Carbon dioxide utilization via carbonate-promoted C-H carboxylation

Article abstract of DOI:10.1038/nature17185

Using carbon dioxide (CO₂) as a feedstock for commodity synthesis is an attractive means of reducing greenhouse gas emissions and a possible stepping-stone towards renewable synthetic fuels. A major impediment to synthesizing compounds from CO₂ is the difficulty of forming carbon-carbon (C-C) bonds efficiently: although CO₂ reacts readily with carbon-centred nucleophiles, generating these intermediates requires high-energy reagents (such as highly reducing metals or strong organic bases), carbon-heteroatom bonds or relatively acidic carbon-hydrogen (C-H) bonds. These requirements negate the environmental benefit of using CO₂ as a substrate and limit the chemistry to low-volume targets. Here we show that intermediate-temperature (200 to 350 degrees Celsius) molten salts containing caesium or potassium cations enable carbonate ions (CO₃^2-) to deprotonate very weakly acidic C-H bonds (pK_a > 40), generating carbon-centred nucleophiles that react with CO₂ to form carboxylates. To illustrate a potential application, we use C-H carboxylation followed by protonation to convert 2-furoic acid into furan-2,5-dicarboxylic acid (FDCA) - a highly desirable bio-based feedstock with numerous applications, including the synthesis of polyethylene furandicarboxylate (PEF), which is a potential large-scale substitute for petroleum-derived polyethylene terephthalate (PET). Since 2-furoic acid can readily be made from lignocellulose, CO₃^2--promoted C-H carboxylation thus reveals a way to transform inedible biomass and CO₂ into a valuable feedstock chemical. Our results provide a new strategy for using CO₂ in the synthesis of multi-carbon compounds.

Full text of DOI:10.1038/nature17185

LETTEr
doi:10.1038/nature17185
Carbon dioxide utilization via carbonate-promoted
C–H carboxylation
aanindeeta Banerjee1, Graham r. Dick1, Tatsuhiko Yoshino1† & Matthew W. Kanan1
Using carbon dioxide (CO2) as a feedstock for commodity synthesis
Carbonate-promoted C–H carboxylation is particularly desirable
is an attractive means of reducing greenhouse gas emissions and a for the synthesis of polymer units from biomass. A longstanding goal
possible stepping-stone towards renewable synthetic fuels1,2. A major of renewable polymer chemistry is a scalable synthesis of FDCA from
impediment to synthesizing compounds from CO2 is the difficulty lignocellulose to replace petroleum-derived terephthalic acid for
of forming carbon–carbon (C–C) bonds efficiently: although polyester synthesis (Fig. 1b)6. In particular, PEF has been reported
CO2 reacts readily with carbon-centred nucleophiles, generating to have superior physical properties to PET, a polymer produced at
these intermediates requires high-energy reagents (such as highly 50 megatons a year as an industrial commodity8. Current approaches
reducing metals or strong organic bases), carbon–heteroatom to FDCA synthesis use dehydration processes to convert hexose sugars
bonds or relatively acidic carbon–hydrogen (C–H) bonds3–5
.
into hydroxymethyl furfural (HMF), which is then oxidized to form
These requirements negate the environmental benefit of using FDCA17. Recent work has improved the efficiency of converting fruc-
CO2 as a substrate and limit the chemistry to low-volume targets. tose to HMF8,18 and it has been estimated that PEF production from
Here we show that intermediate-temperature (200 to 350 degrees fructose would emit 50% less CO2 than the established process for PET
Celsius) molten salts containing caesium or potassium cations production7. However, producing FDCA on a scale commensurate
enable carbonate ions (CO32–) to deprotonate very weakly acidic with terephthalic acid and achieving maximal reduction in CO2 emis
-
C–H bonds (pKa >40), generating carbon-centred nucleophiles sions will require using lignocellulose as the feedstock. Converting
that react with CO2 to form carboxylates. To illustrate a potential lignocellulose into HMF is very challenging because the hexoses are
application, we use C–H carboxylation followed by protonation to incorporated into intractable cellulose fibres17. Researchers have
convert 2-furoic acid into furan-2,5-dicarboxylic acid (FDCA)—a recently reported a high-yielding conversion of lignocellulose to HMF
highly desirable bio-based feedstock6 with numerous applications, in ionic liquids19,20 and an efficient process that converts lignocellulose
including the synthesis of polyethylene furandicarboxylate (PEF), into sugar monomers21. Despite these advances, an economical, large-
which is a potential large-scale substitute for petroleum-derived scale lignocellulose-to-HMF process has not been demonstrated17,22
.
polyethylene terephthalate (PET)7,8. Since 2-furoic acid can readily In contrast, the conversion of lignocellulose to furfural is a decades-old
be made from lignocellulose9, CO32–-promoted C–H carboxylation industrial process that is currently performed on a scale of about 400
thus reveals a way to transform inedible biomass and CO2 into a kilotons a year9,22. Furthermore, several methods are available for
valuable feedstock chemical. Our results provide a new strategy for oxidizing furfural to 2-furoic acid23,24. At present, however, the avail-
using CO2 in the synthesis of multi-carbon compounds.
able methods for converting 2-furoic acid into FDCA are inefficient,
The chemistry described here is inspired by ribulose-1,5-bisphosphate unselective, and consume stoichiometric amounts of energy-intensive
carboxylase/oxygenase (RuBisCO), which effects C–C bond formation reagents25,26. CO32–-promoted C–H carboxylation could be used to
in the Calvin cycle by deprotonating a C–H bond of ribulose-1,5- convert 2-furoic acid into FDCA and thereby open a new route to
bisphosphate and exposing the resulting carbon-centred nucleophile PEF using a lignocellulose-derived monomer that is already produced
– 10
to CO2 to form a carboxylate (C–CO2 ) . Emulating this strategy industrially.
synthetically requires deprotonating un-activated C–H bonds using a
The CO32–-promoted C–H carboxylation reaction required
simple base that does not have a large CO2 footprint. To meet these for FDCA synthesis is the conversion of furan-2-carboxylate into
requirements, we envisioned a CO32–-promoted C–H carboxylation furan-2,5-dicarboxylate (FDCA2–) (Fig. 2a). Assuming it is similar
reaction, wherein CO32– reversibly deprotonates a C–H bond to gen- to an un-substituted furan27, the pKa of the C–H at the 5 position of
erate HCO3– and a carbon-centred nucleophile that reacts with CO2 to furan-2-carboxylate is ~35. Deprotonation of this C–H has previously
form C–CO2– (Fig. 1a). HCO3– decomposition results in a net consump- required lithium diisopropylamide or n-butyllithium26. We hypothe-
tion of one-half equivalents of CO32– and CO2 per C–CO2– produced. sized that CO32– would deprotonate furan-2-carboxylate if the reaction
–
The cycle could be closed by protonating C–CO2 with strong acid and were performed in a molten salt with a high concentration of alkali cat-
using electrodialysis to regenerate the acid and base11,12, effecting a ions to stabilize the conjugate base by ion pairing. To test this hypoth-
net transformation of C–H and CO2 into C–CO2H without using any esis, we attempted C–H carboxylation with mixtures consisting of an
other stoichiometric reagents. Alternatively, CO2-promoted esterifica- alkali metal salt of 2-furoic acid and an alkali metal carbonate. With
tion could be used to convert the carboxylate into an ester (C–CO2R) these components, the reaction was found to proceed efficiently when
and regenerate CO32– directly13. Previously, researchers have shown that caesium ions (Cs+) were used (Fig. 2a and Extended Data Table 1a).
Cs2CO3 can deprotonate alkynyl14, allylic15, and activated heteroaryl When 1mmol of caesium furan-2-carboxylate and 0.55mmol Cs2CO3
C–H bonds with pKa values of up to 27 (ref. 16) in organic solvents at were heated at 260°C under a CO2 flow of 40mlmin–1 in a tube furnace,
elevated temperature (where pKa is the negative base-10 logarithm of FDCA2– was formed in 76% yield after 12h (Extended Data Fig. 1).
the acid dissociation constant). However, using CO32–-promoted C–H The mass balance was composed of unreacted starting material and
carboxylation for commodity synthesis requires deprotonating C–H decomposition products including acetate. Reactions performed
bonds that are considerably less acidic.
in a Parr reactor showed improved yields and less decomposition.
1Department of Chemistry, Stanford University, Stanford, California, USA.
†Present address: Faculty of Pharmaceutical Sciences, Hokkaido University, Kita-ku, Sapporo 060-0812, Japan.
1 0 M a r c h 2 0 1 6
| V O L 5 3 1 | N a T U r E | 2 1 5
© 2016 Macmillan Publishers Limited. All rights reserved

reSeArCH Letter
a
C–CO2M
½ H2O
HCl
CO32– - promoted C–H carboxylation
ROH
½ CO2
C–H + M2CO3
M+C– + MHCO3
M+C– + CO2
C–CO2M
C–H
½ CO2
C–CO2H
MHCO3
½(M2CO3 + CO2 + H2O)
C–CO2R
½ H2O
Protonation and CO32– regeneration
MCl
½ M2CO3
C–CO2H + MCl
C–CO2M + HCl
Electrodialysis
MCl + H2O
HCl + MOH
MOH + ½ CO2
½(M2CO3 + H2O)
HCl
½(CO2 + H2O)
CO2 - promoted esterification
Net transformations
C–CO2R + MHCO3
C–CO2H
C–CO2M + CO2 + ROH
C–H + CO2
MHCO3
½(M2CO3 + CO2 + H2O)
C–CO2R + H2O
C–H + CO2 + ROH
b
Fructose
O
O
O
O
O
Difficult
Oxidation
HO
OH
HO
OH
FDCA
HMF
O
Lignocellulose
Industrial
process
O
O
O
Oxidation
O
O
O
CO2
HO
H
OH
OH
Furfural
2-furoic acid
FDCA
Figure 1 | CO2 utilization cycle. a, C–H deprotonation by CO32– forms
a carbon-centred nucleophile (C–) that reacts with CO2 to form a C–C
bond (red box and arrows). Protonation and electrodialysis (black box
and arrows) produces a carboxylic acid. Alternatively, CO2-promoted
esterification (grey box and arrows) produces an ester. Net transformations
are shown in the blue box. b, Potential application for FDCA synthesis
from lignocellulose. Current approaches rely on converting lignocellulose
into HMF, a difficult reaction that has not been commercialized.
C–H carboxylation enables the use of furfural, which has been produced
industrially from lignocellulose for many decades. M, metal; R, alkyl group.
In 1-mmol-scale reactions at 200°C under a pressure of 8bar CO2, caesium furan-2-carboxylate, 1 mmol CD3CO2Cs, and 1.1 mmol
FDCA2– was formed in a 77% yield after 2h and 89% yield after 5h, Cs2CO3 was heated under N2 in the Parr reactor to 200 °C for 1 h.
with only 5% decomposition products. In 10-mmol-scale reactions 1H, 2H, and 13C nuclear magnetic resonance (NMR) spectra of the
under similar conditions, FDCA2– was formed in a 78% yield after 5h crude product mixture showed hydrogen/deuterium (H/D) scram-
and 81% yield after 10h (Extended Data Fig. 3). Further increasing bling between acetate and the 5-position of furan-2-carboxylate and,
the reaction time did not appreciably increase the FDCA2– yield, while to a lesser extent, the 3 and 4 positions (Extended Data Figs 6 and 7).
increasing the temperature diminished the yield because of increased The H content remaining in furan-2-carboxylate indicated that the
decomposition. Increasing the CO2 pressure slowed the reaction by exchange was ~60% complete. High-resolution mass spectrometry
–
sequestering CO32– in the form of HCO3 . Finally, a 100-mmol-scale showed the fully protonated, and singly, doubly, and triply deuter-
reaction was performed under ~1bar CO2 in a rotating round-bottom ated furan-2-carboxylate (Fig. 2d). When a 1:1 mixture of caesium
flask in a 260°C bath. After 48h, FDCA2– was formed in a 71% yield. furan-2-carboxylate and CD3CO2Cs was heated in the absence of
The scaling behaviour suggests that the reaction takes place at the Cs2CO3 at 200 °C, ~15% H/D exchange was observed, with nearly
molten salt–CO2 interface. The reaction slows and the yield decreases all of the exchange occurring at the 5 position (Fig. 2c and Extended
somewhat as the scale is increased because the surface area-to-volume Data Fig. 8). Thus, at 200°C in a molten salt, a carboxylate is able to
ratio decreases. Improved yields and rates are anticipated with reactors deprotonate the 5 position of furan-2-carboxylate, and CO32– is able
that disperse the salt more effectively.
to deprotonate all positions. The selectivity seen in the carboxylation
No FDCA2– was observed when carboxylation reactions were reaction suggests a greater abundance of the carbanion that leads to
attempted using mixtures of furan-2-carboxylate and CO32– with FDCA2–.
alkali cations other than Cs+. These salt mixtures required temper-
Additional substrates were evaluated to gain further insight into
atures much higher than 200 °C to melt, which resulted in decom- the CO32–-promoted C–H carboxylation reaction. Heating caesium
position reactions. The restriction to Cs+ can be lifted, however, thiophene-2-carboxylate with 0.55 equivalents of Cs2CO3 under
by incorporating another carboxylate salt to attain a semi-molten flowing CO2 at 325 °C resulted in a 71% yield of thiophene-2,5-
solution. For example, heating potassium furan-2-carboxylate with dicarboxylate after 4h (Extended Data Fig. 2). This substrate required
0.5 equivalents of K2CO3 and 1 equivalent of potassium isobu- a higher temperature than furan-2-carboxylate to form a semi-molten
tyrate at 320 °C under 40 ml min–1 CO2 resulted in 62% potassium solution. To see whether C–H carboxylation is possible for substan-
FDCA2– (Extended Data Fig. 5a). Similar results were obtained with tially weaker acids, we evaluated the reactivity of benzoate. A previous
potassium acetate as an additive. Thus, C–H carboxylation does not study reported the carboxylation of caesium benzoate with Cs2CO3 at
require Cs+ per se, but caesium salts do typically have lower melting extreme CO2 pressure (400bar) at 380°C via electrophilic activation
points.
of the aryl ring with a CO32––CO2 complex29. The results with
2–
While the carboxylation results are consistent with the mechanistic furan-2-carboxylate indicate that benzoate deprotonation by CO3
scheme in Fig. 1a, there are other possible mechanisms that do not is feasible, which would enable carboxylation under milder condi-
involve a carbanion intermediate. To probe the acid–base properties tions at low CO2 pressures. Remarkably, heating caesium benzoate
of furan-2-carboxylate independently, an isotope exchange experi- with 0.55 equivalents of Cs2CO3 to 320°C under 8bar CO2 resulted in
ment was performed between furan-2-carboxylate and acetate a combined yield of 66% for a mixture of phthalates and tri- and tetra-
2–
(Fig. 2c). (The pKa of acetate is >33; ref. 28). A mixture of 1mmol carboxylates (Fig. 3a and Extended Data Fig. 4). The ability of CO3
2 1 6
| N a T U r E | V O L 5 3 1 | 1 0 M a r c h 2 0 1 6
© 2016 Macmillan Publishers Limited. All rights reserved

Letter reSeArCH
2 Cs+
1.05 equiv.
Cs2CO3
a
O
O
O
O
Cs+
0.55 equiv.
Cs2CO3
CO2
O
O
O
–
–
–
O
+
OH
O
O
Variables: T,
150 °C
FDCA2–
p
CO , time
Furan-2-carboxylate
2
Starting Decomposition
Entry
Scale
T
Time
FDCA2–
materials
products
p
CO2
1
2
3
4
5
6
7
1 mmol
1 mmol
260 °C
260 °C
200 °C
200 °C
195 °C
195 °C
260 °C
Flowing
Flowing
8 bar
6 h
12 h
2 h
57%
76%
77%
89%
78%
81%
71%
26%
8%
17%
16%
5%
1 mmol
18%
6%
1 mmol
8 bar
5 h
5%
10 mmol
10 mmol
100 mmol
8 bar
5 h
11%
8%
11%
11%
26%
8 bar
10 h
48 h
1 bar
3%
2 K+
b
c
Flowing CO2
320 °C
K+ isobutyrate
O
O
O
0% without K+ isobutyrate
62% with K+ isobutyrate
K+
0.55 equiv.
K2CO3
O
O
–
–
–
O
+
O
O
8 h
O
O
0.55 equiv.
Cs2CO3
Cs+
Cs+
O
O
O
O
H
Cs+
–
O
O
O
Cs+
H/D
H/D
–
–
O
O
+
+
–
D3C
D3C
(D/H)3C
O
O
N2, 200 °C
1 h
H
H/D
H
~60% exchange
O
O
Cs+
Cs+
H
Cs+
–
O
N2, 200 °C
1 h
Cs+
H/D
–
–
O
O
+
+
–
(D/H)3C
O
O
H
H
H
H
~15% exchange
With Cs2CO3
Without Cs2CO3
d
D1
112.0145
111.0083
111.0084
100
100
80
80
60
40
20
D1
112.0144
60
D2
113.0207
40
D3
114.0270
20
D2
113.0179
110.0 111.0 112.0 113.0 114.0
110.0 111.0 112.0 113.0 114.0
m/z
m/z
Figure 2 | C–H carboxylation of furan-2-carboxylate. a, Carboxylation of exchange between deuterated acetate and all positions of furan-2-
caesium-furan-2-carboxylate to caesium FDCA2–
.
2 is the pressure of
carboxylate. In the absence of Cs2CO3, partial exchange is observed at the
5 position. The extent of exchange is calculated based on the integration of
the 1H NMR (see Extended Data Figs 6b and 8a). d, Mass spectrometry of
the exchange reactions in c. D1, D2, and D3 correspond to singly, doubly
and triply deuterated furan-2-carboxylate. m/z, mass to charge ratio.
pCO
carbon dioxide at the reaction temperature T. 1.25 equivalents of Cs2CO3
were used for entry 7. b, Carboxylation of potassium furan-2-carboxylate
to potassium FDCA2– using potassium isobutyrate as a low-melting
co-salt. c, Evidence for C–H deprotonation. Cs2CO3 catalyses isotopic
to deprotonate the C–H bonds of a phenyl ring was tested inde- to 12% of the Cs2CO3 at 350°C, and 19% at 360°C. In addition to ben-
pendently by heating a mixture of 12C6D5CO2Cs, 13C6H5CO2Cs, and zene carboxylation, isobutyrate carboxylation to dimethyl malonate
0.55 equivalents of Cs2CO3 to 320°C under N2 for 30min. 1H NMR and decomposition to formate and acetate also occurred under these
analysis of the products revealed H/D scrambling at all positions on conditions (Extended Data Table 1b). The carboxylation of benzene
the benzoate ring (Fig. 3b). No H/D exchange was observed in the is more challenging than benzoate because there is a larger entropic
absence of Cs2CO3 (Extended Data Fig. 9).
penalty and the solubility of benzene in the carboxylate salt is likely
2–
The results with benzoate suggest that benzene would undergo C–H to be very low. Nevertheless, these results demonstrate that CO3
-
carboxylation if exposed to CO32– in a molten salt. Heating Cs2CO3 promoted hydrocarbon carboxylation is possible.
under benzene and CO2 at temperatures up to 380 °C resulted in
Scalable CO32–-promoted C–H carboxylation requires facile product
no reaction. The lack of reactivity can be attributed to the fact that isolation and highly efficient recovery of the alkali cation. Researchers
Cs2CO3 does not melt. To provide a molten component, reactions were have recently reported the conversion of carboxylates to methyl esters
performed in the presence of caesium isobutyrate. Heating 1.5mmol by reaction with CO2 and methanol at 160–200°C (ref. 13). Similar
Cs2CO3 and 1mmol caesium isobutyrate to 340–380°C under 31bar conditions were found to effect double esterification of FDCA2– to
CO2 and 42 bar benzene resulted in the formation of benzoate, dimethyl furan-2,5-carboxylate (DMFD) in moderate yield (Fig. 4a).
phthalates, and benzene tricarboxylates (Fig. 3c and Extended Data Heating 1mmol of caesium FDCA2– to 200°C in 100ml anhydrous
Fig. 5b). The amount of benzene carboxylation products corresponded methanol under 45bar CO2 for 30 min resulted in 50% yield of DMFD,
1 0 M a r c h 2 0 1 6
| V O L 5 3 1 | N a T U r E | 2 1 7
© 2016 Macmillan Publishers Limited. All rights reserved

reSeArCH Letter
Figure 3 | C–H carboxylation of benzoate and
benzene. a, Carboxylation of caesium benzoate;
yields are given with respect to benzoate.
a
O
O
O
Phthalate: 11%
Tricarboxylates: 12%
–
–
8 bar CO2
0.55 equiv.
Cs2CO3
O
O
+
Cs+
Isophthalate: 21% Tetracarboxylates: 1%
Terephthalate: 21%
320 °C, 5 h
b, CO32–-catalysed H/D isotope exchange between
differentially labelled benzoates. Exchange is seen
in the appearance of 1H peaks associated with
12C–H bonds. No exchange is observed in the
absence of Cs2CO3. c, Carboxylation of benzene in
the presence or absence of caesium isobutyrate. The
yield refers to the amount of CO32– consumed for
the formation of benzene carboxylation products.
equiv., equivalents; p.p.m., parts per million.
O
–
b
H
D
O
O
O
O
Cs+
Cs+
H/D
13C
H/D
H/D
12C
H/D
Cs+
Cs+
0.55 equiv.
Cs2CO3
H
H
D
H/D
H/D
H/D
H/D
–
–
–
–
O
O
O
O
13C
12C
D
+
+
320 °C
D
H
D
H/D
H/D
2 bar N2, 1 h
H
Starting mixture
1H NMR
Product mixture
1H NMR
8.5
8.0
7.5
p.p.m.
7.0
6.5
8.5
O
8.0
7.5
p.p.m.
7.0
6.5
c
O
–
–
Phthalates,
tricarboxylates,
tetracarboxylates
340–380 °C
O
O
+
+ CO2 + Cs2CO3
40 bar 30 bar 1.5 mmol
Cs+ isobutyrate
O
8h
O
–
Total yield Benzoate Phthalates (Tri+tetra)carboxylates
T
Without Cs+ isobutyrate ≤380 °C
With Cs+ isobutyrate 350 °C
360 °C
0%
––
––
––
12%
19%
46 μmol
39 μmol
41 μmol
69 µmol
17 μmol
35 μmol
36% of the mono-ester, 5-(methoxycarbonyl)furan-2-carboxylate carboxylation/esterification cycles were performed (Fig. 4b). 1mmol
(MMFD), and 14% remaining FDCA2–. Subjecting a 1:1 mixture of 2-furoic acid and 1.05 equivalents of Cs2CO3 were subjected to
of DMFD and Cs2CO3 to the same conditions resulted in a similar a one-pot, two-step sequence of carboxylation and esterification.
product mixture, indicating that the reaction is under thermo- Extraction of the crude product mixture afforded 388μmol of DMFD.
dynamic control. To test whether Cs2CO3 can be recovered, sequential The residual material containing FDCA2– and Cs2CO3 was combined
Figure 4 | Product isolation and Cs+ recovery.
a
MMFD
O
DMFD
O
2 Cs+
O
a, CO2 promotes the esterification of FDCA2– to
MMFD and DMFD. The table shows the results of
four 1-mmol-scale reactions performed in 100 ml
anhydrous methanol starting with either caesium
FDCA2– or DMFD. b, Two sequential carboxylation/
esterification sequences with Cs2CO3 recycling. The
total carboxylation yield (91%) is similar to the yield for
an individual experiment, indicating complete recovery
of Cs2CO3 after the first sequence. c, Isolation of FDCA
by protonation with strong acid. CsCl is retained in the
aqueous solution.
CO2
CH3OH
O
CO2
CH3OH
O
O
O
O
O
Cs+
–
–
–
O
O
O
OCH3
H3CO
OCH3
CsHCO3
2 CsHCO3
Product mixture
p
Starting material
CO2
T
FDCA2–
14%
24%
66%
5%
MMFD
36%
37%
28%
29%
DMFD
50%
39%
6%
FDCA2–
FDCA2–
FDCA2–
200 °C
200 °C
180 °C
200 °C
45 bar
24 bar
43 bar
45 bar
DMFD + Cs2CO3
66%
b
O
2 Cs+
O
O
O
O
O
1. 1.05 mmol Cs2CO3
O
O
+
–
–
O
OH
O
H3CO
OCH3
2. Carboxylation
3. Esteriꢀcation
1 mmol
Cs2CO3
388 μmol isolated
Recycled
No additional Cs2CO3
O
O
2 Cs+
O
O
O
O
O
OH
1. 388 μmol
O
+
–
–
O
O
H3CO
OCH3
2. Carboxylation
3. Esteriꢀcation
Cs2CO3
458 μmol isolated
417 μmol isolated
c
2 Cs+
O
O
O
O
O
3N HCl
H2O
O
+
2 CsClaq
–
–
O
O
HO
OH
>99% Cs+ recovery
Crude C−H
carboxylation product
Precipitate
>99% recovery
2 1 8
|
N a T U r E
|
V O L 5 3 1
|
1 0 M a r c h 2 0 1 6
© 2016 Macmillan Publishers Limited. All rights reserved

Letter reSeArCH
9. Lange, J. P., van der Heide, E., van Buijtenen, J. & Price, R. Furfural—a promising
platform for lignocellulosic biofuels. ChemSusChem 5, 150–166 (2012).
10. Cleland, W. W., Andrews, T. J., Gutteridge, S., Hartman, F. C. & Lorimer, G. H.
Mechanism of Rubisco: the carbamate as general base. Chem. Rev. 98,
549–562 (1998).
11. Mani, K. Electrodialysis water splitting technology. J. Membr. Sci. 58,
117–138 (1991).
12. Davis, J. R., Chen, Y., Baygents, J. C. & Farrell, J. Production of acids and bases
for ion exchange regeneration from dilute salt solutions using bipolar
membrane electrodialysis. ACS Sustainable Chem. Eng. 3, 2337–2342 (2015).
13. Barve, P. P., Kamble, S. P., Joshi, J. B., Gupte, M. Y. & Kulkarni, B. D. Preparation of
pure methyl esters from corresponding alkali metal salts of carboxylic acids using
carbon dioxide and methanol. Ind. Eng. Chem. Res. 51, 1498–1505 (2012).
14. Dingyi, Y. & Yugen, Z. The direct carboxylation of terminal alkynes with
carbon dioxide. Green Chem. 13, 1275–1279 (2011).
15. Kudo, K., Ikoma, F., Mori, S. & Sugita, N. Synthesis of glutaconic acid salt from
cesium 3-butenoate with carbon dioxide. J. Jpn. Petrol. Inst. 38, 48–51 (1995).
16. Vechorkin, O., Hirt, N. & Hu, X. Carbon dioxide as the C1 source for direct C-H
functionalization of aromatic heterocycles. Org. Lett. 12, 3567–3569 (2010).
17. van Putten, R. J. et al. Hydroxymethylfurfural, a versatile platform chemical
made from renewable resources. Chem. Rev. 113, 1499–1597 (2013).
18. Román-Leshkov, Y., Chheda, J. N. & Dumesic, J. A. Phase modifiers promote
efficient production of hydroxymethylfurfural from fructose. Science 312,
1933–1937 (2006).
19. Binder, J. B. & Raines, R. T. Simple chemical transformation of lignocellulosic
biomass into furans for fuels and chemicals. J. Am. Chem. Soc. 131,
1979–1985 (2009).
20. da Costa Lopes, A. M. & Bogel-Łukasik, R. Acidic ionic liquids as sustainable
approach of cellulose and lignocellulosic biomass conversion without
additional catalysts. ChemSusChem 8, 947–965 (2015).
21. Luterbacher, J. S. et al. Nonenzymatic sugar production from biomass using
biomass-derived gamma-valerolactone. Science 343, 277–280 (2014).
22. Sheldon, R. A. Green and sustainable manufacture of chemicals from biomass:
state of the art. Green Chem. 16, 950–963 (2014).
23. Hoydonckx, H., Van Rhijn, W., Van Rhijn, W., De Vos, D. & Jacobs, P. Furfural
and derivatives. Ullmann’s Encyclopedia of Industrial Chemistry 16, 285–313
http://onlinelibrary.wiley.com/doi/10.1002/14356007.a12_119.pub2/full
(Wiley-VCH, 2007).
24. Taarning, E., Nielsen, I. S., Egeblad, K., Madsen, R. & Christensen, C. H.
Chemicals from renewables: aerobic oxidation of furfural and
hydroxymethylfurfural over gold catalysts. ChemSusChem 1, 75–78 (2008).
25. Thiyagarajan, S., Pukin, A., van Haveren, J., Lutz, M. & van Es, D. S. Concurrent
formation of furan-2, 5-and furan-2, 4-dicarboxylic acid: unexpected aspects
of the Henkel reaction. RSC Adv. 3, 15678–15686 (2013).
26. Fischer, R. & Fišerová, M. One-step synthesis of furan-2, 5-dicarboxylic acid
from furan-2-carboxylic acid using carbon dioxide. ARKIVOC Online
J. Org. Chem. 4, 405–412 (2013).
27. Fraser, R. R., Mansour, T. S. & Savard, S. Acidity measurements in THF.
V. Heteroaromatic compounds containing 5-membered rings. Can. J. Chem.
63, 3505–3509 (1985).
28. Renaud, P. & Fox, M. A. An electrochemical characterization of dianions:
dilithiated carboxylic acids. J. Am. Chem. Soc. 110, 5705–5709 (1988).
29. Kudo, K. et al. Carboxylation of cesium 2-naphthoate in the alkali metal molten
salts of carbonate and formate with CO2 under high pressure. J. Jpn.
Petrol. Inst. 38, 40–47 (1995).
with a second aliquot of 388μmol 2-furoic acid and carried through
a second sequence of carboxylation and esterification. Extraction
yielded 417μmol DMFD, while analysis of the unreacted material indi-
cated 458μmol of a mixture of FDCA2– and MMFD (Extended Data
Fig. 10). The total amount of carboxylation products (DMFD +
FDCA2– + MMFD) was 1.26mmol, which is 91% of the total amount
of 2-furoic acid substrate (1.39mmol) used for the two carboxylation/
esterification sequences. These results indicate that the Cs2CO3
produced from esterification of caesium FDCA2– can be reused in a
subsequent C–H carboxylation without loss of yield, which in princi-
ple enables a cycle that converts 2-furoic acid into DMFD using only
CO2 and methanol as stoichiometric reagents. It may be possible to
improve esterification yields under milder conditions by removing
water in situ30 or using an alternative solvent.
As an alternative to esterification, treatment of crude caesium
FDCA2– from a C–H carboxylation reaction with 3 N HCl affords
immediate precipitation of FDCA, leaving CsCl in the aqueous solu-
tion with >99% Cs+ recovery (Fig. 4c). To complete the cycle, bipolar
membrane electrodialysis11,12 could be used to convert CsCl into HCl
and CsOH solutions. HCl is recycled for the protonation step, while
CsOH is reacted with 2-furoic acid and CO2 to generate the starting
material for C–H carboxylation. The energy requirement for convert-
ing aqueous alkali chloride solutions into HCl and alkali hydroxide
solutions is ~0.08kWh per mole of alkali chloride11,12, which would
correspond to ~1 kWh per kilogram of FDCA. While additional
energy would be required for water removal in each cycle, using highly
concentrated solutions would minimize this requirement. The overall
process would convert 2-furoic acid into FDCA without using any
organic solvents or product distillation steps.
Our results demonstrate a very simple strategy for engaging CO2
in C–C bond formation that does not require synthetic or biological
catalysts. The ability to deprotonate unactivated C–H bonds opens the
possibility of using this approach to prepare numerous high-volume
targets. In particular, combining carboxylation with hydrogenation
reactions may enable the synthesis of multi-carbon alcohols and
hydrocarbons using CO2 and renewable H2.
Online Content Methods, along with any additional Extended Data display items and
Source Data, are available in the online version of the paper; references unique to
these sections appear only in the online paper.
received 10 September 2015; accepted 29 January 2016.
1. Mikkelsen, M., Jorgensen, M. & Krebs, F. C. The teraton challenge. A review of
fixation and transformation of carbon dioxide. Energ. Environ. Sci. 3, 43–81
(2010).
2. Aresta, M., Dibenedetto, A. & Angelini, A. Catalysis for the valorization of
exhaust carbon: from CO2 to chemicals, materials, and fuels. Technological
use of CO2. Chem. Rev. 114, 1709–1742 (2014).
3. Mander, L. N., Adreatta, J. R. & Darensbourg, D. J. Carbon dioxide. Encyclopedia
of Reagents for Organic Synthesis (e-EROS) http://onlinelibrary.wiley.com/
doi/10.1002/047084289X.rc011.pub2/full (2008).
4. Cai, X. & Xie, B. Direct carboxylative reactions for the transformation of
carbon dioxide into carboxylic acids and derivatives. Synthesis 45,
3305–3324 (2013).
5. Liu, A. H., Yu, B. & He, L. N. Catalytic conversion of carbon dioxide to carboxylic
acid derivatives. Greenhouse Gas. Sci. Technol. 5, 17–33 (2015).
6. Werpy, T. et al. Top value added chemicals from biomass. In Results of
Screening for Potential Candidates from Sugars and Synthesis Gas Vol. 1, 26–28,
http://www.nrel.gov/docs/fy04osti/35523.pdf (US DOE, 2004).
7. Eerhart, A. J. J. E., Faaij, A. P. C. & Patel, M. K. Replacing fossil based PET with
biobased PEF; process analysis, energy and GHG balance. Energ. Environ. Sci.
5, 6407–6422 (2012).
30. Diban, N., Aguayo, A. T., Bilbao, J., Urtiaga, A. & Ortiz, I. Membrane reactors
for in situ water removal: a review of applications. Ind. Eng. Chem. Res. 52,
10342–10354 (2013).
Acknowledgements We thank Stanford University and the Henry and Camille
Dreyfus Foundation for support of this work through a Teacher-Scholar Award
to M.W.K. G.R.D. gratefully acknowledges a fellowship through the Stanford
Center for Molecular Analysis and Design, and T.Y. acknowledges a Postdoctoral
Fellowship for Research Abroad through the Japan Society for the Promotion
of Science. We thank T. Veltman for installation of the Parr reactor, S. Lynch
for assistance with 2H NMR, and J. Du Bois for discussions. High-resolution
mass spectrometry was performed at the Vincent Coates Foundation Mass
Spectrometry Laboratory, Stanford University Mass Spectrometry.
Author Contributions M.W.K. and A.B. conceived the project. A.B., G.R.D. and T.Y.
performed the experiments. M.W.K., A.B. and G.R.D. wrote the paper. All authors
contributed to the analysis and interpretation of the data.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare competing financial interests:
details are available in the online version of the paper. Readers are welcome to
comment on the online version of the paper. Correspondence and requests for
materials should be addressed to M.W.K. (mkanan@stanford.edu).
8. de Jong, E., Dam, M., Sipos, L. & Gruter, G. Furandicarboxylic acid (FDCA),
a versatile building block for a very interesting class of polyesters. Biobased
Monomers Polymers Mater. 1105, 1–13 (2012).
1 0 M a r c h 2 0 1 6
| V O L 5 3 1 | N a T U r E | 2 1 9
© 2016 Macmillan Publishers Limited. All rights reserved

reSeArCH Letter
MethOdS
the caesium furan-2-carboxylate and Cs2CO3 was attached to the rotovap. The
Materials. Caesium carbonate (99.995%, trace metal basis), 2-thiophenecarboxylic joint was taped with black electrical tape and fitted with a green Keck clip, and
acid (99%), benzoic acid-(phenyl-13C6) (99 at% 13C), and tetrabutylammonium the entire apparatus was the backfilled with CO2 gas. A short piece of tubing was
bromide (TBABr, 99%) were purchased from Sigma Aldrich; caesium carbonate connected in such a way as to deliver a slow stream of cooling air to the taped
(≥99%, for large-scale reactions) and furan-2,5-dicarboxylic acid (99.6%) were joint to prevent melting. The reaction was then dipped into the eutectic salt bath
purchased from Chem Impex International; dimethyl furan-2,5-dicarboxylate and rotated slowly for 48h with a gentle flow of CO2 through the bubbler of the
(99%) was purchased from Astatech; benzoic-d5 acid-(phenyl-d5) (98%) was Schlenk line. Over the course of the reaction, the solid initially melts, then turns
purchased from Cambridge Isotope Laboratories; 2-furoic acid (98%), benzoic blackish-brown and solidifies. Once complete, the reaction was slowly cooled to
acid (99%) and anhydrous methyl alcohol (99.8%) were purchased from Acros room temperature and detached from the rotovap. Disodium tartrate dihydrate
Organics; sodium L-(+)-tartrate dihydrate (99.7%), benzene (HPLC grade) (10 mmol, 2.31g, 0.1 equiv.) was added followed by 200ml of deionized water.
was purchased from Alfa Aesar; carbon dioxide (99.99%) was purchased from An aliquot of the resulting solution was evaporated in vacuum then dissolved
Praxair. The methanol was further dried over 3Å molecular sieves before using. in D2O. A 1H NMR was obtained with the following distribution of products:
Reagent-grade benzene was dried by passing through an activated alumina caesium furan-2,5-dicarboxylate (71%), caesium malonate (2%), and caesium
column on an Innovative Technology PureSolv solvent purification system. N,N- acetate (11%) (NMR yields). A repeat of the experiment gave the following dis-
dimethylformamide (DMF, 99.8%) was purchased from Fisher Scientific and dried tribution: caesium furan-2,5-dicarboxylate (69%), caesium malonate (7%), and
by passing through an activated molecular sieve column. All other chemicals were caesium acetate (8%).
used as received without further purification.
The product was isolated from the reaction mixture using the following pro-
Equipment. Experiments under flowing CO2 were performed in a Thermo cedure. First, the reaction was filtered through a pad of celite to remove insoluble
Scientific Lindberg/Blue M tube furnace. Experiments under pressurized CO2 material. The resulting solution was then acidified below a pH of 2 with sulfuric acid
or N2 were performed in a 300-ml high-temperature, high-pressure Parr reactor (15ml concentrated acid). (Caution: this reaction is exothermic and effervescent.)
(model 4561-HT-FG-SS-115-VS-2000-4848) equipped with a glass liner.
The product precipitated from the solution and was collected on a Büchner fun-
Structural analysis. 1H-NMR, 2H-NMR and 13C-NMR spectra were recorded nel. The product was then dissolved in 1 litre of methanol and decolorized with
at 23°C on a Varian Unity Inova 600MHz spectrometer, a Varian Unity Inova activated carbon. The activated carbon was removed by filtering the solution
500MHz spectrometer, a Varian Direct Drive 400MHz spectrometer, a Varian through a pad of celite. The resulting solution was concentrated in vacuum and
Mercury 400 MHz spectrometer, or a Varian Unity Inova 300 MHz spectro- then triturated with 500ml of ethyl acetate. The product was collected on a Büchner
meter. 1H chemical shifts (δ) are reported in parts per million downfield from funnel, washed with ethyl acetate, and then dried in vacuum to afford a white
tetramethylsilane and referenced to residual protium in the NMR solvent (D2O, crystalline powder (10.35g, 66%).
2
δ = 4.79, CDCl3, δ = 7.26). H chemical shifts (δ) are reported in parts per C–H carboxylation of furan-2-carboxylate using K2CO3. 2-furoic acid (56mg,
million downfield from tetramethylsilane and referenced to deuterium in the NMR 0.50mmol, 1.0 equiv.), potassium isobutyrate (63mg, 0.50mmol, 1.0 equiv.), and
solvent (D2O, δ=4.71). 13C chemical shifts (δ) are reported in parts per million K2CO3 (73mg, 0.53mmol, 1.05 equiv.) were dissolved in the minimum of water
downfield from tetramethylsilane and referenced to carbon resonances in the NMR in a quartz boat, and evaporated to dryness by heating at 150°C for 2h under a
solvent (CDCl3 δ=77.16, centre line), to added methanol (δ=49.00), or to added stream of N2. The sample was heated to 320°C under CO2 flowing at 40 ml min−1
tetramethylsilane (δ=0.00).
in the tube furnace for 8 h. The solid product mixture was dissolved in D2O
High-resolution mass spectra were recorded on a Waters Aquity UPLC and and analysed using 1H NMR as described above. The yield of potassium FDCA2–
Thermo Exactive Orbitrap mass spectrometer by direct injection electrospray was 62%, and the conversion of furan-2-carboxylate was 76%. A representative
ionization–mass spectrometry (ESI-MS).
Preparation of reactant mixtures consisting of caesium carboxylate+0.55
spectrum is shown in Extended Data Fig. 5a.
A carboxylation reaction was also performed with 2-furoic acid (56 mg, 0.50
equivalents Cs2CO3. The carboxylic acid (2-furoic acid, thiophene-2-carboxylic mmol, 1.0 equiv.), potassium acetate (49mg, 0.50mmol, 1.0 equiv.), and K2CO3
acid or benzoic acid) and 1.05 equivalent of Cs2CO3 were dissolved in the mini- (73mg, 0.53mmol, 1.05 equiv.). The sample was heated to 300°C under CO2 flow-
mum amount of deionized water and evaporated to dryness by heating at 150°C for ing at 40 ml min−1 in the tube furnace for 8h. The yield of potassium FDCA2– was
at least 2h. The solid mixture was cooled and used directly for C–H carboxylation 57%, and the conversion of furan-2-carboxylate was 96%.
reactions. We note that residual moisture in the reactant mixture reduces the yield Attempted carboxylation of caesium furan-2-carboxylate in DMF. A two-necked
of the carboxylation reaction.
25-ml round-bottomed flask was equipped with a reflux condenser, gas adaptor,
C–H carboxylation under flowing CO2. The appropriate amount of reactant mix- PTFE-coated stir-bar, and a septum. The flask was charged with 2-furoic acid
ture was weighed out into a quartz boat and placed in the tube furnace. The furnace (115mg, 1.03mmol, 1.03 equiv.), Cs2CO3 (555mg, 1.70mmol, 1.70 equiv.), and
was heated to the desired temperature under CO2 flowing at 40mlmin−1 for a water (2 ml). Once the resulting acid–base reaction subsided, the reaction vessel
given period of time. The sample was cooled to ambient temperature, dissolved in was heated to 150°C in an oil bath under a stream of N2. After 30 min the solution
D2O and filtered through a 0.2μm polytetrafluoroethylene (PTFE) syringe filter had dried out, and the reactor was then placed under vacuum and heated to 175°C.
to prepare samples for NMR analysis. The product yields were calculated from At the same time, a heat gun was used to dry the rest of the apparatus. The reactor
integration of the 1H NMR peaks using sodium L-(+)-tartrate dihydrate as an was then cooled to 125°C and back-filled three times with CO2. Using a syringe,
internal standard. Representative spectra and data are shown in Extended Data dry DMF (2 ml) was added to the reaction. The reaction was then stirred for 12h at
Figs 1 and 2 and Extended Data Table 1a.
125°C under a CO2 atmosphere (1 bar). After 12h, the reaction was concentrated
C–H carboxylation under pressurized CO2. The reactant mixture was charged under vacuum. NMR analysis of the residue in D2O indicated no conversion of
into the Parr reactor equipped with a glass liner. The reactor was sealed and then the starting material.
evacuated and backfilled with CO2 three times. It was then filled to a final CO2 C–H carboxylation of benzene. Into a 20ml vial we added a mixture of caesium
pressure at ambient temperature corresponding to a pressure of 8 bar at the desired carbonate (489 mg, 1.5 mmol, 1.0 equiv.) and caesium isobutyrate (220 mg,
reaction temperature. The reactor was heated to the desired temperature, main- 1.0mmol, 0.67 equiv.). This vial was placed into the Parr reactor, which was sealed
tained at that temperature for a given period of time, and then cooled to ambient and then evacuated and backfilled with CO2 three times. Anhydrous benzene was
temperature and depressurized. The solid product mixture was dissolved in D2O injected into the reactor in an amount ranging from 10ml to 35ml depending on
and analysed using NMR as described above. Representative spectra and data are the desired final partial pressure of benzene. The reactor was then pressurized
shown in Extended Data Figs 3 and 4 and Extended Data Table 1a.
with CO2 and heated to the desired temperature. The partial pressure of CO2 at the
100-mmol-scale synthesis of furan-2,5-dicarboxylic acid at 1 atm of CO2. To a final temperature was calculated assuming ideal behaviour. The partial pressure of
1-litre round-bottomed flask we added the 2-furoic acid (100mmol, 11.21g, 1.0 benzene was calculated by subtracting the CO2 pressure from the measured reactor
equiv.) and Cs2CO3 (125mmol, 40.73g, 1.25 equiv.) followed by 100ml of deion- pressure. After the indicated period of time (Extended Data Table 1b), the reactor
ized H2O. The addition of water results in an acid–base reaction that liberates was cooled to ambient temperature and depressurized. The crude product was dis-
CO2. (Caution: this reaction is exothermic and effervescent.) Once the reaction solved in D2O and analysed using NMR as described above. In addition to benzene
was complete, the water was removed in vacuum on a rotary evaporator (rotovap) carboxylation products, isobutyrate decomposition products were observed, which
at 75°C and at 100mTorr and 230°C. The resulting off-white solid was scraped included formate, acetate, acetate carboxylation products (malonate and methane
and ground into a fine white powder. In a fume hood, a reactor was assembled tricarboxylate), and insoluble char. Control experiments were performed in the
consisting of a rotovap with P2O5 in the collection flask connected to a Schlenk absence of either benzene or gaseous CO2. In the absence of benzene, only a trace
line and a eutectic salt bath (48.7 mol% NaNO3, 51.3 mol% KNO3). The eutectic amount of formate was observed and the main product was insoluble char from
salt bath was heated to 260°C, and the 1-litre round-bottomed flask containing isobutryrate decomposition. In the absence of gaseous CO2, a small amount of
© 2016 Macmillan Publishers Limited. All rights reserved

Letter reSeArCH
benzoate was observed, which is attributed to the formation of CO2 in situ from Extended Data Fig. 6a. This material was heated in the Parr reactor to 200°C under
the decomposition of HCO3– that is formed by deprotonation of benzene with 2 bar of N2 for 1h. After cooling to room temperature, the material was dissolved
CO32–. Representative spectra and data are shown in Extended Data Fig. 5b and in D2O and analysed by NMR. Integration of the furan-2-carboxylate peaks in the
Extended Data Table 1b.
1H NMR spectrum using an internal standard indicated that the H/D scrambling
For comparison, an experiment was performed in the absence of caesium was ~15% complete with substantial scrambling at the 5 position and almost no
isobutyrate. A 20ml vial with caesium carbonate (326mg, 1.0mmol) was placed scrambling at the 3 and 4 positions. The presence of D at the 5 position was evident
into the Parr reactor, which was sealed and then evacuated and backfilled with in the 2H NMR spectrum, the peak splitting of the 13C NMR spectrum, and in the
CO2 three times. 35ml of anhydrous benzene was injected into the reactor. The high-resolution mass spectrum (Fig. 2d and Extended Data Fig. 8).
reactor was pressurized with 15 bar of CO2 and heated to 340°C or 380°C for 3h
The mass spectrometry sample was prepared by adding 6 N HCl dropwise to
and 8h respectively. The reactor was then cooled to ambient temperature and the NMR sample until the clear solution turned into a suspension. The water was
depressurized. The crude product was dissolved in D2O and analysed using NMR removed under vacuum and the residue was suspended in 2.5ml of methanol.
as described above. No reaction was observed in this case.
The suspension was allowed to stand till the solid particles settled. An aliquot of
Thermal annealing of caesium furan-2-carboxylate and CO2 in the absence of the supernatant liquid was further diluted with methanol and analysed directly
Cs2CO3. The Parr reactor was equipped with an oven-dried glass liner and charged by mass spectrometry.
with caesium furan-2-carboxylate (244mg, 1.0mmol). The reactor was sealed and Benzoic acid isotope exchange experiment. Benzoic acid-(phenyl-13C6) (7.8mg,
then evacuated and backfilled with CO2 three times. It was pressurized to 5 bar 60.9μmol), benzoic-d5 acid-(phenyl-d5) (7.8mg, 61.3μmol) and Cs2CO3 (41.5mg,
CO2 and then heated to 200°C, at which point the CO2 pressure was 8 bar. After 127.4μmol) were dissolved in 1 ml of deionized water and evaporated to form a
2h, the reactor was cooled to room temperature, vented, and disassembled. The dry powder that consists of caesium benzoate-(phenyl-13C6), caesium benzoate-
residue was dissolved in D2O and analysed using NMR. No FDCA2– was formed (phenyl-d5) and 0.55 equivalents of Cs2CO3. This material was heated in the Parr
in the reaction.
reactor to 320°C under 2 bar of N2 for 1h. After cooling to room temperature,
Recovery of caesium by acidic precipitation at the 10-mmol scale. The general the material was dissolved in D2O and analysed by NMR. The 1H spectra of the
procedure outlined previously (see sections ‘Preparation of reactant mixtures reactant mixture and the product mixture are shown in Fig. 3.
consisting of caesium carboxylate + 0.55 equivalents Cs2CO3,’ and ‘C–H car
-
A control experiment was performed to test whether Cs2CO3 is necessary for
boxylation under pressurised CO2’) was followed using 2-furoic acid (10mmol, isotopic scrambling. Benzoic acid-(phenyl-13C6) (7.8mg, 60.9μmol), benzoic-d5
1.12g, 1.00 equiv.) and Cs2CO3 (10.5mmol, 3.42g, 1.05 equiv.). Once the reaction acid-(phenyl-d5) (7.8 mg, 61.3 μmol) and Cs2CO3 (19.8 mg, 60.7 μmol) were
had completed and cooled to room temperature, the resulting solid was treated dissolved in 1ml H2O and evaporated to form a dry powder that consists of the
with 7 ml of 3 N HCl. The FDCA immediately precipitated from the reaction caesium benzoate salts. After heating to 320°C under 2 bar N2 for 1 h, no H/D
mixture as an off-white solid. The suspension was filtered through a glass frit exchange was observed by 1H NMR (Extended Data Fig. 9).
(medium porosity) and washed with a minimum of deionized water (3 ×0.5 Esterification of caesium FDCA2–. Caesium FDCA2– (420mg, 1.0mmol) was
ml). The filtrate was then transferred to a 100-ml round-bottomed flask and the charged into the Parr reactor equipped with an oven-dried glass liner. The reactor
filter cake was transferred, washing with methanol, to a separate, tared, 100-ml was sealed and then evacuated and backfilled with CO2 three times. Anhydrous
round-bottomed flask. The flask containing the filter cake was evaporated to methanol (100ml) was injected into the reactor. The reactor was then pressur-
dryness to afford a yellow solid. The solid was analysed using 1H NMR in ace- ized with either 28.5 bar or 15 bar CO2 and heated to 200 °C or 180 °C. After
tone-d6. NMR analysis indicated a crude isolated yield of 81% for FDCA, along 30 min, the reactor was cooled to ambient temperature, vented, and disassem-
with 8% residual unreacted 2-furoic acid. The flask containing the filtrate was bled. The reaction mixture was transferred to a 250-ml round-bottomed flask
evaporated to dryness to afford 3.74 g yellow solid. The solid was analysed using and the methanol was removed under vacuum on a rotary evaporator at 45°C.
1H NMR in D2O with an internal standard to quantify organic contaminants. The residue was washed twice with 5 mL CHCl3 to dissolve the DMFD. The
2-furoic acid and FDCA were present in an amount corresponding to <0.4% of combined CHCl3 washes were evaporated to afford DMFD as a white powder.
the mass of the solid. Based on this analysis, the caesium was recovered in >99% The material was dissolved in CDCl3 and analysed by 1H NMR with TBABr as
yield as the CsCl salt.
an internal standard. The remaining residue that was not dissolved in the CHCl3
Recovery of caesium by acidic precipitation at the 100-mmol scale. The washes was dissolved in CD3OD and analysed by 1H NMR using TBABr as an
general procedure outlined previously (see section ‘100-mmol-scale synthesis internal standard. This material consists of FDCA2−, MMFD, and a small amount
of furan-2,5-dicarboxylic acid at 1 atm of CO2’) was followed using the fol- of additional DMFD.
lowing quantities of 2-furoic acid (100mmol, 11.23 g, 1.00 equiv.) and Cs2CO3 Hydrolysis of dimethyl furan-2,5-dicarboxylate (DMFD). DMFD (184 mg,
(105mmol, 34.36g, 1.05 equiv.). Once the reaction had completed and cooled to 1.0mmol, 1 equiv.) and Cs2CO3 (326mg, 1.0mmol, 1 equiv.) were charged into a
room temperature, the resulting solid was treated with 110 ml of 2 N HCl. The Parr reactor equipped with an oven-dried glass liner. The reactor was sealed and
FDCA immediately precipitated from the reaction mixture as an off-white solid. then evacuated and backfilled with CO2 three times. Anhydrous methanol (100ml)
This reaction was filtered through a glass frit (medium porosity) and washed was injected into the reactor. The reactor was pressurized with 28.5 bar CO2 and
with a minimum of deionized water (3×5 ml). The filtrate was then transferred heated to 200°C The total pressure at 200°C was 105 bar and the calculated CO2
to a tared 500-ml round-bottomed flask and the filter cake was transferred and pressure was 45 bar. After 30 min, the reactor was cooled down to ambient tem-
washed with methanol into a separate, tared, 500-ml round-bottomed flask. The perature then vented and disassembled. The reaction mixture was transferred to a
filter cake and filtrate were massed and analysed as described above, yielding 250-ml round-bottomed flask and the methanol was removed under vacuum on
the following results: 69% crude isolated yield for FDCA; >98% recovery of a rotary evaporator at 45°C. The residue was processed and analysed by 1H NMR
caesium as CsCl.
as described above for the esterification of FDCA2–
.
H/D isotope exchange between furan-2-carboxylate and acetate-d3. 2-Furoic Recycling Cs2CO3. The Parr reactor was equipped with an oven-dried glass liner
acid (112mg, 1.0mmol), acetic acid-d4 (58μl, 1.0mmol) and Cs2CO3 (682mg, and charged with caesium furan-2-carboxylate (244mg, 1.0mmol, 1.0 equiv.) and
2.1mmol) were dissolved in the minimum amount of deionized water and evapo- Cs2CO3 (179mg, 0.55mmol, 0.55 equiv.). The reactor was sealed and then evac-
rated to form a dry powder that consisted of 1 mmol caesium furan-2-carboxylate, uated and backfilled with CO2 three times. The reactor was pressurized to 5 bar
1 mmol caesium acetate-d3 and 1.1 mmol of Cs2CO3. This material was heated CO2 and then heated to 200°C, at which point the CO2 pressure was 8 bar. After
in the Parr reactor to 200°C under 2 bar of N2 for 1 h. After cooling to room 5 h, the reactor was cooled and vented then evacuated and backfilled with N2. To
temperature, the product mixture was dissolved in D2O and analysed by NMR. remove the water by-product, 10ml of dry methanol was injected into the reactor
The integration of the furan-2-carboxylate peaks in the 1H NMR spectrum using to dissolve the reaction mixture. The methanol was removed by heating the reac-
an internal standard indicated that the H/D scrambling was ~60% complete with tor to 150°C under vacuum. Subsequently, N2 gas was flowed over the reaction
substantially more scrambling at the 5 position than at the 3 and 4 positions. The mixture for 8 h by keeping the gas release valve of the reactor open. The reac-
presence of D at all positions was evident in the 2H NMR spectrum, the peak tor was cooled to ambient temperature and 90 ml of dry methanol was injected
splitting of the 13C NMR spectrum, and in the high-resolution mass spectrum into it. The vessel was pressurized with 28.5 bar CO2 and heated to 200°C. After
(Fig. 2d and Extended Data Figs 6b, 7a and 7b).
30 min, the reactor was cooled to ambient temperature, depressurized and opened.
For comparison, an experiment was performed in the absence of Cs2CO3. The reaction mixture was diluted with 65ml deionized water and extracted with
2-Furoic acid (112mg, 1.0mmol), acetic acid-d4 (58μl, 1.0mmol) and Cs2CO3 CHCl3 (2×65 ml). The combined organic layers were dried over Na2SO4 and
(312 mg, 0.96 mmol) were dissolved in minimum amount of deionized water concentrated under vacuum to afford the DMFD as a white solid and the yield
and evaporated to form an oil that consisted of caesium furan-2-carboxy- was determined from 1H NMR using TBABr as an internal standard. The aqueous
late and caesium acetate-d3. A 1H NMR spectrum of this mixture is shown in extract was concentrated under reduced pressure to approximately a 2 ml solution
© 2016 Macmillan Publishers Limited. All rights reserved

reSeArCH Letter
and transferred to the same glass liner used in the first cycle. To the solution was Caesium benzoate:
added 2-furoic acid equivalent to the amount of DMFD produced in the first 1H-NMR (500 MHz, D2O) δ 7.80 (d, J = 5 Hz, 2H), 7.41 (t, J = 7.5 Hz, 1H),
cycle. The liner was resealed inside the reactor and heated to 150°C under an 7.36 (t, J=7.5 Hz, 2H)
atmosphere of N2. The reactor was subsequently evacuated and backfilled with 13C-NMR (125MHz, D2O) δ 176.0, 137.0, 131.9, 129.5, 129.0
N2, kept under a stream of N2 at 150°C for 8h, and then cooled to ambient tem- Caesium phthalate:
perature. The reaction mixture was subjected to a second cycle of carboxylation 1H-NMR (500 MHz, D2O) δ 7.45 (dd, J = 5.7, 3.3 Hz, 2H), 7.37 (dd, J = 5.7,
followed by esterification following the same procedures described above. At the 3.3 Hz, 2H)
completion of the cycle, the solution was diluted with 65 ml deionized water and 13C-NMR (125MHz, D2O) δ 177.8, 138.5, 129.5, 127.8
extracted with CHCl3 (2×65 ml). The combined organic layers were dried over Caesium isophthalate:
Na2SO4 and concentrated under vacuum to afford the DMFD (Extended Data 1H-NMR (500 MHz, D2O) δ 8.25 (S, 1H), 7.89 (d, J = 7.5 Hz, 2H), 7.39
Fig. 10a). Evaporation of the aqueous extract yielded the crude mixture of (t, J=7.7 Hz, 1H)
Cs2FDCA, caesium 5-(methoxycarbonyl)furan-2-carboxylate and Cs2CO3. The 13C-NMR (125MHz, D2O) δ 175.5, 137.1, 132.1, 129.7, 129.0
amount of unreacted Cs2FDCA and caesium 5-(methoxycarbonyl)furan-2- Caesium terephthalate:
carboxylate were quantified by 1H NMR using sodium L-(+)-tartrate dihydrate 1H-NMR (500MHz, D2O) δ 7.80 (s, 4H)
as an internal standard (Extended Data Fig. 10b).
13C-NMR (125MHz, D2O) δ 175.5, 139.4, 129.3
NMR spectra. The NMR peaks in the spectra for the carboxylation reactions were Caesium benzene-1,2,3-tricarboxylate:
assigned to different products by comparison to spectra for pure caesium salts 1H-NMR (500MHz, D2O) δ 7.47 (d, J=7.7 Hz, 2H), 7.30 (t, J=7.6 Hz, 1H)
obtained independently from the pure carboxylic acids. The resonances for these 13C-NMR (125MHz, D2O) δ 177.3, 177.1, 138.3, 137.5, 128.7, 127.8
compounds are provided below for reference.
Caesium furan-2,5-dicarboxylate:
Caesium benzene-1,3,5-tricarboxylate:
1H-NMR (500MHz, D2O) δ 8.40 (s, 3H)
1H-NMR (400MHz, D2O) δ 6.95 (s, 2H)
13C-NMR (100MHz, D2O) δ 166.1, 150.4, 115.8
Caesium acetate:
13C-NMR (125MHz, D2O) δ 175.0, 137.5, 132.2
Caesium benzene-1,2,4-tricarboxylate:
1H-NMR (500MHz, D2O) δ 7.92 (s,1H), 7.84 (d, J=8 Hz, 1H), 7.48 (d, J=8 Hz, 1H)
13C-NMR (125 MHz, D2O) δ 177.6, 177.1, 175.1, 141.2, 138.1, 137.1, 129.9,
128.3, 127.6
1H-NMR (400MHz, D2O) δ 1.87 (s, 3H)
13C-NMR (100MHz, D2O) δ 181.4, 23.6
Caesium malonate:
Caesium benzene-1,2,4,5-tetracarboxylate:
1H-NMR (400MHz, D2O) δ 3.09 (s, 2H)
13C-NMR (100MHz, D2O) δ 177.29, 49.03
Caesium thiophene-2,5-dicarboxylate:
1H-NMR (300MHz, D2O) δ 7.4 (s, 2H)
13C-NMR (75MHz, D2O) δ 169.5, 144.8, 131.0
1H-NMR (500MHz, D2O) δ 7.51 (s, 2H)
13C-NMR (125MHz, D2O) δ 176.8, 138.6, 126.6
Dimethyl 2,5-furandicarboxylate:
1H-NMR (300MHz, CDCl3) δ 7.20 (s, 2H), 3.91 (s, 6H)
13C-NMR (75MHz, CDCl3) δ 158.5, 146.7, 118.6, 52.5
© 2016 Macmillan Publishers Limited. All rights reserved

Letter reSeArCH
Extended Data Figure 1 | NMR spectra for the carboxylation of caesium furan-2-carboxylate under flowing CO2. a, 1H NMR (300 MHz) and
b, 13C NMR (100 MHz) in D2O of the crude product mixture after the reaction of 1 mmol caesium furan-2-carboxylate and 0.55 mmol Cs2CO3
under CO2 flowing at 40 ml min–1 at 260 °C for 12 h. f1 indicates the chemical shift, δ.
© 2016 Macmillan Publishers Limited. All rights reserved

reSeArCH Letter
Extended Data Figure 2 | NMR spectra for the carboxylation of caesium thiophene-2-carboxylate. a, 1H NMR (300 MHz) and b, 13C NMR (100 MHz)
in D2O of the crude product mixture after the reaction of 1 mmol caesium thiophene-2-carboxylate and 0.55 mmol Cs2CO3 under CO2 flowing at
40 ml min–1 at 325 °C for 12 h.
© 2016 Macmillan Publishers Limited. All rights reserved

Letter reSeArCH
Extended Data Figure 3 | NMR spectra for the carboxylation of caesium
furan-2-carboxylate in the Parr reactor. a, 1H NMR (300 MHz) in D2O
of the crude product mixture after the reaction of 1 mmol caesium
furan-2-carboxylate and 0.55 mmol Cs2CO3 under 8 bar CO2 at 200 °C
for 5 h. b, 1H NMR (300 MHz) in D2O of the crude product mixture after
the reaction of 10 mmol caesium furan-2-carboxylate and 5.5 mmol
Cs2CO3 under 8 bar CO2 at 200 °C for 10 h.
© 2016 Macmillan Publishers Limited. All rights reserved

reSeArCH Letter
Extended Data Figure 4 | NMR spectra for the carboxylation of caesium benzoate. a, 1H NMR (300 MHz) and b, 13C NMR (100 MHz) in D2O of the
crude product mixture after the reaction of 1 mmol caesium benzoate and 0.55 mmol Cs2CO3 under 8 bar CO2 at 320 °C for 5 h.
© 2016 Macmillan Publishers Limited. All rights reserved

Letter reSeArCH
Extended Data Figure 5 | NMR spectra for the carboxylation of
potassium furan-2-carboxylate and benzene. a, 1H NMR (600 MHz)
in D2O of the crude product mixture after the reaction of 0.5 mmol
potassium furan-2-carboxylate, 0.5 mmol potassium isobutyrate and
0.28 mmol K2CO3 under CO2 flowing at 40 ml min–1 at 320 °C for 8 h.
b, 1H NMR (600 MHz) of the crude product mixture after the reaction in
D2O of a 1.5 mmol of caesium carbonate and 1 mmol caesium isobutyrate
under 42 bar benzene and 31 bar CO2 at 350 °C for 8 h.
© 2016 Macmillan Publishers Limited. All rights reserved

reSeArCH Letter
Extended Data Figure 6 | 1H NMR spectra for H/D exchange between furan-2-carboxylate and deuterated acetate in the presence of Cs2CO3.
a, 1H NMR (400 MHz) in D2O of a 1:1 mixture of caesium furan-2-carboxylate and CD3CO2Cs. b, 1H NMR (400 MHz) in D2O of the crude product
mixture after the reaction of a 1:1 mixture of caesium furan-2-carboxylate and CD3CO2Cs with 0.55 equivalents Cs2CO3 at 200 °C under 2 bar N2 for 1 h.
© 2016 Macmillan Publishers Limited. All rights reserved

Letter reSeArCH
Extended Data Figure 7 | Additional NMR spectra for H/D exchange between furan-2-carboxylate and deuterated acetate in the presence
of Cs2CO3. a, 13C NMR (75 MHz) and b, 2H NMR (92 MHz) in D2O of the crude product mixture after the reaction of a 1:1 mixture of caesium
furan-2-carboxylate and CD3CO2Cs with 0.55 equivalents Cs2CO3 at 200 °C under 2 bar N2 for 1 h.
© 2016 Macmillan Publishers Limited. All rights reserved

reSeArCH Letter
Extended Data Figure 8 | NMR spectra for H/D exchange between furan-2-carboxylate and deuterated acetate in the absence of Cs2CO3.
a, 1H NMR (400 MHz) and b, 2H NMR (92 MHz) in D2O of the crude product mixture after the reaction of a 1:1 mixture of caesium furan-2-carboxylate
and CD3CO2Cs at 200 °C under 2 bar N2 for 1 h.
© 2016 Macmillan Publishers Limited. All rights reserved

Letter reSeArCH
Extended Data Figure 9 | No H/D exchange is observed between differentially labelled caesium benzoates when heated to 320 °C in the absence
of Cs2CO3.
© 2016 Macmillan Publishers Limited. All rights reserved

reSeArCH Letter
Extended Data Figure 10 | NMR spectra for the Cs2CO3 recycling experiment. a, 1H NMR (400 MHz) in CDCl3 of the DMFD isolated after the
second carboxylation/esterification sequence. b, 1H NMR (400 MHz) in D2O of the material recovered from the aqueous phase after the second
carboxylation/esterification sequence.
© 2016 Macmillan Publishers Limited. All rights reserved

Letter reSeArCH
extended data table 1 | Additional C–h carboxylation data
a, C–H carboxylation of caesium furan-2-carboxylate. b, C–H carboxylation of benzene. Formate, acetate, and acetate carboxylation products (‘pdts’) (malonate
and methane tricarboxylate) arise from decomposition of isobutyrate.
© 2016 Macmillan Publishers Limited. All rights reserved