demonstrate the need for the development of integrated
continuous synthesis-purification processes.[2a-b] It has recently
been realized that traditional downstream separation processes
can account for as much as two third of the total manufacturing
costs, and eventually contribute half of the industrial
[a]
[b]
[c]
T. Fodi, C. Didaskalou, Dr. G. Szekely
School of Chemical Engineering, The University of Manchester
The Mill, Sackville Street, Manchester, M13 9PL, United Kingdom
E-mail: gyorgy.szekely@manchester.ac.uk
T. Fodi, Dr. J. Kupai, Prof. P. Huszthy
Department of Organic Chemistry and Technology
Budapest University of Technology and Economics
Szent Gellert ter 4, Budapest, 1117, Hungary
T. Fodi, Dr. G.T. Balogh
Compound Profiling Laboratory, Gedeon Richter Plc.
PO Box 27, Budapest, 1475, Hungary
†
These authors equally contributed.
Figure 1. Completive relationship between flow synthesis and nanofiltration.
Supporting information for this article is given via a link at the end of
the document.
This article is protected by copyright. All rights reserved.

ChemSusChem
10.1002/cssc.201701120
FULL PAPER
Table 1. Prior research on continuous hybrid processes comprising of flow reactors and nanofiltration. PEEK: polyether ether ketone; PFR: plug flow reactor;
CSTR: continuous stirred-tank reactor; PBR: packed bed reactor; IPA: isopropyl alcohol; DMF: N,N-dimethylformamide.
Role of
membrane
#
Publication
Membrane
Reaction
Reactor
Solvent
Catalyst
Catalyst retention
&
Solvent exchange
Livingston,[2a] 2016
PEEK &
Duramem 150
Homogeneous
Pd-complex
1
Heck coupling
PFR-m-CSTR
DMF  EtOH
®
Ormerod,[9f] 2016
Inopor® TiO
PEEK
Homogeneous
Pd-complex
2
3
4
Catalyst removal
Catalyst recycle
Catalyst retention
2
Suzuki coupling
Coil reactor
PBR
EtOH, IPA
Asymmetric
hydrogenation
Toluene,
tert-butanol
Homogeneous
Ru-complex
Jensen,[14] 2015
Livingston,[15] 2013
Homogeneous
Pd-complex
PEEK
Heck coupling
PFR-m-CSTR
DMF
Heterogeneous
trialkyl amine
base
Solvent & reagent
recycle
Duramem® 150
5
Our work
Michael addition
PBR
Acetone
Michael addition of nitromethane to trans-chalcone was
selected as a model reaction (Scheme 1) since polymer
tested in numerous solvents at varying temperatures (Table 2).
Ethyl acetate and acetone resulted in 100% conversion at 50 C.
supported organocatalysts attracted a great deal of attention in
continuous-flow Michael addition.[16] Furthermore, OSN was
Table 2. Parameter screening for reaction optimisation in batch mode.
used for homogeneous catalyst recycling in
addition. Hodge et al. reported the first application of a weak
anion-exchange resin AmberlystTM A21 as a polymer supported
catalyst to promote Michael addition in continuous-flow
conditions,[1c] which triggered the catalyst screening for the
present work. The most efficient catalysts were tested in a
a
Michael
Catalyst
Solvent
Temperature (°C)
Conversion (%)
[
17]
MeOH
EtOH
30/40/50/60/70
30/40/50/60/70
30/40/50/60/70
30/40/50/60/70
30/40/50
8/17/32/51/56
23/34/41/53/61
11/18/20/21/27
42/67/81/92/96
77/89/100
3 2
CH NO
IPA
continuous-flow packed bed reactor. Subsequently
a
EtOAc
nanofiltration membrane unit was connected to the reactor in
order to allow in situ solvent and reagent recycle, and
subsequently improve the sustainability of the synthesis.
Acetone
MeCN
MeOH
EtOH
30/40/50
78/91/100
30/40/50/60
30/40/50/60/70
30/40/50/60/70
30/40/50/60/70
30/40/50/60
30/40/50/70
30/40/50
61/76/91/100
6/15/29/52/61
13/19/37/58/73
7/15/23/32/44
55/78/90/100
71/84/97/100
76/87/100
3 2
CH NO
IPA
EtOAc
Acetone
MeCN
30/40/50/60
74/83/96/100
Scheme 1. Michael addition of nitromethane reagent (5 equivalents, 60 mM)
to chalcone substrate (1 equivalent, 12 mM) catalysed by polymer supported
trialkyl amine base, yielding 4-nitro-1,3-diphenylbutan-1-one product. The
Rapid decline in the catalytic activity of AmberlystTM A21 was
observed once translated to flow reactor conditions. Although
–1
molecular weights in g·mol unit are given for each compound in parentheses.
1
00% conversion was initially achieved in 9.3 min residence
−1
−1
Results and Discussion
time which represents 42 g∙L ∙h space-time yield for the
reactor, it dropped below 50% within 2 days of operation.
Figure 2 and Figure 3 show the infrared spectra for catalyst
inactivation and the underlying mechanism, respectively. The
C-N stretching vibrations of aliphatic amine end-groups of trialkyl
amine catalysts were observed as medium bands at 1203 cm-1
(Figure 2, Panels A & C). After catalyst inactivation these bands
disappear and the N-O stretching vibrations of aliphatic nitro
group occur at 1576 cm-1 (asymmetrical) and 1371 cm-1
(symmetrical) as intense bands (Figure 2, Panels B & D). The
Michael addition is promoted by the basic character of the
Optimisation of the Flow Reactor
The stable, long-term operation of the continuous hybrid process
requires an efficient catalyst/solvent system for the chosen
Michael addition. These atom-efficient reactions are usually
promoted by acids, bases, metal salts or organocatalysts. With
application of solid base catalysts, formation of undesirable side
products resulting from polymerization, bis-addition and self-
condensation could be excluded and salt formation consequent
to the neutralization of soluble bases with acids could be
avoided. AmberlystTM A21, a polymer supported weak anion
N,N-dimethyl N-benzylamino group (pK
a
= 8.8) of the catalyst
= 10.2). The basic
and the acidic feature of the reagent (pK
a
exchange resin featuring
reported to be a cheap, commercial catalyst for Michael
additions.[1c] Consequently, the performance of this catalyst was
a
dialkylbenzylamine base was
tertiary amino group deprotonates nitromethane via an
equilibrium reaction which results in the formation of a nitronate
intermediate. This ambident nucleophile attacks the soft
This article is protected by copyright. All rights reserved.

ChemSusChem
10.1002/cssc.201701120
FULL PAPER
electrophilic centre of the substrate. Under mild conditions (low
temperature and/or short residence time, low amount of
nitromethane) this nucleophilic addition is kinetically favourable
due to the HSAB theory. However, after longer residence time,
the excess of nitromethane can promote a side-reaction. The
nitronate group can attack the electrophilic α-position of the resin,
which results in the formation of a protonated dimethylamino
group. This good leaving group can undergo a nucleophilic
substitution on the resin (Figure 3). In case of the AmberlystTM
A21, the benzyl group favours this reaction, because it uses the
90% and 100% conversion was obtained at 50 C with the use
of 3, 4 and 5 equivalents of reagent, respectively. Consequently,
50 C and 5 equivalents of reagent were selected for the
development of the continuous hybrid process.

system of the benzene ring for conjugation with the p orbital
stabilizing the transition state. On the other hand, the
AmberliteTM IRA67 lacks this stabilizing conjugation in the
transition state which hinders this side-reaction. The underlying
theory supports the fact that inactivation of the latter resin
occurred at 60 C, which is 10 C higher than the threshold
inactivation temperature for the former resin.
TM
Figure 3. Mechanism of nitromethane-induced degradation of Amberlyst
TM
A21 (A) and Amberlite IRA67 (B) after the Michael addition at 50 C and
60 C, respectively. See Figure 2 for the corresponding infrared spectra.
Since the reagent is applied in high excess and the substrate
is the limiting species, the reaction can be assumed to follow
pseudo-first order kinetics.[6b] Consequently, the total reaction
rate is described in Equation 1, and the application of t = 0, C =
ꢁ
C
, C = Csubsꢀraꢀe boundary conditions gives Equation 2.
subsꢀraꢀe
푑퐶ꢃꢄꢅꢃꢆꢇꢈꢆꢉ
′
푟 = −푟푠푢푏푠푡ꢂ푎푡푒 = −
= 푘 ꢊ
Eq. 1
Eq. 2
푠푢푏푠푡ꢂ푎푡푒
푑푡
′
ꢁ
ln ꢊ푠푢푏푠푡ꢂ푎푡푒 = −푘 ꢋ + ln ꢊ
푠푢푏푠푡ꢂ푎푡푒
where r is the total reaction rate, k’ is the pseudo–first order
–
1
reaction rate coefficient (s ) and C is the concentration of
–
1
substrate (mol·L ). The experimental results in Figure 5 show
excellent correlation with the first-order kinetic model validating
the assumed linear relationship between ln Csubsꢀraꢀe and t. Refer
to the Supporting Information for the conversions at different
flow rates. The reaction rate coefficient was determined at
different temperatures in order to estimate the apparent
a
activation energy (E ) and pre-exponential factor (A) for the
reaction (Figure 5B) using the Arrhenius equation:
Eꢏ
ꢎ
kꢌ = Aꢍ RT
Eq. 3
–1
–1
where R is the ideal gas constant of 8.314 J∙K ∙mol and T is
Figure 2. Nitromethane-induced degradation of the trialkyl amine base
the temperature in K. The apparent activation energy was found
TM
catalysts: infrared spectra of Amberlyst A21 before (A) and after (B) the
–1
–1
to be 65 kJ·mol , which falls within the 19-76 kJ·mol range
reported for the same Michael addition catalysed by cinchona-
urea based organocatalysts.[ The obtained kinetic parameters
can be used for process modelling towards automation.
Michael addition at 50 C, and AmberliteTM IRA67 before (D) and after (E) the
Michael addition at 60 C. For B) and D) the experiments were carried out in a
18]
–1
flow reactor at 1 mL·min flow rate in acetone for 48 h.
The AmberliteTM IRA67 catalyst was further characterised in
the flow reactor regarding the effect of temperature and reagent
excess on the conversion (Figure 4). After 8 h of continuous
operation stable 80% and 88% reaction conversion was
achieved during 7 days of runtime at 30 C and 40 C,
respectively. Full conversion was reached at 50 C to 70 C
within 4 h of operation. However, catalyst inactivation was
observed within 2-4 days at 60-70 C. The attainable conversion
also heavily depends on the amount of reagent employed: 70%,
Membrane screening for the optimisation of the separation
The role of the nanofiltration unit coupled to the flow reactor is
two-fold: i) recycle the solvent and the reagent, and
simultaneously ii) concentrate the product in continuous
operation. Duramem 150, GMT-oNF-2, SolSep NF030705, and
polybenzimidazole
(PBI)
solvent-resistant
nanofiltration
membrane at 10-40 bar pressure were screened (Figure 6).
This article is protected by copyright. All rights reserved.

ChemSusChem
10.1002/cssc.201701120
FULL PAPER
Figure 5. Reaction kinetics for catalysis with AmberliteTM IRA67: A) the first
order kinetic plot at various temperatures, and B) the Arrhenius plot and
Figure 4. Optimisation of reaction conditions for catalysis with AmberliteTM
IRA67: A) the temperature-dependency, and B) the effect of reagent excess
on the reaction conversion at 50 C. All experiments were performed at
calculated apparent activation energy (E
the Michael addition.
a
) and pre-exponential factor (A) for
–1
1
mL·min flow rate in acetone.
Continuous-flow reactor/nanofiltration hybrid process
The hybrid process comprised of flow reactor and
Solute rejections (Equation 4) were obtained taking into account
the retentate and permeate concentrations. Permeance
a
a
(
Equation 5) showing the volume of liquid that permeates the
membrane (Vpermeate) per membrane area (A ), per unit of time
t), and per applied pressure (p) was also measured. High
nanofiltration unit which were filled with acetone at the start of
the continuous operation (Figure 7). The feed solution
containing 9.3 mM substrate and 5 equivalents of reagent in
acetone was allowed to pass through the packed bed flow
reactor having 8.2 grams of catalyst. The flow rate was set to
1 mL∙min allowing 100% substrate conversion in 6.25 min
residence time. Consequently, the crude product stream
comprised of 1 equivalent of product and 4 equivalents of
reagent. This stream was fed into the nanofiltration unit having
0.021 m2 membrane area and 35 mL volume. A recirculation
pump set at 1 L∙min−1 ensured homogeneous solute
concentration in the membrane loop. The membrane split the
crude mixture into a concentrated, product-rich retentate stream
m
(
product rejection prevents its undesired recycle into the reactor
and subsequent accumulation which in turn can lead to
productivity decrease. On the contrary, low reagent rejection is
needed to i) minimise product contamination in the retentate
stream leading to higher product purity, and ii) minimising waste
generation by maximising the reuse of the reagent. Duramem
−
1
1
50 membrane operating at 40 bar pressure was selected for
the continuous process because these conditions exhibited the
highest rejection for the product (100%), while still maintaining
low rejection for the reagent (12.2%).
(
10% of feed flow rate), and a reagent-rich permeate stream
퐶푝ꢉꢇ푚ꢉꢈꢆꢉ
푆표푙ꢐꢋꢑ 푟ꢑ푗ꢑ푐ꢋ푖표푛 = 100 × (1 − 퐶ꢇꢉꢆꢉꢒꢆꢈꢆꢉ) = (%)
Eq. 4
Eq. 5
(90% of feed flow rate). During the start-up of the continuous
system the permeate stream was discarded (Figure 7A). Once
steady-state permeate concentration was reached, the recycle
푉푝ꢉꢇ푚ꢉꢈꢆꢉ
푃ꢑ푟ꢓꢑꢔ푛푐ꢑ = 퐴푚 × 푡 × ꢕ = (퐿 ∙ ꢓꢎ2 ∙ ℎꢎꢖ ∙ ꢗꢔ푟ꢎꢖ
)
This article is protected by copyright. All rights reserved.

ChemSusChem
10.1002/cssc.201701120
FULL PAPER
of the permeate stream through a dynamic mixing chamber was
commenced (Figure 7B). In order to compensate the changes
was discarded (Figure 7A) and afterwards it was recycled to the
flow reactor through a dynamic mixing chamber (Figure 7B). The
product leaves the flow reactor and enters the nanofiltration unit
(
i.e. concentration and flow rate) induced by the permeate
−
1
recycle, both the flow rate and concentrations of the feed were
adjusted accordingly, and subsequently the steady-state
operation was maintained. The new feed flow rate was set to
at 2.5 g∙L
concentration, and then gets concentrated to
−
1
25 g∙L , hence the considerably longer equilibrium time of 25 h.
The hybrid process showed stable performance over 6 weeks of
continuous operation achieving catalyst turnover number (TON)
−
1
0
.091 mL∙min , while the feed concentrations were changed to
−
1
102 mM substrate and 1.4 equivalent of reagent.
of 32, catalyst turnover frequency (TOF) of 0.032 h ,
−
1
−1
productivity of 3.75 g∙L ∙h and 91.5% product purity. It should
be noted that membrane performance could change and fouling
could occur over time, which can be mitigated by operating well
below the solubility limit of the solutes.[ During the 6 weeks of
operation neither catalyst deactivation nor membrane fouling
were observed.
19]
Figure 6. Rejection profiles for Duramem 150 (A), GMT-oNF-2 (B), SolSep
NF030705 (C), and PBI (D) membranes at 10-40 bar pressure. The feed
–
1
solution comprised of the product, the substrate and the reagent at 1 g·L
Figure 7. Schematic process configuration and conditions for the start-up (A)
and continuous steady-state (B) operation of the hybrid process. The jacketed
flow reactor operated at 50 C allowing 100% conversion of the substrate. The
nanofiltration unit was continuously operated at 40 bar splitting the crude
mixture into a concentrated, product-rich retentate stream, and a reagent-rich
permeate stream. Initially the permeate stream was discarded (A) and once
steady-state was reached, the recycle of the permeate stream was
commenced (B), and the feed concentration and flow rate were adjusted.
concentration in acetone. The permeance values expressed in L·m–2·h ·bar–1
are given in the boxes. Refer to the Supporting Information for the membrane
screening process scheme.
–1
The concentration profiles of the solutes in the retentate and
permeate streams are shown in Figure 8. The experimental
data shows good correlation with the predicted curves. Refer to
the Supporting Information for the detailed mathematical
framework describing the process modelling. The equilibrium
time is defined as the time needed to reach 98% of the steady-
state concentration of a solute, and it was found to be 2.5 h for
the reagent in the permeate stream. This is the threshold to start
recycling the solvent and the reagent. Until 2.5 h the permeate
The effects of various system parameters
— namely
conversion, product and reagent rejections, retentate/permeate
flow rate ratio, and membrane loop volume/area — on solvent
consumption, product purity, productivity and equilibrium time
This article is protected by copyright. All rights reserved.

ChemSusChem
10.1002/cssc.201701120
FULL PAPER
were investigated (Figure 9). The membrane loop volume/area
does not affect the solvent consumption and the product purity
but its decrease results in the improvement of both productivity
and equilibrium time. Although minimisation of this volume is
practically difficult using flat sheet membranes (laboratory scale),
the membrane modules used in industrial applications have
−
2 [20]
relatively low values of 0.8-1.2 L∙m . The increase in product
rejection is beneficial for all process metrics. On the contrary,
the decrease in reagent rejection enhances product purity and
shortens equilibrium time, but neither solvent consumption nor
productivity is affected. Similar to product rejection, the higher
the conversion, the better the process metrics become except
for the equilibrium time which remains constant. Decreasing
retentate/permeate flow rate ratio results in favourable increase
in purity and decrease in solvent consumption while the
productivity and equilibrium time are not affected. Figure 9F
shows that high-purity product can be obtained by minimising
the excess of the reagent and the rejection of the reagent.
Furthermore, the lower the excess of reagent, the less influence
the reagent rejection has on the product purity. Refer to the
Supporting Information for the detailed sensitivity analysis.
Figure 8. Concentration profile in the nanofiltration unit. The curves show the
simulated values and the symbols show the experimental data. The system
was operated in the start-up process configuration (Figure 7A) until steady-
state concentration was reached (2.5 h) for the reagent in the permeate
stream. At that point the recycle of the permeate stream was commenced.
Sustainability assessment of the hybrid process
The carbon footprint contribution of the thermostat for the
jacket of the flow reactor is 2257 kg·kg–1 accounting for 78% and
97% of the total carbon footprint in the absence and presence of
the nanofiltration unit, respectively (Figure 10B). In line with the
6th principle of green chemistry,[23] such high contribution calls
for energy integration in a manufacturing plant. The obtained
values for the E-factor demonstrate that the in situ solvent and
reagent recycle can turn the process from the non-green to
green category based on the industrial classification by
Sheldon.[24] The carbon footprint of the continuous hybrid
process rapidly decreases over time and reaches a constant
value of 2323 kg·kg–1 within less than 2 days of operation, and it
takes 10 h to become more environmentally benign than the
benchmark process (Figure 10C). The carbon footprint reduction
depends on the solvent used in the process (Figure 10D).
Consequently, the carbon footprint reduction potential of
nanofiltration was assessed considering the top solvent wastes[
generated in the pharmaceutical sector. Depending on the
carbon footprint potential of each solvent,[ the implementation
of a nanofiltration unit could result in 17 to 31% overall reduction.
The proposed process configuration enables further process
window expansion. First, the continuous reaction-separation
methodology can be used for the sought-after continuous flow
multi-step organic synthesis.[2,26] Second, the pressurized
system allows the reaction to be realised at elevated
temperatures up to supercritical conditions which is of recent
interest.[
The effect of the nanofiltration-assisted in situ solvent and
reagent recycle on the sustainability of the continuous-flow
synthesis was assessed. Swapping from batch to continuous
processing, in particlular flow synthesis can considerably reduce
the environmental burden of chemical manufacturing.[21a-d]
Nonetheless, the use of organic solvents was identified as one
of the largest contributor to waste generation in flow
synthesis.[
21a,22a-b]
The E-factor and carbon footprint, defined in
Equation 6 and Equation 7, have pivotal roles in driving resource
efficiency and waste minimisation in the chemical industries.[10]
ꢘ푔 푤푎푠푡푒 푔푒ꢙ푒ꢂ푎푡푒푑
퐸 − 푓ꢔ푐ꢋ표푟 = ꢘ푔 ꢚ푠ꢛꢜ푎푡푒푑 ꢕꢂꢛ푑푢ꢝ푡
Eq. 6
Eq. 7
푒푞푢ꢚ푣푎ꢜ푒ꢙ푡 ꢘ푔 ꢛꢟ 퐶푂ꢠ
ꢊꢔ푟ꢗ표푛 푓표표ꢋꢞ푟푖푛ꢋ = ꢘ푔 ꢚ푠ꢛꢜ푎푡푒푑 ꢕꢂꢛ푑푢ꢝ푡
5]
The E-factor ignores any recycled streams and reused
reactants or reagents as those are not considered waste. On the
other hand, it neglects the energy required for the process.
Consequently, carbon footprint, taking into account waste
generation and energy consumption, was also derived to obtain
a holistic view of the environmental burden of the hybrid process
25]
(
Figure 10). The coupling of the nanofiltration unit to the flow
reactor allows the recycle of 90.9% of solvent and 90% of
reagent, corresponding to 287 kg solvent and 0.814 kg reagent
per kg of product, respectively (Figure 10A). The solvent
consumption is the main contributor to the E-factor, and the
nanofiltration unit reduces waste generation and raw material
27]
usage by 90.7%. The application of the nanofiltration unit is Conclusions
sustainable,[ and in this particular case it poses only 0.4% of
7]
the carbon footprint taking into account both the energy
requirements for the pumps and the disposal cost of the
membranes. Refer to the Supporting Information for the
break-down of the waste generation and energy consumption.
Increasing demand for more sustainable manufacturing has led
to an increasing interest in the area of continuous processing
within the pharmaceutical sector. Flow synthesis allows
continuous production of fine chemicals in a safe, compact and
This article is protected by copyright. All rights reserved.

ChemSusChem
10.1002/cssc.201701120
FULL PAPER
Figure 9. Sensitivity analysis. The variation of six system parameter settings, namely membrane loop volume/area (A), product rejection (B), reagent rejection (C),
conversion (D), retentate/permeate flow rate ratio (E), and molar equivalents of the excess of reagent (F), and their effects on four process metrics, namely
product purity (%), solvent consumption (kg·kg–1), productivity (g∙L ∙h ) and equilibrium time (h). The arrow heads ( ) indicate the actual parameter value used
for the continuous hybrid process presented in Figure 7B and Figure 8. For the sensitivity analysis of each parameter, the remaining parameters were kept
constant using the same value as for the continuous hybrid process. The excess of reagent (F) influences only the purity process metric.
−1
−1
Figure 10. The effect of in situ solvent and reagent recycle on the E–factor (A) and the carbon footprint (B-C) of the continuous flow-synthesis. The main
contributors of the environmental burden are the solvent waste and the energy to operate the thermostat for the jacket of the flow reactor. The rest of the
contributors have cumulative contributions lower than 0.5%. Refer to the Supporting Information for the actual values for all contributors. Based on the acetone
solvent recovery flow rate and the density of the different solvents, the potential of nanofiltration for the carbon footprint reduction can be estimated[25] for the top
solvent wastes[5] generated in the pharmaceutical sector (D).
controlled environment but subsequent downstream processing
remains challenging. Concentration and purification of the
product is usually done in batch operation, and these separation
processes can account for as much as 80% of the total
manufacturing costs. In this work, a continuous hybrid process
comprising of a flow reactor and a subsequent nanofiltration unit
for in situ solvent and reagent recycle has been developed. The
nanofiltration unit is straightforward to implement and control in a
continuous operation. A predictive in silico tool — where the
main input parameters are the membrane permenace and solute
rejections — was developed to understand the boundaries of
process conditions. The hybrid process was continuously
operated over 6 weeks in situ recovering about 90% of the
solvent and the excess of the reagent. Consequently, the
This article is protected by copyright. All rights reserved.

ChemSusChem
10.1002/cssc.201701120
FULL PAPER
E-factor and the carbon footprint were reduced by 91% and 19%, Continuous-flow reactions
respectively. The nanofiltration unit concentrated the product
11 times and increased the purity from 52.4% to 91.5%. In line
A Supelco® stainless steel column of 10 mm × 250 mm
TM
coupling of nanofiltration to continuous-flow reactors can
significantly improve the sustainability of flow synthesis through
in situ solvent and reagent recycle, as well as concentration and
purification of the product. The proposed process configuration
enables further process window expansion for continuous flow
multi-step organic synthesis.
dimension was loaded with a solid base Amberlyst
A21
(13.1 g) free base form or AmberliteTM IRA67 (8.1 g) free base
form, respectively. Prior to use the filled column was washed
with the reaction solvent (50 mL) with a flow rate of 1.0 mL∙min−1
to remove air and unwanted materials from the surface of the
solid base. A feed solution containing trans-chalcone (12 mM)
and nitromethane (1-5 equiv., 12-60 mM) was pumped through
the column at 1-10 mL∙min−1 flow rate corresponding to
6
.25-0.625 min residence time at 30-70 C temperature. A
Experimental Section
Lauda Alpha RA12 thermostat was used to control the
temperature with 0.1 C accuracy in the jacketed flow reactor.
The energy consumption of the equipment was measured with a
Fluke 1736 power logger with resolution 10 mA and accuracy
Materials and general methods
Chemicals (reagent grade) and solvents (analytical grade) were
purchased from Sigma-Aldrich (UK) and were used without
further purification. In the present study three commercially

0.1%.
®
Membrane screening
available OSN membranes were used: SolSep NF010206,
GMT-oNF-2® and DuraMem150® (DM150) from SolSep BV,
Borsig Membrane Technology GmbH and Evonik-MET,
The polybenzimidazole (PBI) membrane was prepared
according to our previously reported protocol.[8f] The feed
solution for the membrane screening comprised of a 1 g∙L−1
mixture of substrate, reagent and product in acetone. The
pressure range for the screening was 10-40 bar and the tests
were carried out in duplicates in a cross-flow nanofiltration rig
described elsewhere.[8f] The feed solution was recirculated for
24 h followed by collection of samples from the permeate and
the retentate streams. The solute rejection values for the
substrate and product were determined by LCMS, while the
GCMS was used for the reagent. Refer to the Supporting
Information for the membrane screening process scheme.
–
1
respectively. 26 wt% polybenzimidazole (MW = 27,000 g.mol )
containing 1.5 wt% lithium chloride (stabiliser) dissolved in N,N-
dimethylacetamide (DMAc) solution was purchased from PBI
Performance Products Inc (USA). Non-woven polypropylene
fabric Novatexx 2471 was sourced from Freudenberg Filtration
Technologies (Germany).
1H and 13C NMR spectra were recorded on a Bruker AV-400
spectrometer. HPLC measurements were carried out on a VWR
HPLC system equipped with a VWR 5160 pump, a VWR 5260
auto sampler and thermostat, a VWR 5430 DAD and an ACE 5
C18 150 × 4.6 mm, 5 µm column. The pump flow rate was set at
1
mL·min−1 and the column temperature was 25 °C. The mobile
phase was a 1:1 mixture of acetonitrile and water with 0.1%
trifluoroacetic acid. Nitromethane was quantified using a Varian Acknowledgements
CP-3800 GC fitted with a Varian Saturn 2200 MS and a Varian
VF-5ms capillary column (0.25 µm, 30 m × 0.25 mm). The GC
oven temperature was 27 °C for 4 min, then increased at
This work was supported by the Hungarian Scientific Research
Fund/National Research, Development and Innovation Office
[OTKA 112289, PD108462]. The authors are grateful to Levente
Cseri and Hai Anh Le Phuong, both from The University of
Manchester, for their technical support and useful discussions.
−
1
15 °C∙min to 200 °C. LCMS measurements were carried out on
an Agilent 1100 HPLC equipped with gradient pump,
autosampler and PDA detector. A triple quadrupole mass
spectrometer with positive electrospray ionisation source was
employed as the MS detector. Infrared spectra were recorded
on a Bruker Alpha-T FT-IR spectrometer.
Conflict of Interest
Batch reactions
The authors declare no conflict of interest.
In a round-bottomed flask equipped with a reflux condenser, a
solution of trans-chalcone (50 mg, 4.8 mM) and nitromethane
Keywords: catalytic membrane reactor • continuous process •
membranes • Michael addition • organic solvent nanofiltration
(
2
5 equiv., 24 mM) was stirred in either methanol, ethanol,
-propanol, ethyl acetate, acetone or acetonitrile solvents (50
[1]
a) B. Gutmann, D. Cantillo, C. O. Kappe, Angew. Chemie - Int. Ed.
2015, 54, 6688–6728; b) I. Mándity, B. Olasz, S. B. Ötvös, F. Fülöp,
ChemSusChem 2014, 7, 3172–3176; c) F. Bonfils, I. Cazaux, P. Hodge,
C. Caze, Org. Biomol. Chem. 2006, 4, 493–497.
mL). In case of nitromethane as both solvent and reagent, trans-
chalcone (50 mg, 4.8 mM) was directly dissolved in
TM
nitromethane (50 mL). To these solutions Amberlyst A21 or
TM
[2]
a) L. Peeva, J. da Silva Burgal, Z. Heckenast, F. Brazy, F. Cazenave, A.
G. Livingston, Angew. Chemie – Int. Ed. 2016, 55, 13774–13777; b) R.
Örkényi, J. Éles, F. Faigl, P. Vincze, A. Prechl, Z. Szakács, J. Kóti, I.
Greiner, Angew. Chemie - Int. Ed. 2017, 129, 1–5.
Amberlite IRA67 (50 mg of each) catalyst was added. After
stirring the reaction mixture at 30-70 °C temperature for 24 h,
conversions were determined by HPLC.
This article is protected by copyright. All rights reserved.

ChemSusChem
10.1002/cssc.201701120
FULL PAPER
[
3]
a) J. F. Kim, G. Szekely, I. B. Valtcheva, A. G. Livingston, Green Chem.
[21] a) L. Vaccaro, C. Petrucci, V. Kozell, E. Ballerini, Flow Tools to Define
Waste/Time/Energy-Minimized Protocols. In Sustainable Flow
Chemistry: Methods and Applications, (Ed.: L. Vaccaro), Wiley-VCH,
Weinheim, 2017; b) C. Petrucci, G. Strappaveccia, F. Giacalone, M.
Gruttadauria, F. Pizzo, L. Vaccaro, ACS Sustainable Chem. Eng. 2014,
2, 2813–2819; c) C. Jiménez-González, P. Poechlauer, Q. B.
Broxterman, B. S. Yang, D. am Ende, J. Baird, C. Bertsch, R. E.
Hannah, P. Dell’Orco, H. Noorman, S. Yee, R. Reintjens, A. Wells, V.
Massonneau, J. Manley, Org. Process Res. Dev. 2011, 15, 900–911; d)
S. L. Lee, T. F. O’Connor, X. Yang, C. N. Cruz, S. Chatterjee, R. D.
Madurawe, C. M. Moore, X. Y. Lawrence, J. Woodcock, J. Pharm.
Innov. 2015, 10, 191–199.
2014, 16, 133–145; b) D. S. Sholl, R. P. Lively, Nature 2016, 532, 435–
437.
[
4]
5]
C. Jiménez-González, A. D. Curzons, D. J. C. Constable, Clean
Technol. Envir. 2005, 7, 42–50.
[
C. S. Slater, M. J. Savelski, W. A. CarSole, D. J. C. Constable in Green
Chemistry in the Pharmaceutical Industry (Eds.: P. J. Dunn, , A. S.
Wells, M. T. Williams) Wiley-VCH, Weinheim, 2010, pp. 49–82.
a) I. Denꢀić, D. Ott, D. Kralisch, T. Noël, J. Meuldijk, M. de Croon, V.
Hessel, Y. Laribi, P. Perrichon, Org. Process Res. Dev. 2014, 18,
[6]
1326–1338; b) P. Yaseneva, D. Plaza, X. Fan, K. Loponov, A. Lapkin,
Catal. Today 2015, 239, 90–96.
[
7]
8]
G. Szekely, M. F. Jimenez-Solomon, P. Marchetti, J. F. Kim, A. G.
Livingston, Green Chem. 2014, 16, 4440–4473.
[22] a) L. Bianchi, E. Ballerini, M. Curini, D. Lanari, A. Marrocchi, C. Oldani,
L. Vaccaro, ACS Sustainable Chem. Eng. 2015, 3, 1873–1880; b) P.
Yaseneva, P. Hodgson, J. Zakrzewski, S. Falß, R. E. Meadows, A. A.
Lapkin, React. Chem. Eng. 2016, 1, 229–238.
[
a) L. S. White, A. R. Nitsch, J. Membr. Sci. 2000, 179, 267–274; b) I.
Sereewatthanawut, F. W. Lim, Y. S. Bhole, D. Ormerod, A. Horvath, A.
T. Boam, A. G. Livingston, Org. Process Res. Dev. 2010, 14, 600–611;
c) E. M. Rundquist, C. J. Pink, A. G. Livingston, Green Chem. 2012, 14,
[23] Anastas, P. T.; Warner, J. C. Green Chemistry: Theory and Practice,
Oxford University Press: New York, 1998, pp. 30.
2
197–2205; d) J. F. Kim, G. Szekely, M. Schaepertoens, I. B. Valtcheva,
M. F. Jimenez-Solomon, A. G. Livingston, ACS Sustainable Chem. Eng.
014, 2, 2371–2379; e) D. Ormerod, B. Noten, M. Dorbec, L.
Andersson, A. Buekenhoudt, L. Goetelen, Org. Process Res. Dev. 2015,
9, 841–848; f) M. Schaepertoens, C. Didaskalou, J. F. Kim, A. G.
[24] R. A. Sheldon, Green Chem. 2007, 9, 1273–1283.
[25] Ecoinvent v2.0 database. Swiss Centre for Life Cycle Inventories,
Dübendorf, Switzerland can be found under: http://www.ecoinvent.org.
[26] a) D. Webb, T. F. Jamison, Chem. Sci. 2010, 1, 675–680; b) T.
Tsubogo, H. Oyamada, S. Kobayashi, Nature 2015, 520, 329–332.
[27] F. Mandrelli, A. Buco, L. Piccioni, F. Renner, B. Guelat, B. Martin, B.
Schenkel, F. Venturoni, Green Chem. 2017, 19, 1425–1430.
2
1
Livingston, G. Szekely, J. Membr. Sci. 2016, 514, 646–658; g) C.
Didaskalou, S. Buyuktiryaki, R. Kecili, C. P. Fonte, G. Szekely, Green
Chem. 2017, 19, 3166–3125.
[9]
a) M. Janssen, J. Wilting, C. Müller, D. Vogt, Angew. Chemie - Int. Ed.
2010, 49, 7738–7741; b) A. Kajetanowicz, J. Czaban, G. R. Krishnan,
M. Malińska, K. Woźniak, H. Siddique, L. G. Peeva, A. G. Livingston, K.
Grela, ChemSusChem 2013, 6, 182–192; c) D. Ormerod, B. Bongers,
W. Porto-Carrero, S. Giegas, G. Vijt, N. Lefevre, D. Lauwers, W.
Brusten, A. Buekenhoudt, RSC Adv. 2013, 3, 21501–21510; d) S.
Ferguson, F. Ortner, J. Quon, L. Peeva, A. Livingston, B. L. Trout, A. S.
Myerson, Cryst. Growth Des. 2014, 14, 617–627; e) J. M. Dreimann, M.
Skiborowski, A. Behr, A. J. Vorholt, ChemCatChem 2016, 8, 3330–
3333; f) D. Ormerod, N. Lefevre, M. Dorbec, I. Eyskens, P. Vloemans,
K. Duyssens, V. Diez De La Torre, N. Kaval, E. Merkul, S. Sergeyev, B.
U. W. Maes, Org. Process Res. Dev. 2016, 20, 911–920; g) P.
Marchetti, L. Peeva, A. Livingston, Annu. Rev. Chem. Biomol. Eng.
2017, 8, 473–497; h) R. Abejón, A. Garea, A. Irabien, Chem. Eng.
Trans. 2015, 43, 1057–1062; i) P. Schmidt, E. L. Bednarz, P. Lutze, A.
Górak, Chem. Eng. Sci. 2014, 115, 115–126.
[
[
10] R. A. Sheldon, Green Chem. 2017, 19, 18–43.
11] a) S. G. Koenig, H. F. Sneddon, Green Chem. 2017, 19, 1418–1419; b)
T. O’Connor, S. Lee, Emerging Technology for Modernizing
Pharmaceutical Production: Continuous Manufacturing (Eds.: Y. Chen,
G.G.Z. Zhang, L. Yu, R.V. Mantri) Academic Press, Boston, 2017, pp.
1031–1046.
[
[
[
12] E. Drioli, A. I. Stankwicz, F. Macedonio, J. Membr. Sci. 2011, 380, 1–8.
13] D. S. Sholl, R. P. Lively, Nat. Mater. 2017, 16, 276–279.
14] E. J. O’Neal, C. H. Lee, J. Brathwaite, K. F. Jensen, ACS Catal. 2015, 5,
2
615–2622.
15] L. Peeva, J. da Silva Burgal, S. Vartak, A. G. Livingston, J. Catal. 2013,
06, 190–201.
16] I. M. Mándity, S. B. Ötvös, G. Szőlősi, F. Fülöp, Chem. Rec. 2016, 16,
018–1033.
17] W.E. Siew, C. Ates, A. Merschaert, A.G. Livingston, Green Chem. 2013,
5, 663–674.
[
[
[
[
[
[
3
1
1
18] B. I. Károlyi, A. Fábián, Z. Kovács, S. Varga, L. Drahos, P. Sohár, A.
Csámpai, Comput. Theor. Chem. 2012, 996, 76–81.
19] L. Peeva, J. da Silva Burgal, I. Valtcheva, A. G. Livingston, Chem. Eng.
Sci. 2014, 116, 183–194.
20] J. Johnson, Design and Construction of Commercial Spiral Wound
Modules. In Encyclopedia of Membrane Science and Technology (Eds.:
E. Hoek, V. Tarabara) Wiley-VCH, Weinheim, 2013, pp. 1–21.
This article is protected by copyright. All rights reserved.

ChemSusChem
10.1002/cssc.201701120
FULL PAPER
Entry for the Table of Contents
FULL PAPER
Flow synthesis allows continuous
production of fine chemicals in a safe,
compact & controlled environment but
subsequent downstream processing
remains challenging. A continuous
hybrid process comprising of a flow
reactor & a subsequent nanofiltration
unit for in situ solvent & reagent
recycle has been developed. Over
Tamas Fodi, Christos Didaskalou,
Jozsef Kupai, Gyorgy T. Balogh, Peter
Huszthy, Gyorgy Szekely*
Page No. – Page No.
Nanofiltration-enabled in situ solvent
and reagent recycle for sustainable
continuous-flow synthesis
6
weeks of continuous operation the
E-factor & the carbon footprint of a
Michael addition were reduced by
91% and 19%, respectively.
This article is protected by copyright. All rights reserved.