Enabling an atom-economic production of chiral amino alcohols by electrodialysis with bipolar membranes

Article abstract of DOI:10.1039/c9gc02460c

Optically pure amino alcohols are widely used as essential building blocks of catalysts and auxiliaries in asymmetric synthesis, especially in pharmaceuticals. Catalytic hydrogenation has been regarded as the most effective route to produce chiral amino alcohols, but it is still unsatisfactory, which mainly manifests in the involvement of extra acids and bases, thus generating a large amount of inorganic salt byproducts, along with a relatively low yield. In this paper, we propose a novel approach that integrates catalytic hydrogenation and electrodialysis with bipolar membranes (EDBM) to produce pure chiral amino alcohols in a completely green way. EDBM tackles the sustainability issue by splitting the resulting salt of the hydrogenation into the corresponding acid and base (amino alcohol solutions), and the acid could be reused in the next batch directly and a closed loop forms. Here, as a model system to demonstrate the feasibility, a biorenewable chiral l-alanine solid and clean H₂ gas are used as the feed stocks and finally converted into a pure l-alaninol product successfully, with no discharge of harmful substances. Furthermore, three typical mineral acids (H₃PO₄, H₂SO₄ and HCl) have been investigated systematically and achieved similar performances. The results of all tests and the cost analysis strongly suggest that this new approach is widely applicable, highly productive and sustainable. It demonstrates a true atomically economic route to produce chiral amino alcohols and entirely meets the requirements of green chemistry.

Full text of DOI:10.1039/c9gc02460c

Green Chemistry
View Article Online
View Journal
PAPER
Enabling an atom-economic production of chiral
amino alcohols by electrodialysis with bipolar
Cite this: DOI: 10.1039/c9gc02460c
membranes†
a,b
Wei-Ting Gao,a Qing Chen,a,b Ming-Gu Du,b Wei-Ming Zhang,
*
c
Chang-Yan Caoc and Wei-Guo Song
*
Optically pure amino alcohols are widely used as essential building blocks of catalysts and auxiliaries in
asymmetric synthesis, especially in pharmaceuticals. Catalytic hydrogenation has been regarded as the
most effective route to produce chiral amino alcohols, but it is still unsatisfactory, which mainly manifests
in the involvement of extra acids and bases, thus generating a large amount of inorganic salt byproducts,
along with a relatively low yield. In this paper, we propose a novel approach that integrates catalytic
hydrogenation and electrodialysis with bipolar membranes (EDBM) to produce pure chiral amino alcohols
in a completely green way. EDBM tackles the sustainability issue by splitting the resulting salt of the hydro-
genation into the corresponding acid and base (amino alcohol solutions), and the acid could be reused in
the next batch directly and a closed loop forms. Here, as a model system to demonstrate the feasibility, a
biorenewable chiral L-alanine solid and clean H2 gas are used as the feed stocks and finally converted into
a pure L-alaninol product successfully, with no discharge of harmful substances. Furthermore, three
typical mineral acids (H3PO4, H2SO4 and HCl) have been investigated systematically and achieved similar
performances. The results of all tests and the cost analysis strongly suggest that this new approach is
widely applicable, highly productive and sustainable. It demonstrates a true atomically economic route to
produce chiral amino alcohols and entirely meets the requirements of green chemistry.
Received 17th July 2019,
Accepted 7th November 2019
DOI: 10.1039/c9gc02460c
rsc.li/greenchem
well as abundant availability, and is far superior to expensive
and hazardous hydride reagents such as NaBH4 and LiAlH4.
1. Introduction
Chiral (optically active) amino alcohols can serve as essential
However, there are still some drawbacks in this hydrogenation
building blocks or auxiliaries in numerous pharmaceuticals, process. Because of the coexistence of acidic (carboxyl) and basic
surfactants, and other fine chemical products, by endowing (amino) groups, amino acids usually exist as zwitterions in a
them with the valuable quality of chirality.1 Although several wide pH range depending on their molecular structures, in
different routes can be used for the synthesis of chiral amino which the carboxyl group is deprotonated, while the amino
alcohols,1–6 the best available choice is the catalytic hydrogen- group is protonated.7 But meanwhile, hydrogenation requires
ation of chiral amino acids for the following reasons.1,7,8 On the carboxylic acid functionality to be undissociated, because
the one hand, chiral amino acids are biorenewable and readily the reduction of carboxylate anions to alkoxide is more endo-
produced on the industrial scale via multiple routes (such as thermic and less thermochemically feasible.7 Therefore, in order
protein hydrolysis, enzymatic catalysis and fermentation),9 to keep the amino acid in the protonated form, it is necessary to
which enables them to be the perfect feedstock. On the other add an extra inorganic acid to the reaction solution.7,10
hand, H2 is an ideal reductant because of its clean nature as
Taking the hydrogenation of L-alanine as an example, the
conventional reaction sequence is shown in Fig. 1a. First,
mineral acid (such as H3PO4, H2SO4 or HCl) is added into the
reaction solution to protonate the carboxyl group of amino
acid. Then, the protonated amino acid can be directly hydro-
genated at elevated temperature (100–140 °C) and H2 pressure
(5–14 MPa) over carbon-supported ruthenium catalysts. With
sufficient mineral acid, the protonated amino alcohol product
will be synthesized with an ultrahigh yield (>95%) and stereo-
chemical retention (enantiomeric excess >99%).7 In order to
isolate the amino alcohol product, excess strong alkali must be
aCollege of Chemistry & Materials Engineering, Wenzhou University,
Wenzhou 325000, P.R. China. E-mail: weiming@iccas.ac.cn; https://selsep.com
bLaboratory of Industrial Electrochemistry & Membrane Processes, Institute of New
Materials & Industrial Technology, Wenzhou University, Wenzhou 325024,
P.R. China
cLaboratory of Molecular Nanostructures and Nanotechnology, Institute of
Chemistry, Chinese Academy of Sciences, Beijing 100190, P.R. China.
E-mail: wsong@iccas.ac.cn
†Electronic supplementary information (ESI) available: Fig. S1–S10 and Tables
S1, S2. See DOI: 10.1039/c9gc02460c
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
Fig. 1 (a) The simplified reaction sequence of catalytic hydrogenation of L-alanine to obtain L-alaninol. (b) The corresponding atomic mass balance.
added into the resulting solution, while a large amount of in-
In this work, we focus on the feasibility of coupling catalytic
organic salt (such as Na3PO4) is generated as a byproduct. The hydrogenation of chiral amino acids and EDBM together to
overall atomic mass balance is presented in Fig. 1b, which realize a green synthesis of chiral amino alcohols. The separ-
clearly shows that the catalytic hydrogenation process is not ation performance and the economic aspects have been care-
atomically economic. The amount of waste is even much fully investigated. Specifically, with the formation of the proto-
higher (2.18 times in weight) than that of the product. To nated amino alcohol, we tried to convert the resulting salt
make matters worse, the waste sludge adsorbs and clads the (protonated L-alaninol dihydrophosphate) into the corres-
amino alcohol products, which makes the distillation of final ponding acid (H3PO4) and base (L-alaninol) solutions, respect-
products difficult and results in significantly lower overall ively via EDBM. Once the mineral acid is removed completely
yield (only 60–70%). For these reasons, priority should be in this way, a pure amino alcohol product would be readily
given to improving the catalytic hydrogenation process for obtained by distillation (water removal), without any byproduct
better sustainability.
salt sludge. The corresponding atomic mass balance of this
Electrodialysis with bipolar membranes (EDBM) is an emer- improved integrated process can be illustrated in Fig. 2b. It is
ging technology that integrates water and salt splitting, which a noticeably green reaction with a perfect atom economy that
converts the salt into the corresponding acid and base utilizes biorenewable amino acids and clean H2 as reactants to
in situ.11 It has been studied and applied in chemistry,12 food produce chiral amino alcohol products. No extra mineral acids
processing,13,14 and environmental protection.15,16 Shen et al. and strong bases are needed. All reactants are converted into
proposed an EDBM system for the processing of L-alaninol sul- final products and water is the only byproduct.
phate solution, and the results demonstrated that it is feasible
and economically attractive.17 However, since the feeding solu-
tion is obtained from a routine reduction of L-alanine by hazar-
dous hydride reagents (NaBH4), there is no chance for a
desired atom economy in their study. Thus, further efforts are
2. Experimental
2.1. Membranes, chemicals and synthetic feed solutions
needed to achieve completely green routes for the industry. For
the mass balance in Fig. 1b, a closed-loop mass flow can be
technically achieved by converting the byproduct salt (Na3PO4)
into the corresponding acid (H3PO4) and base (NaOH) and
reusing them as reactants. Furthermore, a better strategy is to
directly convert the amino alcohol salt (anions and protonated
amino alcohol cations) into the corresponding acid and base,
and the improved reaction sequence is shown in Fig. 2a.
Four different membranes were used in this study, including
Fumasep FBM bipolar membranes (BPMs, FuMA-Tech GmbH,
Germany), Neosepta AHA anion exchange membranes (AEMs,
ASTOM Corporation, Japan), Selemion CMD heavy cation
exchange membranes (CEMs, Asahi Glass Engineering Co.,
Ltd, Japan), and Tingrun JAM AEMs (Beijing Tingrun
Membrane Technology Co., Ltd, China). The detailed specifica-
tions of these membranes are listed in Table 1. The L-alaninol
Fig. 2 (a) The improved reaction sequence of catalytic hydrogenation of L-alanine to obtain L-alaninol by EDBM. (b) The corresponding atomic
mass balance.
Green Chem.
This journal is © The Royal Society of Chemistry 2019

View Article Online
Green Chemistry
Paper
Table 1 Main characteristics of the membranes used in the
experiments
especially as the feed solution in the EDBM process. Therefore,
the synthetic L-alaninol salt solution (L-alaninol 2.00 M +
H3PO4 2.05 M) was used in the following tests, which makes
the current work more operationally feasible. Furthermore,
some other mineral acids (such as H2SO4 and HCl) have also
been investigated here besides H3PO4. The initial pH of the
reactant mixture was 1.53 and 0.88 for H2SO4 (2.00 M L-alanine
+ 1.03 M H2SO4) and HCl (2.00 M L-alanine + 2.05 M HCl),
respectively. And the pH of both increased to about 3.80 after
the hydrogenation.
Area resistance
Blasting
strength (MPa)
Thickness
(mm)
Membrane
(Ω cm2)
pH
AHA
FBM
JAM
4.1
≥0.90
0.22
0–14
1–14
0.20–0.25
0.16–0.23
0.40
4–8
17
>0.25
0.90
CMD
(Shanghai Bangcheng Chemistry Co., Ltd, China) used in this
study was a colourless liquid, with a purity of more than 99%.
Other chemicals involved were of analytical grade, including
L-alanine, H3PO4, H2SO4 and HCl (Sinopharm Chemical
Reagent Co., Ltd, China). Deionized water (>5 MΩ cm) was
used for all tests. The pKa values of all corresponding species
listed in Table 2 were adopted from the handbook,18 except
for L-alaninol, whose pKa values were calculated by using
ChemBio3D software.
The protonated L-alaninol salt solution (the base stream of
EDBM) was obtained by the direct hydrogenation of a mixture
of L-alanine and mineral acid over the home-made Ru–C cata-
lysts (Ru nanoparticles loaded on graphene aerogels), which is
a successive study of our wettability-induced hydrogenation on
superaerophilic Pd–C catalysts.19 The initial pH of the reactant
mixture (2.00 M L-alanine + 2.05 M H3PO4) was 2.07, and the
corresponding conductivity was 29.42 mS cm−1. The hydrogen-
ation reaction was conducted at 140 °C for 2.0 hours, and
during that time the H2 pressure was maintained at 8.0 MPa in
the autoclave. When the reactants were cooled to room temp-
erature and the pressure was discharged, the pH of the result-
ing solution increased to 3.82 and the conductivity increased
to 30.73 mS cm−1. The HPLC (High Performance Liquid
Chromatography) analysis indicates that nearly all L-alanine
had converted into L-alaninol (>98%), and the chirality of the
product was nearly 100% preserved. The similar catalytic per-
formance of hydrogenation, with the ultrahigh conversion and
chiral retention, can also be readily obtained through other
Ru–C catalytic systems.1,8,10,20–25
2.2. EDBM system and operation protocols
The overview photograph of the EDBM system used in this
study is presented in Fig. 3a. The core module is a homemade
EDBM stack, equipped with spacers of 130 mm (length) ×
86 mm (width) × 0.5 mm (thickness). The spacer consists of a
polyethylene (PE) sheet and a polypropylene (PP) turbulence
accelerating mesh net, with a tortuous flow geometry of
22 mm (width) × 274 mm (length) × 0.5 mm (thickness), and
the effective area is 60 cm2. The inner configuration of the
EDBM stack is illustrated in Fig. 3b, and the photograph can
be found in Fig. S1 in the ESI.† There are 12 cell pairs in the
stack, which are all formed by BPMs, AEMs and the adjacent
turbulence accelerating spacers. The total effective membrane
area is 720 cm2 for both BPMs and AEMs. Selemion CMD was
selected here as the anode membrane, because it is durable to
Since the residual L-alanine and byproduct impurities are
virtually inexistent in the resulting solution, the simple
mixture of L-alaninol and mineral acid may well mimic the real
composition. In addition, the volume of samples obtained by
hydrogenation in our lab is rather limited (∼300 mL in a
batch), which cannot meet the need of the following tests,
Table 2 The pKa values of all corresponding species involved in this
study
Species
pKa values
L-Alaninola
L-Alanine
H3PO4
H2SO4
HCl
pKa 9.54, pKa 14.74
1
2
pKa 2.34, pKa 9.87
1
2
pKa 2.16, pKa 7.21, pKa 12.32
1
3
pKa −3.0, pKa2 1.99
Fig. 3 (a) Overview photograph of the EDBM system. (b) Inner
configuration of the stack and principle of the mass transfer. Symbols:
1
2
pKa −5.9
BP, bipolar membrane; A, anion exchange membrane; C, cation
a Calculated by using ChemBio3D 14.0 software.
2−
exchange membrane; HxPO4y−, H2PO4− and/or HPO4
.
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
the oxidants (such as Cl2 and O2) which are generated on the 2.3. Back diffusion investigations
anode surface. Furthermore, it is very thick (∼400 μm), which
The diffusivities of L-alaninol and mineral acids across the
membranes were evaluated with a home-built diffusion cell
(briefly illustrated in the insets of Fig. 4), which has been well
documented in our previous studies.26,27 The effective area
was 12.0 cm2 inside the cell. At the beginning of each test, the
left and right cells were both filled with the corresponding
solutions, and the initial volume was 50.0 mL. The solutions
in both cells were circulated independently by a peristaltic
pump (BT-100, Baoding Longer Precision Pump Co., Ltd,
China) instead of stirring. Both the cell and the solutions were
also immersed in a water bath to maintain the temperature at
25.0 °C. The diffusion test was stopped precisely after
120.0 min, and the solutions in both cells were drained and
collected carefully. The mass of the resulting solutions was
weighed, and the density was also measured, in order to
obtain a precise volume after the test. Then, the amount of the
mass transfer across the membrane could be determined by
the volume variations and the concentration analysis of the
species.
can minimize the cross leakage between the electrode rinse
and the acid stream. H3PO4 solution (0.30 M) was used as the
electrode rinse in this study. On the one hand, it can avoid the
contamination of foreign ions completely. On the other hand,
it is not consumed at all and discharges no poisonous gas.
The system was operated in a batch mode. Before the
EDBM test, acid solution (0.050 M H3PO4), base solution (syn-
thetic mixture of 2.00 M L-alaninol and 2.05 M H3PO4) and
electrode rinse (0.30 M H3PO4) had been filled in the corres-
ponding jacketed tanks, with a volume of 600 mL, 850 mL and
850 mL, respectively. All solutions were stirred vigorously by
using magnetic bars, and their temperatures were initially
maintained at 25.0 °C with a water bath (under the work-
bench). During the test, all streams were driven in a cocurrent
manner by using a four-channel metering pump (LEAD-2,
Baoding Longer Precision Pump Co., Ltd, China), and the flow
rate of each stream was adjusted to 240 mL min−1. The corres-
ponding flow velocity was 3.03 cm s−1 for both acid and base
streams, and the resident time was 9.04 s inside the stack.
During the EDBM operation, the synthetic feed solution was
basified in the base chamber to obtain L-alaninol solution,
and the acid was accumulated in the acid chamber to acquire
phosphoric acid, as shown in Fig. 3b. The stack current,
voltage and accumulated charge transfer were all recorded
automatically by using a homemade computer system based
on the National Instruments DAQmx platform. The pH, con-
ductivity, and temperature of both acid and base solutions
were also monitored. The samples from all tanks were taken at
a pre-determined interval.
2.4. Analytical methods
The total P concentration in samples was analyzed by induc-
tively coupled plasma optical emission spectrometry (ICP-OES,
PerkinElmer Optima 8000, USA). The L-alaninol concentration
was determined by using a total organic carbon analyzer
(TOC-VCPH, Shimadzu, Japan). The Cl− and SO4 concen-
2−
trations were determined by potentiometric titration (ZDJ-4B
automatic titrator from INESA Scientific Instrument, China),
with 0.100 M AgNO3 and 0.050 M Pb(NO3)2 as standard solu-
tions. When too low to be determined reliably in some
samples, they were analyzed by ion chromatography (IC,
Dionex ICS-600, Thermo Fisher Scientific, USA) instead.
The removal ratio R of a specific species from the base
stream (also known as the recovery ratio for the favorable mass
transfer, or leakage ratio for the unfavorable mass transfer)
can be calculated by the following formula:
The EDBM tests of L-alaninol salt with other mineral acids
(H2SO4 and HCl) were performed similarly. For H2SO4, the
initial base solution was a mixture of 2.00 M L-alaninol and
1.03 M H2SO4, and the initial acid stream was 0.025 M H2SO4.
While for HCl, the initial base solution was a mixture of 2.00
M L-alaninol and 2.05 M HCl, and the initial acid stream was
0.050 M HCl. The rinse solution was 0.30 M H2SO4 for both
H2SO4 and HCl. All other operational protocols were identical
as in the case of H3PO4.
c0V0 ꢀ ctVt
R ¼
ꢁ 100%
ð1Þ
c0V0
Fig. 4 Diffusion fluxes and volume change ratios (i.e. (Vt − V0)/V0, termed ΔVolume) across various membranes in different solutions with H3PO4
and L-alaninol. In the initial (a) and final (b) stages of the EDBM process.
Green Chem.
This journal is © The Royal Society of Chemistry 2019

View Article Online
Green Chemistry
Paper
where c is the concentration (mol L−1) of the species con- meation of organic ions across the ion exchange membranes
cerned and V is the volume (L) of the base solution. The sub- (IEMs) is usually prevented in practical operations.28–31 In this
scripts “0” and “t” represent the sampling points, namely study, a two-compartment EDBM configuration with only BPM
“before the test” and “after the test”, respectively.
and AEM (in the repeating cell pairs) is employed to comple-
The ultimate goal is to remove the mineral acid (such as tely avoid the transportation of protonated amino alcohol
H3PO4) from the amino alcohol salt solution; however, the cations across CEMs, and the detailed configuration of the
amino alcohol also migrates with the acid, which causes the stack is also shown in Fig. 3b.
loss of the final product. The ratio of acid removal and product
The membrane screening. During the EDBM process, H3PO4
loss depends on the selectivity of the EDBM process, which is (as separated phosphate HxPO4y− and H+) is driven by the elec-
termed permselectivity
formula:
P
and defined as the following tric field and migrates from the base stream to the acid stream
against the concentration difference, and that is the favored
mass transfer in the current study. However, the spontaneous
RB
RA
PAB
¼
ð2Þ
back diffusion of H3PO4 orients to the reversed direction
(down the concentration difference). Furthermore, as this con-
centration difference increases, the back diffusion will go up
proportionally as well, which may deteriorate the efficiency of
the process significantly. Therefore, a series of tests have been
carried out to investigate the back diffusion and to screen the
suitable membranes. As given in Fig. 3b, there is only one
piece of CEM (as the anode membrane) in the EDBM stack,
which is very thick (∼400 μm) and the cross leakage is greatly
ð3Þ limited. The repeating cell pairs consisted of BPMs and AEMs,
and the area of these membranes is much larger. As a result,
where RA and RB are removal ratios of species A and B, respect-
ively, and PBA represents the permselectivity of species B relative
to species A.
The cumulative current efficiency η of the process is calcu-
lated from the amount of the recovered acid and the total
charge during the test:
zðc′tV′t ꢀ c′0V′0Þ
η ¼
ꢁ 100%
Q
where V′ is the volume (L) of the acid solution; c′ is the concen- the back diffusions across BPM and AEM membranes have
tration (mol L−1) of the anions (total phosphate, SO4 or Cl−) been intensively studied.
2−
in the acid solution, and Q is the cumulative charge trans-
Two different diffusion tests were conducted here to fully
ferred in moles of electrons (also known as Faraday). z is the understand the effects during the whole EDBM process. The
stoichiometric coefficient of the acid, which is 3, 2 and 1 for first one was to mimic the diffusion in the initial stage (H2O
H3PO4, H2SO4 and HCl, respectively.
and salt), in which the acid solution is 0.050 M H3PO4, and the
The cumulative charge transferred Q and the energy con- base solution is composed of 2.00 M L-alaninol and 2.00 M
sumption E are calculated from the following formulas:
H3PO4. Dilute acid was used instead of pure water as the initial
acid solution, because the conductivity of the pure water is too
low for the practical EDBM tests. The second test was to mimic
the final stage (acid and base), in which the acid solution is
2.05 M H3PO4, and the base solution is composed of 2.00 M
L-alaninol. The compositions of the two final solutions are
both estimated based on 100% acid removal efficiency. The
schematic illustrations of these diffusion tests are shown in
the insets of Fig. 4, respectively. In this study, three different
kinds of membranes (FBM, AHA and JAM) were selected and
evaluated intensively in these synthetic solutions.
ð
t
Q ¼ NIdt=F
ð4Þ
ð5Þ
0
ð
t
E ¼ UIdt
0
where U and I are the stack voltage (V) and current (A), respect-
ively; F is the Faraday constant (26.8 A h mol−1); t is the test
time; and N is the number of repeating cell pairs, which is 12
in this study.
The diffusion flux J (mol m−2 h−1) is a reasonable indicator
for the permeability (or the rate of leakage) of the species
across the membrane, which can be calculated from the fol-
lowing formula, and here the positive diffusion direction is
defined as from the acid to the base solution:
3. Results and discussion
3.1. Design of the EDBM system
The stack configuration. Technically speaking, there are
three practicable cell configurations (BPM + AEM + CEM, BPM
+ CEM and BPM + AEM) in the design of an EDBM stack, all of
cat Vta ꢀ ca0V0a ctbVtb ꢀ cb0V0b
J ¼ ꢀ
¼
ð6Þ
At
At
which can produce an acid and/or base from the corres- where A is the effective membrane area (m2), t is the time for
ponding salts.11 As a kind of organic species, protonated the diffusion test (h), V is the solution volume (m3), and c is
amino alcohol has a much higher affinity with the membrane the concentration (mol m−3) of the species in solution. The
matrix than that of the inorganic ions. This high affinity will subscripts “0” and “t” represent the sampling points, namely
cause a strong mass transfer resistance in the membrane, “before” and “after” the test, respectively. The superscripts “a”
which not only increases the energy consumption but also and “b” represent the acid solution and the base solution,
shortens the life span of the membranes. Hence the per- respectively.
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
The diffusion fluxes of the species across each membrane severe, which results in a higher net water transport (a more
in the initial (H2O and salt) and final (acid and base) stage of positive volume variation).
the EDBM process are given in Fig. 4a and b, respectively.
3.2. The EDBM tests for L-alaninol phosphate
Meanwhile, the volume change ratio, which can be used to
indicate the water transport, is also plotted here and rep- The synthetic L-alaninol phosphate solutions with a pH ∼3.80
resented by ΔVolume. In addition, the detailed compositions were prepared from pure L-alaninol, H3PO4 and deionized
of the solutions in both acid and base cells are also marked in water. The typical titration curve of L-alaninol solution (∼2.0
the insets. The diffusion flux in the final stage (Fig. 4b) was M) with H3PO4 standard solution (∼2.0 M) is presented in
significantly higher than that in the initial stage (Fig. 4a) for Fig. S2a,† in which the volume of the titrant has been normal-
all membranes, which indicates that the leakages of all mem- ized to the molar amount of acid consumed by per mole of
branes are much more serious in the final stage. This phenom- L-alaninol for a clear comparison. There are two end points on
enon is straightforward, because the driven force of mixing the pH curve, of which the normalized amounts of the titrant
acid (H3PO4) and base (L-alaninol) is much stronger than that are 0.50 and 1.00, and the corresponding pH values are 8.02
of mixing salt and water. Therefore, the result in Fig. 4b is the and 4.25, respectively. This indicates that the reaction occurs
key to understanding the overall diffusion throughout the in two steps. Combining this titration curve with the pKa
whole process. First, the water transport across the membranes values of H3PO4 in Table 2, it can be deduced that the two end
is not an issue here, because the volume change ratios are very points correlate with the stoichiometric hydrogen phosphate
low (−1%–1%) for all tests. Second, the acid leakage across (HPO42−) and dihydrogen phosphate (H2PO4−), respectively.
AEMs (AHA and JAM in Fig. 4b) is much more severe than The actual pH of the feed solution (3.80) is a little lower than
BPMs (FBM in Fig. 4b), which means that the bottleneck of the that of the second end point (4.25), which means that the feed
EDBM efficiency mostly lies in AEMs, since BPMs perform is composed of L-alaninol dihydrogen phosphate and a slight
better with a lower leakage in the system. Furthermore, the excess of free H3PO4. The remaining free acid guarantees an
diffusion flux across AHA is less than that of JAM, which indi- effective and complete hydrogenation of the carboxyl group
cates that AHA is superior to JAM because of the lower leakage into the alcoholic hydroxyl group.
rate. Therefore, FBM and AHA were chosen for the following
tests.
When a DC electric field is applied on the EDBM stack,
water molecules are split in the BPMs, and the resulting H+
Intriguingly, it is shown clearly that the water transport and OH− migrate into the acid and base streams respectively,
direction across AEMs (AHA and JAM) is opposite to that of as shown in Fig. 3b. The OH− ion reacts with the protonated
BPM (FBM), which can be deduced directly from the positive/ L-alaninol to form the deprotonated L-alaninol product.
negative sign of the ΔVolume value. The total water transport Meanwhile, the phosphate ions (as H2PO4− at the initial stage,
2−
here is composed of the water osmosis and the water transfer and HPO4 later) are transported from the base to the acid
associated with hydrated ions. Since BPMs are fabricated from stream across the AEMs, which combine with the H+ ions to
stacked AEM and CEM layers, they exhibit a very low per- obtain free H3PO4. Since the initial pH of the base stream is
meability for charged ions because of the Donnan exclusion. slightly acidic (pH 3.80), which means that the concentration
Consequently, the water osmosis dominates for BPMs. In this of free OH− ions is much lower than that of phosphate ions
situation, the osmotic pressure of the acid solution (2.05 M (∼2 M), the shuttle of OH− across the AEMs is rather limited.
H3PO4, pKa = 2.16) is higher than that of the base solution As a result, the feed solution (L-alaninol dihydrogen phos-
1
(2.00 M L-alaninol, pKb = 4.46), which is due to its slightly phate) will be split into the corresponding acid (H3PO4) and
1
higher original concentration, as well as the higher degree of base (L-alaninol) solutions during the EDBM process.
dissociation. This difference in osmotic pressure causes the
Online data profiles. The EDBM test of L-alaninol phosphate
water transport from the base to the acid solution (negative solution was repeated three times in this study. Utilizing our
water transport for FBM in Fig. 4b). However, the mass transfer powerful online data monitoring capability, main parameters,
is more complicated for AEMs, which are permselective to including stack voltage and current, pH, conductivity and
anions theoretically. Besides the negatively charged anions, H+ temperature of acid and base solutions, have been fully
can also migrate across AEMs because of proton shuttling, recorded and are plotted in Fig. 5. The results indicate that the
which means that the permeability of free acid is relatively reproducibility of this work is excellent, and the curves almost
high for AEMs. Since the ions in this study are all hydrated, a replicate each other for the three tests. It is also implied that
great amount of water is transferred together from the acid cell the steady state operations have been achieved and the mem-
to the base cell (positive as defined). As a result, the overall branes are well compatible with the chemicals involved, which
water transport here is a compromise between the water is promising to practical applications.
osmosis (negative direction) and the water transfer associated
In all experiments, the EDBM stack was operated in a
with ions (positive direction, the net effect of the diffusion of voltage–current double clamp mode, in which the voltage was
H3PO4 and L-alaninol). For both AHA and JAM membranes, limited within 35 V and the current within 3.0 A (i.e. corres-
the latter predominates, and then the overall water transport ponding current density of 50 mA cm−2 in the stack). As
direction reverses compared with FBM. Furthermore, the acid shown in Fig. 5a, the stack voltage decreases slightly from ∼35
transport across JAM caused by the back diffusion is more V to 30.5 V, and then back to ∼35 V. The corresponding stack
Green Chem.
This journal is © The Royal Society of Chemistry 2019

View Article Online
Green Chemistry
Paper
stack resistance is as high as 17.0 Ω, due to the rather low con-
ductivity of the initial acid solution. However, the concen-
tration of the acid solution increases quickly as the EDBM
process proceeds. As a result, the overall resistance decreases
and the stack current increases quickly. After the current
reaches the limit (3.0 A), the stack voltage starts to decrease
since the stack resistance continuously decreases to 10.1 Ω at
150 min. The minimum stack resistance appears here, because
the conductivities of the acid as well as the base solutions are
high at this moment. After this point, the conductivity of the
base solution drops quickly, which leads to an increase in
both stack resistance and voltage again. At the last stage of the
test (after 330 min), the stack resistance rises steeply, which
results from the low conductivity of the base solution. After
the voltage reaches its limit (35 V), the stack current drops dra-
matically from 3.0 to ∼1.25 A owing to the soaring stack resis-
tance (Fig. 5b).
The temperatures of all solutions were initially maintained
at 25.0 0.5 °C by using a water bath before the EDBM tests.
However, because a large amount of heat was generated
during the EDBM operations, the solution temperatures varied
accordingly. The temperature variations of acid and base solu-
tions are plotted by dashed and solid lines respectively in
Fig. 5c. The temperature increases fast in the first 12 min
(from 25.5 °C to 29 °C), because the electric energy is partially
converted into heat in the EDBM process. After 12 min, the
temperature stabilizes at 29 1 °C in all tests. Similarly, near
the end there is also an inflection point, where the tempera-
ture falls rapidly after 351 min, which is due to the decline of
the stack power input caused by the increasing stack resis-
tance. The cumulative charge transferred in moles of electrons
is readily calculated from the stack current profiles (versus
time) in Fig. 5a according to eqn (4), and the results are also
given in Fig. 5b. Basically it increases linearly from 0 to
∼8.3 mol electrons, since the stack current is quite stable (∼3.0
A) for the majority of the duration of the experiment.
Fig. 5 Online data profiles of the EDBM process. (a) Stack voltage and
current curves. (b) Stack resistance and charge transferred. (c) The
temperature and the EDBM energy consumption curves. (d) The pH vari-
ations of both acid and base solutions. (e) The conductivity curves of
both acid and base solutions.
current rises rapidly from 2.03 A to 3.0 A in the very initial Furthermore, the stack energy consumption can be easily
stages of all tests, then stabilizes at 3.0 A for the majority of obtained likewise, according to the stack voltage and current
the duration of the experiment and drops fast to 1.25 A at last. (versus time) profiles. It also increases linearly with time, and
Although these change patterns root in the variations of the the total energy is about 0.589 kW h for each batch.
stack resistance, in return the stack resistance can be readily
As presented in Fig. 5d, the pH increases from 3.75 to 11.72
calculated from the stack voltage and current according to in the base solution and decreases from 1.88 to 0.75 in the
Ohm’s law, and the results are plotted in Fig. 5b. In addition, acid solution, which is because the base solution is basified
the conductivities of both acid and base solutions in all tests and H3PO4 accumulates in the acid solution during the EDBM
are also given in Fig. 5e, because the fluctuation of stack resis- test. The pH variation of the base solution perfectly matches
tance mainly depends on the conductance of the solutions the titration curve in Fig. S2a,† but in a reversed manner. In
since the membranes have already been fixed. The conduc- order to clearly explicate the phenomena and the root causes
tivity of the acid solution increases monotonically from 6.5 to beneath, the whole test is divided into three parts here. In the
83 mS cm−1, while that of the base solution slightly increases first half (0–163 min), OH− ions enter and H2PO4 ions leave
−
from 28.4 to 30.8 mS cm−1 within the first 150 min, and then the base chamber (see Fig. 3b). However, because there are
declines quickly to 0.87 mS cm−1 at the end of the test. still large amounts of H2PO4 remaining, OH− ions react with
−
−
Consequently, a combination of this affects the final stack re- them (smallest pKa value) until all the H2PO4 ions convert
2−
sistance, jointly with the membrane resistance (Rstack = 1/Gacid into HPO4 (at 163 min), with an increase in pH from 3.75 to
+ 1/Gbase + Rmembr). The resistance in the electrode compart- ∼8.10. During this period, the molar concentration of anions
ments is neglected here, because it is much lower than that of declines with the monovalent-to-divalent conversion, because
the multiple (12) repeating cell pairs. At the beginning, the the equivalent concentration remains constant here, due to
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
the electroneutrality restriction, since the concentration of were precisely recorded by the scale outside the tanks before
cations (protonated L-alaninol) basically remains unchanged. It each sampling point (Fig. 6a). The initial volume of the base
is worth noting that the equivalent limiting conductivity of solution (850 mL) is larger than that of the acid solution
−
HPO42− (51.8 S cm2 mol−1) is greatly higher than that of H2PO4
(600 mL). This was deliberately designed to concentrate the
(31.8 S cm2 mol−1),32 which can explain why the removal of recovered mineral acids (such as H3PO4) in a limited amount
H2PO4 in this period can raise the conductivity of the base of water, maintaining the water balance when reusing them.
−
solution as shown in Fig. 5e. In the second part (163 min– The volume increases from 600 to 790 mL in the acid tank,
380 min), the further basification involves more OH− in while and decreases from 850 to 605 mL in the base tank. This sig-
2−
HPO4 out, and the experimental phenomena are much more nificant water transport from the base to the acid stream is
3−
straightforward. As shown in Table 2, PO4 (pKa 12.32) is a caused by the electro-osmosis, namely the motion of water
stronger base than pure L-alaninol (pKa 9.54). Thus, OH− ions molecules across the membrane associated with the migration
preferentially react with protonated L-alaninol to produce pure of ions under an electric field (basically phosphate anions in
2−
neutral L-alaninol, while HPO4 ions leave the base chamber. Fig. 3b). The water transport number (tw) is a practical indi-
As a result, the solution conductivity tails off. At the last stage cator to present the water flux, which is defined as the number
(after 380 min), the conductivity of the final base solution is as of the moles of water transported as 1 F of electric charge
low as 0.87 mS cm−1, which is very close to that of the pure passes through.33,34 It can be estimated from the slope of
L-alaninol solution with equivalent concentration (0.839 mS water transference from the base chamber to the acid
cm−1 for 2.1 M L-alaninol). This low conductivity implies that chamber. As being calculated from Fig. 6a, the water transfer-
the residual phosphate is negligible in the final base solution.
ence rate is 26.7 1.0 mL H2O/F (average of the base solution
Offline sample analysis. About the offline sample analysis, and the acid solution), so consequently tw is 1.48 0.06 mol
the samples (1.5 mL) were taken from both acid and base H2O/F here. It is lower than that in the literature (usually
tanks when the charge reached every 1.2 mol electrons, and all 4–8),33,34 which may be mitigated by the water osmosis. Since
the sampling points are also marked on the charge curves in the concentration of the base solution is significantly higher
Fig. 5b. These samples were carefully analyzed and the results than that of the acid solution, the osmotic water transport is
are presented in Fig. 6 in detail. The volumes of both solutions from the acid chamber to the base chamber, which is exactly
the opposite of electro-osmosis. The densities of all samples
were carefully measured as well, and the results are plotted in
Fig. 6b. The density of the acid solution rises smoothly from
1.016 to 1.115 g mL−1, and that of the base solution declines
from 1.128 to 1.000 g mL−1. The change patterns also indicate
that the solute concentration increases in the acid solution
while it decreases in the base solution, because it is positively
correlated with the density.
The phosphate concentration was precisely determined by
measuring the total P with ICP-OES, and the results are given
in Fig. 6c and exhibit a perfect reproducibility in the three
repeated tests. It decreases gradually in the base solution as
the charge increases, and the residual concentration
(<0.001 mol L−1) is negligible in the final product. Meanwhile,
in the acid solution, it increases significantly from 0.050 to
2.14 M in average, which is even higher than the initial acid
concentration (∼2.05 M) in the reactants of hydrogenation.
Therefore, this recovered acid solution can be reused directly
for the next batch of hydrogenation because of its high concen-
tration. The concentration of L-alaninol in all samples was
determined by using a TOC analyzer, and the results are
plotted in Fig. 6d. It indicates that L-alaninol leaks partially
(also known as back diffusion) from the base solution to the
acid solution, because its concentration in the acid solution
increases from 0 to 0.22 mol L−1 in average. In the base solu-
tion it also increases from 2.01 to 2.30 mol L−1 at first
(0–4.8 mol e), which is caused by the volumetric shrinkage of
the base solution (as shown in Fig. 6a). But afterwards
(4.8–8.3 mol e), the leakage of L-alaninol is more impact, and
Fig. 6 Offline sample analysis results during the EDBM process. (a)
Volume variations of acid and base solutions. (b) Density variations of
acid and base solutions. (c) Concentration of total P variations of acid
the concentration stops increasing and even decreases slightly
at last (2.30 to 2.25 mol L−1 in average).
and base solutions. (d) Concentration of L-alaninol variations of acid and
base solutions.
Green Chem.
This journal is © The Royal Society of Chemistry 2019

View Article Online
Green Chemistry
Paper
OH− ions react with the phosphates in different forms; and on
the other hand, they compete with phosphates as well when
migrating through the AEM. For the above reasons, a weaken-
ing EDBM performance was observed.
Finally, as demonstrated in Fig. 3b, after the whole EDBM
process the mass transfer in the three chambers (base
chamber, acid chamber and electrode chamber) can be sum-
marized as follows: (1) for the base chamber, by removing
the water with a rotary evaporator, the pure L-alaninol
product was obtained in the form of a slightly viscous liquid,
with a concentration of 11.74 0.12 M (average of the three
batches) according to titration (12.09 M for pure L-alaninol).
(2) For the acid chamber, the concentration of the recovered
phosphoric acid (2.14 0.02 M) in the acid tank is higher
than the initial concentration (2.05 M) and can be reused
directly. (3) For the electrode chamber, the samples of H3PO4
rinse were also taken before and after each test, and the data
show that the volume and composition of the rinse solution
remain stable. Specifically, the reduction in volume (due to
the water electrolysis) is less than 10 mL for each batch and
no leakage of L-alaninol was detected (TOC < 30 mg L−1),
Fig. 7 (a) The removal ratio of total phosphate and L-alaninol. (b) The
corresponding permselectivity of total phosphate over L-alaninol, as well
as the cumulative current efficiency.
The removal ratios of total phosphate (favored) and which means that the rinse can also be reused. On the
L-alaninol (unfavored) are both calculated according to eqn (1) whole, all the chemicals involved here either convert into
based on the volume and concentration data in Fig. 6, and final products or can be recycled directly, with no waste dis-
plotted in Fig. 7a. For total phosphate, it increases smoothly charge at all, and thus meet the requirements of green
from 0% to over 99.9%, which indicates that the final base chemistry.
solution is a virtually pure L-alaninol product. The removal
Brief cost analysis. The total cost of a membrane-based
ratio of L-alaninol rises from 0% to 11.2% on average, which is system is usually composed of capital and operational
not negligible but still acceptable. Furthermore, this part of costs.35–37 The former involves the investment for the equip-
L-alaninol will not be lost and discharged as waste, because it ment, and the latter involves the energy cost. The cost of an
can be returned to the hydrogenation process when the result- EDBM system usually suffers from high prices of IEMs,
ing acid is reused. The permselectivity of phosphate over especially BPMs. Thus, it is reasonable to estimate the total
L-alaninol is readily calculated by eqn (2) and shown in Fig. 7b. equipment cost based on the membrane prices. In addition,
The selectivity is higher at the start of all tests, which is as the stack power is commonly considered as the total power in
high as 16.9 in average, and then decreases monotonically to an EDBM system, because it is much higher than the energy
about 8.7 at the end of the tests. According to eqn (3), the consumptions of other auxiliary instruments, such as
cumulative current efficiency η is calculated and also declines pumps.35 A rough cost model is established in Table 3, and
from 90% to 61% monotonically. It is worth noting that η is an the normalized cost is $1.32 kg−1 pure L-alaninol product,
apparent value here for the phosphate system, since the stoi- including both capital (54%, $0.715 kg−1) and energy (46%,
chiometric coefficient z is simplified to be 3 (the actual z is not $0.606 kg−1) costs. Meanwhile, the reclaimed H3PO4 (1.75 kg
determinable), under an assumption that all phosphates kg−1 L-alaninol) can bring a profit of $1.67 kg−1 according to
3−
migrate across the AEM as PO4 ions. The actual current the market price of phosphoric acid ($810 per ton, as 85 wt%
efficiency may be lower, because the phosphates also migrate solutions).38 Taken together, the EDBM process in this study is
−
in other forms (such as H2PO4 and HPO42−). In contrast, the quite economically competitive, since it is already profitable in
situations are much simpler and more straightforward for the terms of recovering H3PO4 alone, not to mention the extra
EDBM processes of both chloride and sulphate systems, since benefits of eliminating waste generations as well as maximiz-
all chlorides exist as Cl− ions (z = 1) and sulphates exist as ing the product yield. Accordingly, this is a sustainable process
SO42− ions (z = 2) in the whole pH range (pH 4–11). The degra- for the production of chiral amino alcohols.
dation of permselectivity as well as current efficiency can be
3.3. L-Alaninol sulphate and chloride systems
explained as follows. First, as described in Fig. 4, the back
diffusion of H3PO4 increases significantly as the free acid Besides H3PO4, some other mineral acids4,39 (such as H2SO4
accumulates in the acid stream, which is the opposite of the and HCl) can also be used in the catalytic hydrogenation of
EDBM process, thus reducing the current efficiency and the amino acid as shown in Fig. 2. In order to demonstrate the
permselectivity. Secondly, OH− generated in bipolar mem- generality of this green production of chiral amino alcohols by
branes also accounts for the degradation of EDBM perform- EDBM, another two synthetic feed solutions (synthetic
ances. On the one hand, besides the protonated L-alaninol, L-alaninol sulphate and chloride solutions) have been carefully
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
Table 3 Cost estimation of the EDBM process with L-alaninol
phosphate
increases first and then decreases, and it is caused by the con-
−
2−
version of the different phosphate species (H2PO4 to HPO4
,
already discussed in section 3.2 in detail). In contrast, without
any species conversation, the patterns of both H2SO4 and HCl
systems are much more straightforward, and the conductivity
decreases monotonically. All the results manifest that EDBM
can be applicable to L-alaninol salt with different acids, such
as H3PO4, H2SO4 and HCl, and for a clearer comparison the
parameters of the three analogous systems are given in
Table 4. Compared with the H3PO4 system, the overall
efficiency is much higher for H2SO4 and HCl, and the oper-
ation time of each batch is also much shorter (164 min and
207 min vs. 382 min). In terms of the reasons, on the one
hand the utilization ratio of protons in the hydrogenation
process is 100% for H2SO4 and HCl, while only 33.3% for
H3PO4. Thus, the H3PO4 consumption is far higher and the
recovery is more difficult (more time/charge needed). On the
Items
Notes
Capital cost
Area of AEM
Area of BPM
Price of AEMs
BPM-price
Membrane cost
Stack cost
0.072
0.072
360
720
77.76
116.64
m2
m2
$ m−2
$ m−2
$
$
1.5 times of
membrane
0.5 times of stack
Peripheral equipment
cost
Total capital cost
58.32
$
$
174.96
Operational cost
Energy
Batch time
Product concentration
Product volume
Batch productivity
0.589
382
2.138
606
kW h
min
mol L−1
mL
other hand, the conductivity of SO4 and Cl− is much higher
1.296
mol
2−
than that of phosphate, which significantly reduces the overall
stack resistance and raises the current. Specifically, the plateau
resistance is about 10.2, 7.7 and 7.9 Ω for H3PO4 (Fig. 5b),
H2SO4 (Fig. S5b†) and HCl (Fig. S8b†). Likewise, the energy con-
sumption and total costs are also calculated (Tables S1 and
S2†). If only judged by Table 4, it seems that H2SO4 and HCl
might be more effective than H3PO4 to produce the pure amino
alcohol by EDBM. However, considering that the catalytic hydro-
genation occurs under high-temperature and high-pressure con-
Overall cost
System lifespan
Total productivity
Capital cost
Energy
16 000
244.6
0.715
6.06
h
kg
2-Year estimation
As pure L-alaninol
$ kg−1
kW h kg−1
$ kg−1
0.606
$0.10/kW h power
price
Capital + energy
Total cost
1.321
$ kg−1
investigated here as well. First, pure L-alaninol solution (∼2.0 ditions, H2SO4 and HCl are much more corrosive to the metallic
M) was titrated with H2SO4 (∼1.0 M) and HCl (∼2.0 M), and (such as Stainless Steel 316) autoclaves than H3PO4. Therefore,
the curves are presented in Fig. S2b and S2c† respectively. As a under these circumstances, H3PO4 should still be chosen as a
strong acid (see the pKa values in Table 2), there is only one superior source of protons for the EDBM process.
end point on each pH curve for both H2SO4 and HCl titrants.
3.4. The integrated green process
The normalized amount is 0.50 for H2SO4 and 1.00 for HCl,
which indicates that they are totally dissociated and all A completely green synthesis process of chiral amino alcohols
protons are utilized. The diffusion of L-alaninol salts and water is proposed in Fig. 8, in which L-alaninol is taken as an
across the three membranes was also studied (Fig. S3† for instance of chiral amino alcohol, and H3PO4 is taken as an
H2SO4 and Fig. S4† for HCl). Very similar to the H3PO4 system, instance of mineral acid. Firstly, biorenewable L-alanine solid,
the leakages of solutes are much more serious in the final Ru–C catalyst powder, and H3PO4 solution are injected into an
stage of the EDBM process. In addition, the acid leakage autoclave and mixed to form a smooth suspension. After that,
across AEMs (AHA and JAM) is significantly more severe than the autoclave is purged with an inert gas first and then filled
that of BPMs (FBM), and the diffusion flux across AHA is lower with high-pressure H2 gas; meanwhile, the reactant suspen-
than that of JAM in all tests. Therefore, FBM and AHA are still sion is heated. During this period, the amino acid is reduced/
the reasonable choices to study here.
hydrogenated by H2 gas via the Ru–C catalysts. When the reac-
Similar EDBM tests have also been conducted for L-alaninol tion is completed and the solution is cooled down, H2 gas is
sulphate (Fig. S5–S7†) and chloride (Fig. S8–S10†) in this vented and the resulting suspension is collected from the
study. The pH variations of the base solutions in H2SO4 bottom of the reactor and sent to a solid–liquid separation.
(Fig. S5d†) and HCl (Fig. S8d†) systems are very different from The solid is the catalyst powder which can be reused directly,
that of the H3PO4 system (Fig. 5d). There is only one inflection and the filtrate is the L-alaninol salt solution. Together with
point (a sudden pH increase from 4 to 8) for the H2SO4 and water, it is fed into an EDBM stack, where it is converted into
HCl systems, while there are two (pH 8.02 and 4.25, respect- the corresponding base (L-alaninol) and acid (H3PO4) solu-
ively) for the H3PO4 system. This can be explained by the fact tions. As discussed above, the recovered phosphoric acid can
that the H2SO4 and HCl are strong acids but H3PO4 is not, be used directly as the feed acid for the next batch. The last
which accords well with their pKa values (listed in Table 2) and step of the synthesis process is distillation. The distillate is
corresponding titration curves (Fig. S2†). Another obvious water that can also be reused directly as EDBM feed, and the
difference appears in the conductivity variations of the base residuum is the pure L-alaninol product. The overall process
solutions. For the H3PO4 system (Fig. 5e), the conductivity completely avoids the involvement of extra acids and bases,
Green Chem.
This journal is © The Royal Society of Chemistry 2019

View Article Online
Green Chemistry
Paper
Table 4 The comparison of different mineral acid systems for EDBM tests
Proton utilization in
hydrogenation
Batch time
(min)
Plateau resistance
(Ω)
Energy consumption
Total cost
Mineral acid
(kW h kg−1
)
($ kg−1
)
H3PO4
H2SO4
HCl
33%
100%
100%
382
164
207
10.2
7.7
7.9
6.06
1.65
2.13
1.321
0.415
0.528
4. Conclusions
In this work, the results demonstrate the effectiveness of a new
approach for the production of chiral amino alcohols for the
first time. Biorenewable amino acids (L-alanine here) and clean
H2 were used as feed stocks and successfully converted into
the pure L-alaninol product by an integrated process com-
prised of catalytic hydrogenation and EDBM. All the chemicals
involved either convert into the final product or are being
recycled directly, without any waste discharge. Three typical
mineral acids (H3PO4, H2SO4 and HCl) have been investigated
systematically here, which demonstrates a great generality of
the approach. According to the results of all tests, the removal
ratios of the acids in L-alaninol solutions are all very high
(>99.9%), and the energy consumption of EDBM processes is
also competitive. Combined with the cost analysis, it is justi-
fied that this new approach is high-productive, cost-effective
and environmentally friendly. It is realized in a true atomically
economic way for the completely green production of chiral
amino alcohol. This work provided another excellent example
that EDBM is versatile to upgrade the existing or emerging
industrial processes, which is promising to build a more sus-
tainable future.
Fig. 8 A green synthesis process of L-alaninol which integrates catalytic
hydrogenation and EDBM.
and eliminates waste generation as well. It strictly conforms to
the standards of the ideal atom economy, which is a good
model of green synthesis enabled by the hybrid of catalytic
hydrogenation and EDBM.
Conflicts of interest
This novel approach has an excellent generality that exhi-
bits three qualities. First, the catalytic hydrogenation method
can be applied to the synthesis of alcohol products from
various alpha amino acids (such as alanine, serine, valine
and proline),1,39 beta and gamma amino acids (such as
3-aminopropanoic acid and 4-aminobutanoic acid),40 and
even ordinary organic acids (such as lactic acid and propionic
acid).21 Secondly, the amino alcohol products, which exist as
cations (protonated form in acidic solutions with mineral
acids), cannot transfer across the membrane theoretically
during the EDBM process (BP + AEM configuration, see
Fig. 3b for details). Accordingly, the various kinds of alcohol
products would give similar performances in the EDBM
process. Finally, based on the systematical investigation of
the three typical mineral acids (H3PO4, H2SO4 and HCl) here,
it is justified that this process will be universal for different
acids. Thus, we have good reason to believe that such a green
synthesis process has a good generality, which is able to
produce various chiral amino alcohols besides L-alaninol. It
is of great significance for the industrial application
prospect.
There are no conflicts to declare.
Acknowledgements
The authors are grateful for financial support from the
National Natural Science Foundation of China (21673159,
21203139 and 21003097). The authors are also thankful to Mr
Mattias Max for assistance in writing improvement and con-
structive discussions.
Notes and references
1 K. P. Pimparkar, D. J. Miller and J. E. Jackson, Ind. Eng.
Chem. Res., 2008, 47, 7648–7653.
2 M. J. McKennon, A. I. Meyers, K. Drauz and M. Schwarm,
J. Org. Chem., 1993, 58, 3568–3571.
3 S. Antons, A. S. Tilling and E. Wolters, US Pat, 6310254,
2001.
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
4 O. Kutzki, K. Ditrich and M. Bartsch, US Pat, 20100240928, 22 R. Alamillo, M. Tucker, M. Chia, Y. Pagán-Torres and
2010. J. Dumesic, Green Chem., 2012, 14, 1413–1419.
5 M. Rignon, J.-C. Catonne, F. Denisard and J. Malafosse, US 23 M. G. Al-Shaal, A. Dzierbinski and R. Palkovits, Green
Pat, 4678549, 1987.
Chem., 2014, 16, 1358–1364.
6 Y. Wei, Z. Tian and J. Zhu, US Pat, 20100168385, 2010.
7 F. T. Jere, D. J. Miller and J. E. Jackson, Org. Lett., 2003, 5,
527–530.
24 V. A. Sifontes Herrera, O. Oladele, K. Kordás, K. Eränen,
J.-P. Mikkola, D. Y. Murzin and T. Salmi, J. Chem. Technol.
Biotechnol., 2011, 86, 658–668.
8 J. Verduyckt, R. Coeck and D. E. De Vos, ACS Sustainable 25 M. Selva, M. Gottardo and A. Perosa, ACS Sustainable Chem.
Chem. Eng., 2017, 5, 3290–3295. Eng., 2013, 1, 180–189.
9 K. Drauz, I. Grayson, A. Kleemann, H. P. Krimmer, 26 H.-F. Xiao, Q. Chen, H. Cheng, X.-M. Li, W.-M. Qin,
W. Leuchtenberger and C. Weckbecker, Amino Acids, in
Ullmann’s Encyclopedia of Industrial Chemistry, 2007, DOI:
10.1002/14356007.a02_057.pub2.
B.-S. Chen, D. Xiao and W.-M. Zhang, J. Membr. Sci., 2017,
537, 111–118.
27 Q. Chen, Y.-Y. Du, K.-M. Li, H.-F. Xiao, W. Wang and
W.-M. Zhang, Mater. Des., 2017, 113, 149–156.
28 H.-J. Lee, M.-K. Hong, S.-D. Han, S.-H. Cho and
S.-H. Moon, Desalination, 2009, 238, 60–69.
29 M. Szczygiełda, J. Antczak and K. Prochaska, Sep. Purif.
Technol., 2017, 181, 53–59.
30 X. Sun, H. Lu and J. Wang, J. Cleaner Prod., 2017, 143, 250–
256.
31 M. Miao, Y. Qiu, L. Yao, Q. Wu, H. Ruan, B. Van der
Bruggen and J. Shen, Sep. Purif. Technol., 2018, 205, 241–
250.
10 F. T. Jere, J. E. Jackson and D. J. Miller, Ind. Eng. Chem.
Res., 2004, 43, 3297–3303.
11 C. Huang and T. Xu, Environ. Sci. Technol., 2006, 40, 5233–
5243.
12 J. R. Davis, Y. Chen, J. C. Baygents and J. Farrell, ACS
Sustainable Chem. Eng., 2015, 3, 2337–2342.
13 S. Mikhaylin, V. Nikonenko, G. Pourcelly and L. Bazinet,
Green Chem., 2016, 18, 307–314.
14 S. Mikhaylin, L. Patouillard, M. Margni and L. Bazinet,
Green Chem., 2018, 20, 449–456.
15 J. Lin, W. Ye, J. Huang, B. Ricard, M.-C. Baltaru, 32 H. Kalka, Table of Diffusion Coefficients, http://www.aqion.
B. Greydanus, S. Balta, J. Shen, M. Vlad, A. Sotto, P. Luis de/site/194, (accessed April 2, 2019).
and B. Van der Bruggen, ACS Sustainable Chem. Eng., 2015, 33 N. P. Berezina and E. N. Komkova, Colloid J., 2003, 65, 1–10.
3, 1993–2001.
34 N. P. Berezina, N. A. Kononenko, O. A. Dyomina and
N. P. Gnusin, Adv. Colloid Interface Sci., 2008, 139, 3–28.
35 H. Strathmann, Desalination, 2010, 264, 268–288.
16 Q. Wang, C. Jiang, Y. Wang, Z. Yang and T. Xu, ACS
Sustainable Chem. Eng., 2016, 4, 5743–5751.
17 J. Shen, J. Lin, J. Yu, K. Jin, C. Gao and B. Van der Bruggen, 36 J.-X. Zhuang, Q. Chen, S. Wang, W.-M. Zhang, W.-G. Song,
Chem. Eng. Process., 2013, 72, 137–143.
18 D. R. Lide, CRC Handbook of Chemistry and Physics, CRC
Press LLC, 84th edn, 2004.
19 Z. Li, C. Cao, Z. Zhu, J. Liu, W. Song and L. Jiang, Adv.
Mater. Interfaces, 2018, 5, 1801259.
20 A. S. Piskun, J. E. de Haan, E. Wilbers, H. H. van de
Bovenkamp, Z. Tang and H. J. Heeres, ACS Sustainable
Chem. Eng., 2016, 4, 2939–2950.
L.-J. Wan, K.-S. Ma and C.-N. Zhang, J. Membr. Sci., 2013,
432, 90–96.
37 Q. Chen, C. Xue, W.-M. Zhang, W.-G. Song, L.-J. Wan and
K.-S. Ma, Ind. Eng. Chem. Res., 2014, 53, 13467–13474.
38 Phosphoric Acid Price Analysis, https://www.echemi.com/
productsInformation/pd20150901061-phosphoric-acid.
html, (accessed Apr. 2, 2019).
39 Z. Qiu and C. Yong, CN Pat, 103896818, 2014.
21 Y. Chen, D. J. Miller and J. E. Jackson, Ind. Eng. Chem. Res., 40 M. Tamura, R. Tamura, Y. Takeda, Y. Nakagawa and
2007, 46, 3334–3340.
K. Tomishige, Chem. Commun., 2014, 50, 6656–6659.
Green Chem.
This journal is © The Royal Society of Chemistry 2019