Electrochemical reduction of carbon dioxide. Characterization of the

Langmuir , 1988, 4 (1), pp 51–57 ... Cite this:Langmuir 4, 1, 51-57. Note: In lieu of an abstract ... Kortlever, Shen, Schouten, Calle-Vallejo, and ...
0 downloads 0 Views 2MB Size
Langmuir 1988,4, 51-57 Table I. Extent of Interfacial Adsorbates 1.0 X

lo-' M LIX

svstem 65 N in heptane/pH 4.0 acetate

buffer average value partial removal range 20-80 m L 40-80 mL 1.0 X M Kelex 100 in xylene/pH 0.9 bisulfate buffer 2.5 X M DPP in CHC13/pH 1.1 bisulfate buffer

% S 89 82 98 90 f 8.0 85

77 85 84

The ability of the MTP.S to "strip" the adsorbed species from the organic phase can then be gauged by the value of the percent stripped, % S, which is

9%

s = 100

Ivvo(A- A,) dV (AU@ - A,) v o

(3)

Equation 3 is readily adapted to cover the case of the modified procedure in which the curves start at a given volume removed (V > 0)

51

where A:,, and A 6 are the absorbances in the absence of and during stirring, respectively, at an organic phase volume of V, - V. From the values in Table I for the three systems studied, it is evident that, as a first approximation at least, the species adsorbed on the interface during the initial phase of the experiment, i.e., before any filtered organic phase is removed, is all retained in the reaction flask. This would be consistent with positioning of the polar end of the adsorbed species in the aqueous side of the interface and of the nonpolar end in the organic side. As the phases separate on the MTPS and the total interfacial area decreases, the adsorbed species are returned to distribute between the phase mixture remaining in the flask. That the adsorbed species does not accompany the organic phase through the MTPS probably can be attributed to an energetically more unfavorable dehydration of the polar portion of the molecule than to the desolvation of the nonpolar portion required when the species leaves the organic interfacial region. Because of the more pronounced response at removal of larger volumes, this kind of experiment is also a valuable variation in those cases in which AA (IA , , - A,) is very small, making it possible to determine interfacial excess from more dilute solutions than heretofore possible.

Acknowledgment. This research was supported by a grant from the National Science Foundation.

Electrochemical Reduction of Carbon Dioxide. Characterization of the Formation of Methane at Ru Electrodes in Carbon Dioxide Saturated Aqueous Solution David P. Summers* and Karl W. Frese, Jr.* SRI International, Menlo Park, California 94025 Received July 29, 1987 The electrochemicalreduction of C02to CHI on electroplated Ru electrodes wm studied. Scanning electron micrographs showed that the surface of such an electrode is made up of fused spheroids of Ru divided by a network of cracks. Cyclic voltammetry performed in the presence and absence of C 0 2 suggested the formation of carbon-containing intermediates that block a portion of the surface hydrogen evolution sites. The variation of the CHI formation rate with pH implied that the rate increase with decreasing pH is due to an increasing hydrogen coverage on the electrode until the coverage becomes so high that sites for CHI formation are blocked. The reduction of C02and not H2C03or HCO, is implicated. The effect of potential on CHI formation indicated that CHI evolution occurs at potentials at which an appreciable hydride coverage exists, also indicating the importance of surface hydrides. The rate of CHI formation increases with temperature, but at T > 85 "C the electrode becomes deactivated because of a surface carbon species. At -0.545 V vs SCE, an activation energy for CHI formation of -35 kJ mol-l was inferred. Electrolyte impurities are implicated as promoters in the formation of CH, in reagent grade sodium sulfate.

Introduction The goals of replacing finite world natural gas reserves and producing fuels from inorganic sources and solar energy have been a motivating force for studying the electrochemical reduction of C02 to CHI. Although initial work focused on semiconductor electrodes in order to capitalize on their potential ability to directly utilize light energy for fuel generation, such efforts have only led to the formation of methanol.' In order to meet the more

demanding requirements of C02 methanation, we turned to metal electrodes because such surfaces have demonstrated ability, in the gas phase, to catalyze methane formation from both CO and COP Metal electrodes could then be coupled to an external photovoltaic cell to drive the endothermic cell reaction composed of COz reduction and H20 oxidation and achieve solar energy storage. We had previously reported that both CHI and CH,OH are formed in the electrochemical reduction of COz in aqueous solution at Ru electrodes.2 Results presented here

(1) (a) Halmann, M.Nature (London) 1978,275,115. (b) Inoue, T.; Fujishima, A.; Hounda, K. Nature (London) 1979,277,637. (c) Canfield, D.;Frese, K. W., Jr. J.Electrochem. SOC. 1983,130,1772. (d) Frese, K. W., Jr.; Canfield, D. J. Electrochem. SOC.1984, 131, 2614.

(2) Frese, K.W., Jr.; Leach, S. J.Electrochem. SOC.1985, 132, 259. (3) Frese, K.W., Jr.; Summers, D. P. In Carbon Dioxide Reduction; ACS Symposium Series; American Chemical Society: Washington, DC, 1986, in press.

0743-7463/88/2404-0051$01.50/0

0 1988 American Chemical Society

Summers and Frese

52 Langmuir, Vol. 4, No. 1, 1988 A

on the formation of CH, from COz (eq 1) substantially COz

+ 8H+ + 8e-

-

CHI + 2Hz0

(1)

-0.297 V vs SCE at pH 4 extend those results and provide detailed information on the effect of pH, electrode potential, temperature, electrolysis time, and electrolyte purity.

Experimental Section Electrolyteswere C0,-saturated (1atm) aqueous solutions of either 0.2 M reagent grade sodium sulfate, 0.2 M 99.999% sodium sulfate,or 0.05 M reagent grade sulfuricacid in distiied deionized

water (Millipore). In the experiments concerning variable pH, a 0.2 M reagent grade sodium sulfate solution was used with the pH adjusted by the addition of reagent grade sulfuric acid or sodium hydroxide prior to saturation with CO, at 1 atm. The pH was then measured before commencement of the electrolysis. The experiment at pH 9 was done in a solution prepared from 0.3 M NaHCOs with the pH adjusted by the addition of sodium hydroxide. All electrolysis experiments were conducted at -60 "C for 5-6 h with reagent grade Na,SO, unless otherwise noted. Electrodes were prepared by plating Ru metal onto spectrompic carbon rods or Ti foil as previously described? Unless otherwise noted all electrodesused were plated onto carbon rods. The geometrical area of the electrodes was 3 cm2 20%. Each entry in the tables and figureswas obtained on differentdays with the electrode kept in ordinary laboratory air overnight between runs. Cyclic voltammetry experiments were performed with a Pine Instrument Corp. Model RDE3 potentiostat. Electrolyseswere performed with an Aardvark Model PEC-1 potentiostat, a Keithley Model 616 digital electrometer, and a microcomputer data acquisition system for measuring current as a function of time. A two-component cell was employed to avoid oxidation of the CO, reduction products. This procedure allowed pH changes of 1-2 units to OCNI during the electrolysis. In other experiments, as noted in the text, the pH was controlled at a constant value with an automatic syringe pump. All electrolyses were carried out with COPgas at 1 atm circulated in a closed system as previously described.'d The circulated CO, was huhhled through the solution, causing gentle agitation. The temperature was controlled by placing the entire system with the exception of the circulating pump in a heated enclosure. In all cases electrolyte volumes were either 25 or 50 mL. Samples were analyzed on a Gowmac Model 750 gas chromatograph with a FID detector. Samples were collected for CH, analysis from the gas phase over solution. A log-log calibration plot of detector response vs CH, concentration in the electrolyte was linear over the range of interest (lO-'-lO+ M). A column of Porapak Q (6 ft)followed by Porapak R (3 ft) at 50 "C was used for CH,/CO analysis. Scanning electron microscopy was done on an International Scientific Instruments SX-40 instrument,and Auger spectra were obtained with a Perkin-Elmer PHI Auger spectrometer. Analog spectra were digitized after collection. Auger samples were removed under potential control, rinsed with water, and allowed to dry before being mounted on a sample holder.

B

*

Results Scanning Electron Microscopy. This paper focuses on results obtained witb electroplated Ru eledrodessimilar to those used previously to reduce C02 to methane? ?Lpieal scanning electron micrographs (SEMs) of such an (4) Solymasi. F.; Erdohelyi, A,; Koeaia, M. J. Chem. Soc., Faraday Tmm. I 1981,77, 1003. (5) Vanniee, M.A. J. Catol. 1975,37,462. (6) Kern, D. J . Chem. Edue. 1960.37.14. (7) wmm, D.;mite,J. surf.sei.1972,90,201 and references therein. (8) (a) Wise, H.;MECarty, J. Surf. Sei. 1988,133,311. (b) Winslow, p.; Bell, A. J. Catol. 1985,94,385. (9) MeKes, D. J. Catol. 1967,8,240. (10) Stonehart, P.: Ross,P. N. Catal. Rev.-Sei. Eng. 1975, 12, 1.

F m I. Scannii electron micrograph of the surfaceof a typical Ru plate electrode used for aqueous CO, reduction at maguifications of (A) 1020X and (B)28OOX. Picture A was taken at an angle to enhance perspective while picture B was taken straight

on.

electroplated Ru electrode are shown in Figure 1. The micrograph in Figure la was taken a t a low angle to give a clear picture of the general surface topography. The surface shown is identical with that of all other electrodes investigated regardless of whether they are fresh from the plating bath or have been used in many different electrolysis experiments. The surface consists of fused spheroids (-3 pm) of Ru formed around sites for nucleation during plating as is commonly found for electrodeposited surfaces. In Figure lb, cracks that are present on the surface can be seen completely splitting many of the Ru spheroids, indicating that they are formed after most of, if not all, the Ru has been d e p i t e d . These cracks are probably related to the strong hydrogen evolution that occurs during plating. Ru is known to absorb hydrogen, which is formed in large quantities during the plating process. Embrittlement during electrolysis and/or stress fractures by rapid hydrogen gas formation could explain the cracks. A SEM of a section of Ru plate that lifted UP during sample preparation showed the thickness of the Plates to be -4 Pm. Inactivity of Substrate Surfaces. Since the Ru plate is cracked, probably allowing electrolyte access to the carbon substrate, the questions of the formation of CH, at the carbon surface and the direct reduction of the a b nsubstrate to m e b e arise. However, it has h e a d y been shown that these same carbon rods are inactive toward methane formation." Also, it was shown previously

Electrochemical Reduction of C02

Langmuir, Vol. 4, No. 1, 1988 53

1;

f

N

. D

reagent 99.99%

:\

0' 0

I

20

40

60

so

'rime (hr)

Figure 3. Average methane formation rate vs total electrolysis time at 60 O C in a 0.2 M Na2S04COz-saturatedsolution at an

initial pH of 4 (closed symbols, reagent grade Na2S0,; open symbols, 99.999% Na2S04).

m nocontrol I

-1.0

I

I I -0.8 -0.6 -0.4 E (V vs SCE)

I -0.2

Figure 2. Cyclic voltammetryat 3 V min-I of a Ru plate electrode at 60 O C in a quiet, 0.2 M Na$04 aqueous solution saturated with Nz or COz.

that CHI could also be formed at electrodes made from Teflon-bonded Ru sponge.2 To further clarify these questions, an electroplated Ru electrode was prepared on a Ti foil substrate and used to reduce a carbon dioxide saturated solution, a system where the only source of carbon was C02. An electrode made from the Ti foil used as a substrate was found to produce no methane at 60 "C, pH 3, and -0.545 V vs SCE. Under identical conditions, an electrode of Ru electroplated onto Ti foil formed methane at a rate of 7.9 X lo* mol cm-2 h-l and a faradaic efficiency of 2.390, indicating that CHI does not arise from the hydrogenation of carbon atoms from the carbon substrate. Thus we can rule out any activity of the carbon rod itself toward CHI formation. Current/Voltage Curves. Figure 2 shows the cyclic voltammetry of an electroplated Ru electrode at 60 "C in the presence and absence of C02in a pH 4.2,0.2 M Na2S04 solution. The first feature of note is the absence, in a C0,-saturated solution, of the anodic peak at -0.9 V due to hydrogen oxidation that is present in a C02-free solution. The second feature is that, as the potential moves into the hydrogen evolution region, currents rise much more slowly in a C02-saturated solution compared to a nitrogen-saturated solution. Both these observations are consistent with a model in which some sites for hydrogen evolution are blocked by C02 reduction intermediates or produck The loss of the hydrogen oxidation peak suggests either that the C02reacts with the hydrogen or that carbonaceous species, such as CO, are formed from COPand reduce the available sites for hydrogen evolution. Methane Formation Rate. A plot of the average rate of CH, formation vs total electrolysis time for many experiments at 60 "C, initial pH 4 (the pH of the electrolyte was not controlled), in both reagent grade and 99.999% Na2S04is shown in Figure 3. The data for a Ru electrode in pure 0.05 M H2S04or 0.05 M H2S04/0.2 M reagent Na2S04 are very similar to those for experiments con(11)SRI International Annual Reports for the Gas Research Institute (1984-1986),SRI Project No. PYH 7142,GRI Contract No. 5083-2600922.

control

0.61

5

0.4

0.2

V."

0

100

200

300

400

Time (rnin)

Figure 4. Plots of normalized current vs time at Ru electrodes in carbon dioxide saturated aqueous 0.2 M NazS04at 60 "C, -0.54 V vs SCE, and an initial pH of 4 with and without holding the pH (with the addition of acid).

Time (hrl

Figure 5. Plot of the amount of methane formed vs time for a Rielectrode in 0.2 M reagent grade Na2S04,pH 3, at -0.48 V vs SCE and I 5 OC.

ducted at an initial pH of 4 with respect to the decline in rate and electrolyte purity. It has been shown previously that current and methane formation rate decrease with electrolysis time.3 This is only partially attributable to a rise in the uncontrolled pH with time. Figure 4 is a plot of normalized current vs time for two experiments at 60 "C, -0.54 V vs SCE, and an initial pH of 4. In one experiment the pH is not controlled and current declines throughout the experiment. In the other experiment the pH is held at 4 and, after an initial drop, the current levels off and no longer decreases. It has been shown recently by Conway and Bail2 that the hydrogen (12)Conway, B.E.;Bai, L.Electrochim. Acta 1986, 31, 1013.

54 Langmuir, Vol. 4, No. 1, 1988

Summers and Frese

Table I. Activity for Methane Formation on Three Ru Electrodes over Multiple Electrolysesa run j : *A cm-2 rate: mol cm+ h-' eff: % Electrode 1 (pH* 4, T = 60 "C) 1 160 4.3 x 10-8 5.7 2 3 4 1 2 3 1

2 3

140 88 87

3.1 X 3.3 x 10-8 4.1 X

Electrode 2 (pHb 2.9, T = 60 "C) 366 14.3 X lo-@ 328 15.9 X 239 14.3 X lo-@ Electrode 3 (pHb 2.7, T = 80 "C) 392 46 X 289 36 X low8 224 24 X

4.7 8.0 10.0 12.2 10.4 12.7 25 27 23

a All electrolysis times are 5-6 h in 0.2 M reagent grade NazS04 at 60 "C and -0.545 V vs SCE with an initial pH of 4. *Initial pH. CAveragecurrent density based on geometrical area. dAverage rate of methane formation. e Faradaic efficiency for methane formation.

evolution kinetics degrade with time at Pt electrodes. Tafel slopes increased by a factor of 2 while exchange current densities decreased 2-fold. A similar fundamental change in the hydrogen evolution kinetics during the early stages of these experiments is thought to occur at our Ru electrodes. Figure 5 shows a plot of total methane formed vs time for the reduction of carbon dioxide in a single experiment at 60 "C, constant pH 4, and -0.54 V vs SCE, and its appears that there is no deactivation of the electrode for at least 4 h. These results show that even though the overall current drops (Figure 4) the methane formation rate is constant during the same period. Investigations of the long-term stability of the methane rate are under way. In Table I data obtained with three Ru electrodes on the effect of multiple electrolyses on the CHI formation rate are presented. It is apparent from the data at 60 OC that if an electrode that has been used to electrolyze a C02-saturatedsolution is reused no measurable decrease in average CH4 formation rate is observed. Thus, at 60 OC, any drop in rate that occurs during an individual experiment is not due to an irreversible deactivation of the catalytic surface. A different conclusion is reached after electrolysis at T > -80 "C (see Table I and Effect of Temperature). An effect of electrolyte purity is also evident in Figure 3, as shown by the 5-6 times lower rate in the high-purity electrolyte (high-purity electrolytes show the same drop in current that is seen in reagent grade electrolytes). The effect appears to be due to the electrodeposition of adventitious impurities. Electrochemical stripping experimenta performed on a solution made from reagent grade NazSO, revealed the presence of Zn,Cu, and As, and Auger spectroscopy of electrodes used to electrolyze such solutions showed Cu, Fe, Ni, and Zn on the surface. An electrode that was used in several COP electrolysis experiments showed surface Cu by Auger spectroscopy (see below). Effect of Added CO. Carbon monoxide is formed in COz reduction experiments with faradaic efficiencies of typically 1%but ranging from 0% to 10%. Because CO is a strongly chemisorbing and sometimes poisoning species on transition-metal surfaces such as Pt, we added CO to the C02 at the beginning of two electrolysis experiments and observed the effect on the methane rate. The data in Table I1 show the effect of added CO on the average rate of CHI formation for two different electrodes. The

Table 11. Effect of CO and H2on the Electrochemical Reduction of CO, to Methanea electrode

gas

1 1 1 1

co co

2 2

H2

3 3

Hz

gas added, mL 0 35 0 50 0 100 0 200

j,b uA cm-2 107 72 79 63 89 104 87 109

rate: mol cm-2 h-' effad% 9.3 X 18.6 4.9 X 14.7 7.5 x 10-8 20.2 3.6 X 12.4 4.8 X 11.5 11.1 5.3 X 4.2 X 10.3 11.5 5.8 X ~~

"All electrolysis times are 5-6 h in 0.2 M reagent grade NaZSO4 at 60 "C and -0.545 V vs SCE with an initial pH of 4. CO and Hz were added displacing an equal volume of C 0 2 from the 1.3-L C 0 2 reservoir (see Experimental Section). Average current density based on geometrical area. CRate of methane formation. Faradaic efficiency for methane formation.

CHI rate was established in the absence of CO, and then a measured amount was added in a succeeding experiment. The 35-mL and 50-mL additions correspond to 1.2 and 1.7 mM gas-phase concentrations, respectively (0.027 and 0.038 atm). These levels are 100-1000 times higher than the levels found in an electrolysis without added CO. The effect of CO is to lower the average current, CHI formation rate, and faradaic efficiency. Normally, lower current increases the faradaic efficiency with only a small effect on the CHI formation rate. Carbon monoxide does indeed inhibit the formation of CH,, but the effect is not very great at the CO concentrations employed, the rate of CH, formation being reduced by about a factor of 2. The inhibition effect may occur by the blocking of sites for C02/H+ reduction by the more strongly bound CO. Carbon monoxide is easily hydrogenated to CHI on Ru surfaces,4but the activation e n e r d is considerablyhigher than with COz. Therefore the lower methane rate in the presence of adsorbed CO may also be due to slower electrochemical methanation kinetics for CO. Effect of Added Hydrogen. The effect of the addition of hydrogen gas on the CH4 formation rate is also shown in Table 11. In a typical experiment after passing 10 C, a partial pressure of 3 X lo4 atm of hydrogen is obtained. If 0.1 L or 0.2 L of hydrogen gas is added, the partial pressure of H2is increased by a factor of -2000 or -4200, respectively. This has the effect of increasing the CH4 formation rate by 10% and 41%, respectively. These data show that increasing the surface coverage of H atoms by increasing the hydrogen partial pressure leads to a significant increase in the rate of CH4 formation, demonstrating that hydrogen coverage is an important factor in the rate of CHI formation. The relatively small increase in CH4formation rate for such a large increase in Hz partial pressure could mean that the solution next to the surface is near saturation with H2. However, the shape of the hydrogen adsorption isotherm is not known, and it is possible that the hydrogen coverage on the surface may not vary greatly over the concentrations of hydrogen gas used. Effect of pH. The effect of the pH on the rate of CH4 formation at constant overpotential of 200 mV for two different electrodes is shown in Figure 6. The data indicate that CH4 can be made at pH values as alkaline as 9.1 with modest rates. This might indicate that direct reduction of bicarbonate ions occurs, but even at an alkaline pH a significant partial pressure of COz is present (at all other pHs solutions were made from gaseous COP and had a constant COz partial pressure of 1 atm; see Experimental Section). Indeed, analysis of the gas over the solution indicated the presence of more COz than our

Langmuir, Vol. 4, No. I, 1988 55

Electrochemical Reduction of CO2 ""I

--

301

120

- loo

A

25-

-

2

6

q

20-

5

e

,

electrode 1 electrode 2

X

x rate

g

-60 -40

+

-

x

10-

o " ' = " " . " " ' " " ' " -0.45

I

01 0

2

6

4

8

PH

8 5

5

W

20 0

-0.60

-0.65

E (v vs SCE)

10

Figure 6. Plots of the rate of methane formation at Ru electrodes as a function of p H in 0.2 M reagent grade Na2S04at 60-63 "C

-0.55

-0.50

-

T

B

+ current

15-

%

80

Figure 7. Plot of methane formation rate and current vs potential for electrochemical reduction of COPat Ru electrodes in 0.2 M Na$Oc initial pH 4,at -0.545 V vs SCE. Hydrogen couple formal potential, -0.36 Q vs

SCE.

and a t a constant overpotential. Table 111. Effect of Potential on t h e Rate of Methane Formation a t Ru Electrodes" j: rate? mol V vs SCE Q,b C PA cm-2 cm-2 h-' eff: 70 0 0 -0.48 1.3 22 10.8 2.0 30 1.6 X -0.50 2.8 x 10-8 10.2 -0.51 3.4 57 4.3 X 15.4 3.3 60 -0.545 10.3 4.8 86 4.1 X -0.545 6.8 X 14.2 -0.60 5.8 102 aIn 0.2 M Na2S04at 60 "C and an initial pH of 4. All electrolysis times were -5 h. bTotal charge passed. CAveragecurrent density based on geometrical area. dAverage rate of methane formation. e Faradaic efficiency for methane formation.

gas chromatograph could measure (-0.1 atm). However, CHI can be formed at pH values (below pH 3) where there is not a significant concentration of bicarbonate ions. Therefore, at acidic pHs at least, reduction of C 0 2 or H&03 and not of HCOy occurs. Also, the concentration of carbonic acid at a C02 partial pressure of 1 atm is probably much too lowe (6 X M) to support the observed CHI formation rates, even if the methane formation was diffusion controlled, further supporting the reduction of aqueous C02. As Figure 6 shows, the CH, formation rate does depend on pH. In the pH region 9 to -3 the rate increases. This effect is rationalized as occurring because of either an increased surface hydride coverage, increasing the rate of hydrogenation of C02 reduction intermediates, or an increased rate of oxygen removal from the surface, favoring the deoxygenation of COz or its intermediates. At pHs less than 2-2.5 the rate begins to decrease. This may be due to a coverage of surface hydrides that is so high that sites for C 0 2 reduction are blocked. Effect of Electrode Potential. Figure 7 and Table I11 summarizethe results concerning the influence of electrode potential on the CHI formation rate in a C02-saturated, 0.2 M Na2S04solution at 60 "C. The data show an apparent linear dependence of the CHI formation rate on potential. The increase in rate with potential is not unexpected since the CHI formation reaction is too slow to be diffusion controlled. The rate drops to zero at -0.48 V vs SCE. With an average pH of -5 and a partial pressure of hydrogen estimated to be 1 X lo+ atm, the reversible potential for hydrogen evolution is -0.36 V vs SCE. Therefore, the potential at which the CHI formation begins is 120 mV cathodic of the formal hydrogen potential, and hence the electrode has an appreciable hydrogen coverage at potentials where CH4 is formed. This

-

Table IV. Effect of Stirring on t h e Rate of Methane Formation at Ru Electrodesa j , b PA cm-2 rate: mol cm-2 h-' eff,d % stir no 91 1.2 x 10-7 27.5 Yes 135 1.3 x 10-7 20.7 no 128 1.3 x 10-7 21.7 Yes 213 1.4 x 10-7 13.6 In 0.2 M Na2S04at 60 "C and an initial pH of 4. All electrolysis times were -5 h. *Average current density based on geometrical area. Average rate of methane formation. Faradaic efficiency for methane formation. Table V. Effect of Temperature on Methane Formation Rate' rate: mol runb T,"C Q," C j , d FA cm-2 cm-2 h-' eff] '70 1 41 4.6 78 1.2 x 10-8 3.4 2.9 X 12.1 4 50 2.8 51 1.6 X lo-@ 7.8 7 60 2.5 45 2 61 3.0 68 5.9 x 10-8 18 8.4 X 31 3 71 3.0 58 8.4 X 42 5 82 2.4 43 3.9 x 10-8 19 6 90 2.4 44

" In 0.2 M Na2S04at -0.545 V and an initial pH of 4. All electrolysis times were 5-6 h, using the same electrode. Order of experiment. Total charge passed. Average current density based on geometrical area. eAverage rate of methane formation. /Faradaic efficiency for methane formation. is consistent with a model for the reduction of COz in which surface carbon intermediates are hydrogenated with surface hydrides in a key step. At more cathodic potentials the hydrogen coverage increases, thereby increasing the rate. Further discussion on this point is presented in the Discussion section. Effect of Stirring Rate. The data in Table IV show the effect of stirring on methane formation at 60 OC, initial pH 4, and -0.545 V vs SCE. In each case the electrolyte was either unstirred, except for the mild agitation caused by the C02 recirculating pumping, or stirred as vigorously as possible with a magnetic stirring bar. Vigorous stirring caused the average current density to increase, most likely due to enhanced hydrogen transport to the electrode surface or into voids near the somewhat porous electrode surface. The methane formation rate, however, does not change, indicating that the rate is not governed by mass transport to the surface and consistent with an activation controlled surface chemical rate-determining step. Effect of Temperature. The effect of temperature on the rate of CH4 formation at a Ru/C electrode (Ru electroplated on carbon) at -0.545 V vs SCE is shown in Figure

56 Langmuir, Val. 4, No. 1, 1988 5

Summers and Frese

I

3

I

. 40

50 60

70

80

90 1W

T I"C1

-20 0.0027

0,0028

0.0029

0.0030

0.0031

0.0032

1/T(K11

Figure 8. Plot of In of methane formation rate vs temperature and faradaic efficiency vs temperature for electrochemical reduction of C 0 2 at Ru electrodes in 0.2 M Na2S04,initial pH 4, at 4 . 5 4 5 V vs SCE (for clarity the faradaic efficiency was plotted only for Ru/C).

8; the data are listed in Table V. The inset to Figure 8 illustrates a plot of faradaic efficiency vs temperature. It is noteworthy that an efficiency of 42% at 80 "C is the highest recorded for CH, formation. The increase in the faradaic efficiency for CHI formation indicates that CHI formation increases faster with temperature than competing reactions, e.g., H2 evolution, and so must have a higher activation energy. The temperature dependence experiments are numbered in the order in which they were conducted. For the first four experiments (each done on successive days at T < 75 "C), the order of the experiments did not affect the performance of the electrode. However, if the electrode is used in electrolyses above -80 "C, it begins to deactivate. The experiment at 90 "C led to a decline in the CH, rate, and, when the electrode was used at 60 "C after the 90 "C experiment, the CHI formation rate was significantly reduced (as was the faradaic efficiency). Thus at T > -80 "C, irreversible deactivation does occur. If an electrode prepared by plating onto a Ti foil (Ru/Ti) is used in a similar set of experiments, then similar behavior is observed (Figure 8) with irreversible deactivation occurring at T > -80 "C. This shows that the deactivation is not caused by migration of carbon atoms onto the surface of the Ru from the carbon substrate (in the case of Ru electroplated onto carbon rods). The same effect is seen from the data on multiple electrolyses in Table I. It can be seen that if an electrode is used several times to electrolyze a C0,-saturated solution at 60 "C no decline in average CHI formation rate is observed, but at 80 "C a clear decline in CHI formation is seen from one experiment to the next. Such behavior is also consistent with a slow deactivation of the electrode surface at higher (>-80 "C) temperatures. Auger Electron Spectroscopy. Before the temperature experiments that were conducted on the Ru/Ti electrode were performed, a section of the electrode (which had been used previously for one electrolysis at 60 "C, pH 4, -0.545 V vs SCE)was removed. The Auger electron spectrum of this section (Figure 9) shows a predominantly Ru surface with traces of Ti and 0. The oxygen may result either from a partial surface coverage of oxygen (either on Ti or Ru) or from traces of 50;- from the electrolyte on the surface (the sulfur Auger peak may be buried under the Ru peak at 150 eV). The Auger electron spectrum of the other portion of the electrode after it was used to study the effect of temperature (and was subsequently deactivated by electrolysis at T 2 80 "C; see Figure 8) is shown in Figure 10. The large peak due to Ru is changed in symmetry and size. The

100

I

I

I

I

200

300

400

500

Energy (ev)

Figure 9. Auger electron spectrum of the surface of a Ru electroplate electrode (Ti substrate) after one electrolysis of a C02-saturated,pH 3,0.2 M Na2S04aqueous solution at -0.48 V vs SCE at 60 "C.

0

W/C 100

200

I

,

300

400

I

500

600

Energy (ev)

Figure 10. Auger electron spectrum of the surface of the same electrode that was sampled for Figure 9 at 60, 70,80, and 90 "C in a C02-saturated,pH 4,0.2 M Na2S04aqueous solution at 4 . 5 4 5 V vs SCE and 60-90 "C (see text).

signal is no longer highly symmetrical, with only one relatively strong positive peak for the highest energy signal, but becomes highly unsymmetrical and exhibits two weak positive peaks (one is a shoulder on the other). These characteristics are indicative of the presence of a large amount of carbon on the surface due to the accidental overlap of the Auger signals of Ru and carbon7 If the electrode that shows the large carbon signal is Ar+ ion sputtered, the carbon signal disappears and the Ru signal regains its normal structure, indicating that the carbon signal is due to carbon on the surface of the Ru. Thus it appears that the deactivation of electrodes at higher temperatures is caused by the formation of surface carbonaceous species. This does not occur at lower temperatures since an electrode used only at 60 "C does not show the presence of large amounts of surface carbon (Figure lo), and no deactivation is seen over repeated electrolysis at 60 "C (see above). The lack of an oxygen signal of similar size indicates the carbon species being formed on the surface is not oxygenated (e.g., Cz03,COH, or CO). Thus,species such as graphitic carbon and/or CH, are implicated. The changes in the Ru signal are the same as those reported3 for Ru/C electrode in similar experiments. It has been postulatedsJ3that the formation of CHI by the heterogeneous catalytic reduction of CO gas with gaseous hydrogen proceeds via carbon atoms. Such a (13) Winslow, P.;Bell, A. T.J. Catal. 1985, 91, 142.

Langmuir, Vol. 4, No. 1, 1988 57

Electrochemical Reduction of COZ mechanism involving dissociative adsorption of CO may also operate during the electrochemical reduction of C02 in aqueous solution. We are led to the conclusion that the deactivation of the electrode occurs because of the polymerization of either surface carbon atoms or CH, species to an inactive form of carbon. This may occur because of a depletion of hydrogen atoms on the surface or an excess coverage of carbon at higher temperatures, or it may be that at higher temperature carbon atoms are more mobile and can move across the surface to combine. Nonoxygen-containing species, such as graphitic carbon or CH,, are most likely to be formed from a mechanism involving the dissociation of carbon dioxide to surface carbon and oxygen on the surface. An Arrhenius plot using the low-temperature data (last three points) gives an activation energy of -36 kJ mol-'. McKeee observed an activation energy of 38 kJ mol-l for the rate of formation of CHI from H2 and CO on unsupported Ru catalysts in the temperature range 25-150 OC. McKee also observed a curvature in his Arrhenius plot similar to that seen in Figure 8 though at a slightly higher temperature.

Discussion A case has beenlo made for a mechanistic commonality between gas/solid and electrocatalytic approaches to similar reactions such as the interaction of hydrogen molecules or CO with Pt surfaces. Unsupported Ru has exceptional activity for methanation and Fischer-Tropsch-type gas/ solid reactions>b In previous work,ll we have found that the electrochemical formation of CHI has only been observed on Ru and not with other materials such as Pt, Mo, C, Pd, Ag, Os, Ni, GaAs, Gap, and Si. Evidently the exceptional character of Ru in gas-phase reactions is carried over in electrochemical systems. It is useful to discuss our electrochemical results vis B vis what is known about the gas/solid methanation reaction. However, the formation of CHI from CO,5 rather than C02,4is much better characterized. Consider what is clear. The pH dependence of the CHI rate at constant overpotential has a pronouhced maximum. A rate-limiting surface process involving H atoms is suggested. That CH, is only formed at Ru electrodes indicates that importance of the surface catalytic properties of Ru; otherwise, other electrode surfaces would lead to methane. There is strong support in the 1iteraturel3for the existence of surface carbon atoms formed from CO dissociation and for hydrogenation of the active form leading to CH4. Although rate limitation by chemical dissociation of CO or C02 is plausible, the maximum in the pH dependence would seem to rule out such a limitation. The coverage with hydrogen atoms would decline for any increase in pH, thus more free sites for oxygen or carbon would be present, and the CHI rate would increase without a maximum, contrary to observation. The C 0 2 reduction current leading to methanol on GaAs also has a similar maxi-

mum,ld and it was concluded that the rate-limiting step is a chemical combination of a surface H atom and a carbon-containing intermediate. Our results here support such a conclusion. The CH4rate did not saturate in the electrode potential range investigated. It may be concluded that the surface is not saturated with intermediates at pH 4-5 at 60 OC. The enhancement in the rate upon addition of hydrogen gas is consistent with (a) an unsaturated surface and (b) the increase in CHI rate for pH