Homogeneous Catalysis in Supercritical Fluids: Hydrogenation of Supercritical Carbon Dioxide to Formic Acid, Aklyl Formates, and Formamides

Article abstract of DOI:10.1021/ja953097b

Rapid, selective, and high-yield hydrogenation of CO2 can be achieved if the CO2 is in the supercritical state (scCO2).Dissolving H2, a tertiary amine, a catalyst precursors such as RuH24 or RuCl24, and a promoting additive such as water, CH3OH, or DMSO in scCO2 at 50 deg C leads to the generation of formic acid with turnover frequencies up to or exceeding 4000 h^-1.In general, experiments in which a second phase was formed by one or more reagents or additives had lower rates of reaction.The high of reaction is attributed to rapid diffusion, weak catalyst solvation, and the high miscibility of H2 in scCO2.The formic acid synthesis can be coupled with subsequent reactions of formic acid, for example, with alcohols or primary or secondary amines, to give highly efficient routes to formate estaers or formamides.With NH(CH3)2, for example 420 000 mol of dimethylformamide/mol Ru catalyst was obtained at 100 deg C.The demonstrated solubility and catalytic activity of complexes of tertiary phosphines in scCO2 suggest that scCO2 could be an excellent medium for homogeneous catalysis and that many phosphine-containing homogeneous catalysts could be adopted for use in supercritical media.

Full text of DOI:10.1021/ja953097b

344
J. Am. Chem. Soc. 1996, 118, 344-355
Homogeneous Catalysis in Supercritical Fluids: Hydrogenation
of Supercritical Carbon Dioxide to Formic Acid, Alkyl
Formates, and Formamides
Philip G. Jessop, Yi Hsiao, Takao Ikariya, and Ryoji Noyori*,†
Contribution from ERATO Molecular Catalysis Project, 1247 Yachigusa, Yakusa-cho,
Toyota 470-03, Japan
ReceiVed September 7, 1995X
Abstract: Rapid, selective, and high-yield hydrogenation of CO2 can be achieved if the CO2 is in the supercritical
state (scCO2). Dissolving H2, a tertiary amine, a catalyst precursor such as RuH2[P(CH3)3]4 or RuCl2[P(CH3)3]4,
and a promoting additive such as water, CH3OH, or DMSO in scCO2 at 50 °C leads to the generation of formic acid
with turnover frequencies up to or exceeding 4000 h-1. In general, experiments in which a second phase was formed
by one or more reagents or additives had lower rates of reaction. The high rate of reaction is attributed to rapid
diffusion, weak catalyst solvation, and the high miscibility of H2 in scCO2. The formic acid synthesis can be coupled
with subsequent reactions of formic acid, for example, with alcohols or primary or secondary amines, to give highly
efficient routes to formate esters or formamides. With NH(CH3)2, for example 420 000 mol of dimethylformamide/
mol of Ru catalyst was obtained at 100 °C. The demonstrated solubility and catalytic activity of complexes of
tertiary phosphines in scCO2 suggest that scCO2 could be an excellent medium for homogeneous catalysis and that
many phosphine-containing homogeneous catalysts could be adopted for use in supercritical media.
Introduction
H2 with scCO2.15 The concentration of H2 in a supercritical
mixture of H2 (85 atm) and CO2 (120 atm) at 50 °C is 3.2 M,
while the concentration of H2 in THF under the same pressure
is merely 0.4 M.16 This property of scCO2 must allow
significant rate enhancement of reactions for which the rate is
greater than zeroth order in H2 concentration. The usefulness
of this property has been noted previously.8,9 The high
concentration of CO2 in scCO2 could also be advantageous in
any reaction which incorporates CO2. It is possible therefore
that the use of scCO2 as a solvent will allow the more
widespread use of CO2 as an organic carbon source; we present
here, and in our preliminary communications12,17,18 and
patents,11,19-21 examples of this strategy.22,23
Carbon dioxide is an abundant and nonharmful source of
carbon for incorporation into organic molecules. The lack of
reactivity of the molecule has prevented CO2 from becoming a
carbon source for more than the production of urea, aspirin,
and carbonates.24 Another possible reaction is the hydrogenation
of CO2 to formic acid, which is rendered thermodynamically
favorable by the addition of a base (eq 1).
Supercritical fluids (SCFs), substances heated beyond their
critical point, have densities and viscosities between those of
liquids and gases. SCFs, especially supercritical carbon dioxide
(scCO2; Tc ) 31 °C, Pc ) 72.9 atm),1 have been used as solvents
for extractions,2,3 chromatography, stoichiometric organic reac-
tions, and heterogeneously catalyzed reactions4-7 but, with few
exceptions,8-14 have not been used for homogeneously catalyzed
reactions. The benefits of SCFs suggest that for some reactions,
the yield, rate, or selectivity will be dramatically enhanced.
Among the chemical advantages, the most pertinent to homo-
geneous catalysis in scCO2 is the miscibility of gases such as
† Permanent address: Department of Chemistry, Nagoya University,
Chikusa, Nagoya 464-01, Japan.
X Abstract published in AdVance ACS Abstracts, December 15, 1995.
(1) (a) Reid, R. C.; Prausnitz, J. M.; Poling, B. E. The Properties of
Gases and Liquids, 4th ed.; McGraw-Hill: New York, 1987. (b) Interna-
tional Tables of the Fluid State; Angus, S., Armstrong, B., de Reuck, K.
M., Eds.; Pergamon Press: Oxford, 1976; Vol. 3 (Carbon Dioxide).
(2) Eckert, C. A.; Van Alsten, J. G.; Stoicos, T. EnViron. Sci. Technol.
1986, 20, 319-325.
(3) McHugh, M. A.; Krukonis, V. J. Supercritical Fluid Extraction;
Butterworths: Boston, 1986.
(4) Subramaniam, B.; McHugh, M. A. Ind. Eng. Chem. Process Des.
DeV. 1986, 25, 1-12.
(5) Brennecke, J. F. In Supercritical Fluid Engineering Science: Fun-
damentals and Applications; Kiran, E., Brennecke, J. F., Eds.; ACS
Symposium Series 514; American Chemical Society: Washington, DC,
1993; pp 201-219.
(6) Shaw, R. W.; Brill, T. B.; Clifford, A. A.; Eckert, C. A.; Franck, E.
U. Chem. Eng. News 1991, 69 (51), 26-39.
(7) Kaupp, G. Angew. Chem., Int. Ed. Engl. 1994, 33, 1452-1455.
(8) Kramer, G. M.; Leder, F. U.S. Patent 3 880 945, 1975.
(9) Rathke, J. W.; Klingler, R. J.; Krause, T. R. Organometallics 1991,
10, 1350-1355.
(10) Vieville, C.; Mouloungui, Z.; Gaset, A. Ind. Eng. Chem. Res. 1993,
32, 2065-2068.
(11) Ikariya, T.; Jessop, P. G.; Noyori, R. Japan Patent Appl. No. 274721,
1993.
(15) Tsang, C. Y.; Streett, W. B. Chem. Eng. Sci. 1981, 36, 993-1000.
(16) Hydrogen and Deuterium; Young, C. L., Ed.; Solubility Data Series;
Pergamon Press: Oxford, 1981; Vol. 5/6.
(17) Jessop, P. G.; Hsiao, Y.; Ikariya, T.; Noyori, R. J. Am. Chem. Soc.
1994, 116, 8851-8852.
(18) Jessop, P. G.; Hsiao, Y.; Ikariya, T.; Noyori, R. J. Chem. Soc., Chem.
Commun. 1995, 707-708.
(19) Ikariya, T.; Jessop, P. G.; Noyori, R. Japan Patent Appl. No. 125401,
1994.
(20) Ikariya, T.; Jessop, P. G.; Hsiao, Y.; Noyori, R. Japan Patent Appl.
No. 125402, 1994.
(21) Ikariya, T.; Hsiao, Y.; Jessop, P. G.; Noyori, R. Eur. Patent Appl.
0 652 202, 1995.
(22) Concurrent but independent work on the use of scCO2 as a medium
and reactant for its reaction with an alkyne, catalyzed by a nickel-phosphine
complex, was the subject of a communication by Reetz et al.23
(23) (a) Reetz, M. T.; Kojnen, W.; Strack, T. Chimia 1993, 47, 493. (b)
Banister, J. A.; Lee, P. D.; Poliakoff, M. Organometallics 1995, 14, 3885,
footnote 5.
(24) Keim, W. In Carbon Dioxide as a Source of Carbon; Aresta, M.,
Forti, G., Eds.; NATO ASI Series, Series C 206; D. Reidel: Dordrecht,
1987; pp 23-31.
(12) Jessop, P. G.; Ikariya, T.; Noyori, R. Nature 1994, 368, 231-233.
(13) Jessop, P. G.; Ikariya, T.; Noyori, R. Science 1995, 269, 1065-
1069.
(14) Burk, M. J.; Feng, S.; Gross, M. F.; Tumas, W. J. Am. Chem. Soc.
1995, 117, 8277-8278.
0002-7863/96/1518-0344$12.00/0 © 1996 American Chemical Society

Homogeneous Catalysis in Supercritical Fluids
J. Am. Chem. Soc., Vol. 118, No. 2, 1996 345
3) catalyzed by RuH2[P(C6H5)3]4, RuCl2[P(C6H5)3]3, and other
catalyst precursors has been reported,25,28,44-47 the highest yield
of DMF being 3400 TON at 130 °C.44
CO2 + H2 + B f [BH][O2CH]
(1)
Although this reaction and variations thereof have been known
for 2 decades,25-27 efficient systems for its homogeneous
catalysis have only recently been reported,28 especially by the
team of Leitner et al.29-33 Surprisingly, the highest conversion
(TON ) 3400)31 has been obtained with water as the solvent
(TON ) turnover number, mol of product/mol of catalyst).34
We have reported in a preliminary communication that the
hydrogenation of CO2 is particularly efficient if the CO2 is in
the supercritical state.11,12 Since then, we have expanded the
range of conditions and additives tested and investigated a
number of other factors which influence rate and yield.
The related syntheses of alkyl formate18,19 and formamides17,20
from scCO2 have also been investigated in our laboratory.
Methyl formate is used for the industrial synthesis of formic
acid and DMF, as well as for other applications.35 It can be
produced by the base-catalyzed carbonylation of methanol with
CO, the currently used industrial process, by methanol dehy-
drogenation36 or by hydrogenation of CO2 in the presence of
methanol (eq 2).28
CO2 + H2 + NHR2 f HCONR2 + H2O
(3)
In this report, we present the details of our research using
scCO2, emphasizing the often dramatic effects of phase changes
on reaction behavior and the strong promoting ability of
additives.
Experimental Section
Materials and Methods. The compounds RuH2[P(C6H5)3]4,48 RuH2-
[P(CH3)3]4 (1),49 RuCl2[P(CH3)3]4 (2),50,51 RuCl(O2CCH3)[P(CH3)3]4
55
(3),52 trans-RuCl2(dmpe)2,53 trans-RuHCl(dmpe)2,54 and [RhCl(nbd)]2
and the carbamates56 were prepared by literature methods (nbd ) 2,5-
norbornadiene; dmpe ) (CH3)2PCH2CH2P(CH3)2). Note that complex
1 is light-sensitive and should be stored in the dark. Ru3(CO)12 (Strem)
was used as received. The catalyst precursors and the alkylphosphines
were stored and handled under argon at all times, except for 2 and 3
which could be weighed under air if afterwards returned to an inert
atmosphere. Liquid reagents and solvents were dried, distilled, and
degassed before use except for DMSO, amines, and water which were
only degassed. MS3A (Nacalai) was activated and the acidic resins
Nafion NR50 (DuPont) and Amberlyst (Aldrich) were washed and dried
under vacuum before use. The H2 gas used was 99.99% purity, zero
grade product of Sumitomo. Two grades of CO2 were used for formic
acid production, 99.99% (Showa Tansan) and normal grade (Fuji), with
comparable results. For the formamide and alkyl formate syntheses,
only the purer grade was used, although the normal grade could again
be satisfactory. CO (Sumitomo Seika) was UHP grade.
CO2 + H2 + ROH catalyst8 HCO2R + H2O
(2)
base
Homogeneous catalysts for this process include RuH2[P-
(C6H5)3]4,26,27,37 other metal-phosphine complexes,26,27,38,39 and
anionic carbonyl complexes.40-42 The temperatures used by
previous investigators were unfortunately high (100-175 °C),
while the highest reported yield was 470 TON, obtained in
methanol solution with inorganic bases.38
Spectroscopic and chromatographic measurements were performed
with a JEOL JNM EX-400 NMR spectrometer, a JASCO FT/IR-5300
spectrometer, and a Shimadzu Parvum GC/MS instrument consisting
of a GC-17A gas chromatograph and a QP-5000 mass spectrometer.
Differential scanning calorimetry (DSC) measurements were performed
under argon with a Rigaku DSC 8230B instrument and a TAS 100
system controller. The supercritical fluid equipment, a diagram of
which is shown in Figure 1, was modified from supercritical chroma-
tography equipment manufactured by JASCO International. The
principal components are a stainless steel 50-, 150-, or 300-mL reactor
vessel, two PU-980 HPLC pumps, an 880-81 back-pressure regulator,
a CO-965 (maximum 80 °C) or 866-CO (100 °C) column oven, and a
magnetic stir plate. One of the pumps was fitted with a liquid CO2
reservoir cooled by a SCINICS CH-201 coolant circulator. The other
pump was used to supply CH2Cl2 during the flushing procedure; the
1/16 in. stainless steel tubing of the equipment was flushed between
reactions with CH2Cl2, CO2, and H2, in that order. For safety reasons,
the back-pressure regulator was set to vent at 40 atm higher than the
desired total pressure and was tested before every reaction.
Dimethylformamide (DMF), an important industrial solvent,
is currently prepared by the sodium methoxide-catalyzed
carbonylation of dimethylamine with CO in methanol.43 The
synthesis of formamides from CO2, H2, and dialkylamines (eq
(25) Haynes, P.; Slaugh, L. H.; Kohnle, J. F. Tetrahedron Lett. 1970,
365-368.
(26) Kolomnikov, I. S.; Lobeeva, T. S.; Vol’pin, M. E. IzV. Akad. Nauk
SSSR, Ser. Khim. 1972, 2329-2330.
(27) Inoue, Y.; Sasaki, Y.; Hashimoto, H. J. Chem. Soc., Chem. Commun.
1975, 718-719.
(28) Jessop, P. G.; Ikariya, T.; Noyori, R. Chem. ReV. 1995, 95, 259-
272.
(29) Graf, E.; Leitner, W. J. Chem. Soc., Chem. Commun. 1992, 623-
624.
(30) Burgemeister, T.; Kastner, F.; Leitner, W. Angew. Chem., Int. Ed.
Engl. 1993, 32, 739-741.
(31) Gassner, F.; Leitner, W. J. Chem. Soc., Chem. Commun. 1993,
1465-1466.
(32) Leitner, W.; Dinjus, E.; Gassner, F. J. Organomet. Chem. 1994,
475, 257-266.
Safety Warning. Operators of high-pressure equipment such as that
required for these experiments should take proper precautions, including
(33) Fornika, R.; Go¨rls, H.; Seemann, B.; Leitner, W. J. Chem. Soc.,
Chem. Commun. 1995, 1479-1481.
(34) A recent publication has described a Ru catalyst system for the
hydrogenation of CO2 in ethanol with TON values of up to 5000 at 150
°C; see: Lau, C. P.; Chen, Y. Z. J. Mol. Catal. A: Chem. 1995, 101, 33-
36.
(35) Reutemann, W.; Kieczka, H. In Ullmann’s Encyclopedia of Indus-
trial Chemistry, 5th ed.; Elvers, B., Hawkins, S., Ravenscroft, M.,
Rounsaville, J. F., Schulz, G., Eds.; VCH: Weinheim, 1989; Vol. A12, pp
13-33.
(36) Weissermel, K.; Arpe, H.-J. Industrial Organic Chemistry, 2nd ed.;
VCH: Weinheim, 1993; pp 39-43.
(37) Hashimoto, H.; Inoue, Y. Japan Kokai Tokkyo Koho 138614, 1976.
(38) Lodge, P. G.; Smith, D. J. H. Eur. Patent Appl. 0 094 785, 1983.
(39) Phala, H.; Kudo, K.; Mori, S.; Sugita, N. Bull. Inst. Chem. Res.,
Kyoto UniV. 1985, 63, 63-71.
(40) Darensbourg, D. J.; Ovalles, C. J. Am. Chem. Soc. 1984, 106, 3750-
3754.
(44) Kiso, Y.; Saeki, K. Japan Kokai Tokkyo Koho 36617, 1977.
(45) Kudo, K.; Phala, H.; Sugita, N.; Takezaki, Y. Chem. Lett. 1977,
1495-1496.
(46) Phala, H.; Kudo, K.; Sugita, N. Bull. Inst. Chem. Res., Kyoto UniV.
1981, 59, 88-105.
(47) Schreiner, S.; Yu, J. Y.; Vaska, L. Inorg. Chim. Acta 1988, 147,
139-141.
(48) Young, R.; Wilkinson, G. Inorg. Synth. 1977, 17, 75-77.
(49) Hartwig, J. F.; Andersen, R. A.; Bergman, R. G. Organometallics
1991, 10, 1875-1887.
(50) Jones, R. A.; Real, F. M.; Wilkinson, G.; Galas, A. M. R.;
Hursthouse, M. B.; Malik, K. M. A. J. Chem. Soc., Dalton Trans. 1980,
511-518.
(51) Schmidbaur, H.; Blaschke, G. Z. Naturforsch. Teil B 1980, 35, 584-
587.
(41) Evans, G. O.; Newell, C. J. Inorg. Chim. Acta 1978, 31, L387-
L389.
(52) Mainz, V. V.; Andersen, R. A. Organometallics 1984, 3, 675-
678.
(42) Darensbourg, D. J.; Ovalles, C.; Pala, M. J. Am. Chem. Soc. 1983,
105, 5937-5939.
(43) Bipp, H.; Kieczka, H. In Ullmann’s Encyclopedia of Industrial
Chemistry, 5th ed.; Elvers, B., Hawkins, S., Ravenscroft, M., Rounsaville,
J. F., Schulz, G., Eds.; VCH: Weinheim, 1989; Vol. A12, pp 1-12.
(53) Chatt, J.; Hayter, R. G. J. Chem. Soc. 1961, 896-904.
(54) Chatt, J.; Hayter, R. G. J. Chem. Soc. 1961, 2605-2611.
(55) Abel, E. W.; Bennett, M. A.; Wilkinson, G. J. Chem. Soc. 1959,
3178-3182.
(56) Theodoropulos, S. Eur. Patent Appl. 0 062 161, 1982.

346 J. Am. Chem. Soc., Vol. 118, No. 2, 1996
Jessop et al.
°C/min, 50:1 split injection, 10 kPa He carrier gas, HCO2CH3 monitored
at 60 m/z, internal standard acetonitrile monitored at 41 m/z).
Synthesis of Formamides. The syntheses of formamides were
performed by the same method as the formic acid synthesis, except
that primary and secondary amines and higher temperatures were used
and neither water nor N(C2H5)3 was added. For experimental conven-
ience, some of the lighter amines were introduced as the corresponding
carbamates rather than the free amines. For example, both dimethyl-
amine and dimethylammonium dimethylcarbamate were tested, with
identical results. Although the carbamates contain CO2, tests showed
that very little conversion to formic acid or formamide takes place
during the temperature equilibration. The product analysis and
identification of the formamides and N(CH3)3 were performed by NMR
in the same manner as for methyl formate. The identification and yield
of DMF were also confirmed, for selected reactions, by GC/MS (30 m
× 0.25 cm TC-wax column, 50-220 °C at 20 °C/min, 50:1 split
injection, 100 kPa He carrier gas, DMF monitored at 73 m/z, internal
standard n-hexadecane monitored at 85 m/z). It was not possible to
obtain confirmation of the identification of N(CH3)3 by GC/MS due to
its high volatility and low yield.
Figure 1. Equipment used for the reactions in scCO2.
but not limited to the use of blast shields and pressure relief
mechanisms, to minimize the risk of personal injury.
Stoichiometric Reaction of RuH2[P(CH3)3]4 with CO2. Carbon
dioxide was bubbled for 1 min through a solution of the complex in
C6D6 prepared under argon. Comparison of the 1H and 31P[1H] NMR
spectra before and after the reaction showed that the only product was
Phase Behavior Observations. The phase behavior of the multi-
component systems was determined visually by use of a 50-mL reactor
equipped with sapphire windows. Quantitative measurements of the
concentration of amine were also performed. In the latter method, a
sample loop was used to transfer a sample from the uppermost phase
of the reactor vessel into a supercritical fluid chromatography system
with a scCO2/methanol mobile phase, a Superpak Crest C18 column,
a JASCO UV-970 UV detector, and an 880-81 back-pressure regulator.
cis-RuH(O2CH)[P(CH3)3]4 in 9% conversion. The complex was not
2
isolated. 1H NMR (C6D6, 400 MHz): δ -8.10 (dq, 1H, JHtransP
)
99.8 Hz, 2JHcisP ) 27.7 Hz, RuH), 0.98 (d, 18H, 2JHP ) 9.8 Hz, P(CH3)3),
2
4
1.22 (d, 18H, JHP ) 13.4 Hz, P(CH3)3), 8.93 (d, 1H, JHtransP ) 4.9
Hz, O2CH). 31P[1H] NMR (C6D6, 162 MHz): δ -13.2 (m, 1P, P trans
2
2
to H), -0.8 (dd, 2P, JPP ) 34.4 Hz, JPP ) 24.3 Hz, mutually trans
P(CH3)3 ligands), 19.5 (td, 1P, 2JPP ) 34.4 Hz, 2JPP ) 18.2 Hz, P trans
to O2CH). Further evidence for the structure assignment was obtained
Hydrogenation of scCO2 to Formic Acid. The oven-dried reactor
(usually 50-mL internal volume) was cooled to room temperature under
vacuum, filled with argon, and then charged with the catalyst precursor
(typically 3 µmol) in an argon-filled glovebag. The reactor was then
evacuated for 10 min under high vacuum and refilled with argon. Water
(2 µL, 0.1 mmol) and N(C2H5)3 (0.7 mL, 5.0 mmol) were injected into
the reactor against a positive argon pressure through a threaded opening
in the top which was plugged at all other times. The reactor was then
attached to the equipment as shown in Figure 1. Pressure testing and
warming to the reaction temperature were performed with a pressure
of 40 atm of H2. After temperature equilibration, which takes ca. 2 h,
the pressure of H2 was topped up to the desired level followed by the
required pressure of CO2. The pressures cited are at reaction temper-
ature. The CO2 was introduced from a cooled (-5 °C) reservoir by
an HPLC pump. The start of the reaction is defined as the time of
CO2 gas introduction. An experiment with a reaction time of 0 h was
performed to confirm that no formic acid is generated during the
prewarming. After the expiration of the desired reaction time, the
reactor was half-submerged in a bath of acetone or alcohol which was
subsequently cooled by addition of dry ice. The use of a liquid nitrogen
bath, which was also effective, is not recommended because it is more
likely to cause weakening of the reactor walls. After the pressure had
reached a steady low value, the H2 gas was vented and the reactor was
slowly warmed, the CO2 venting into a fume hood as it sublimed. The
formic acid to amine molar ratio (hereafter referred to as AAR) and
the yield of formic acid were determined from 1H NMR spectra of the
CD3OD solutions with 0.05 mL of CHCl3 as an internal standard at 24
°C with 16 scans at 5-s intervals and a pulse width of 5.2 µs. The
accuracy of this method was confirmed by analyzing known mixtures
of formic acid and N(C2H5)3 in CD3OD with an average error of 3%
in the AAR. The identification and yield of formic acid were
confirmed, for selected reactions, by GC/MS (30 m × 0.25 cm TC-
wax column, 100-200 °C at 20 °C/min, 100:1 split injection, 50 kPa
He carrier gas, HCO2H monitored at 46 m/z, internal standard
n-hexadecane monitored at 85 m/z).
1
by noting the close similarity of the H NMR (hydride region) and
31P[1H] NMR spectra with those of the known complex cis-RuH(OC6H4-
p-CH3)[P(CH3)3]4.57
Results
Observations of Phase Behavior. Reactions in SCFs are
strongly affected by phase changes. Although the phase
behavior of pure CO2 (Figure 2, top row) is known,1b that of
the multicomponent systems described here has not been
published. For this reason, under conditions relevant to all
reactions, the number of phases present was determined by
visual observation of the CO2/H2/additive mixture in a 50-mL
steel vessel equipped with sapphire windows (Figures 2 and
3). The results are summarized below, and the case of N(C2H5)3
at 50 °C is illustrated in Figure 4.
In the presence of only H2 (80 atm, 50 °C; Figure 4a), 20
mL of N(C2H5)3 forms a liquid phase of approximately the same
volume as the originally charged amine; the gas phase contains
only a small concentration of the amine, as determined by
chromatographic testing. In the presence of only CO2, the liquid
phase expands to 24 mL at 30 atm or 34 mL at 60 atm (Figure
4b). This swelling of the liquid phase must be a result of
dissolution of considerable amounts of CO2 into that phase
(Figure 2, middle row). At pressures above 80 atm, 20 mL of
N(C2H5)3 is completely miscible with scCO2 in the absence of
H2 in a 50-mL vessel at 50 °C (Figure 4c). Visual inspection
showed the existence of a single phase, while chromatographic
testing of the uppermost regions of the reactor interior confirmed
complete dissolution of the amine.
In the presence of both H2 (85 atm) and CO2 (total pressure
210 atm) at the same temperature, the situation is dramatically
different. With amounts of N(C2H5)3 below or equal to ∼4
mL (29 mmol), only a single phase is visible (Figure 4d),
although for amounts between 2 and 4 mL waviness is observed,
indicating the existence of a density gradient (Figure 4e). The
Synthesis of Alkyl Formates. The alkyl formate syntheses were
performed by the same method as the formic acid synthesis, except
that higher temperatures were used and alcohols rather than water were
added. The reaction mixtures were analyzed in the manner described
for the production of formic acid except that the deuterated solvent
was CDCl3 and the internal standard the methyl peak of toluene.
Additional confirmation of the identification and yield of methyl formate
was obtained, for selected reactions, by GC/MS (30 m × 0.25 cm TC-
wax column, 40 °C for 3 min followed by ramping to 100 °C at 20
(57) Osakada, K.; Ohshiro, K.; Yamamoto, A. Organometallics 1991,
10, 404-410.

Homogeneous Catalysis in Supercritical Fluids
J. Am. Chem. Soc., Vol. 118, No. 2, 1996 347
Figure 4. Schematic illustration of the phase behavior of the CO2/
H2/N(C2H5)3 system in a 50-mL reaction vessel at 50 °C, showing the
major components of each phase and the relative volumes.
mL. Chromatographic testing of the upper gaslike phase shows
that it contains very little amine. Thus, H2 and possibly CO2
are the major components of this phase. It is likely that the
large amount of CO2 in the lower phase increases the solubility
of H2 in that phase relative to the solubility of H2 in liquid amine
alone.
Figure 2. Photographs showing the interior of the window-equipped
reactor vessel (see Figure 3) under various conditions: (top left) gently
boiling liquid (lower phase) and gaseous (upper phase) CO2 at 30 °C,
(top right) scCO2 (32 °C), (middle left) N(C2H5)3 (14 mL) under 1 atm
of CO2 gas, (middle right) the same amount of amine under increasing
pressure of CO2, showing the turbulence and the increase in volume
of the liquid phase due to dissolution of CO2 into the liquid amine,
(bottom left) single-phase reaction mixture of H2 (85 atm), N(C2H5)3
(5 mmol), H2O (0.1 mmol), and 3 (3 µmol) in scCO2 (total 216 atm)
at 50 °C just after the start of the reaction, and (bottom right) the same
reaction mixture after the reaction is complete. The drops are liquid
HCO2H/N(C2H5)3 adduct, the bulk of which rests at the bottom of the
reactor and is not visible in the photograph.
Similar observations were made with THF (15 mL), CH3CN
(15 mL), CH3OH (10 mL), and DMSO (0.5 mL), all at 50 °C,
CH3OH (10 mL) at 80 °C, and THF (10 mL) at 100 °C. In
each case two phases are clearly observed and the volume of
the lower phase is significantly larger than the amount of liquid
solvent added, suggesting that a large amount of CO2 is present
in the lower phase. Only with water (10 mL) is the volume of
the lower phase equal to the volume of water added, suggesting
that the amount of CO2 dissolved in the water is small.58,59
Homogeneous Hydrogenation of scCO2 to Formic Acid.
The hydrogenation of scCO2 to formic acid proceeds rapidly
with the use of trimethylphosphine complexes of ruthenium(II)
as catalyst precursors (Table 1). In a typical reaction, N(C2H5)3
(5.0 mmol), the catalyst precursor (3 µmol), and water (0.1
mmol) were kept at 50 °C in a supercritical mixture of H2 (85
atm) and CO2 (total pressure 200-210 atm). At the start of
the reaction only one phase is present (Figure 2, bottom left),
but as the reaction proceeds, the liquid product, an adduct of
N(C2H5)3 and HCO2H, precipitates (Figure 2, bottom right).
Table 1 summarizes the results of the catalyst screening,
indicating the AAR, the yield of formic acid, the TON, and the
rate expressed in terms of the turnover frequency (TOF )
turnovers/h ) mol of product/mol of catalyst/h).
The most active of the catalyst precursors tested in scCO2
are RuH2[P(CH3)3]4 (1), RuCl2[P(CH3)3]4 (2), and RuCl(O2-
CCH3)[P(CH3)3]4 (3). Reactions catalyzed by 1 and 3 do not
Figure 3. Cutaway drawing of the 50-mL window-equipped reaction
vessel, showing (a) sapphire windows, (b) stir bar, (c) seal, (d) inlet/
outlet for gases, and (e) inlet for liquid reagents, usually plugged. The
other 50-mL vessels are similar except that they lack the window
assemblies. Photographs taken through windows “a” are shown in
Figure 2.
density gradients can be eliminated by vertical shaking but not
by horizontal stirring alone. Amounts of N(C2H5)3 above 4 mL
cause the formation of two phases, the lower of which is always
considerably larger in volume than the charged amine. For
example, with 20 mL of N(C2H5)3, the volume of the lower
phase is 40 mL, suggesting that both amine and CO2 are in that
phase (Figure 4f). With 5 mL of amine, the lower phase is 17
exhibit an induction period. With 1, the reaction is fastest in
(58) Takenouchi and Kennedy59 showed that the maximum concentration
of CO2 in liquid water at 100 bars is e1.4 mol % and varies little with
temperature over the temperature range tested (110-350 °C).
(59) Takenouchi, S.; Kennedy, G. C. Am. J. Sci. 1964, 262, 1055-1074.

348 J. Am. Chem. Soc., Vol. 118, No. 2, 1996
Jessop et al.
a
Table 1. Effect of Catalyst Precursor on Yield and Rate of Hydrogenation of scCO2
catalyst precursor
RuH2[P(C6H5)3]4
RuH2[P(C6H5)3]4
additiveb
catalyst (µmol)
time (h)
yield (mmol)
AARc
TONd
TOFe (h-1
)
H2O
CH3OH
H2O
H2O
CH3OH
H2O
H2O
H2O
H2O
H2O
2.8
4.5
2.2
3.2
3.4
2.7
3.2
3.4
3.2
1.8
10
1
0.5
1
3
0.5
1
16
1
5
15
17
143
0.22
3.5
3.0
6.0
2.6
0.63
8.1
3.6
0
0.044
0.69
0.60
1.2
0.51
0.13
1.6
0.72
0
0.08
0.029
0.18
80
770
1400
1900
760
230
2600
1100
0
80
1500
1400
630
1500
230
160
1100
0
RuH2[P(CH3)3]4 (1)
RuH2[P(CH3)3]4 (1)
RuH2[P(CH3)3]4 (1)
RuCl2[P(CH3)3]4 (2)
RuCl2[P(CH3)3]4 (2)
RuCl(O2CCH3)[P(CH3)3]4 (3)
f
trans-RuCl2(dmpe)2
f
trans-RuHCl(dmpe)2
0.42
0.14
0.9
230
14
15
0.8
Ru3(CO)12
Pd/C
H2O
H2O
10g
a Conditions: 50 °C, 80-85 atm of H2, total pressure 200-210 atm, 50-mL reaction vessel, 5.0 mmol of N(C2H5)3. Data rounded to two
significant figures. b Additive ) 0.1 mmol of water (single-phase system) or 250 mmol of CH3OH (two-phase system). c Formic acid to amine
mole ratio. d Turnover number ) mol of product (formic acid)/mol of catalyst. TON is a unitless parameter. e Turnover frequency ) TON/h. Units
are h-1 f dmpe ) (CH3)2PCH2CH2P(CH3)2. g 31 mg of Pd/C, 10 wt % Pd.
.
Table 2. Effect of the Base on the Yield of Hydrogenation of
a
scCO2
base
catalyst time
yield
base
(mmol) (µmol) (h) (mmol) AAR TON
none
K2CO3
KOH
[NH4][O2CNH2]
N(C2H5)3
N(C2H5)3
N(C2H5)3
N(C2H5)3
0
2.5
5.4
2.4
13.2
9.0
2.9
2.5
3.2
2.7
3.2
3.4
13
16
15
15
16
20
84
47
0
1.2
0
140
260
0.75
0.099 0.021
8.1
12.0
18.9
24.4
39
5.0
1.6
1.2
0.63
0.63
2600
4400
6000
7200
10.0
30.2b
40.3b
a Conditions: 50 °C, 50-mL reaction vessel, 80-85 atm of H2, total
pressure 200-210 atm, catalyst precursor 2, 0.1 mmol of water. b In a
150-mL reaction vessel at 220 atm of total pressure.
Figure 5. Dependence of formic acid yield on reaction time.
Conditions: 85 atm of H2, total pressure 200-210 atm, 50 °C, 5.0
mmol of N(C2H5)3, 2.5-3.2 µmol of Ru catalyst, 0.1 mmol of H2O.
precursors have equal activity (Table 1). This suggests that the
difference in activity between the two complexes in scCO2 is a
result of differing solubilities rather than electronic or steric
effects.
Because of the nature of the equipment, we were not able to
rule out the possibility that the catalyst decomposed to form
some catalytically active solids. Blank tests without catalyst
were performed repeatedly throughout the study to confirm that
the reactor walls or any species thereon were not catalytically
active.
Effect of the Base. The presence of a base is crucial for
favorable thermodynamics.28 In the supercritical system, the
yield of formic acid is high in the presence of N(C2H5)3, while
in the absence of base, no formic acid is obtained (Table 2).
Use of the solid bases K2CO3, KOH, or [NH4][O2CNH2] also
allows production of formic acid, albeit with low yields. A
combination of N(C2H5)3 and KOH is no more effective than
the same amount of amine alone. The amine acts as a sink for
the acid and is saturated at an AAR of ∼1.7, the highest AAR
value observed in this system. Leitner et al.29 observed 1.6-
1.8 in DMSO and about 1.0 in water.31 The ratios in nonprotic
solvents are higher than 1 because carboxylic acids and
N(C2H5)3 form stable 2:1 adducts, 4, in nonprotic solvents or
in the absence of any solvent.60,61
the first hour, with a TOF of 1400 h-1, dropping to one-half
this rate in the second hour (Figure 5). After the third hour, no
further increase in yield is observed. On the other hand, the
reactions catalyzed by 2 exhibit an induction period of ca. 1 h
during which the yellow color of the catalyst disappears. The
TOF in the subsequent 2 h is over 1000 h-1. After 5 h, the
hydrogenation catalyzed by 2 is complete, as indicated by an
AAR ratio of 1.6-1.7.
As the reaction proceeds, the conditions change from basic
(AAR < 1) to acidic (AAR > 1). The yield of formic acid
obtained with 1 is lower than with 2, with AAR values of up to
1.2 for the former catalyst, far below the values of 1.6-1.7
obtained with 2. This could be due to catalyst instability in
the acidic conditions present after AAR reaches 1.0. To test
this possibility, experiments were performed in which formic
acid (0.9 AAR) was added before the start of the reaction. The
AAR climbed to 1.6 with catalyst 1 or 1.7 with catalyst 2 within
4 h. This demonstrates that both catalyst precursors still have
activity even in the acidic conditions present at AAR > 1.
The other catalyst precursors tested were inferior in activity.
Gray material was found on the walls of the reactor after a Rh
catalyst precursor, [RhCl(nbd)]2/(c-C6H11)2PCH2CH2P(c-C6H11)2,
was tested. The scCO2-soluble catalyst precursors Ru3(CO)12
and trans-RuHCl(dmpe)2 and the heterogeneous catalyst Pd/C
have low activity, while soluble trans-RuCl2(dmpe)2 is com-
pletely inactive.
The solubility of Ru complex 2 was demonstrated qualita-
tively by passing a scCO2 solution of the complex through a
fine filter and a back-pressure regulator at 50 °C and 120 atm
and collecting the solid which precipitated at the vent. The 1H
NMR spectrum of the collected material was identical with that
of the starting material.
The amount of N(C2H5)3 added has a strong effect on the
rate of reaction. In the 50-mL vessel, the optimum amount of
(60) Barrow, G. M.; Yerger, E. A. J. Am. Chem. Soc. 1954, 76, 5211-
5216.
(61) Wagner, K. Angew. Chem., Int. Ed. Engl. 1970, 9, 50-54.
In scCO2, catalyst precursor 1 is far more active than
RuH2[P(C6H5)3]4, although in liquid CH3OH the two catalyst

Homogeneous Catalysis in Supercritical Fluids
J. Am. Chem. Soc., Vol. 118, No. 2, 1996 349
With CH3OH as the promotor, as with water, the rate is
greater if only a single phase is present (>4000 h-1) than if
enough CH3OH is used to form a second phase (1500 h-1).
However, it is important to note that the second phase formed
is actually a mixture of CH3OH and CO2. Depending on the
compositions and properties of the phases, it could be that this
lower phase is supercritical, rather than liquid, while the upper
phase would be predominantly gaseous H2. For this reason,
the term “two-phase system” may be more accurate than “liquid
CH3OH”. The initial rate of the single-phase reaction with CH3-
OH is too high to measure with the existing equipment;63 the
reaction is complete within 0.5 h. The initial rate must be
greater than 4000 h-1
.
Figure 6. Effect of the amount of N(C2H5)3 on the initial rate of formic
acid production as measured during the first 0.5 or 1 h. Conditions:
50 °C, 3 µmol of RuH2[P(CH3)3]4 (1), 0.1 mmol of H2O, 85 atm of H2,
total pressure 200-210 atm.
The trend of single-phase systems being faster that two-phase
systems might also be true for the case of DMSO as the additive,
but the reactions are too fast to measure.63 Although the
solubility of H2 in pure DMSO is probably very low, the
solubility of H2 in the DMSO/CO2 mixture which exists under
the reaction conditions could be considerably higher. The
effectiveness of DMSO as a medium for subcritical CO2
hydrogenation has been observed previously.32,33,64
N(C2H5)3 is 5.0 mmol (Figure 6). The rate at 15.0 mmol of
amine, in the region of density gradients but below the solubility
limit (∼30 mmol), is 60 times lower than with 5.0 mmol and is
almost as low as the rate obtained in liquid amine (72 mmol)
under otherwise identical conditions. Vertical shaking during
the reaction time, which eliminates the density gradients, does
not affect the rate. It is clear that the rate of the reaction can
be affected not only by the kinetic dependence on amine
concentration but also by phase behavior considerations. For
example, if the system is close to the mixture critical point at
5.0 mmol of amine, then a favorable clustering effect62 could
be the explanation for the high rate at that concentration.
Because the phase behaviors of this CO2/H2/N(C2H5)3 three-
component mixture and the CO2/H2/N(C2H5)3/HCO2H four-
component mixture have not been mapped, only speculation is
possible. Increasing the amount of amine and switching to a
proportionately larger reactor allow increased yields of up to
7200 TON, although with reduced AAR values.
Effect of Water and Other Additives. The rate of the
hydrogenation is improved by the addition of promoting
additives. These additives can be used either in small amounts
so that they can dissolve entirely in the scCO2 phase or in large
amounts so that they will cause the formation of a second phase.
The initial rate was determined by measuring the yield after
short reaction times with Ru complex 1 and a variety of additives
(Table 3).
In the absence of any promoter other than 1 and the amine,
the initial turnover frequency of formic acid production is 680
h-1 at 50 °C. With water as an additive (0.1 mmol, completely
dissolved in scCO2), the rate increases to 1400 h-1. However,
with 10 mL (560 mmol) of added water, a second (aqueous)
phase forms and the rate drops to 34 h-1. The slow rate of the
CO2 hydrogenation in the presence of liquid water was
unexpected because the catalyst should not dissolve in that
phase. A possible reason for the low rate could be reaction of
the water with CO2 and amine to form carbonates which would
render the amine insoluble in scCO2 and soluble in water (eq
4).
Reactions with two phases with THF or CH3CN as the
additive, even with water added as a promoter, have low rates
of reaction. In fact, CH3CN seemed to act as an inhibitor even
when it did not form a second phase, possibly because it binds
too strongly to the Ru center and impedes the catalytic cycle.
Other additives such as ethylenediamine, P(CH3)3, and carbon
monoxide (Table 3) have an inhibiting effect on the reaction.
However, a large amount (3.9 mmol) of P(CH3)3 was required
in order to decrease the rate of reaction; use of only 39 µmol
(20 equiv) had essentially no effect.
At 50 °C, the effect of additives on the total yield (Table 4)
is not as dramatic as their effect on the rate. Slightly higher
yields of formic acid are obtained in a scCO2/CH3OH single-
phase system (1.8 AAR) than in a scCO2/CH3OH biphasic
system (1.6 AAR) or scCO2/H2O single-phase system (1.6
AAR). In contrast, the yield at 80 °C is highly dependent on
the choice of additive, as will be described below.
Effect of H2 Pressure on Yield. With a total pressure of
200-210 atm, the pressure of hydrogen was varied to determine
the effect of this variable on the yield of overnight reactions
catalyzed by 2 in a 50-mL reactor. Pressures of 60-85 atm
give the optimum AAR values of 1.6-1.7. The yield of formic
acid is independent of the amount of catalyst over a range of
3-10 µmol (at 85 atm, 5.0 mmol of amine), indicating that the
reaction reaches equilibrium. At 34 atm of H2, however, a low
yield is obtained (AAR ) 0.3 after 18 h), while no formic acid
is obtained in the absence of H2.
An attempt at the transfer hydrogenation of scCO2 by
2-propanol (18 mmol of 2-propanol, 4.9 µmol of 1, 5.0 mmol
of N(C2H5)3, 130 atm of CO2, no H2, 18 h, 50 °C) was not
successful even though the reaction is enthalpically neutral (eq
5, B ) amine base). Neither formic acid, propyl formate, nor
acetone was detected in the reaction mixture. RuH2[P(C6H5)3]4
is a known hydrogen transfer catalyst in other systems.65
R3N + CO2 + H2O h [R3NH][HCO3]
(4)
(63) Because these reactions are too fast, it was not possible to obtain
reliable measurements of the true initial rate, even by using smaller amounts
of catalyst.
(64) The observations that high yields of formic acid can be obtained in
DMSO solution and that low yields are obtained at higher temperatures
have been explained as an effect of entropy,32 although this is unlikely to
be correct because in general the position of an equilibrium depends on
enthalpy rather than entropy.
Attempts at using bases insoluble in scCO2 have yielded
unsatisfactory results (Table 2); a scCO2-dissolved base is
preferred.
(62) Combes, J. R.; Johnston, K. P.; O’Shea, K. E.; Fox, M. A. In
Supercritical Fluid Technology: Theoretical and Applied Approaches to
Analytical Chemistry; Bright, F. V., McNally, M. E. P., Eds.; ACS
Symposium Series 488; American Chemical Society: Washington, DC,
1992; pp 31-47.
(65) Johnstone, R. A. W.; Wilby, A. H.; Entwistle, I. D. Chem. ReV.
1985, 85, 129-170.

350 J. Am. Chem. Soc., Vol. 118, No. 2, 1996
Jessop et al.
a
Table 3. Effect of Temperature, Water, and Other Additives on the Initial Rate of Hydrogenation of scCO2
T (°C)
additive (mmol)
water (mmol)
phasesb
catalyst (µmol)
time (h)
yield (mmol)
AAR
TOF (h-1
)
15
50
50
50
50
50
50
50
50
50
50
50
50
50
50
50
80
80
80
0.1
0
2
1
1
1
2
1
1
1
d
2
1
2
1
2
1
2
1
1
2
3.2
2.2
2.7
2.2
2.9
2.5
1.9
2.9
3.4
2.0
3.2
3.2
2.5
3.4
2.9
2.5
3.4
2.5
2.5
16
1
0.5
1
0.068
1.5
1.9
0.013
0.30
0.38
0.60
0.020
0.012
0.25
0.11
0.065
0.033
0.23
0.21
1.0
0.51
0.97
0.99
0.17
0.85
1.4
1.3
680
0.1
0.1
560
0.1
0.1
0.1
0
0.1
0.1
0.1
0
0
0
0
0.1
0
1400c
1400
34
3.0
1
0.10
0.059
1.3
0.53
0.32
0.17
1.2
1.1
5.0
2.6
4.9
CO (1 atm)
P(CH3)3 (0.039)
P(CH3)3 (39)
0.5
0.5
0.5
0.6
1
1
1
0.5
0.5
0.5
0.5
0.5
1
48
1300
360
190
84
360
330
H2N(CH2)2NH2 (0.068)
THF (190)e
CH3CN (6.7)
CH3CN (290)e
CH3OH (13)
>4000f
1500
>3300f
>4000f
500
CH3OH (250)g
DMSO (1.6)
DMSO (140)g
5.0
0.85
3.9
CH3OH (13)
1600
2600
CH3OH (250)g
0
1
6.5
a Conditions: Ru complex 1, 5.0 mmol of N(C2H5)3, 80-85 atm of H2, total pressure 200-210 atm, 50-mL reaction vessel. b Number of phases
visible at the start of the reaction. c Average of two runs. d Phase behavior not determined. e 15 mL. f Reactions with AAR ) 1 are essentially
complete; thus the initial rate must be higher than the rate shown.55 g 10 mL.
Table 4. Production of Formic Acid and Methyl Formate at
a
Various Temperatures in scCO2
TON
T
additive N(C2H5)3 time
(°C) additive (mmol) (mmol)
(h) AAR HCO2H HCO2CH3
50 H2O
0.1
13
250
80b
0.1
5.0
5.0
5.0
16 1.6
15 1.8
15 1.6
60 0.70
15 0.04
2600
3300
2900
6700
61
0
150
100
270
0
50 CH3OH
50 CH3OH
50 CH3OH
80 H2O
80 CH3OH
80 CH3OHc
80 CH3OH
80 CH3OH
80 CH3OH
100 CH3OH
30b
5.0
5.0
5.0
13
1
0.22
360
0
13
16 0.89
64 0.66
1100
6800
2100
2200
250
330
3500
8
890
150
80b
30b
250
250
13
5.0
5.0
5.0
1
1.1
Figure 7. Temperature dependence of the rate of formic acid
production. Conditions: 5.0 mmol of N(C2H5)3, 3 µmol of catalyst (L
) P(CH3)3), 0.1 mmol of H2O, 85 atm of H2, total pressure 200-210
atm, 0.5 or 1 h.
19 1.2
16 0.17
a Conditions: 50-mL reaction vessel, no water added except where
indicated, 2.5-3.5 µmol of Ru complex 2, 80-85 atm of H2, total
pressure 200-210 atm. b In a 300-mL reaction vessel. c 4.2 µmol of
2.
CO2 + B + (CH3)2CHOH f [BH][O2CH] + (CH3)2CO
(5)
Effect of Temperature. The greatest rate of reaction with
catalyst 1 and H2O as promoter was observed at 50 °C. The
rates were calculated from the yield of formic acid after reactions
of short duration, usually 0.5 h. The results (Figure 7) show
that at temperatures lower than 50 °C, the rate is very low.
Pretreatment of the catalyst precursor with H2 at 50 °C did not
increase the rate of the reaction at 40 °C. Visual inspection of
a mixture of N(C2H5)3 (0.7 mL, 5.0 mmol), CO2 (120 atm),
and H2 (80 atm) showed density gradients at temperatures below
about 45 °C but not a defined second phase. While it is normal
for a reaction rate to decrease with decreasing temperature, the
steepness of the observed drop in rate suggests that phase
behavior rather than reaction kinetics is responsible. At 15 °C
the rate is extremely slow, probably due to the phase separation
which exists at this temperature. Above 50 °C, the rate
decreases with increasing temperature. Note that at these higher
temperatures, the yield after 15-16 h also drastically decreases
with temperature (Table 4). For example, the final yield of
formic acid obtained after 15-16-h reactions with catalyst
precursor 2 and H2O as promoter is 8.1 mmol (AAR ) 1.6) at
50 °C but only 0.17 mmol (AAR ) 0.04) at 80 °C. Addition
of methanol to the system at higher temperatures greatly
increases both the rate and the final yield (Tables 3 and 4) to
the extent that the reaction is far too fast to determine the initial
Figure 8. Time dependence of the product yields after reactions at 80
°C in a 50-mL reaction vessel with 80 atm of H2, 130 atm of CO2, 5.0
mmol of N(C2H5)3, 3-4 µmol of 2, and 13 mmol of CH3OH.18
rates of formic acid production. In contrast, adding THF (5
mmol) does not increase the yield of the reaction (0.005 mmol
of formic acid after 15 h at 80 °C).
Production of Methyl Formate. The hydrogenation of
scCO2 in the presence of CH3OH produces methyl formate in
excellent yield in addition to formic acid (eq 2). The presence
of a base such as N(C2H5)3 is again necessary. During the
reaction with catalyst 2 at 80 °C, after an induction period of
about 1 h, formic acid is produced rapidly. The concentration
of formic acid reaches an equilibrium value of 0.80-0.89 mol/
mol of the amine by the end of the second hour (Figure 8). The
amount of formic acid is constant after this time. Methyl
formate is produced more slowly. These results clearly show

Homogeneous Catalysis in Supercritical Fluids
J. Am. Chem. Soc., Vol. 118, No. 2, 1996 351
Table 5. Acetic Acid Thermal Esterification with Methanol in the
Presence or Absence of Triethylaminea
esterification of acid (%)
additive (g)
gas
with amine
without amine
CO2
18
25b
28
25
24
20
argon
argon
argon
argon
argon
88
pentane (0.6)
MS 3A (0.5)
Amberlyst (0.1)
Nafion 511 (0.1)
100
100
a Conditions: 44 h, 100 °C, 5.0 mmol of CH3CO2H, 13 mmol of
CH3OH, either 5.0 or 0 mmol of N(C2H5)3, 1 atm of argon or 150 atm
of CO2. MS 3A is a molecular sieve, while Amberlyst and Nafion
511 are acidic resins. b This value decreases to 16% in the presence of
10 mmol of amine.
Figure 9. Effect of the choice of base on the yield of formic acid and
methyl formate. Conditions: 0.72 mmol of base, 13 mmol of CH3OH,
3-5 µmol of 2, 80 °C, 16 h.
a
Table 6. Production of Formamides from Amines and scCO2
TON
amine
(mmol)
catalyst
(µmol)
time
(h)
amine
amide
HCO2H
that methyl formate is synthesized in a two-step pathway: Ru-
catalyzed hydrogenation of scCO2 to formic acid (eq 1) followed
by thermal esterification to methyl formate (eq 6).
NH(C6H11)2
NH(i-C3H7)2
NH(C2H5)2
NH2(n-C3H7)
NH(CH3)2
NH(CH3)2
NH(CH3)2
5.0
5.0
5.0
5.0
7.5
3.4
2.3
2.3
3.4
2.9
3.8
2.5
15
23
13
5
5
15
22
0
0
820
1400
1600
950
620
190
0
260
HCO2H + CH3OH f HCO2CH3 + H2O
(6)
2100
1500
9900b
7.5
The maximum yield of methyl formate is obtained at ca. 80 °C
due to slow thermal esterification at lower temperatures and
low catalytic activity for hydrogenation at higher temperatures.18
Increasing the amount of CH3OH causes a second phase to
form. The yields of formic acid and methyl formate and, to a
lesser extent, the selectivity increase (Table 4). Decreasing the
amount of CH3OH added to the reaction causes dramatically
decreasing yields of both formic acid and methyl formate.
Against expectations, however, the selectivity for methyl formate
increases.18
Base is added to increase the yield of the formic acid in the
hydrogenation step, but the tests described below show that base
inhibits the subsequent thermal esterification. Thus the role of
base is critical to the yield and selectivity for formate ester
(Figure 9). Reducing the amount of amine and keeping the
concentration of alcohol constant does not increase the selectiv-
ity for the ester but only decreases the final yield.18 Substituting
N-methylpiperidine for N(C2H5)3 has little effect on the yield
or selectivity. With weak bases or in the absence of any base,
very low yields of methyl formate are produced with complete
selectivity. In the absence of base and with catalyst 1, no methyl
formate is obtained.
Normally, esterification of carboxylic acids by alcohols is
catalyzed by an acid. However, this synthesis of methyl formate
requires an uncatalyzed thermal esterification in the presence
of an excess of base, a process which is not used in organic
chemistry. To confirm that the reaction can proceed in the
presence of base, formic acid (4 mmol), CH3OH (13 mmol),
and N(C2H5)3 (5 mmol) were heated to 80 °C in a closed vessel
under argon. After 15 h, 7% of the acid had been esterified to
methyl formate. In order to determine the factors which
influence esterification in basic solutions, we have tested the
thermal esterification of CH3OH and acetic acid under argon
without solvent (eq 7, Table 5). Acetic acid was chosen because
it is less volatile and more resistant to decomposition than formic
acid.
25.6b
0
a Conditions: catalyst precursor 2, 80 atm of H2, 130 atm of CO2,
100 °C, 50-mL reaction vessel. b 150-mL reaction vessel.
completely eliminated. Molecular sieves are also ineffective
in the presence of amine. The conversion to ester is slightly
lower in scCO2 than under argon.
The esterification of formic acid (eq 6) is reversible. An
equimolar mixture of methyl formate and water heated to 100
°C for 28 h under argon and in the absence of any catalyst gives
7% conversion to formic acid and CH3OH. In the presence of
catalyst precursor 2, the same amount of CH3OH is observed,
but formic acid is not observed, presumably due to catalyzed
decomposition of the formic acid to CO2 and H2. These results
suggest that improved selectivity for ester formation could be
achieved by effective removal of the water.
Production of Formamides from Primary or Secondary
Amines. In the presence of diethylamine, dimethylamine, or
n-propylamine instead of N(C2H5)3, the hydrogenation of scCO2
with catalyst 2 at 100 °C produces N-substituted formamides
(eq 3) instead of or in addition to the ammonium formate salts
(Table 6). In the case of dimethylamine, 99% conversion to
and selectivity for DMF can be obtained. Bulky dialkylamines
such as dicyclohexylamine and diisopropylamine produce only
ammonium formate salts, which suggests that the rate of
dehydration of such salts to the corresponding formamides is
strongly influenced by steric factors. Surprisingly, ammonia,
introduced as the carbamate (Table 7), gave only low conversion
to the formamide, although the reason for its poor reactivity is
more likely to be its poor solubility in scCO2. A byproduct
from the reaction of NH(CH3)2 was detected as a singlet in the
1H NMR spectrum with a chemical shift identical with that of
N(CH3)3. Yields of this product ranged up to 2 mol % but were
usually negligible.
Dialkylamines react reversibly with CO2 to form dialkylam-
monium dialkylcarbamates (eqs 8 and 9).56,66-68
CH3CO2H + CH3OH f CH3CO2CH3 + H2O
(7)
Triethylamine acts as an inhibitor, as expected, but even
excess amine does not completely prevent esterification. In the
absence of amine, acidic resins are catalysts for the esterification,
increasing the conversion from 88% to 100%. In the presence
of amine, however, the effectiveness of the acidic resins is
(66) Wright, H. B.; Moore, M. B. J. Am. Chem. Soc. 1948, 70, 3865-
3866.
(67) Fields, S. M.; Grolimund, K. J. High Resolut. Chromatogr.,
Chromatogr. Commun. 1988, 11, 727-729.
(68) Takeshita, K.; Kitamoto, A. J. Chem. Eng. Jpn. 1988, 21, 411-
417.

352 J. Am. Chem. Soc., Vol. 118, No. 2, 1996
Jessop et al.
Table 7. Production of Formamides from Ammonium Carbamates,
a
H2, and scCO2
TON
carbamate reactor
PH
PCO time
2
2
amine
NH3
(mmol) vol (mL) (atm) (atm) (h)
amide HCO2H
5
5
5
5
5
50
50
50
50
50
80
80
80
80
80
130 20
130 12
130 15
500 1700
NH2CH3
1800
2000
20
410
NH2C2H5
NH(CH3)2
NH(CH3)2
NH(CH3)2
NH(CH3)2
NH(CH3)2
NH(CH3)2
NH(CH3)2
NH(CH3)2
NH(CH3)2
NH(CH3)2
130
130
0.5
820 2300
4
2600
76
250
4
5
5
50
50
80b 130
4b
80
80
80
80
130 14
130 22
130 19
130 18
2800
90
31
79
210
1900
5
150
150
150
300
50
25 000
62 000
280
680
Figure 10. Composition of the product mixture as a function of reaction
time during the synthesis of DMF from NH(CH3)2 (10 mmol, charged
as the carbamate), 80 atm of H2, and 130 atm of scCO2 at 100 °C
catalyzed by 2.5 µmol of 2.
150 000 4000
420 000
1000 1600
320 160
80c 130c 70
0
86
50
57
0
1
3
5
50
a Conditions: 2-3 µmol of catalyst 2 amines or ammonia charged
as the carbamate, 100 °C. b 1 atm of CO also present. c Extra H2 and
CO2 added several times during the reaction.
higher yields should be possible. The high yield reactions in
Table 7 have overall rates69 of up to 8000 h-1
.
By performing a series of reactions (5.0 mmol of dimethyl-
ammonium dimethylcarbamate) with varying reaction times, a
profile of the reaction was obtained (Figure 10).17 The data
show that formic acid is generated very quickly but reaches a
maximum (6 mmol) at about 0.5 h, after which it declines to
0.2 mmol (3%). Under other conditions or with longer reaction
times, the amount of formic acid remaining is lower or even
undetectable. DMF is produced more slowly than formic acid
but reaches 90% conversion by about 5 h. This reaction profile
is consistent with the catalytic hydrogenation of scCO2 to the
dimethylammonium salt of formic acid followed by dehydration
to DMF. Tests with formic acid and dimethylammonium
dimethylcarbamate showed that the dehydration can be achieved
at 100 °C without catalyst (eq 10). The reaction of 5 mmol
each of formic acid and carbamate, formic acid being the
limiting reagent, gave 100% conversion to DMF at 100 °C after
17 h under 3 or 130 atm of CO2. The same reaction at 150 °C
without solvent has been reported in the patent literature.56
NHR2 + CO2 h R2NCO2H
(8)
(9)
R2NCO2H + NHR2 h [NH2R2][O2CNR2]
These reactions have a number of consequences for the
formamide synthesis. The solubility of nonbulky primary and
secondary alkylamines in CO2 is low67 because of carbamate
formation.66,68 Thus a liquid (R ) CH3) or solid (R * CH3)
carbamate phase, [NH2R2][O2CNR2], is present from the start
of the formamide synthesis. Tests at room temperature and
pressure showed that the Ru catalyst is insoluble in the liquid
carbamate salt. Thus, the initial reaction is believed to take
place in the scCO2 phase. We speculate that the catalyst remains
in the supercritical phase throughout the reaction,17 which may
be the reason for the high rate and yields (see below).
Gaseous amines such as NH(CH3)2 can be charged to the
reactor more conveniently in the form of the liquid or solid
carbamate salts, which can be handled at room temperature.
DSC experiments under argon showed that the temperatures of
the onset of decomposition for the solid carbamates of NH3,
NH2CH3, and NH2CH2CH3 are 38, 52, and 48 °C, respectively.
For the DMF synthesis, the use of dimethylammonium di-
methylcarbamate is experimentally easier than the use of cooled
liquid dimethylamine, although the same results are obtained
by either method. In particular, the time profile shown in Figure
10 for reactions of dimethylammonium dimethylcarbamate is
identical with that obtained with dimethylamine as the starting
material. The results obtained with several carbamates are
shown in Table 7.
By using the carbamate as the source of NH(CH3)2 and by
using a large carbamate to catalyst ratio, an exceedingly high
efficiency of conversion of H2, CO2, and NH(CH3)2 can be
obtained. For example, with 79 mmol of carbamate (158 mmol
of amine) in a 150-mL reactor, the catalytic efficiency was
62 000 TON with 99% conversion of the amine and 99%
selectivity for DMF. If very large amounts of carbamate are
used, H2 becomes the limiting reagent. In order to avoid this
problem and the related problem of dropping scCO2 pressure,
extra H2 and CO2 were periodically added during large scale
reactions, although the H2/CO2 ratio could not be monitored.
By this method, DMF was obtained with 420 000 TON (71%
yield based on charged amine) in a 300-mL reactor. Noteworthy
in this latter reaction was the absence of any formic acid among
the products, despite the presence of unreacted amine. This
suggests that the hydrogenation step slowed down late in the
reaction and became the rate-determining step, probably because
of insufficient H2 or CO2 pressure. With proper monitoring of
the H2/CO2 partial pressures and controlled makeup gas, even
2HCO2H + [(CH3)2NH2][O2CN(CH3)2] f
2HCON(CH3)2 + 2H2O + CO2 (10)
The reverse reaction, hydrolysis of DMF to dimethylammonium
formate, was not observed under these conditions; DMF and
water (10 mmol each) heated to 100 °C in the presence or
absence of 3 did not react within 20 h. Therefore the conversion
of CO2, H2, and NH(CH3)2 to DMF is irreversible under these
conditions.
The initial rates of production of formic acid and DMF over
the first hour at the conditions of Figure 10 are 2300 and 1300
h-1, respectively. Addition of 10 mL of THF, which is too
much to dissolve in the scCO2/H2 mixture, causes the initial
rates to be reduced to 390 and 740 h-1. The dramatic decrease
in the rate of hydrogenation may indicate that the THF induced
the catalyst to dissolve in the liquid phase.
In the absence of Ru catalyst, only trace amounts of formic
acid and DMF are observed. Using no CO2 except for that
contained in the carbamate results in an extremely slow reaction.
An atmosphere of CO strongly inhibits the DMF synthesis. The
synthesis of DMF from scCO2 also proceeds at the lower
temperature of 75 °C, although the rate of the dehydration step
is decreased (1500 mol of HCO2H and 2400 mol of DMF per
mol of 2 or 62% conversion of NH(CH3)2 to DMF after 5 h).
The use of 8 atm of D2 gas and a reaction time of only 75
min produced DMF-d1 in 3% conversion and 80% isotopic
(69) If the same rate were obtained in a flow system, the space-time
yield would be 9 kg of DMF/g of Ru catalyst/L (reactor volume)/h. Because
such figures are usually calculated from flow rather than batch reactions,
this value should only be considered an indicator of high productivity.

Homogeneous Catalysis in Supercritical Fluids
J. Am. Chem. Soc., Vol. 118, No. 2, 1996 353
purity. The mass spectrum of the product DMF suggests that
the deuteron was at the formyl position rather than in a methyl
group. An isolated single peak at 58 m/z (molecular ion minus
CH3) in DMF-d0 was shifted to a single peak at 59 m/z in the
spectrum of DMF-d1 rather than to two peaks at 58 and 59 m/z,
as would have been expected if the deuteron resided in the
methyl groups.
Stoichiometric Reaction of RuH2[P(CH3)3]4 with CO2.
Complex 1 in C6D6 is converted by CO2 (1 min bubbling at 1
atm) in 9% yield to a monohydride complex with an NMR
spectrum consistent with cis-RuH(O2CH)[P(CH3)3]4. The extent
of the conversion was unaffected by the presence or absence of
4 equiv of free P(CH3)3.
Reasons for the High Rate of Reaction in scCO2. Possibly
the most important reason for the high rate obtained in scCO2
is the complete miscibility15 of H2 with scCO2 compared to the
low solubility of H2 in most organic solvents.16 The miscibility
of H2 allows all of the CO2, H2, N(C2H5)3, and catalyst to be in
the same phase, which is not possible with subcritical conven-
tional solvents. This strategy requires the designing of a catalyst
which is highly soluble in scCO2; this point will be discussed
in further detail below. Less obvious in the present system but
potentially as important to the high rate are other factors such
as a weaker coordination sphere around the catalyst, rapid
diffusion in the supercritical phase, and elimination of the
problem of slow mass transfer between the gaseous and liquid
phases.3
cis-RuH2[P(CH3)3]4 + CO2 h cis-RuH(O2CH)[P(CH3)3]4
The solubilization of transition metal-phosphine complexes
in scCO2 is necessary if these ubiquitous and useful catalysts
are to be employed in that medium. Transition metal complexes
which have been demonstrated to be soluble in scCO2 include
those with carbonyl, cyclopentadienyl, porphyrin, acetylaceto-
nate, and other chelating ligands but not, previous to our study,
any with phosphine ligands. While the complexes RuH2-
[P(C6H5)3]4 and RuCl2[P(C6H5)3]3 are known to be active for
CO2 hydrogenation in organic liquids,73-75 we anticipated that
they would have low solubility in nonpolar scCO2. To increase
the solubility of the complexes in scCO2, the trimethylphosphine
analogues were used. The success of this strategy is shown by
the greater rate of the reaction catalyzed by 1 in scCO2 compared
to that by RuH2[P(C6H5)3]4 (Table 1). The rate difference is
not due to electronic differences; the trimethylphosphine and
triphenylphosphine catalyst precursors have equal activity in
liquid methanol (Table 3). The solubility test performed on
the stable catalyst precursor 2 demonstrated conclusively that
trimethylphosphine complexes are soluble in scCO2. Quantita-
tive studies of the solubility of phosphines and their complexes
would be of great value to the study of homogeneous catalysis
and transition metal chemistry in scCO2.
(11)
The ratio of cis-RuH(O2CH)[P(CH3)3]4 to 1 was decreased by
subsequent bubbling of H2 for 1 min through the solution,
presumably by the reverse of reaction 11. No formic acid was
detected.
Insertion of CO2 has been observed previously with the related
complex cis-RuH2[P(C6H5)3]4,70-72 although in that case CO2
insertion was accompanied by phosphine dissociation, giving
RuH(O2CH)[P(C6H5)3]3.
Discussion
Synthetic Utility of the Hydrogenation of scCO2. The
homogeneous hydrogenation of CO2 is rapid and efficient if
the CO2 is in the supercritical state. The yield of formic acid,
in terms of TON or AAR, is excellent, and the selectivity for
formic acid is 100%. The hydrogenation of scCO2 in the
presence of CH3OH and N(C2H5)3 catalyzed by 3 at 80 °C is
also particularly efficient, with the highest yield of methyl
formate being 3500 TON (Table 4). This is 1 order of
magnitude greater than any previous result at any temperature.28
Using the new supercritical method, we were also able to
produce DMF with overall rates of up to 8000 h-1 and with
TON values of up to 420 000. This TON exceeds by 2 orders
of magnitude the highest previously reported value for the
synthesis of DMF using CO2.28,44 Thermodynamically, the
DMF synthesis is strongly favorable, and the dehydration step
is irreversible under these conditions. However, only one of
the previous studies succeeded in obtaining close to quantitative
yields of DMF from NH(CH3)2 by this reaction, and that
accomplishment required reaction temperatures of 150-170
°C.45 By the new method we have obtained 99% conversion
at the relatively low temperature of 100 °C. The advantage in
the use of scCO2 is particularly evident at short reaction times,
when rapid synthesis of the formate salt is a prerequisite for
prompt formation of DMF.
Effect of Product Precipitation. The number of phases
present during the reaction is not only a function of temperature,
pressure, and the amounts of additives but also of the reaction
time. Under the standard conditions used, only a single-phase
exists at the start of the reaction (Figure 2, bottom left). Because
the product, probably [NH(C2H5)3][O2CH]‚HCO2H,60,61 pre-
cipitates from the scCO2 over time, two phases exist during the
later stages of the reaction (Figure 2, bottom right). In a
window-equipped reactor vessel, the first visible signs of liquid
formation come after 10 min at 50 °C with catalyst precursor
1. The change from a single phase to two phases during the
reaction could significantly alter the catalytic activity. For
example, experiments with very large amine to catalyst ratios
performed in a 300-mL reactor resulted in TON values of up
to 7200 but AAR far lower than those found for the standard
experiments, even though the reaction times were greater. This
indicates that equilibrium is not reached, probably because the
reaction is much slower under such conditions. It is possible
that the reaction slows down because the catalyst dissolves in
the liquid phase which builds up as the product precipitates.
The rest of the reaction may take place in that phase, with the
reduced rates associated with liquid phase reactions.
The high TON values obtained in the formic acid, methyl
formate, and DMF syntheses do not necessarily demonstrate
that the equilibria of these reactions have been shifted by the
use of supercritical conditions. Rather, they demonstrate that
the Ru catalysts have long lifetimes under these conditions and
that the reaction rate is so rapid that very high TON can be
attained within reasonable reaction times. Thus, high productiv-
ity could be obtained if these syntheses were applied in
continuous flow reaction systems.
One of the functions of the CH3OH or DMSO promoters may
be as cosolvents acting to increase the solubility of the catalyst
in the supercritical phase. This would be particularly important
during the later stages of the reaction if the promoter were able
(70) Komiya, S.; Yamamoto, A. J. Organomet. Chem. 1972, 46, C58-
C60.
(71) Kolomnikov, I. S.; Gusev, A. I.; Aleksandrov, G. G.; Lobeeva, T.
S.; Struchkov, Yu. T.; Vol’pin, M. E. J. Organomet. Chem. 1973, 59, 349-
(73) Inoue, Y.; Izumida, H.; Sasaki, Y.; Hashimoto, H. Chem. Lett. 1976,
863-864.
(74) Yamaji, T. Japan Kokai Tokkyo Koho 140948, 1981.
(75) Drury, D. J.; Hamlin, J. E. Eur. Patent Appl. 0 095 321, 1983.
351.
(72) Komiya, S.; Yamamoto, A. Bull. Chem. Soc. Jpn. 1976, 49, 784-
787.

354 J. Am. Chem. Soc., Vol. 118, No. 2, 1996
Jessop et al.
in 7 (or 8) by molecular hydrogen forms formic acid, regenerat-
ing the catalytic species 6. Hydrogenolysis, conceivably the
rate-determining step, would be considerably accelerated under
the supercritical conditions because of the high concentration
of H2.
It should be mentioned that the Ru-H species may react
directly with uncoordinated CO2 with the aid of a metal-
coordinated protic reagent.79 Water has been shown to be
necessary in amounts of at least 1 mol/mol of catalyst for
reaction 1 catalyzed by a Pd-phosphine complex.73 We have
found that the use of water, CH3OH, or DMSO as promoters
causes an increase in rate over that in the absence of any such
promoter (Table 3). The highest turnover rates, over 4000 h-1
,
were obtained with CH3OH or DMSO promoters. The ef-
fectiveness of DMSO may be at least partly a result of water
dissolved therein. CH3OH was also effective at 80 °C, at which
temperature the water-promoted reaction was inefficient. There
are many mechanisms for CO2 hydrogenation which could be
consistent with a promoting effect of water or alcohol,28
including the possibility79 that the promoter binds to the metal
and stabilizes the activated complex by hydrogen bonding during
the CO2 insertion. A possible transition state is illustrated by
structure 9.
Figure 11. Possible mechanism of CO2 hydrogenation (X ) H or Cl,
R ) H or CH3, L ) P(CH3)3).
to prevent or retard the dissolution of the catalyst in the liquid
product phase.
Hydrogenation Mechanism. It is not possible with the
present data to determine the mechanism of the CO2 hydrogena-
tion. However, there are some general comments which can
be made.
Most mechanisms for the hydrogenation of CO2 to formic
acid require metal hydride active species.28 Although catalyst
precursors RuCl2[P(CH3)3]4 (2) and RuCl(O2CCH3)[P(CH3)3]4
(3) do not contain hydride ligands, the conversion of these
catalyst precursors to cis-RuHCl[P(CH3)3]4 should be facile in
the presence of H2 and base.76,77 This process is likely the cause
of the induction period when 2 is used as a catalyst precursor
(Figure 5). The further conversion of cis-RuHCl[P(CH3)3]4 to
RuH2[P(CH3)3]4 (1) is possible. It is interesting to compare
the catalytic activity of 2 to a similar dichloro complex which
is unable to undergo the conversion to a chloro hydrido complex.
For example, trans-RuCl2(dmpe)2 can not be converted to a
hydride complex by H2 and base.78 Without any hydride ligand,
the complex can not undergo CO2 insertion or catalyze CO2
hydrogenation: The observed catalytic activity was zero. In
contrast, the complex trans-RuHCl(dmpe)2 is catalytically active.
These considerations demonstrate that a hydride ligand in the
catalyst is a prerequisite for catalytic activity.
A vacant site made available by the dissociation of a
phosphine ligand may be a prerequisite for the CO2 insertion
step or the subsequent hydrogenolysis step.77 Although a few
equivalents of added P(CH3)3 have no effect on the insertion
of CO2 into RuH2[P(CH3)3]4 or on the hydrogenation of scCO2,
a large excess of P(CH3)3 has an inhibiting effect on the catalytic
hydrogenation, which is evidence for a rate-determining step
that requires concomitant or prior phosphine dissociation. The
same conclusion is suggested by the low rate of hydrogenation
with trans-RuHCl(dmpe)2, which is attributed to the inability
of the chelating diphosphine ligands to easily dissociate. The
data mentioned above are consistent with but not proof of
phosphine dissociation in the mechanism.80
Several mechanisms for the catalytic hydrogenation of CO2
by hydride catalysts have been suggested, most of these based
on CO2 insertion into the metal-hydride bond.28 This reaction
One of the alternative mechanisms for CO2 hydrogenation,28
the reverse water-gas shift reaction giving CO and H2O
followed by their recombination to produce formic acid, can
clearly be rejected. CO was found to have a strong inhibiting
effect on the reaction, which suggests that a reverse water-gas
shift reaction does not occur to any significant extent during
the hydrogenation of scCO2 and that the reverse water-gas shift
reaction does not form part of the catalytic cycle.
Measurement of the detailed kinetics of the reaction and in
particular the effect of altered concentrations of CO2 or H2 on
the rate of reaction would not necessarily lead to an unambigu-
ous identification of the mechanism. Adjusting the amounts
of the reagents has such a strong and at present unpredictable
effect on the phase behavior that a proper kinetic study would
be both difficult and inadvisable before the phase behavior is
better understood.
70,71
is known for the closely related complex RuH2[P(C6H5)3]4
and was also observed by us when that complex was dissolved
in scCO2 in the absence of amine and H2. In-situ NMR studies
described in the Results showed that CO2 insertion into the
Ru-H bond of complex 1 is also possible, giving RuH(O2CH)-
[P(CH3)3]4 (eq 11). Ru(O2CH)Cl[P(CH3)3]4 would be the active
species formed from 2 or 3 via RuHCl[P(CH3)3]4 and is expected
to be as active as 3.
Figure 11 illustrates a possible mechanism for the hydrogena-
tion under the current reaction conditions, where water or
alcohol is acting as a promoter. In this mechanism, a phosphine
ligand in the Ru hydride 5 is replaced by ROH (alcohol or water)
to generate the chain carrier 6. Subsequent insertion of CO2
into the Ru-H bond occurs to give the formato complex 7,
which may be in equilibrium with RuX(O2CH)[P(CH3)3]4 (8).
The latter complex (X ) H) was detected in stoichiometric tests,
as mentioned above. Hydrogenolysis of the Ru-O2CH bond
Reaction Pathways to Alkyl Formates and Formamides.
The reaction profile for the methyl formate synthesis (Figure
8)18 is consistent with a two-step pathway of CO2 hydrogenation
(76) Hallman, P. S.; McGarvey, B. R.; Wilkinson, G. J. Chem. Soc. A
1968, 3143-3150.
(79) Tsai, J.-C.; Nicholas, K. M. J. Am. Chem. Soc. 1992, 114, 5117-
5124.
(80) Alternative explanations for the observations are possible: The large
excess of phosphine could be acting as additional base, which would cause
a decrease in the rate, and the poor activity of trans-RuHCl(dmpe)2 could
be due to its trans geometry, which may not be favorable to CO2 insertion.
(77) Jessop, P. G.; Morris, R. H. Coord. Chem. ReV. 1992, 121, 155-
284.
(78) Bautista, M. T.; Cappellani, E. P.; Drouin, S. D.; Morris, R. H.;
Schweitzer, C. T.; Sella, A.; Zubkowski, J. J. Am. Chem. Soc. 1991, 113,
4876-4887.

Homogeneous Catalysis in Supercritical Fluids
J. Am. Chem. Soc., Vol. 118, No. 2, 1996 355
(eq 1) followed by esterification (eq 6), a mechanism which
has been proposed previously.39,40 The formic acid intermediate
is observed in our system because the first step is very fast under
these conditions. The second step, the esterification, is rate
determining. There is a possibility that another mechanism not
involving formic acid is responsible for the production of methyl
formate. However, this is unlikely because independent tests
confirmed that formic acid is thermally esterified by methanol
under these conditions even in the presence of an amine.
Methanol is required not only as an esterification reagent in
the second step but as a remarkably effective promoter of the
hydrogenation step as well. The kinetic data at 50 and 80 °C
clearly show significantly increased rates of formic acid
production in the presence of CH3OH.
homogeneously catalyzed hydrogenation in scCO2 is both
possible and highly efficient, satisfying the second requirement.
The ease by which scCO2 can be hydrogenated suggests that
other reactions for the activation of CO2 or other small molecules
could be efficiently performed by homogeneous catalysis in
SCFs.
Conclusions
We have demonstrated that scCO2 is an excellent and
promising reaction medium for a homogeneous transition metal-
catalyzed reaction. The following more specific conclusions
can also be made.
Transition metal complexes of trialkylphosphines are soluble
and catalytically active in scCO2 for the homogeneous hydro-
genation of scCO2. The activity of an analogous triphenylphos-
phine complex was lower, possibly because of lower solubility.
The hydrogenation catalyzed by Ru complex 1 or 2 is both fast
and efficient. Yields of up to 1.7 mol of formic acid/mol of a
tertiary amine and up to 7200 mol/mol of catalyst can be
obtained. The high efficiency and rate of reaction may be due
to a number of factors, including the high miscibility of H2 with
scCO2, a weaker coordination sphere around the catalyst, rapid
diffusion in the supercritical phase, and elimination of mass
transfer between the gaseous and liquid phases.
The rate of formic acid production is greater if the system is
homogeneous and supercritical at the start of the reaction. In
general, experiments in which one or more reagents form a
second phase had lower rates of reaction. At 50 °C, the highest
rate of reaction, 4000 h-1, was obtained with methanol or
DMSO additives and represents a considerable advance on the
rate mentioned in our preliminary communication.12
The formic acid synthesis can be coupled with subsequent
reactions of formic acid, for example, with alcohols or secondary
amines, to give highly efficient “one-pot” routes to formate
esters or formamides. The highest TON value obtained was
420 000 for the synthesis of DMF. The overall rate of DMF
formation was up to 8000 h-1 at 100 °C.
Complete understanding of the effect of phase behavior on
the rate of reaction could only come with a mapping of the
phase diagrams of the ternary system CO2/H2/N(C2H5)3 and the
quaternary system CO2/H2/N(C2H5)3/HCO2H, a task which can
not be performed with our equipment.
The reaction profile17 of the DMF synthesis (Figure 10) is
consistent with DMF formation via dehydration of dimethyl-
ammonium formate (eq 12), a mechanism which has been
proposed previously.45,46
CO2 + H2 + NH(CH3)2 f [NH2(CH3)2][O2CH] f
HCON(CH3)2 + H2O (12)
The formate salt intermediate was observed because the second
step is rate determining under these conditions. The experiment
with D2 gas, which gave DCON(CH3)2, is also consistent with
the mechanism shown in eq 12. The same result is inconsistent
with a mechanism of the reverse water-gas shift reaction
followed by NH(CH3)2 carbonylation. These conclusions as-
sume that H/D exchange reactions between D2 and NH(CH3)2
did not occur to any appreciable extent before the DMF was
produced. The reaction was stopped at 3% conversion in an
attempt to avoid any such exchange processes. The inhibition
of the DMF synthesis by an atmosphere of CO (Table 7) seems
to rule out the carbonylation mechanism. The extremely low
yield of formic acid in this reaction shows that the CO primarily
inhibits the hydrogenation step, rather than the dehydration step,
presumably by converting the Ru catalyst to less active carbonyl
complexes.
Outlook. scCO2 has been used successfully as an inert
medium for a number of stoichiometric reactions of metal
complexes81,82 and even a few homogeneously catalyzed reac-
tions.13 However, the high reactivity of scCO2 that we have
now demonstrated may have consequences for the ability of
scCO2 to serve as an inert medium for catalysis. The insertion
of CO2 into M-H, M-OR, or M-NR2 bonds (M ) metal
complex), which is expected to be facile in scCO2, could alter
the catalytic activity of homogeneous catalysts which contain
these bonds. Such insertion reactions could lead to catalytically
inactive species, suppressing the catalysis, or to CO2 incorpora-
tion reactions where none such was expected. scCO2 then
should no longer be considered “inert” but, depending on the
context, would be better described as “reactive”.
The processes described in this report, and any other processes
based on homogeneous catalysis in scCO2, are particularly well
suited to industrial application. The high rates of reaction
demonstrated in this report are ideal for larger scale continuous
flow systems. scCO2 could also be used to extract the nonpolar
catalyst from the polar liquid product stream and thus allow
efficient recycling. It is hoped that the high efficiency of the
reactions described herein will encourage the use of CO2-based
processes as replacements for those based on toxic carbon
monoxide.
Industrial application of CO2 fixation by hydrogenation is
feasible if two conditions can be met: An economical local
source of hydrogen is available, and a highly efficient catalyst
system can be found. The results described herein prove that
Acknowledgment. The authors gratefully acknowledge the
experimental work of Dr. K. Aoyagi and Miss M. Kunieda of
ERATO and Mr. H. Onishi of the Aichi Institute of Technology.
Prof. A. Yamamoto of Waseda University and Dr. S. Hashiguchi
and Dr. S. C. A. Nefkens of ERATO provided helpful
discussions, while Dr. Y. Yamauchi of JASCO International
provided technical assistance.
(81) Howdle, S. M.; Poliakoff, M. In Supercritical Fluids Fundamentals
for Application; Kiran, E., Levelt Sengers, J. M. H., Eds.; NATO ASI Series,
Series E 273; Kluwer Academic: Dordrecht, 1994; pp 527-537.
(82) Jessop, P. G.; Ikariya, T.; Noyori, R. Organometallics 1995, 14,
1510-1513.
JA953097B