Organometallics 196)6,14, 4937-4943
4937
Electrochemical Reduction of CO2 Catalyzed by a Dinuclear Palladium Complex Containing a Bridging Hexaphosphine Ligand: Evidence for Cooperativity Bryan D. Steffey, Calvin J. Curtis, and Daniel L. DuBois* National Renewable Energy Laboratory, Golden, Colorado 80401 Received April 27, 1995@ The complexes [P~z(CH~CN)Z(~HTP)I(BF~)~ and [Pdz(PEt3)z(eHTP)](BF4)4(where eHTP is bis(bis((diethylphosphino)ethyl)phosphinane,(EtzPCH2CHZ)zPCHzP(CH2CH2PEtz)z) were prepared and characterized. [P~z(CH~CN)Z(~HTP)](BF~)~ catalyzes the electrochemical reduction of COZ to CO in acidic dimethylformamide solutions. The rate of this reaction exhibits a biphasic dependence on acid, with a first-order dependence at low acid concentrations and zero-order dependence at acid concentrations greater than 0.06 M. At high acid concentrations the rate-limiting step is first order in catalyst and first order in COS. When compared to the kinetic properties of previously studied mononuclear complexes, these data suggest both palladium atoms are involved in COZreduction. The closely related complex [Pdz(PEt3)z(eHTP)](BF4)4undergoes two reversible two-electron reductions. The mixedvalence intermediate resulting from the first two-electron reduction is unstable and undergoes rapid decomposition.
Introduction Structural and spectroscopic studies have shown that COZ binding can be enhanced by cooperative effects between two metals and between a metal and its 1igand.l-l0 The two metals involved may be a transition metal and an alkali metal, two transition metals, or a transition metal and a main group metal. Bimetallic activation has also been implicated in the electrocatalytic reduction of COZto COll and the coupling of COz and alkynes.lZ In these catalytic reactions, a transition metal and a magnesium ion are involved, but only the transition metal is recycled in a catalytic fashion. The @Abstractpublished in Advance ACS Abstracts, September 15, 1995.
( l ) ( a ) Pinkes, J . R.; Steffey, B. D.; Vites, J. C.; Cutler, A. R. Organometallics 1994, 13, 21. (b) Vites, J. C.; Steffey, B. D.; Giuseppetti-Dery, M. E.; Cutler, A. R. Organometallics 1991, 10, 2827. ( c ) Steffey, B. D.; Vites, J. C.; Cutler, A. R. Organometallics 1991,10,3432. (2) (a) Gibson, D. H.; Ye, M.; Richardson, J. F. J. Am. Chem. SOC. 1992, 114, 9716. (b) Gibson, D. H.; Mehta, J. M.; Ye, M.; Richardson, J . F.; Mashuta, M. S. Organometallics 1994, 13, 1070. (3) (a)Pilato, R. S.;Geofioy, G. L.; Rheingold, A. L. Organometallics 1990.9. 2333. (b) Housmekerides. C. E.: Ramaae. D. L.: Kretz, C. M.: Shontz,’J. T.; Pilato, R. S.; Geoffroy, B. L.; Rheiigold, A. L.; Haggerty; B. S. Inorg. Chem. 1992,31, 4453. (4) (a) Fujita, E.; Creutz, C.; Sutin, N.; Brunschwig, B. S. Inorg. Chem. 1993,32,2657. (b) Szalda, D. J.; Chou, M. H.; Fujita, E.; Creutz, C. Inorg. Chem. 1992, 31, 4712. ( c ) Fuiita, E.; Szalda, D.; Creutz, C.; Sutin. R. J . Am. Chem. Soc. 1988.116.4870. (5)’Tanaka, H.; Tzeng, B.-C.; Nagao, H.; Peng, S.-M.; Tanaka, K. Inorg. Chem. 1993,32, 1508. (6) (a)Gamborotta, S.;Arena, F.; Floriani, C.; Zanazzi, P. F. J. Am. Chem. 1983, Chem. SOC.1982, 104, 5082. (b) Floriani, C.; Pure Appl. .. 55, 1. (7) Bianchini, C.; Meli, A. J . Am. Chem. Soc. 1984, 106, 2698. (8) Schmidt, M. H.; Miskelly, G. M.; Lewis, N. S. J.Am. Chem. SOC. 1990,112,3420. (9) (a)Darensbourg, D. J.; Wiegreffe, H. P.; Wiegreffe, P. H. J.Am. Chem. SOC.1990, 112, 9252. (b) Darensbourg, D. J.; Hanckel, R. K.; Banch, C. G.; Pala, M.; Simmons, D.; White, J. N. J. Am. Chem. Soc. 1985, 107, 7463. ( c ) Darensbourg, D. J.; Pala, M. J . Am. Chem. SOC. 1985,107, 5687. (10) (a)Henary, M.; Zink, J. I. J.Am. Chem. SOC.1988, 110, 5582.
(b) Lundquist, E. G.; Folting, K.; Huffman, J. C.; Caulton, K. G. Inorg. Chem. 1987,26,205. (11)Hammouche, M.; Lexa, D.; Momenteau, M.; SavBant, J.-M. J. Am. Chem. SOC.1991, 113, 8455.
magnesium ions act as stoichiometric reagents. Catalytic reactions observed for Ni(cyclam)+(where cyclam is 1,4,8,1l-tetraazacyclotetradecane) adsorbed on Hg may also proceed through bimetallic intermediates,13J4 but no direct evidence has been presented, and dinuclear complexes based on Ni(cyclam)2+have shown no cooperative effects.15 Bimetallic copper complexes and trimetallic nickel clusters have been reported to catalyze the electrochemical reduction of C0z.l6 Mechanistic studies of these bimetallic complexes are consistent with a transition state containing one COZper two copper sites which may indicate a possible cooperative interaction. Unfortunately, no analogous mononuclear complexes exhibiting catalytic activity are available for comparison with these dinuclear complexes. Recent examples of bimetallic activation of other substrates include C-H activation of methane,17phosphonate and phosphate ester hydrolysis,18 and hydroformylation of 01efins.l~ Structural studies and van der Waals energy calculations carried out by Stanley and co-workers on dipalladium complexes of the bridging hexaphosphine ligand eHTP [(E~QPCHZCHZ~PCH$’(CHZCHZPE~Q~I have shown that the preferred structure is a conformation with the palladium atoms nearly eclipsed as shown in (12) (a)Dbrien, S.; Duiiach, E.; PBrichou, J . J.Am. Chem. Soc., 1991, 113,8447. (b) DBrien, S.; Duiiach, E. PBrichou, J. J. Organomet. Chem. 1990.385. C43. (13) Beley, M.; Collin, J-P.; Ruppert, R.; Sauvage, J.-P. J.Am. Chem. Soc. 1986,108, 7461. (14) (a)Balazs, G. B.; Anson, F. C. J.Electroanal. Chem: 1992,322, 325. (b) Fujihira, M.; Hirata, Y.; Suga, K. Ibid. 1990, 292, 199. (15) Collin, J. P.; Jouaiti, A,; Sauvage, J.-P. Inorg. Chem. 1988,27, 1986. (16) (a)Haines, R. J.;Wittrig, R. E.; Kubiak, C. P. Imrg Chem. 1994, 33,4723. (b) Ratcliff, K. S.; Lentz, R. E.; Kubiak, C. P. Organometallics, 1992, 11, 1986. (17) Zhang, X.-X.;Wayland, B. B. J. Am. Chem. Soc. 1994,116,7897. (18) (a) Hance. D. H.: Czarnik. A. W. J.Am. Chem. Soc. 1993,115, 12165. (b) Tsoubouchi, A.; Bruice, T. C. J . Am. Chem. SOC.1994,116, 11614. (19) Laneman, S. A,; Stanley, G. G. Adu. Chem. Ser. 1992, No. 230, 349. ~
0276-7333/95/2314-493~~~9.00/0 0 1995 American Chemical Society
Steffey et al.
4938 Organometallics, Vol. 14, No. 10, 1995
structure 1.20 Work from our laboratory has demon-
1
2
L=NCCH,
3
strated that the closely related [Pd(triphosphine)(CH3CN)I(BF& complexes catalyze the electrochemical reduction of C02 to CO in acidic dimethylformamide or acetonitrile solutions.21 These catalysts exhibit second order rate constants as high as 300 M-l s-l and can be very selective with current efficiencies exceeding 95% for the production of CO. Our results and those of Stanley and co-workers suggested that the complex [Pda(CH&N)2(eHTP)](BF4)4,2, in which the chloride ligands of 1 are replaced with acetonitrile, might exhibit cooperative effects for C02 reduction by forming intermediates containing a bridging C02 ligand. This paper describes studies of 2 as a catalyst for electrochemical CO2 reduction. Results Synthesis and Characterization of Complexes. The ligand eHTP was prepared as described in the literature.22 Reaction of 1equiv of eHTP with 2 equivs of [Pd(CH&N)&BF41z resulted in formation of [Pdz(C&CN)2(eHTP)I(BF4)4,2. The 31P NMR spectrum of 2 (20)Saum, S.E.;Laneman, S. A.; Stanley, G. G. Inorg. Chem. 1990, 29,5065. (21)(a) DuBois, D.L.; Miedaner, A. J . Am. Chem. SOC.1987,109, 113.(b) DuBois, D. L.; Miedaner, A,; Haltiwanger, R. C. J.Am. Chem. SOC.1991,113,8753.( c ) Bernatis, P.R.; Miedaner, A.; Haltiwanger, R. C.; DuBois, D. L. Organometallics 1994,13,4835. (d)Miedaner, A,; Curtis, C. J.; Barkley, R. M.; DuBois, D. L. Inorg. Chem. 1994,33, 5482. (22)Askham, F. R.;Stanley, G. C.: Marques, E. C. J. Am. Chem. SOC.1985,107,7423.
consists of two singlets at 63.8 and 110.2 ppm assigned to the terminal, PT,and central, Pc, phosphorus atoms, respectively, of the coordinated hexaphosphine ligand. These assignments are based on the relative intensities of these resonances and on the similarity of the chemical shifts to those observed for monomeric analogues.21The presence of coordinated acetonitrile is indicated by CN stretching and combination bands at 2295 and 2323 cm-' in the infrared spectrum.23 The labile acetonitrile of 2 is readily replaced by triethylphosphine forming [Pd2(PEt&(eHTP)I(BF4)4,3. The 31PNMR spectrum of 3 is complex as shown by the top trace in Figure 1. This spectrum has been analyzed as an AA'XX'M2M2' spin system using the parameters listed in the Experimental Section. This analysis indicates a 47 Hz coupling between the two central phosphorus atoms, Px and Px, and a 9 Hz coupling between PAand Px. Similar four-bond coupling constants have been observed in phosphido-bridgeddimers of Pd and Pt.24 The other coupling constants and chemical shift values of 3 are close to those observed for other square planar [Pd(triphosphine)(PEt3)1(BF4)2 complexes.21 On the basis of their spectroscopic data both 2 and 3 are assigned the bridged square-planar structures shown above. Electrochemical Studies. Cyclic voltammograms of 2 under different conditions are shown in Figure 2. Cyclic voltammogram a is of the complex under a nitrogen atmosphere. With the exception of the cathodic peak observed at -1.94 V, only a featureless cathodic current is observed with an onset potential of approximately -1.1V. These same features are observed for all scan rates between 0.05 and 100 VIS. A similar cyclic voltammogram is observed in the presence of CO2 as shown by trace b. Under an atmosphere of CO, the current onset of the first wave and the cathodic peak at -1.95 V are shifted to more positive potentials. Controlled potential electrolysis of dimethylformamide solutions of 2 at -1.3 V under Nz, C02, and CO atmospheres resulted in the passage of 2.0 f: 0.2 Faradayslmol of 2 or 1.0 Faradaylmol of palladium. The product was the same in all cases. The 31P NMR (23)Storkhoff, B. N.;Lewis, H. C., Jr. Coord. Chem. Rev. 1977,23, 1.
(24)Glaser, R.;Kountz, D. J.; Waid, R. D.; Gallucci, J. C.; Meek, D. W. J . Am. Chem. SOC.1984,106,6324.
Electrochemical Reduction of C02
Organometallics, Vol. 14,No. 10, 1995 4939 l
2.0
t v1
8
J
I
/
2e
(
l
(
l
~
4.0
T 2.0 l
-4.0
~
l
l
l
J
I
I
-2.0
I
1
I
I
J
1
-1.0
-1.5
J
I
I
1
I
-0.5
Volts vs. Ferrocene
P'
'
00 0 02
0.04
4.0
2.0
!
I
0.0
-2.0
4
(
6.0
f
g
I
0.0
0.06
008
0.10
0 12
WjF,l M
F
Figure 3. Peak current as a function of acid concentration for solutions saturated with COz (open circles)and nitrogen (solid circles).The dashed line represents the best fit curve using eq 1 (kl = 31 M-l s-l and k2 = 5.9 x lo4 M-2 s-l1, and the solid line represents the best fit curve using eq 2 (k1= 45 M-' s-l and kz = 380 M-' s-l). As discussed in the text, the values obtained for k l and k2 should not be interpreted as true rate constants.
-2.0 -4.0
-6.0 -1.6
-1.4
-1.2
-1.0
-0.8 -0.6
Volts vs. Ferrocene
Figure 2. Cyclic voltammograms of a 1.3 x M solution of [P~Z(CH~CN)~~HTPI(BF& in dimethylformamide: (a) under nitrogen (solid line), (b) saturated with CO2 at 620 mmHg (dashed line), (c) saturated with CO at 620mmHg (+++I. Cyclic voltammograms of same solution containing 0.076 M HBF4 (d) under nitrogen (solid line) and (e>saturated with COZ (dashed line). The supporting electrolyte was 0.3M NEt4BF4, all scan rates were 0.1V/s, and the working electrode was glassy carbon. spectrum of the product can be analyzed t o a first approximation as an AA'BBXX'spin system with additional small coupling. This pattern is similar to that observed for the Pd(1) dimer [Pd(etpE)lz(BFd)z(where etpE is bis((diethylphosphinoethyl)phenylphosphine, (EtzPCHzCH2)zPPh) whose structure has been confirmed by an X-ray diffraction study.21b Cyclic voltammograms of the reduction product of 2 in acetonitrile exhibit an irreversible cathodic wave at -1.95 V and an irreversible anodic wave at f0.54 V. The analogous complex [Pd(etpE)lz(BF& has cathodic and anodic waves at -1.86 and f0.58 V. On the basis of the coulometric, spectroscopic,and cyclic voltammetric data, the cathodic waves at -1.94 V of cyclic voltammograms a and b of Figure 2 are assigned to the reduction of a Pd(1)-Pd(1) bond in a tetrameric ( n = 2 ) or oligomeric species such as 4. The observation that this peak persists a t high scan rates indicates that this species forms very rapidly upon reduction of 2.
L
Jn 4
Cyclic voltammograms d and e of Figure 2 are of 2 in the presence of 0.076M HBF4 under nitrogen and C02
4.0
2.0
A t
4
J
02
04
06
08
IO
12
14
[Cat] mM
Figure 4. Plots showing the dependence of the catalytic current on COz (top graph) and catalyst concentrations (bottom graph). atmospheres, respectively. An increase in the cathodic current is observed in the presence of COz compared to nitrogen. This increase is attributed to the reduction of COz to CO. Electrolysis at -1.3 V of a dimethylformamide solution of 2 containing 0.1 M HBF4 and 0.18 M COZproduced CO with a current efficiency of 85%. The other product formed during electrolysis is hydrogen, which accounts for the remaining charge consumed. A turnover number of 8 was calculated based on the moles of CO produced per mol of catalyst. The catalyst was considered deactivated when the current had decayed to 10% of its original value. The dependence of the catalytic current on substrate and catalyst concentrations are shown in Figures 3 and 4. A plot of the catalytic current versus acid concentration (Figure 3,open circles) is biphasic. The catalytic current initially increases with acid concentration and
4940 Organometallics, Vol. 14,No. 10,1995
Steffey et al.
then becomes independent of it. As discussed in more detail later, this behavior is consistent with a first-order dependence of the catalytic rate on acid a t low acid concentrations and a zero-order dependence at high acid concentrations. The solid circles in Figure 3 show the currents observed for solutions of different acid concentrations in the absence of COz. At all acid concentrations a significant current enhancement is observed for solutions purged with C02 compared to those purged with Nz.Plots of the catalytic current versus the square root of the C02 concentration and versus the catalyst concentration are shown in Figure 4. The square root dependence of the catalytic current on C 0 2 concentration is consistent with a rate-determining step that is first order in C02.25 The linear dependence of the current on catalyst concentration implies a first order dependence on the catalyst concentration as well. The catalytic current was decreased by approximately 20% when the C02 pressure was held constant at 1 atm and the partial pressure of CO was increased to 0.5 atm (scan rate 0.1VIS). "his indicates that the product, CO, inhibits the reaction. Reaction of CO with 2 in undried dimethylformamide-d7 produces H2, CO, and 4 over a period of 1-2days with no detectable intermediates as shown in eq 1. The production of 4 was monitored by
CO + 2 + H,O
-
CO,
+ 4 + H,
(1)
31PN M R spectroscopy, and the formation of H2 and CO2 was measured by gas chromatography. Reaction 1 is promoted by addition of water, and it is retarded in the presence of 0.07 M HBF4. Although reaction 1 occurs in acidic DMF solutions under conditions identical t o those for catalytic reactions, the rate is slow with less than 1 equiv of HZproduced in 24 h. These results suggest that hydrogen produced during the electrochemical reduction does not arise from the water gas shift reaction and that the formation of 4 is competitive with hydrogen production. Ideally the catalytic current should be independent of the scan rate.25a However, the catalytic current exhibited a square root dependence on the scan rate for scan rates between 0.05 and 100 VIS a t room temperature as shown in Figure 5 . The catalytic current also remained scan rate dependent for solutions cooled to -50 "C. cyclic voltammograms of 3 a t two different scan rates are shown in Figure 6. At 10 VIS (voltammogram a), two reversible two-electron reductions are observed at -1.12 and -1.44 V. These two waves are assigned t o the two Pd(II/O) couples expected for 3. Analogous [Pd(triphosphine)(PEt3)I(BF4)2complexes are known to undergo reversible two-electron reductions.21 The difference in the redox potentials of these two waves (0.32 V) indicates that there is a significant interaction between the two palladium sites. For comparison, AEu2 is 0.11 V for a dinuclear palladium complex containing the bridging tetrakis(dipheny1phosphino)benzene ligand.26 On the basis of the difference in the redox potentials of these two couples, a comproportionation constant of 7 x 1O'O can be calculated for formation of the mixedvalence complex from Pd(II)-Pd(II) and Pd(0)-Pd(0) (25) (a) Saveant, J. M.; Vianello, E. Electrochimica Acta 1965,10, 905.(b) Saveant, J.-M.; Su, K. B. J.Electoanal. Chem. 1985,196, 1. ( c ) Nadjo, L.; SavBant, J.-M.; Su, K B. Ibid. 1986,196, 23. (26)Wang, P.-W.; Fox,M. A.Inorg. Chem. 1984,33,2938.
I
I
I
I
I
I
I
4
I
I
I
I
1
I
1
1
80.0
#
60.0
8
T
40.0
'
20.0
4.0
2.0
6.0
10.0
8.0
V'fl
Figure 6. Plot of the catalytic current vs the square root of the scan rate. The solid triangles represent the experimentally observed data. The open circles represent the currents calculated assuming the catalytic and decomposition rates given in the text. The solid line represents the best linear fit of the experimental data. 40.0 20.0
&
g
0.0
B
p
-20.0
-40.0
1'
'
I
'
" I
'
'
I
'I
2.0
t
UI
0
0.0
-2.0
-4.0
Figure 6, Cyclic voltammograms of 2.24 x M solutions of [Pdz(PEt&eHTPI(BF&in dimethylformamide under nitrogen at scan rates of 10 V I S (top graph) and 0.05 V I S (bottom graph).
dimers.27 The mixed-valence complex [Pd(O)Pd(II)(PEt&(eHTP)I(BF& is unstable with a lifetime of a few tenths of a second based on the scan rate dependence of cyclic voltammograms of 3. Cyclic voltammogram b of Figure 6 was recorded at a scan rate of 0.05 VIS. The two reversible waves obsenred in voltammogram a have become irreversible, and wave c3 has increased at the expense of cp. On the basis of a linear plot of i , versus the square root of the scan rate between 0.05 and 1.0 VIS, the first reduction wave, CI, is diffusion controlled. Wave c1 also has an E,-E,Iz value of 29 mV at a scan rate of 50 mVIs. This value agrees well with the 25 mV value expected for a reversible two-electron reduction (27) Heinze, J. Angew. Chem., Znt. Ed.Engl. 1984,23,831.
Electrochemical Reduction of CO, followed by a fast chemical reaction.28 Addition of excess triethylphosphine (0.5 M) does not change the cyclic voltammogram observed. This result excludes the possibility of a reversible loss of triethylphoshine being the rate determining step in the decomposition of the mixed-valence complex. Controlled potential electrolysis of complex 3 carried out a t -1.3 V in acetonitrile resulted in the passage of 2.0 Faradaysjmol which is consistent with a two-electron reduction of 3.
Discussion [Pdz(CH&N)2(eHTP)I(BF4)4,2, is readily prepared from eHTP and [Pd(CH&N)41(BF&. Substitution of the labile acetonitrile ligand of 2 with triethylphosphine produces [Pd2(PEt3)2(eHTP)](BF4)4,3. The coupling observed by 31PNMR between the two central phosphorus atoms of 3, z J ~ =t 43 Hz, and between the triethylphosphine ligand and the central phosphorus atom of the second square-planar unit, 4 J =~9 Hz, suggests a significant interaction between the two square-planar units. The 300 mV difference between the two Pd(II/O) couples observed by cyclic voltammetry for 3 is also consistent with a strong interaction between the two palladium sites. The increased current observed in the presence of COz for acidic solutions of 2 suggested that 2 catalyzes the electrochemicalreduction of C02 (Figure 2 trace e). This was confirmed by controlled potential electrolysis experiments carried out at -1.3 V and identification of CO as the major reduction product by gas chromatography. The dependence of the catalytic current on acid concentration (Figure 3)is consistent with two different rate-determining steps. For monomeric [Pd(triphosphine)(CH3CN)I(BF4)2complexes, two rate determining steps were also observed.21 In that case, a linear dependence of the catalytic current on acid concentration was observed for solutions with low acid concentrations. Because of the square root dependence of the catalytic current on substrate c ~ n c e n t r a t i o na, ~linear ~ dependence is consistent with a rate-determining step that is second order in acid, which implies a transition state involving two protons. The proposed rate-determining step at low acid concentrations is the cleavage of the C-0 bond to form CO and water. At higher acid concentrations the current was independent of acid, first order in catalyst, and exhibited a square root dependence on the concentration of CO2. This is consistent with a rate determining step that is first order in catalyst and C02. On the basis of this and other electrochemical data, the reaction of a Pd(1) species with C02 was proposed as the rate determining reaction at high acid concentrations. For the mononuclear complexes, the kinetic data were fit to eq 2, where ic is the
catalytic current, i d is the peak current due to reversible reduction of the catalyst, u is the scan rate, n is the number of electrons involved in catalyst reduction, K 1 and kz are rate constants €or the two rate-determining (28) Andrieux, C. P.; Savbant, J. M. In Investigation of Rates and Mechanisms of Reactions; Bemasconi, C. F., Ed.; Wiley: New York, 1986; Vol. 6, 41E, Part 2, 331.
Organometallics, Vol. 14,No. 10, 1995 4941 steps discussed above, and u is a factor which depends on whether the second electron transfer t o the catalyst occurs at the electrode surface or in solution. Using eq 2, the best fit curve t o the data shown in Figure 3 is shown by the dashed line. The observed data do not fit this model well. A better fit is obtained using eq 3,
which assumes a first-order dependence on acid at low acid concentrations. The solid line shown in Figure 3 is the best curve obtained assuming a first-order dependence on acid. The data are clearly much more consistent with this model. The presence of only one proton in the activated complex of the dimeric catalyst suggests the possibility of a transition state with a bridging C02 molecule as shown in eq 4. In structure
5
5, one proton of the mononuclear transition state has been replaced by the second palladium atom of the dimer. This would account for the biphasic dependence of catalytic current of 2 on acid concentration and also for the difference in the acid dependence observed for monomeric catalysts and 2. At high acid concentrations, the first-order dependence of the rate-limiting step on the concentrations of 2 and COZis consistent with two palladium atoms and one C 0 2 molecule in the transition state. This stoichiometry could reflect a cooperative binding as shown by structure 6 in eq 5 , or it could simply be the result of
3 L=NCCH3
6
deactivation of the second palladium due t o reduction of the first. To distinguish between these two possibilities, an effort was made to measure the rate of this reaction. Determination of catalytic rate constants using eq 3 requires a reliable value for the peak current under noncatalytic conditions. As can be seen from cyclic voltammogram a of Figure 2, determination of an appropriate peak current for 2 is difficult because no peaks are observed. Complex 3 was prepared to provide a good estimate of this value. In previous work, [Pd(triphosphine)(PEta)](BF& complexes have exhibited reversible two-electron reductions, and two reversible two-electron reductions are observed for 3. Dividing the peak currents observed for the two-electron reductions of 3 by 2.83 29 should provide a reasonably good estimate of the current expected €or a reversible one electron (29) Bard, A. J.; Faulkner, L. R. EZectrochemicaZ Methods; Wiley: New York, 1980; p 218.
Steffey et al.
4942 Organometallics, Vol. 14, No. 10,1995
4 ' 0 x lo'
3.0x 104
4
1
k ' 2 ' o IO' x'o'~
0
li/ 1
/ 0.0
20.0
40.0
60.0
80.0
Voltds
Figure 7. Plot of the rate constant kl,calculated according to eq 2, as a function of the scan rate. reduction of 2, because 2 and 3 are expected to have very similar diffusion coefficients. Using the value of i d obtained in this manner, eq 2 can be used t o fit the data shown in Figure 3. The solid curve shown is the fit obtained using values of 45 and 380 M-l s-l for k l and K2. These values do not reflect true rate constants for this reaction because they are scan rate dependent. However, the dependence of the catalytic currents on substrate and catalyst concentrations is consistent with the proposed rate-determining steps. Ideally the catalytic current should not depend on the scan rate. However, because of the square root dependence observed for the catalytic current (Figure 51, the apparent rate constants calculated using eq 3 exhibit a linear dependence on the scan rate as shown in Figure 7 for kl. This linear dependence could arise from (1)a slow catalytic reaction, (2) depletion of the substrates, (3) rapid decomposition of the catalyst, or (4) inhibition by the product CO. A slow catalytic reaction can be ruled out on the basis of the large catalytic currents observed and the fact that enhanced currents are observed in the presence of C02 even at scan rates as high as 100 V/s. A diffusion-controlled wave with small or no current enhancement in the presence of CO2 is expected for a slow catalytic reaction at high scan rates. Depletion of substrates can be discounted because other monomeric catalysts show much larger currents that are scan rate independent under identical conditions. The third possibility, decomposition of the catalyst, is a reasonable possibility. Controlled potential electrolysis experiments indicate the turnover number for 2 is low, 8, and on the basis of 31PNMR spectra of the spent reaction mixture, the major decomposition product is the tetrameric or oligomeric compound 4 discussed above. The formation of 4 must occur very rapidly, because the reduction wave assigned t o this product is observed at scan rates as high as 100 VIS under C02 and N2 atmospheres. Finally, the peak current of cyclic voltammograms simulated assuming a decomposition reaction with a rate competitive with the catalytic reaction has a square root dependence on scan rate as shown in Figure 5. The calculated peak currents are the results of simulations assuming a rate constant for k1 (eq 2) of 1 x lo5 M-l s-l and a first-order rate constant for the decomposition of the catalyst of 3 x l o 4 s-l. These values are not unique in fitting the data but only serve to illustrate that the assumption of a competing decomposition reaction is consistent with the ob-
served behavior of the catalytic current on scan rate.30 Carbon monoxide inhibits the catalytic current as discussed above. This product inhibition could also contribute to the nonideal behavior of the catalytic current. Although CO and H2O react with 2 to form 4, H2, and C02, reaction 1is slow and unlikely to contribute to the observed current inhibition. The reaction of reduced forms of 2 with CO appears to be more rapid as indicated by the shift in the first reduction wave of 2 to more positive potentials in the presence of CO as shown by trace c of Figure 2. It is therefore concluded that it is a reduced form of 2 which binds CO sufficiently rapidly to produce inhibition. However, any CO complexes which form are not stable since reduction of 2 in the presence of CO produces 4. Attempts to slow the decomposition reaction by lowering the temperature to -50 "C and reducing the catalyst concentration were unsuccessful, because the catalytic current still exhibited a scan rate dependence under these conditions. In principle, at sufficiently high scan rates the decomposition of the catalyst should become insignificant and a true catalytic current could be measured. However, this does not occur for scan rates as high as 100 V/s as shown in Figure 7. Consequently a true rate constant cannot be determined, but a lower limit of at least 25 x lo3 M-l s-l can be inferred for k1 from this data. This rate constant is 3 orders of magnitude larger than the 25 M-' s-l value expected for a mononuclear [Pd(triphosphine)(CH&N)](BF4)2 complex with a redox potential of -1.18 V. This result supports a cooperative effect between the two metal sites of 2 during the electrochemical reduction of C02 to co.
Summary and Conclusions Complex 2 catalyzes the reduction of C02 to CO in acidic DMF solutions. Three observations suggest a cooperative interaction between the two palladium sites and CO2 during the catalytic cycle. (1) At low acid concentrations, the first-order dependence of the catalytic rate on acid concentration observed for 2 compared to a second-order dependence for mononuclear complexes is consistent with the second palladium atom of 2 binding to an oxygen atom of C02 as shown in structure 6. (2) At high acid concentrations, the dependence of the catalytic current on the concentrations of C02 and catalyst is consistent with two palladium atoms per CO2. (3) The catalytic rate of COS reduction for 2 is at least 3 orders of magnitude greater than that expected for a mononuclear [Pd(triphosphine)(CH&N)I(BF4)2 complex having the same potential. These observations suggest that dinuclear complexes such as 2 can provide a useful approach for dramatically enhancing catalytic rates of C 0 2 reduction. On the other hand, the very high rate of decomposition of these catalysts must be overcome before advantage can be taken of these higher catalytic rates. Research is in progress to address this issue.
Experimental Section Materials and Physical Methods. Reagent-grade diethyl ether and toluene were purified by distillation from sodium (30)Simulations of cyclic voltammograms under catalytic conditions were carried out using the program Digisim (Bioanalytical Systems, Inc.).
Electrochemical Reduction of COz benzophenone ketyl. Acetonitrile was distilled from CaHz under nitrogen. Reagent-grade dimethylformamide from Burdick and Jackson was deoxygenated by purging with nitrogen and stored in an inert-atmosphere glovebox. Acetone-de was dried over 4-A molecular sieves, vacuum transferred, and stored in a glovebox. Triethylphosphine was obtained from Strem Chemicals, Inc. and used without further purification. The ligand eHTP and [Pd(CH&N)41(BF& were prepared by literature method^.^^,^^ NMR spectra were obtained on a Varian Unity 300 MHz NMR spectrometer operating at 299.95,121.42, and 75.43 MHz for 'H, 31P,and 13C nuclei, respectively. Chemical shifts for 'H NMR spectra are reported in ppm relative to tetramethylsilane using solvent peaks as secondary references. 31PNMR chemical shifts are referenced to external H3P04. Infrared spectra were obtained using a Nicolet 510 P spectrometer on samples prepared as Nujol mulls or dichloromethane solutions. Coulometric measurements were carried out at 25-30 "C using a Princeton Applied Research Model 173 potentiostat equipped with a Model 179 digital coulometer and a Model 175 universal programmer. The working electrode was constructed from a reticulated vitreous carbon rod with a 1-cm diameter and length of 2.5 cm (100 poredin., The Electrosynthesis Co., Inc.). The counter electrode was a glassy carbon rod, and a Pt wire immersed in a 50:50 mixture of permethylferrocene/permethylferrocenium was used as a reference e l e c t r ~ d e . The ~~ electrode compartments were separated by Vycor disks (7-mm diameter). Measurements of current efficiencies for gas production were carried out in a sealed flask (120 mL). In a typical experiment, a 1.0 x M solution of 2 in dimethylformamide (10.0 mL) was saturated with COz (0.18 M at 620 mmHg33 by purging the solution for approximately 30 min. HBF4 (50 pL of a 7.6 M aqueous solution) was added via syringe and the solution electrolyzed at -1.3 V. The electrolyses were considered complete when the current had decayed to approximately 5-10% of the initial value. During the course of the electrolysis, gas aliquots were withdrawn for gas chromatographic analysis. The gases were analyzed at 200 "C using a l / ~ in. x 15 ft stainless steel column packed with 60/80 carboxen 1000 support (Supelco). Cyclic voltam(31)(a) Sen, A.; Ta-Wang, L. J. Am. Chem. SOC.1981,103,4627. (b) Hathaway, B. J.; Holah, D. G.;Underhill, A. E. J. Chem. SOC.1962, 2444. (32)Bashkin, J. K Kinlen, P. Inorg. Chem. 1990,29,4507. (33)Solubilities ofInorganic and Organic Compounds;Stephen, H.; Stephen T., Eds.; Pergamon: New York, 1958;Vol. 1, p 1063.
Organometallics, Vol. 14, No. 10, 1995 4943 metry and chronoamperometry experiments were carried out using a Cypress Systems computer-aided electrolysis system. The working electrode was a glassy carbon disk of approximately 1-mm diameter. The counter electrode was a glassy carbon rod, and the reference electrode was a Pt wire immersed in a permethylferrocendpermethylferrocenium solution. Ferrocene was used as an internal standard, and all potentials are reported vs the ferrocene/ferrocenium couple.34 All solutions for cyclic voltammetry and coulometric experiments were 0.3 N NEt4BF4 in dimethylformamide. Synthesis. [Pd2(CH&N)deHTP)I (BFd.4, 2. A solution of eHTP (1.00 g, 1.84 mmol) in toluene (20 mL) was added t o a solution of [Pd(CH&N)4](BF& (1.63 g, 3.68 mmol) in acetonitrile (50 mL), and the orange reaction mixture was stirred for 1 h. The solvent was removed with a vacuum t o produce a yellow solid which was washed with ether and dried. The product was recrystallized from a mixture of acetonitrile and ether (yield 1.8 g, 83%). Anal. Calcd for CZ&&F~~NZP~Pdz: C, 29.35; H, 5.44; N, 2.36; P, 15.66. Found: C, 29.76; H, 5.59; N, 2.07; P, 15.65. lH NMR (acetone-&): PCHZP,3.65 ppm (t); PCHzCHzP, 3.1 and 2.4 ppm (m); PCHzCH3, 2.4 ppm 1.30 ppm (m). 31PNMR (m); CH3CN, 1.98 ppm (s); PCHZCH~, (acetone&): Pc, 110.2 ppm (SI; PT,63.8 ppm (s). [Pdz(PEts)a(eHTP)](BF4)4,3. A solution of eHTP (0.25 g, 0.46 mmol) in toluene (5 mL) was added t o a solution of [Pd(CH&N)4](BF4)2 (0.41 g, 0.92 mmol) in acetonitrile (25 mL), and the orange reaction mixture was stirred for 1h. Triethylphosphine (O.llg, 0.95 mmol) was added to the reaction mixture, which was stirred for an additional 2 h. The solvent was removed in a vacuum, and the orange solid which resulted was washed repeatedly with ether and dried (0.34 g, 55%). Anal. Calcd for c42Hg&F16P8Pd2: C, 35.96; H, 6.47; P, 17.66. Found: C, 37.11; H, 6.47; P, 16.74. 'H NMR (acetone-&): PCHzP, 3.63 ppm (t); PCHZCHZPand PCH2CH3, 3.4-2.0 ppm (m);PCHzCH3,1.35 ppm (m). 31PNMR (acetone-&): PA,14.63 ppm; PM,55.4 ppm; Px, 112.4 ppm; Jpx, 326 Hz; J M , 9 Hz; JAM,31 Hz; J m , 11.7 Hz; Jxx., 42.7 Hz.
Acknowledgment. This work was supported by the United States Department of Energy, Office of Basic Energy Sciences, Chemical Sciences Division. OM950306S (34)(a) Gagne, R. R.; Koval, C. A.; Lisensky, G. C. Inorg. Chem. 1980,19,2854.(b) Gritzner, G.; Kuta, J. Pure Appl. Chem. 1984,56, 462.
