Rh-Catalyzed Hydrogenation of CO2 to Formic Acid in DMSO-based Reaction Media: Solved and Unsolved Challenges for Process Development

Article abstract of DOI:10.1002/adsc.201801098

Process concepts have been conceived and evaluated for the amine-free homogeneous catalyzed hydrogenation of CO₂ to formic acid (FA). Base-free DMSO-mediated production of FA has been shown to avoid the formation of stable intermediates and presumably the energy-intensive FA recovery strategies. Here, we address the challenges in the development of an overall process: from catalyst immobilization to the FA isolation. The immobilization of the homogeneous catalyst was achieved using a multiphasic approach (n-heptane/DMSO) ensuring high retention of the catalyst (>99%) and allowing facile separation of the catalyst-free product phase. We show that the strong molecular interactions between DMSO and FA on the one hand shift the equilibrium towards the product side, on the other hand, lead to the formation of an azeotrope preventing a simple isolation step by distillation. Thus, we devised an isolation strategy based on the use of co-solvents and computed the energy demands. Acetic acid was identified as best co-solvent and its compatibility with the catalyst system was experimentally verified. Overall, the outlined process involving DMSO and acetic acid as co-solvent has a computed energy demand on a par with state-of-the art amine-based processes. However, the insufficient chemical stability of DMSO poses major limitations on processes based on this solvent. (Figure presented.).

Full text of DOI:10.1002/adsc.201801098

Title: Catalytic Hydrogenation of CO2 to Formic Acid in DMSO-based
Reaction Media: Solved and Unsolved Challenges for Process
Development
Authors: Christian Jens, Martin Scott, Bastian Liebergesell, Christian
Westhues, Pascal Schäfer, Giancarlo Franciò, Kai Leonhard,
Walter Leitner, and André Bardow
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: Adv. Synth. Catal. 10.1002/adsc.201801098
Link to VoR: http://dx.doi.org/10.1002/adsc.201801098

10.1002/adsc.201801098
Advanced Synthesis & Catalysis
Very Important Publication
FULL PAPER
DOI: 10.1002/adsc.201((will be filled in by the editorial staff))
Rh-Catalyzed Hydrogenation of CO2 to Formic Acid in DMSO-
based Reaction Media: Solved and Unsolved Challenges for
Process Development
Christian M. Jens,a# Martin Scott,b# Bastian Liebergesell,a Christian G. Westhues,b
Pascal Schäfer,a Giancarlo Franciò,b Kai Leonhard,a Walter Leitnerb,c* and André
Bardowa*
a
Chair of Technical Thermodynamics, RWTH Aachen University, Schinkelstrasse 8, 52062 Aachen, Germany
[Tel.: +49 241 80 95380; Fax: +49 241 80 92255; andre.bardow@ltt.rwth-aachen.de]
Institut für Technische und Makromolekulare Chemie, RWTH Aachen University, Worringerweg 2, Aachen, 52074,
b
Germany
c
Max Planck Institute for Chemical Energy Conversion, Stiftstr. 34 – 36, Mülheim an der Ruhr, 45470, Germany
[Tel.: +49 208 306 3614; Fax: +49 208 306 3951; leitner@cec.mpg.de]
Received: ((will be filled in by the editorial staff))
Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/adsc.201######.((Please
delete if not appropriate))
Abstract. Process concepts have been conceived and on the other hand, lead to the formation of an azeotrope
evaluated for the amine-free homogeneous catalyzed preventing a simple isolation step by distillation. Thus, we
hydrogenation of CO₂ to formic acid (FA). Base-free devised an isolation strategy based on the use of co-solvents
DMSO-mediated production of FA has been shown to avoid and computed the energy demands. Acetic acid was
the formation of stable intermediates and thus the energy- identified as best co-solvent and its compatibility with the
intensive FA recovery strategies. Here, we address the catalyst system was experimentally verified. Overall, the
challenges in the development of an overall process: from outlined process involving DMSO and acetic acid as co-
catalyst immobilization to the FA isolation. The solvent has a computed energy demand on a par with state-
immobilization of the homogeneous catalyst was achieved of-the art amine-based processes. However, the insufficient
using a multiphase approach (n-heptane/DMSO) ensuring chemical stability of DMSO poses major limitations on
high retention of the catalyst (>99%) and allowing facile processes based on this solvent.
separation of the catalyst-free product phase. We show that
the strong molecular interactions between DMSO and FA on
the one hand shift the equilibrium towards the product side,
Keywords: Carbon Dioxide Hydrogenation; Biphasic
Catalysis; PC-SAFT; COSMO-RS; DMSO decomposition
sources like power plants, fermentation and biogas
facilities, or steel industry.[4] Additionally, FA could
be used as chemical storage of hydrogen originating
from surplus renewable energy, thereby facilitating
Introduction
Formic acid (FA, HCOOH) is a commodity chemical
product with a variety of agricultural and industrial
applications.[1] The annual production amounts to
697 kt (2013)[2] with main uses for silage, in animal
feed fermentation, and food preservation, for leather
and textile treatment as well as in the chemical,
rubber and steel industry.[1b] These markets are
strongly influenced by the growth of the population
and the demand of FA is thus expected to rise. FA is
produced in a two-step process from carbon
monoxide as carbon source, which is typically
obtained from fossil resources. The use of waste
carbon dioxide as starting material for FA production
would be a step towards a more sustainable chemical
production reducing the environmental impact in
terms of carbon footprint and fossil resource
depletion.[2a, 3] The competitiveness of such a process
is maximized when CO₂ is harvested at point
the transition from fossil to a circular energy
5]
economy.[1b,
Thus,
the
reversible
hydrogenation/dehydrogenation of CO₂ is a widely
discussed option for interfacing the energy with the
chemical sector.[6] Compared to other storage
materials,[5a, 7] FA has a relatively low mass fraction
of hydrogen (4.4 wt%). However, this drawback
might be offset as FA can be directly used in certain
applications, e.g. in form of FA fuel cells
(DFAFCs).[7-8] Furthermore, FA has been suggested
as a practical carbon monoxide surrogate and stable
formal intermediate of the reverse water gas shift
reaction (rWGS).[8-9] Latest research has shown that
the decarbonylation/dehydration of FA can be carried
out with a wide range of commercially available
catalysts such as transition metal doted zeolites.[10]
While the use of FA as CO storage system has
1
This article is protected by copyright. All rights reserved.

10.1002/adsc.201801098
Advanced Synthesis & Catalysis
received less attention up to now, the mass fraction of
60.8 wt% makes FA attractive as a liquid CO-source
that can be readily transported and distributed.[11]
determined experimentally and the effect of hydrogen
bonding was analyzed in detail by DFT
calculations.[15c, 17]
It might be expected that the reduced number of
components of the binary FA/solvent mixture as
compared to the ternary FA/base/solvent mixture
could simplify the downstream processing of the
catalytic CO2 hydrogenation in DMSO. However, no
overall process schemes and no experimental or
process simulation data for solvent-using processes
are currently available to corroborate this assumption.
Figure 1. BASF process for the hydrogenation of CO2 to
FA using NHex3 and a biphasic system.
Among the concepts for integrated hydrogenation
and FA isolation, the process developed by BASF
was studied in detailed and operated at a mini plant
scale (Figure 1).[12] It is thus used as benchmark for
the analysis in the present paper. This process uses an
apolar amine (NHex3) as stabilizing agent forming an
adduct with FA and as catalyst phase at the same time,
whereas a polar diol is used as the product phase.
Because of some catalyst leaching in the product
phase, an additional extraction with NHex3 is
necessary before the product mixture can be
subjected to thermal cleavage of the FA-NHex3
adduct and FA distillation. Most recently, it has also
been demonstrated that aqueous amine solutions, as
used for CO2 scrubbing from flue gases, can directly
be hydrogenated in a biphasic system with excellent
catalyst retention and recyclability leading to the
attractive prospect of an integrated CO2-capture and
utilization process.[13] Nonetheless, the separation
step(s) and isolation of pure FA from amine-based
processes are complex and energy demanding
motivating the search for alternative solutions.[2a]
Sufficient stabilization to shift the equilibrium
towards FA can also be provided by a suitable solvent
avoiding the use of an amine. For instance, water
stabilizes FA thanks to the hydrogen bonding,[14]
resulting in a major enthalpic contribution from
solute-solvent interaction. At the same time, solvation
of the reactive gases lowers the entropic penalty in
the Gibbs free energy of the transformation. Similarly,
hydrogen bonds between FA and dipolar aprotic
solvents such as dimethyl sulfoxide (DMSO) are
particularly effective in stabilizing FA.[15] Recently,
Moret, Dyson, and Laurenczy reported the
hydrogenation of CO₂ in DMSO under relatively
mild conditions obtaining FA concentrations up to the
range of 2 M, the highest FA concentration for base-
free systems currently known.[16] For the solvation
process, the thermodynamic boundaries were
Figure 2. Base-free process designed in this work using a
biphasic system (n-heptane/DMSO) and a co-solvent
(AcOH).
Herein, we present the first detailed process design
for a base-free FA synthesis based on DMSO as a
stabilizing solvent (Figure 2). The hydrogenation of
CO2 to FA has been carried out in a biphasic system
n-heptane/DMSO using Rh-catalysts with tailored
apolar phosphine ligands. Facile separation of a
catalyst-free product phase suitable for FA isolation
and catalyst reutilization has been demonstrated. For
the downstream processing, DMSO is shown to form
an azeotrope with FA. To overcome the DMSO-FA
azeotrope, the use of a co-solvent was found to be
key for isolation of FA via a single column
distillation. The use of an in silico solvent selection
procedure
via
the
quantum-mechanics-based
thermodynamic model COSMO-RS allowed the
identification of acetic acid as suitable co-solvent.
The tolerance of acetic acid in the hydrogenation step
was experimentally verified and useful FA
concentrations have been achieved. The detailed
analysis of the devised process scheme predicts an
energy demand comparable with benchmark amine-
18]
based processes.[12,
However, the insufficient
chemical stability of DMSO during the FA isolation
step has been recognized as major limitation.
Results and Discussion
Hydrogenation of CO2 to FA in a DMSO/n-alkane
biphasic system
2
This article is protected by copyright. All rights reserved.

10.1002/adsc.201801098
Advanced Synthesis & Catalysis
The hydrogenation of CO2 to FA in DMSO is an
equilibrium reaction (Scheme 1). Catalysts promoting
the conversion of CO2 to FA under hydrogen pressure
are usually capable to catalyze the back reaction, i.e.
the decomposition of FA to hydrogen and carbon
dioxide, once the overpressure is released and
moderate temperature are applied.[19]
excess of either tris(4-octylphenyl)phosphine (L1) or
trioctylphosphine (L2) to [Rh(COD)Cl]2 (COD = 1,5-
cyclooctadiene). The obtained catalyst, dissolved in
n-heptane, was loaded into an autoclave containing
DMSO. The autoclave was pressurized with CO₂
and H2, stirred and heated until constant pressure was
reached. After cooling to r.t. and depressurization, the
FA containing DMSO phase was removed via syringe
and replaced by a fresh batch of DMSO. The
autoclave was then pressurized again for a further
reaction using with the same catalyst phase. Rh- and
P-content of each collected product phase were
analyzed by ICP-MS for quantifying the catalyst
leaching (Table 1).
Scheme 1. Hydrogenation of carbon dioxide in DMSO.
For instance, when the Wilkinson catalyst
[Rh(PPh3)3Cl] is added to a 1.4 M solution of FA in
DMSO, the FA concentration decreases to 0.7 M
within 16°h at a temperature of 60 °C. As a
consequence, a distillation of FA from FA/DMSO
would result in a significant decomposition of the
synthesized FA, if the catalyst is still present in the
mixture. Currently, all systems, which use DMSO as
a solvent for FA synthesis, do not include a catalyst
removal from the product mixture.[15c]
Table 1. Rh-catalyzed hydrogenation of CO2 to FA in
n-heptane-DMSO: recycling experimentsa
Leaching
t
cFA
P
[%]
Rh
[%]
Ligand Run [h]c) TON [mol/L]
L1
L2
1b)
2
3.8
5.0
129
150
1.29
1.40
1.22
0.42
0.29
0.31
0.11
0.30
0.35
We
propose
here
a
biphasic
system
3
12.7
117
396d
DMSO/n-heptane for the hydrogenation of CO2 to FA.
This approach allows facile catalyst recycling and the
generation of a catalyst-free DMSO-FA product
phase, which is essential for an effective isolation
(Figure 3).
1.30e 1.02d 0.76d
1b)
2
0.3
1.1
129
144
1.34
1.37
1.32
1.27
1.21
0.87
0.89
0.92
0.91
0.77
0.07
0.06
0.17
0.14
0.12
3
1.9
140
4
5.0
132
5
20.3
131
675d)
1.30e) 4.36d) 0.56d)
a)
Reaction conditions: CO₂ (40 bar), H₂ (80 bar), 60
°C, [Rh(COD)Cl]2 (4.97 µmol), ligand (59.6 µmol, 6.5
eq), n-heptane (2.0 mL), DMSO (2.0 mL);
b) trioctylamine (59.6 µmol) was added; c) the gas-
uptake was monitored with a digital pressure gauge and
as soon as constant pressure was observed, the reaction
was regarded as completed; d) total amount over all
runs; e) average over all runs.
The obtained results demonstrate that the envisaged
biphasic approach is principally suited for catalyst
immobilization and recycling. The catalyst based on
the more basic alkylphosphine L2 is more active than
that based on the triphenylphosphine derivative L1.
High FA concentrations between 1.22 and 1.40 mol/L
were obtained in each run. In both cases, the catalyst
activity declines with each run, but can be
compensated by extended reaction times to adjust the
equilibrium. The leaching values indicate that more
than 99% of the metal is retained in the n-heptane
phase even after five cycles. The excess of 3.5
equivalents of the corresponding phosphorus ligand
may be responsible for a slightly higher phosphorus
leaching (cf. table 1, entry 4 and 10). Despite the high
catalyst retention, the increase of reaction time i.e. the
decrease of catalyst activity indicates a significant
Figure 3. Formation of the Wilkinson-Type pre-catalyst
(upper part) and depiction of the investigated biphasic
catalytic approach (bottom).
To ensure high solubility and effective retention of
the catalyst in n-heptane, Wilkinson-type complexes
modified with apolar phosphine ligands were used.
The pre-catalysts were formed in situ by adding an
3
This article is protected by copyright. All rights reserved.

10.1002/adsc.201801098
Advanced Synthesis & Catalysis
catalyst deactivation during the recycling experiments. Figure 4. Isobaric (p = 100 mbar) vapor-liquid equilibrium
Catalyst decomposition due to adventitious oxygen
during the recycling procedure and/or ligand
displacement by minor amounts of decomposition
products of DMSO (e.g. Me₂ S, vide infra) may be
responsible for the observed decreased activity.
Importantly, after separation of the catalyst phase by
decantation, the product mixture DMSO/FA was
heated to 60 °C over a period of 21 h. No decrease of
FA concentration was detectable furnishing the proof
that such mixture can be processed as such in the
downstream isolation step.
(VLE) for the mixture of DMSO and FA. The points show
the experimental data. The continuous lines show the
modeled vapor-liquid equilibrium with the equation of
state PC-SAFT, where the PC-SAFT parameters have been
regressed against all available experimental data. These
data include the vapor-liquid equilibrium and reaction
yields. Measurement and regression details can be found in
the electronic supporting information (ESI). Upper figure:
The diagram for the entire FA concentration range. Lower
figure: Zoom for the FA mole fraction range x/yFA = 0−
0.3.
Separation step: process design and solvent
selection
Entrainer strategy
We first focused on the FA purification from a
DMSO/FA model mixture via distillation, which is
the most applied separation technology in the
chemical industry.[20] COSMO-RS predicts the
existence of a temperature-maximum azeotrope
between DMSO and FA. As no experimental data for
the mixture of DMSO and FA are reported in
literature, we performed vapor-liquid equilibrium
measurements (fig. 4) which confirmed the existence
of the predicted azeotrope (see ESI for
experimental detail). This azeotrope must be
overcome to purify FA for any DMSO-based
process.
Initially, we considered a traditional entrainer
strategy to overcome the azeotrope.[21] As no
entrainers are known for the separation of FA and
DMSO, we employed phase-equilibrium data from
the literature[22] to identify benzene as a suitable
entrainer. Although the experimental phase-
equilibrium data gives us the confidence that the
separation will be feasible, multiple columns and
recycles are needed in the entrainer strategy because
both FA and DMSO should be recovered in pure
form (process flowsheet and process details can be
found in ESI). In alternative, assuming that the
pressure could be further increased so that the FA
yields is above the azeotrope concentration, just the
amount of FA exceeding the azeotrope composition
could be recovered using a single distillation column
and the remaining azeotropic FA/DMSO mixture
returned to the hydrogenation reactor. The use of the
azeotropic DMSO-FA mixture (~24 mol. % FA) as
reaction medium would inevitably lead to a lower
conversion of the gaseous reactants CO₂ /H₂ . A
lower conversion will in turn result in an increased
need for gas recycle and related compression cost,
since FA and CO₂ /H₂ have to be separated by
depressurization (fig. 2) rendering such a process
highly uneconomically. Multiple recycles also
complicate the operations when fluctuating H₂ is
used as raw material. Additionally, benzene is an
unfavorable entrainer due to its hazardous “nature”.
Therefore, the entrainer strategy was not pursued any
further and an alternative approach using co-solvents
was investigated for overcoming the azeotrope
DMSO/FA.
390
380
370
360
350
340
330
Measured liquid FA mole fraction
320
Measured gas FA mole fraction
Simulated (PC-SAFT) liquid FA mole fraction
310
Simulated (PC-SAFT) gas FA mole fraction
300
0,0
0,2
0,4
0,6
0,8
1,0
x,y Formic acid
394
Measured liquid FA mole fraction
Measured gas FA mole fraction
Simulated (PC-SAFT) liquid FA mole fraction
Simulated (PC-SAFT) gas FA mole fraction
Co-solvent strategy
393
392
391
390
The use of a co-solvent should allow the recovery of
all FA in pure form in a single distillation column.
The co-solvent is first added to the product feed of
DMSO and FA prior to the distillation (point 2 –
fig. 5/6). The addition of the co-solvent dilutes the
DMSO/FA feed to prevent the necessity of crossing
an azeotropic point/line, thereby allowing to obtain
FA in pure form at the top of the single distillation
column (dashed line - fig. 6). Recovery of pure FA at
the top of the distillation column (point 1 - fig. 5/6)
means that DMSO diluted with co-solvent is the
bottom product of the distillation (point 3 – fig. 5/6).
0,00
0,05
0,10
0,15
0,20
0,25
0,30
x,y Formic acid
4
This article is protected by copyright. All rights reserved.

10.1002/adsc.201801098
Advanced Synthesis & Catalysis
The bottom product of the distillation is recycled to
the reactor to avoid further distillation columns and
recycles. Therefore, in the co-solvent strategy, the FA
synthesis is performed in a mixture of DMSO and co-
solvent. Thus, in the co-solvent strategy, DMSO and
the co-solvent fulfil distinct roles: While the presence
of DMSO in the reactor sustain the FA yield, the co-
solvent enables the single-column recovery of pure
FA.
Obviously, the co-solvent strategy only works with
suitable co-solvents. Thus, we identified 12 potential
co-solvents by screening more than 5000 possible
solvents with the predictive thermodynamic model
COSMO-RS.[23] The screening criteria and list of co-
solvents can be found in the ESI. These 12 co-
solvents include carboxylic acids and aromatics such
as acetic acid and 4-fluorophenol, respectively. For
these co-solvents, experimental phase-equilibrium
data only exist for the mixtures DMSO-acetic acid
and for acetic acid-FA.[24]
and feed (Point 2) in the distillation column (cf. fig. 5).
The distillation feed (point 2) is on the side of the
distillation-boundary line that allows the recovery of pure
FA at the top of the distillation column (point 1).
Energetic evaluation of the process strategies
Objective function and assumptions
The energetic requirement for each solvent-process-
strategy combination has been calculated considering
both the electricity and the heating demand of the
entire process. To account for these two contributions
on a common basis, we measured the demand in
exergy, which is the part of energy that can be turned
into work.[25] Thus, exergy demand is the equivalent
work demand. Furthermore, exergy has also been
proposed as an assessment criterion for the
environmental impact of a process.[26] The overall
exergy demand (E) contains two contributions:
E = EEL + EQ
where the exergy demand for electricity EEl is
necessary for the pumps and compressors, and the
exergy demand for heat EQ is mainly needed for the
distillation (fig.°5). Exergy calculation details can be
found in the ESI.
To obtain the exergy demand, we performed
process simulations with Aspen Plus v.8.4 using
rigorous unit-operation models. We minimized the
exergy demand by optimizing the pressure in the
reactor and the amount of co-solvent added to the
process. The catalyst phase is not considered in the
simulation due to the low mutual solubility between
the phases; E.g. there is a liquid-liquid equilibrium
between FA and n-heptane.[27]
Figure 5. Process flowsheet of the co-solvent process
strategy. Compressed CO2 and H2 enter the reactor
together with the recycled DMSO/co-solvent mixture.
After the reactor, the unreacted gases are recovered by
depressurization and recycled (not shown for clarity)
before FA is purified by distillation. From the bottom of
the distillation column, the DMSO/co-solvent mixture is
recovered and recycled. The arrows show where electricity
(EEl) and heat (EQ) are added to the process. The numbers
1, 2 and 3 allow to relate the positions in the distillation
column to the residue-line diagram (fig. 6).
Two accuracy levels were used in the simulations:
On the first level (filled points in fig. 7), the FA yield
is obtained by assuming that the reaction equilibrium
is reached.[1b, 28] For this simulation level, all pure
component thermodynamic data (vapor pressure, heat
capacity, enthalpy of vaporization) are taken from the
Aspen Plus database, which contains experimental
data. The mixture thermodynamic data (activity
coefficient, Henry coefficient) are taken from
29]
literature if available[24,
or, if not, they are
Figure 6. Qualitative residue-line diagram for the ternary
mixture of DMSO, FA and an example co-solvent. The
dashed line connects the top (Point 1), bottom (Point 3)
calculated with COSMO-RS[30]. On the basis of these
data, a first level exergy demand was predicted. On
this level, 5000 co-solvents have been simulated and
also the entrainer-based process for comparison. One
of the lowest exergy demands was obtained for the
co-solvent acetic acid (filled circle, fig.°7). Acetic
acid is non-toxic, available at commercial scale and,
thus, a convenient possible co-solvent. Moreover,
previous studies have shown that the presence of a
small portion (up to 1.33 mol/L) of AcOH and water
(5 mol %) in DMSO can lead to improved catalyst
performance in the hydrogenation of CO2 as
compared to neat DMSO.[15c]
The DMSO/acetic-acid process was then selected
to be simulated on a second more accurate level
5
This article is protected by copyright. All rights reserved.

10.1002/adsc.201801098
Advanced Synthesis & Catalysis
(open symbols in fig. 8). On this level, the modelling
of the gas solubility is improved by employing the
thermodynamic model PC-SAFT in the simulation of
the reaction. Here, the PC-SAFT parameters that
influence the hydrogen bonding in the mixture (e.g.
kij) are regressed against available experimental
reaction yield and vapor-liquid equilibrium data[24, 29]
(including measurements from this work). To cross
the distillation boundary line at a practically useful
pressure of 100 mbar, a mixture is necessary with a
reduces the exergy demand by 45 % compared to the
entrainer strategy using benzene (triangles, fig. 7).
Interestingly, the second level process simulation
based on experimental mixture data results in an even
lower exergy demand for the DMSO/acetic-acid
process than the first level calculation. The reason for
the exergy reduction is that the second level process
simulation uses the experimental reaction yield which
is higher than the yield predicted with COSMO-RS at
the first level (fig. 7), since COSMO-RS
underestimates the hydrogen-bond strength between
DMSO and FA. This underestimation can be seen by
ratio nDMSO:nAcOH below 7.2 : 1. Using a nDMSO:nAcOH
:
ratio of 5 : 1 and the Ru/PTA complex reported by
Moret et al. as the catalyst under standard conditions
(CO2: 50 bar, H2: 50 bar, T = 60 °, t = 120 h), a FA
concentration of 0.85 mol/L was obtained. As
correctly predicted by COSMO-RS, the presence of
acetic acid halves the FA concentration, i.e. from
1.90 mol/L to 0.85 mol/L (fig. 8).
comparing the measured enthalpy of mixing 푯퐞퐌퐱퐩
=
[17]
⁄
ퟏퟐ 퐤퐉 퐦퐨퐥
with the COSMO-RS predicted
퐌
⁄
enthalpy of mixing 푯퐞퐱퐩 = ퟐ. ퟓ 퐤퐉 퐦퐨퐥. Therefore,
a similar decrease in the exergy demand, as observed
for the co-solvent acetic acid, is also likely for other
co-solvents suggesting that the co-solvent strategy is
promising in general. The entire list of identified co-
solvents can be found in the ESI.
10
Noteworthy, the co-solvent strategy with
DMSO/acetic acid has the lowest exergy demand E,
not because the exergy demand for electricity EEl or
for heat EQ are, singularly taken, the lowest ones, but
because the co-solvent acetic acid leads to best
compromise between the exergy demands for both
contributions. The exergy demand for electricity EEl is
required to increase the yield, since electricity is
necessary for compressors and pumps (fig. 5) that
elevate the pressure in the reactor. Similarly, the
exergy demand for heat EQ is required for FA
purification, since heat is added in the distillation
column. Thus, the co-solvent process based on acetic
acid has a low total exergy demand because acetic
acid balances the trade-off between increasing the
yield (EEl) and the FA purification (EQ). This trade-off
is reached by using the strengths of both DMSO and
the co-solvent: DMSO facilitates the highest known
base-free FA yield, while acetic acid allows efficient
purification in a single distillation column (fig. 5/6).
To avoid excessive dilution of DMSO, the amount of
co-solvent is determined by the position of the binary
DMSO/co-solvent azeotrope (fig. 6). Thus, the
position of this binary azeotrope seems to be key with
regard to the determination of the co-solvent's exergy
demand. These complex relationships highlight the
need for a process-level analysis to identify efficient
solvents.
It has to be stressed that single column purification
requires that the co-solvent acetic acid must be
recycled together with DMSO to the hydrogenation
reactor (fig. 5/6). In order to fulfil this requirement
the catalyst system has to be compatible with AcOH
in the two phase system. This prerequisite has been
verified using L2 as the ligand in n-heptane/DMSO-
AcOH (again with a nDMSO:nAcOH ratio of 5 : 1) under
otherwise identical conditions as in Table 1. In this
experiment, a FA concentration of 1.10 mol/L was
obtained, a value just slightly lower than that
achieved in the absence of AcOH (1.30 mol/L). The
obtained FA concentration of 1.10 mol/L suggests
that an even more favorable exergy balance might be
/
/
Entrainer strategy with benzene
Co-solvent strategy with DMSO
NHex3-using benchmark process
Iso exergy lines Heat
1
Co-solvent acetic acid
0,1
0,1
1
10
EEl / kWh/l Formic acid
Figure 7. Exergy demand of identified solvent-strategy
combinations and benchmarks. Each point is the exergy
demand of a DMSO/co-solvent combination or benchmark
in their respective strategy. The lines are iso-exergy-
demand lines. The filled points show the exergy demand
from the first level that employs predictions with COSMO-
RS. The hollow points show the exergy demand obtained
from the more accurate second level process simulations
employing experimental mixture data. Calculation details
can be found in the ESI.
For comparison, both the entrainer-using process and
the base-using process by the BASF have been
simulated on the same level of detail. All data sources
and further details regarding the simulations can be
found in the ESI.
Exergetic ranking of solvent-process strategy
combinations
The lowest exergy demand was achieved by the co-
solvent strategy that utilizes a mixture of DMSO and
acetic acid (fig. 7). This process uses DMSO and the
co-solvent acetic acid in an optimized molar ratio of
nDMSO : nAcOH = 7 : 1. The DMSO/acetic-acid process
6
This article is protected by copyright. All rights reserved.

10.1002/adsc.201801098
Advanced Synthesis & Catalysis
achieved than calculated with the conservative FA
concentration of 0.85 mol/L.
Finally, we compared the DMSO/acetic acid co-
solvent strategy with the BASF process.[12] This
process uses the base trihexylamine (NHex3) and a
used. The distillation was carried out with an oil bath
temperature of 115 °C at a pressure of 100 mbar.
Before the vacuum pump, a cooling trap was
employed. Two fractions were collected (fraction 1:
130 mg, Thead = 42 °C; fraction 2: 1.879 g, Thead
=
FA-NHex3 adduct is formed upon CO2 hydrogenation. 48 °C) containing a total of approx. 1.0 mol% FA
Later, the FA-NHex3 adduct is thermally split and
two distillation columns are used for purifying FA
(details in the ESI).[18] Thus, the co-solvent strategy
reduces the necessary equipment and simplifies the
process (cf. fig. 5), by removing one distillation
column as well the catalyst extraction step carried out
in a dedicated high pressure apparatus (cf. fig. 1).
Moreover, the exergy demand of this NHex3-using
process was found by simulations to be 20 % higher
than exergy demand of the co-solvent strategy with
DMSO and the co-solvent acetic acid (fig. 8).
However, it should be noted that there are
uncertainties connected to the simulations of the
NHex3-using process due to missing experimental
data. Nevertheless, our findings support the
perception that processes based on solvent-
stabilization have the potential to be energetically
competitive with processes using base-stabilization
for FA synthesis.
from the initial 7 mol%. Both fractions contained no
DMSO confirming that the predicted boundaries
could be overcome and DMSO-free FA could be
retrieved as top product in the distillation (cf. fig. S8).
However,
a variety of DMSO decomposition
products were observed in all fractions and in the
cooling trap. The S-containing species MeSH, Me2S2,
2,4-dithiapentane were identified as major
components by GC-MS (ESI, fig. S9 to S14). Indeed,
during the distillation, a strong sulfide smell was
perceived.[31] The decomposition of DMSO has been
topic of previous investigations and several pathways
lead to the observed decomposition products.[32]
While highly pure DMSO can be heated to its boiling
point of 189 °C with only minor decomposition,
small impurities can lead to an increased
decomposition.[33] In the present case, the acidic
conditions strongly facilitate the decomposition of
DMSO (Scheme 2).
[16]
Scheme 2. Possible decomposition pattern of DMSO.
No decomposition products were present in the
bottom liquid (cf. figure S4) probably because of the
high volatility of the formed compounds. Thus, the
mass balance could not be closed and no extra effort
was undertaken.
Figure 8. Comparison of the experimentally measured FA
yield (blue), with the predictions from COSMO-RS (red)
and regression with PC-SAFT (black). The error bars Conclusion
originate from reproducibility. The PC-SAFT parameters
were regressed against both VLE data and the reaction
yields.
A process for the synthesis and isolation of
formic acid from CO2/H2 was developed. The
synthetic step relies on the metal-catalyzed
hydrogenation of CO2 in the presence of DMSO
as stabilizing medium for FA. A biphasic system
DMSO/n-heptane with tailored Wilkinson
Rh-complexes was devised allowing for catalyst
immobilization and metal-free product phase.
Since DMSO/FA is shown to form an azeotropic
mixture, simple distillation is not a viable option.
Both an entrainer and a co-solvent strategy were
considered for the isolation of FA via distillation.
The energy demand (exergy) for the most
promising systems was computed in detail.
Eventually, the co-solvent strategy based on
acetic acid resulted in the lowest exergy demand.
FA Isolation via Distillation
To experimentally prove the isolation of FA
according to the proposed co-solvent approach, we
performed a batch distillation experiment with a
model product phase mixture consisting of 7 mol-%
FA, 76 mol-% DMSO and 16 mol-% acetic acid
(cf. fig. 8). The distillation was carried out with a
laboratory distillation equipment using a 250 mL
round bottom flask containing 210 mL / 216 g of the
model product phase mixture (7 mol-% FA, 76 mol-
% DMSO and 16 mol-% AcOH). A 45 cm long filled
column equipped with 0.8×0.8 mm glass rings was
7
This article is protected by copyright. All rights reserved.

10.1002/adsc.201801098
Advanced Synthesis & Catalysis
These results were achieved even assuming the
most conservative experimental value of FA
concentration of 0.85 mol/L obtained in a
monophasic DMSO-AcOH mixture (nAcOH:nDMSO
ratio of 1:5) with a literature known catalyst.
Notably, higher FA concentration of 1.10 mol/L
was obtained in the hydrogenation step using the
biphasic system n-heptane/DMSO-AcOH and L2
as the ligand rendering the co-solvent strategy
even more competitive. The distillation of a
representative product mixture containing DMSO,
acetic acid and FA confirmed that DMSO-free
FA can be obtained. However, extensive
decomposition of DMSO was observed as well.
Thus, although the co-solvent process described
here theoretically provides even higher efficiency
compared to the state-of-the-art amine-based
process, the insufficient chemical stability of
DMSO during the distillation step presents a
major limitation for the downstream processing.
In the light of these results, alternatives to DMSO
as stabilizing solvent for the hydrogenation of
carbon dioxide to FA need to be identified.
(300 MHz, CDCl3, 25°C): δ = 7.26 - 7.17 (m, 2H, Ar-H),
7.16 - 7.09 (m, 2H, Ar-H), 2.50 (t, 3JHH = 8.0 Hz, 2H, CH2),
1.70 - 1.49 (m, 2H, CH2), 1.37 - 1.17 (m, 10H, CH2), 0.87
3
(t, JHH = 6.6 Hz, 3H, CH3) ppm. 13C-NMR (100 MHz,
1
CDCl3, 25°C): δ = 143.5 (s, Ar-Cq), 134.4 (d, JCP = 9.2
2
Hz, Ar-Cq), 133.7 (d, JCP = 19.5 Hz, Ar-CH), 128.6 (d,
3JCP = 7.1 Hz, Ar-CH), 31.9 (s, CH2), 31.4 (s, CH2), 29.5
(s, CH2), 29.4 (s, CH2), 29.3 (s, CH2), 22.7 (s, CH2), 14.2 (s,
CH3) ppm. 31P-NMR (160 MHz, CDCl3, 25°C): δ = 7.8 (s,
PAr) ppm.
Catalytic Hydrogenation of CO2 in DMSO/n-heptane:
In a Schlenk-tube, a catalyst stock solution was prepared
by suspending [Rh(COD)Cl]2 (49.7 µmol) in n-heptane
(10 mL) and adding L1 or L2 (6.5 equivalents). The
mixture was stirred for 15 to 30 min at room temperature.
In a typical experiment, the catalyst stock solution (2 mL)
was transferred into a 10 mL window autoclave and
DMSO (2 mL) was added (in the first experiment of a
recycling series NOct₃ (59.6 µmol) was added). The
autoclave was pressurized at r.t. under stirring with CO2
(40 bar) and H2 (80 bar) to a total pressure of 120 bar. The
autoclave was placed on a heating plate (60 °C) and stirred.
The pressure was monitored with a digital pressure gauge.
When no more pressure decrease was observed, the
autoclave was cooled to r.t. and the pressure was released
to approx. 1-2 bar overpressure. The remaining
overpressure was released under the use of Schlenk-
technique. The product containing DMSO phase was
withdrawn via syringe and weighed. 1,4-Dioxane or
mesitylene was added as a reference and a sample of the
Experimental Section
General Information: All procedures using air-sensitive
compounds were carried out under an inert argon
atmosphere and application of standard Schlenk-technique.
Unless otherwise noted, all chemicals were obtained from
commercial sources and used as supplied. NMR-Spectra
were recorded on an AV-400 or a DPX-300 (Bruker).
Chemical shifts are reported as δ in ppm relative to
tetramethylsilane (1H and 13C) and 85% phosphoric acid
(31P). Hereby, the residual solvent peak of the
corresponding deuterated solvents was used as internal
standard for 1H and 13C NMR spectra. GC-MS
measurements were carried out using a Variant 500-MS by
Agilent with an 60m CP-WAX-52-CB Column. ICP-MS
was carried out on an 8800 ICP-MS Triple Quad by
Agilent with water as matrix.
1
product phase was analyzed by H-NMR spectroscopy. A
fresh batch of DMSO was loaded into the autoclave and
pressurized described as above. This procedure was
repeated until a significant decrease of catalyst activity was
observed i.e. until the pressure decrease was significantly
slower.
Acknowledgements
The work has been carried out within the project “Carbon2Chem”
(03EK3042C). The project was funded by the German Federal
Ministry of Education and Research (BMBF). Kai Leonhard
thanks the “Tailor-Made Fuels from Biomass” cluster of
excellence for financial support. Furthermore, the authors want
to thank Sebastian Kaminski for valuable discussions, Sonja
Kröngen und Jannik Häßy for assistance with the vapor-liquid-
equilibrium experiments and Julia Nießen for the ICP-
measurements.
Synthesis of tris(4-octylphenyl)phosphine, L1:[34]
A
100 mL Schlenk flask equipped with a magnetic PTFE bar
was loaded with magnesium turnings (240 mg, 10.0 mmol,
4.0 eq) and dried in vacuo while stirring vigorously, hereby
activating the magnesium. THF (8 mL) was added.
1-Bromo-4-octylbenzene (2.16 g, 8.00 mmol, 1.90 mL,
3.2 eq) was added dropwise and the reaction mixture was
stirred for 1.5 h at room temperature. During this time, the
colorless suspension underwent a color change to dark
green. At 0 °C phosphorous trichloride (0.343 g,
2.50 mmol, 0.220 mL, 1.0 eq) was added slowly. The
reaction mixture was stirred overnight at r.t. while the dark
green suspension turned yellow. The mixture was
quenched with a saturated degassed aqueous ammonium
chloride solution (10 mL). After extracting with n-pentane
(5 × 5 mL) the reaction mixture was purified via filter
column loaded with aluminum oxide and silica. After
removing the solvent in vacuo the product L1 was obtained
References
[1]
[2]
a) W. Reutemann, H. Kieczka, in Ullmann's
Encyclopedia of Industrial Chemistry, 3. ed., Wiley-
VCH, Weinheim, 2016, pp. 1-22; b) J.
Klankermayer, S. Wesselbaum, K. Beydoun, W.
Leitner, Angew. Chem. Int. Edit. 2016, 55, 7296–
7343.
a) M. Aresta, A. Dibenedetto, A. Angelini,
Chemical Reviews 2013, 114, 1709-1742; b) M.
1
as pale yellow oil (1.33 g, 2.23 mmol, 89%). H-NMR
8
This article is protected by copyright. All rights reserved.

10.1002/adsc.201801098
Advanced Synthesis & Catalysis
Pérez-Fortes, J. C. Schöneberger, A. Boulamanti, G.
Harrison, E. Tzimas, Int. J. Hydrogen Energy 2016,
41, 16444-16462.
[18] T. Schaub, D. M. Fries, R. Paciello, K. D. Mohl, M.
Schäfer, S. Rittinger, D. Schneider, United States
Patent 2014, US, 8791297.
[3]
a) A. Sternberg, C. M. Jens, A. Bardow, Green
Chem. 2017, 19, 2244-2259; b) J. Artz, T. E. Muller,
K. Thenert, J. Kleinekorte, R. Meys, A. Sternberg,
A. Bardow, W. Leitner, Chem. Rev. 2018, 118, 434-
504.
[19] a) T. C. Johnson, D. J. Morris, M. Wills, Chem. Soc.
Rev. 2010, 39, 81-88; b) M. Grasemann, G.
Laurenczy, Energy Environ. Sci. 2012, 5, 8171–
8181.
[20] Ö. Yildirim, A. A. Kiss, E. Y. Kenig, Sep. Purif.
Technol. 2011, 80, 403-417.
[21] J. Stichlmair, Chem. Ing. Tech. 1988, 60, 747–754.
[22] a) L. Negadi, D. a. A.-K. Vellut, A., J. Jose,
ELDATA: The International Electronic Journal of
Physico-Chemical Data 1998, 4, 173–178; b) R.
Parashar, M. A. Ali, S. K. Mehta, J. Chem.
Thermodyn. 2000, 32, 711–716.
[23] a) A. Klamt, G. Schüürmann, J. Chem. Soc., Perkin
Trans. 2 1993, 0, 799-805; b) A. Klamt, V. Jonas, T.
Bürger, J. C. W. Lohrenz, J. Phys. Chem. A 1998,
102, 5074-5085.
[4]
[5]
N. von der Assen, L. J. Müller, A. Steingrube, P.
Voll, A. Bardow, Environmental Science
Technology 2016, 50, 1093-1101.
&
a) W. Leitner, Angew. Chem. Int. Ed. 1995, 34,
2207-2221; b)Y. Himeda, European Journal of
Inorganic Chemistry 2007, 2007, 3927-3941; c) K.
Müller, K. Brooks, T. Autrey, Energy & Fuels 2017,
31, 12603-12611.
a) K. Sordakis, A. Tsurusaki, M. Iguchi, H.
Kawanami, Y. Himeda, G. Laurenczy, Green Chem.
2017, 19, 2371-2378; b) J. Eppinger, K.-W. Huang,
ACS Energy Letters 2016, 2, 188-195.
X. W. Yu, P. G. Pickup, J. Power Sources 2008,
182, 124-132.
K. Sordakis, C. Tang, L. K. Vogt, H. Junge, P. J.
Dyson, M. Beller, G. Laurenczy, Chem. Rev. 2018,
118, 372-433.
[6]
[24] S. Miyamoto, S. Nakamura, Y. Iwai, Y. Arai, J.
Chem. Eng. Data. 2001, 46, 405–409.
[25] M. J. Moran, H. N. Shapiro, D. D. Boettner, M. B.
[7]
[8]
Bailey,
Fundamentals
of
Engineering
Thermodynamics, 7. ed., John Wiley & Sons,
Hoboken, 2010.
[9]
C. M. Jens, K. Nowakowski, J. Scheffczyk, K.
Leonhard, A. Bardow, Green Chem. 2016, 18,
5621-5629.
[26] a) J. Dewulf, M. E. Bösch, B. D. Meester, G. van
der Vorst, H. van Langenhove, S. Hellweg, M. A. J.
Huijbregts, Environ. Sci. Technol. 2007, 41, 8477–
8483; b) A. Zvolinschi, S. Kjelstrup, O. Bolland, H.
J. van der Kooi, J. Ind. Econ. 2007, 11, 85–98; c) M.
E. Bösch, S. Hellweg, M. A. J. Huijbregts, R.
Frischknecht, Int. J. Life Cycle Ass. 2007, 12, 181–
190.
[27] T. Horlenko, United States Patent 1973, US,
3718545.
[28] J. de Vries, C. J. Elsevier, in The Handbook of
Homogeneous Hydrogenation, 1. ed., Wiley-VCH,
Weinheim, 2008, pp. 489-511.
[29] a) J. O. Spano, C. K. Heck, P. L. Barrick, J. Chem.
Eng. Data. 1968, 13, 168–171; b) A. Jónasson, O.
Persson, P. Rasmussen, G. S. Soave, Fluid Phase
Equilib 1998, 152, 67–94; c) C. Hung Yu, J. Rong
Feng, L. Ming Jer, L. Ho Mu, J. Supercrit. Fluid.
2008, 44, 273–278; d) Y. Peng, L. Ping, S. Lu, J.
Mao, Fluid Phase Equilib 2009, 275, 27–32.
[30] a) C. M. Jens, K. Nowakowski, J. Scheffzcyk, K.
Leonhard, A. Bardow, Green Chem 2016, 18,
5621–5629; b) J. Scheffczyk, C. Redepenning, C. M.
Jens, B. Winter, K. Leonhard, W. Marquardt, A.
Bardow, Chem. Eng. Res. Des. 2016, 115, 433-442.
[31] These types of compounds are commonly used as
olfactory leak indicators for gas pipe and can be
perceived by the human nose at very low
concentrations.
[32] a) W. J. Kenney, J. A. Walsh, D. A. Davenport, J.
Am. Chem. Soc. 1961, 83, 4019-4022; b) V. J.
Traynelis, W. L. Hergenrother, J. Org. Chem. 1964,
29, 221-222; c) D. L. Head, C. G. McCarty,
Tetrahedron Letters 1973, 14, 1405-1408; d) B. T.
Brandes, D. K. Smith, Process Saf. Prog. 2016, 35,
374-391.
[10] a)Y. Himeda, Green Chem. 2009, 11, 2018-2022; b)
A. Boddien, F. Gärtner, D. Mellmann, P. Sponholz,
H. Junge, G. Laurenczy, M. Beller, CHIMIA
International Journal for Chemistry 2011, 65, 214-
218; c) A. Sternberg, A. Bardow, Energy Environ.
Sci. 2015, 8, 389-400; dP. Losch, A.-S. Felten, P.
Pale, Advanced Synthesis & Catalysis 2015, 357,
2931-2938.
[11] W. Supronowicz, I. A. Ignatyev, G. Lolli, A. Wolf,
L. Zhao, L. Mleczko, Green Chem 2015, 17, 2904–
2911.
[12] T. Schaub, R. A. Paciello, Angew. Chem. Int. Edit.
2011, 50, 7278–7282.
[13] a) J. Kothandaraman, A. Goeppert, M. Czaun, G. A.
Olah, G. K. S. Prakash, Green Chem. 2016, 18,
5831-5838; b) M. Scott, B. Blas Molinos, C.
Westhues, G. Franciò, W. Leitner, ChemSusChem
2017, 10, 1085-1093.
[14] S. M. Lu, Z. J. Wang, J. Li, J. L. Xiao, C. Li, Green
Chem. 2016, 18, 4553-4558.
[15] a) E. Graf, W. Leitner, Journal of the Chemical
Society, Chemical Communications 1992, 623-624;
b) W. Leitner, E. Dinjus, F. Gassner, J. Organomet.
Chem. 1994, 475, 257–266; c) K. Rohmann, J.
Kothe, M. W. Haenel, U. Englert, M. Hölscher, W.
Leitner, Angew. Chem. Int. Ed. 2016, 55, 8966-
8969; d) N. V. Kolesnichenko, N. N. Ezhova, E. V.
Kremleva, E. V. Slivinskii, Russian Chemical
Bulletin 2004, 53, 2542-2545.
[16] S. Moret, P. J. Dyson, G. Laurenczy, Nat Commun
2014, 5, 4017.
[17] C. Fink, S. Katsyuba, G. Laurenczy, PCCP 2016,
18, 10764-10773.
9
This article is protected by copyright. All rights reserved.

10.1002/adsc.201801098
Advanced Synthesis & Catalysis
[33] Y. Babasaki, Y. Iizuka, A. Miyake, Journal of
Thermal Analysis and Calorimetry 2015, 121, 295-
301.
[34] S. Franks, F. R. Hartley, J. Chem. Soc., Perkin
Trans. 1 1980, 2233-2237.
10
This article is protected by copyright. All rights reserved.

10.1002/adsc.201801098
Advanced Synthesis & Catalysis
FULL PAPER
Catalytic Hydrogenation of CO2 to Formic Acid:
Solved and Unsolved Challenges for Process
Development
Adv. Synth. Catal. Year, Volume, Page – Page
Christian M. Jens,a# Martin Scott,b# Bastian
Liebergesell,a Christian G. Westhues,b Pascal
Schäfer,a Giancarlo Franciò,b Kai Leonhard,a
Walter Leitnerb,c* and André Bardowa*
11
This article is protected by copyright. All rights reserved.