J . Phys. Chem. 1991,95,9451-9464
9457
Surface Kinetics Studled by Using Ion-Stimulated Desorption of Neutral Molecules: Time-Resolved Decomposition of CH,OH Adsorbed on Ni( 110) John J. Vajo,**+James H. Campbell,$ and Christopher H. Beckert Molecular Physics Laboratory, SRI International, Menlo Park, California 94025, and Department of Chemistry, Stanford University, Stanford, California 94305 (Received: April 17, 1991; In Final Form: June 12, 1991)
The time-resolved decomposition of methanol on Ni( 1 10) has been studied by using a pulsed Ar' beam at 2 keV followed by laser ionization at 193 nm of the desorbed (sputtered) neutral species and reflecting time-of-flight mass spectrometry of the resulting photoions. Methanol decomposition was monitored during temperature-programmed reaction (TPR) following molecular adsorption at 120 K for initial coverages of 0.09 and 0.29 monolayer (ML) with respect to the number of exposed nickel atoms, and at 170 K for 0.04,0.15, and 0.29 ML. Cleavage of the 0-H bond in molecularly adsorbed methanol and decomposition of methoxy to carbon monoxide and hydrogen were clearly resolvable even though the products for both of these reactions remained adsorbed on the Ni(l10) surface. Reaction occurred exclusively via a methoxy intermediate that formed via 0-H bond cleavage during TPR from 140 to 240 K. Methoxy decomposed to adsorbed carbon monoxide and hydrogen at 240-290 K. No evidence for reactions involving C-O bond cleavage in either molecularly adsorbed methanol or methoxy was seen. For an initial coverage of 0.29 ML, methanol desorption occurred together with decomposition and -0.08 ML of methanol desorbed during TPR. Methoxy decomposition and 0-H bond cleavage kinetics were monitored primarily using the decrease in the surface concentrations of methoxy and molecular methanol, respectively, with increasing temperature during TPR. The kinetics for both of these reactions appeared similar for the range of initial methanol coverages studied. Based on analysis of the kinetics using model calculations, details of the mechanism for both methoxy formation and decomposition on Ni( 1 10) are discussed.
1. Introduction Kinetic information concerning elementary surface chemical transformations is essential to a fundamental understanding of gassurface heterogeneous reactions at the molecular level. For instance, delineating the kinetics for the individual reaction steps that comprise an overall catalytic reaction cycle can serve as a basis for a microscopic description of heterogeneous catalysis and thereby rational catalyst design and modification. However, measurements of the kinetics of surface chemical reactions are technically difficult, and only recently have experimental techniques become available with the requisite sensitivity and generality to monitor surface reactions in real time. These techniques include time-resolved electron energy loss spectroscopy (TREELS),'>* secondary ion mass spectrometry (SIMS)?" laser-induced thermal desorption (LlTD),F7 IR reflection-absorption spectrascopy,8-I0 and time-resolved Auger and photoelectron spectroscopies."J2 In our laboratory, we have used pulsed ion-stimulated desorption followed by laser ionization time-of-flight (TOF) mass spectrometric analysis of the sputtered neutral atoms and molecules to study surface reaction kinetics." This technique is known as surface analysis by laser ionization (SALI).I4 The difficulty in monitoring surface reaction kinetics originates from several factors including the low concentration of surface species, usually much less than a monolayer (ML). In addition, to extract meaningful kinetic information, the signal that characterizes a particular technique must be relatable to the surface concentration, or at least the relative concentration, of the species being studied. Quantification of surface concentrations is often required for several species simultaneously under conditions of changing surface composition and total coverage. Finally, interrogation of the surface chemistry must not significantly alter the kinetics. Each experimental approach mentioned above has inherent limitations as well as advantages based on these criteria. Discussions of these limitations and advantages may be found elsewhere.13Js Recently, we have examined methanol adsorption and decomposition at a coverage of 0.09 ML on Ni( 110) by using SALI.I3 Adsorption at 170 K occurred with a probability near unity and *Current address: Hughes Research Laboratories, M.S. RL62, 301 1 Malibu Canvon Rd. Malibu. CA 90265. 'SRI lnt&national. *Stanford University.
0022-3654/91/2095-9457$02.50/0
was dissociative, forming an adsorbed methoxy radical and a hydrogen adatom. At 120 K, adsorption was molecular. Temperature-programmed reaction (TPR) of adsorbed methanol with a heating rate of 0.26 K/s indicated that 0-H bond cleavage occurred from -140 to 240 K. Decomposition of adsorbed methoxy to adsorbed carbon monoxide and hydrogen was observed at 240-290 K with a maximum rate at -260 K. Assuming a preexponential factor of i09-iOi3 s-I, an activation energy for methoxy decomposition of 12.5-17.2 kcal/mol was calculated based on the Redhead equation.16 These results agree with those of previous studies for the formation of methoxy from methanol on Ni(1 10),1917-19Ni(100),20J' and Ni( 11 1).22-26 By using EELS, 0-H bond cleavage in mo(1) Richter, L. J.; Gurney, B. A.; Villarrubia, J. S.;Ho, w. Chem. Phys. Lerr. 1984, 113, 185. (2) Ho, W. J. Electron Specfrosc. Relur. Phenom. 1987, 45, I. (3) White, J. M. Appl. Surf. Sci. 1986, 26, 392. (4) Radloff, P. L.; White, J. M. Acc. Chem. Res. 1986, 19, 287. (5) Hall, R. B.; Desantolo, A. M. Surf. Sci. 1984, 137, 421. (6) Hall, R. B. J. Phys. Chem 1987, 91, 1007. (7) Deckert, A. A.; Brand, J. L.; Mak, C. H.; Kochler, B. G.; George, S. M. J. Chem. Phys. 1987,87, 1936. (8) Chenery, D. H.; Chesters, M. A.; McCash, E. M. Sur$ Sci. 1988,198. 1.
(9) Dubois, L. H.; Ellis, T. H.; Zegarski, B. R.; Kevan, S.D. Surf Sci.
1986, 172, 385.
(10) Burrows, V. A.; Sundaresan, S.;Chabal, Y. L.; Christmann, S. B. Surf. Sci. 1986, 180, 1 IO. ( I 1) Balooch, M.; Olander, D. R.; Abrefah, J.; Siekhaus, W.J. Surf. Sci. 1985, 149, 285. (12) Rubloff, G. W.Surf. Sci. 1979, 89, 566. (13) Vajo, J. J.; Campbell, J. H.; Becker, C. H. J. Vuc. Sci. Technol. A 1989, 7, 1949. (14) Becker, C. H.;Gillen, K. T. A w l . Chem. 1984, 56. 1671. (15) Ho, W . J . Phys. Chem. 1987, 91, 766. (16) Redhead, P. A. Vucuum 1962, 12, 203. (17) Richter, L. J.; Ho, W. J . Chem. Phys. 1985, 83, 2569. (18) Bare, S. R.; Stroscio, J. A.; Ho,W. Surf. Sci. 1985, 150, 399. (19) Richter, L. J.; Ho, W . J . Vac. Sci. Technol. A 1985, 3, 1549. (20) Johnson, S.;Madix, R. J. Surf. Sci. 1981, 103, 361. (21) Hall, R. B.; Desantolo, A. M.; Bares, S. J. Surf. Sci. 1985, 161, L533. (22) Gates, S. M.; Russell, J. N., Jr.; Yates, J. T., Jr. Surf. Sci. 1985. 159,
233. (23) Gates, S. M.; Russell, J. N., Jr.; Yates, J. T., Jr. J. Curd. 1985, 92, 25.
0 1991 American Chemical Society
9458 The Journal of Physical Chemistry, Vol. 95, No. 23, 1991 Excimer Laser
'
- k
Vajo et al. 1 meter
Arlon
1
u Computer
Figure I. Schematic diagram of the apparatus for stimulated desorption studies showing the configuration of the sample, electron and Art sources, and time-of-flight (TOF) mass spectrometer. Pulsed sources of ions or electrons (typically 5 ws) with a repetition rate of 20-100 Hz desorb both neutral molecules and secondary ions from the sample. Neutral molecules are ionized 1 mm away from the surface by a focused excimer laser operating at 193 nm and firing at the end of the electron or ion pulse. The photoions are then analyzed by the TOF mass spectrometer.
-
-
lecular adsorbed methanol on both Ni( 1 IO)'* and Ni( 1 1 was observed after heating to 180 K. In a preliminary TREELS study, 0-H bond cleavage on Ni(l10) was observed with a maximum rate at 180 K using a heating rate of 1.9 K/s following saturation exposure at 130 K.I7 On Ni( IOO), isothermal LITD measurements have determined that 0-H bond cleavage occurs with an activation energy of 9 kcal/mol at low coverages.21 Although kinetic parameters for 0-H cleavage were not measured directly in the EELS and SAL1 studies, an activation energy of 9 kcal/mol is consistent with the observations for Ni(l1 and Ni( 1 10).13*'7 The similarity of the kinetics for the formation of adsorbed methoxy on Ni( 1 1 I), Ni( IOO), and Ni( 110) demonstrates that for low coverages this reaction on nickel is not particularly sensitive to surface structure. Decomposition of adsorbed methoxy has also been studied on by using TREELS and on Ni(l11) by using both Ni( 1 isothermal measurements% and temperature-programmed reaction in a constant flux of methanol.22 On Ni( 1 IO), a preexponential factor of s-I and an activation energy of 16 f 2 kcal/mol were measured by using the method of heating rate variationI7 at coverages below 0.2 ML. Our estimate of 12-17 kcal/mol agrees well with these values. By using second harmonic generation to monitor the concentration of methoxy following temperature jumps to between 240 and 260 K, a preexponential factor of 10'2.4s-' and an activation energy of 17.0 f 0.3 kcal/mol were measured for methoxy decomposition on Ni(l1 Methoxy decomposition on Ni( 11 I ) occurs at 230-290 K during heating from 187 K in a methanol flux of 4 X IOL2 cm-2 s-1.22 This temperature range for decomposition of methoxy is similar to that seen on Ni( 1 IO). In the present study, our previous work on the decomposition of methanol adsorbed on Ni( 1IO) is extended to include initial coverages from 0.04 to 0.3 ML. At the highest coverage studied, desorption of molecular methanol competes with methanol decomposition. However, the kinetics for both the formation and decomposition of adsorbed methoxy are similar over the range of coverages studied. Based on these results, details of the de(24) Gatcs, s. M.;Russell, J. N., Jr.; Yates, J. T., Jr. SurJ Sci. 1984, 146,
199.
(25) Demuth, J. E.; Ibach, H. Chem. Phys. Leu. 1979, 60, 395. (26) Hall, R. 9.: De Santolo, A. M.;Grubb, S. G. J . VUC.Sci. Techno/. A 1987, 5, 865.
composition of adsorbed methoxy and 0-H bond cleavage in molecularly adsorbed methanol during temperature-programmed reaction on Ni( 110) are discussed. 2. Experimental Procedures The measurements were performed in a two-level ultrahighvacuum system designed specifically for studies using electronor ion-induced desorption of neutral species together with TOF mass spectrometry. Facilities are contained in the upper level for low-energy electron diffraction (LEED), Auger electron spectroscopy (AES) using a cylindrical mirror analyzer, and electron-impact quadrupole mass spectrometry. The lower level contains pulsed electron and ion sources, laser entrance and exit windows, a laser ionization reflecting TOF mass spectrometer, and a doser assembly. As shown in Figure 1, the electron and ion sources are located on either side of the TOF mass spectrometer, while the laser windows are aligned so that photoionization occurs -3 mm in front of the TOF optics and 1 mm in front of the sample. Neutral species desorbed by pulsed electrons or ions, or present in the background gas, were ionized by a focused, pulsed, untuned excimer laser operating at 193 nm and -50 mJ/pulse, yielding approximately 3 X lo9 W/cmZ. A 40-cm focal length quartz lens that is located outside the vacuum system focuses the laser beam. The laser is triggered at the end of the electron or ion pulse. The path of the ions formed by laser ionization is shown schematically in Figure 1. Following ionization, an ion optics column accelerates, focuses, and deflects the ions into the TOF mass spectrometer, which operates at a drift potential of -1 keV. An electrostatic mirror reflects ions formed by laser ionization, while any secondary ions pass through the reflector and are not detected.14 Reflected ions are detected by microchannel plates. Typical flight times following ionization for m / e = 12 and 64 are 10 and 30 ps, respectively. The signal is amplified (1SX) and recorded by a 100-MHz transient recorder interfaced to a DEC LSI 1 1 /23 laboratory computer via CAMAC. The vacuum pump chamber, which has a base pressure of 2 X 10-loTorr, is pumped by a 200 L/s ion pump and a titanium sublimation pump (TSP). In addition, the ion source is differentially pumped by a 50 L/s turbomolecular pump, and the drift region of the TOF mass spectrometer is pumped by a TSP. Differentially pumping the ion source limits the pressure rise at the sample to 240 K,methanol desorption (eq 5) and CH30(a) decomposition (eq 2) occur concurrently. If the decline in I( 13) below -240 K is attributed to desorption of methanol, then for dinit= 0.29 ML and 8 = 0.26 K/s approximately 0.08 ML of methanol desorbs during TPR. Desorption of methanol has been reported previously during TPR on Ni( 1 10) following relatively high initial c o ~ e r a g e s . ~Methanol ~ . ~ ~ desorption was seen following exposures >1 L at 80 K (equivalent to -0.3 ML if the adsorption probability was unity); however, the quantity of methanol that desorbed was not reported.I8 In another study on Ni( 1 IO), 0.06 f 0.03 ML of methanol was estimated to desorb during TPR following a saturation exposure at 170 K.” In these previous studies, methanol desorption at T 2 200 K was considered to be recombinative;I7J8 Le., methanol desorption occurred via the reverse reaction in eq 1 followed immediately by eq 5. A line shape analysis of the methanol thermal desorption spectra yielded a desorption energy of IO kcal/mol.I8 Subsequently, heating rate variation studies and second-order line fits to thermal desorption spectra gave a desorption energy of 14 f 4 k ~ a l / m o l . ’However, ~ the kinetic analysis for this desorption feature may not be straightforward considering that the recombinative desorption of methanol on Ni( 11 1) occurs during 0-H bond cleavage (eq Thus, any kinetic analysis of methanol desorption may depend sensitively on the kinetics of eq I . Moreover, CH,O(a), which has a temperature-dependent surface
coverage, may affect the methanol desorption energy via metal-adsorbate or adsorbate-adsorbate interactions. In general, the mechanism given solely by eqs 1-5 and additionally the kinetics for the formation and decomposition of CH30(a), which are discussed below, are generally consistent, as discussed elsewhere,I3 with previous studies of methanol decomposition on Ni( l l and Ni( 1 I O ) l 7 in which CH30H, CO, and H2 were the only final products observed. 4.2. Issue of C-O Bond Cleavage. For the range of coverages studied here, there was no evidence for reactions involving C-O bond cleavage in either CH30H(a) or CH30(a) unlike previous reports on a polycrystalline Ni Pd( 11 1),30*3’and oxygenprecovered Pt( 11 On a polycrystalline Ni foil, CH3(a) and OH(a) were reported to form via C-O bond cleavage immediately upon adsorption of methanol at temperatures 1 140 K from a flux of 10I6cm-2 For a methanol flux of 1014cm-2 s-I, only the reactions described by eqs 1-5 were observed.29 In contrast, on Ni( 11 1) no isotopic exchange in the product CO was detected from a mixture of I3CH3l60Hand 12CH3180Hduring TPR.32 This implies that the C-O bond remains intact during methanol decomposition on Ni(l1 I ) . In light of these results, it was suggested that the C-0 bond cleavage mechanism, observed on polycrystalline Ni foil, occurs on disordered areas, on more open crystalline areas such as (100) and (1 10) areas, or at grain boundaries on the foil surface.33 On Pd( 11 I), it was reported that the net reaction 2CH30H(a) CH30(a) CH3(a) H,O(a) occurs during TPR via C-O bond cleavage.m Since H2 and C O were the only products observed during methanol decomposition on Pd( l l l), molecularly adsorbed water, produced during C-0 bond cleavage, was proposed to decompose to H(a) and O(a) at 225 K. At higher temperatures, recombinative desorption of O(a) and C(a) (from decomposition of CH3(a)) yielded CO. However, the assertion that H20(a) decomposes on Pd(ll1) is contrary to the established behavior of molecular adsorbed water on many transition metals including Pd(l1 The potentially unique behavior of water on Pd( 1 1 1) during methanol decomposition remains to be examined in detail. In the present study, if indeed C - O bond cleavage were occurring, the ISD signals for m / e = 12, 15, and 16 could be indicative of C(a), CH,(a), and O(a), respectively, Le., possible products of C-O bond cleavage. However, the variation in these signals is consistent with the variation in the signals for m/e = 13, 28, and 32. Therefore, these signals are predominantly fragments of CH30H, CH30, or CO and simply further illustrate the mechanism depicted in eqs 1-5. In addition, as mentioned above, CO(a) is formed during decomposition of CH,O(a). Finally, no water was detected either on the Ni( 1 10) surface by using ISD measurements or desorbing from the surface in separate thermal desorption measurements. 4.3. Formation of Adsorbed Metboxy. Formation of CH30(a), via eq 1, occurs during TPR from 140 to 240 K. This IS an extremely large temperature range for a simple unimolecular process with /3 = 0.26 K/s. To illustrate this point, eq 1 was used to model Z(32) for dinit = 0.09 ML using a rate constant of the form k , = y l exp[-EI/RT] where the preexponential factor, vl, and the activation energy, E l , were assumed to be coverage independent. The results of the calculations with v I = IO9 and IO2 S-I and E l = 10.3 and 3.8 kcal/mol, respectively, are shown as curves a and b, respectively, in Figure 5 together with Z(32) for einit= 0.09 ML. For v I = lo9 s-I, which is a low but physically
-
+
+
-
(29) Steinbach, F.; Krall, R.; Cai, J.-X.; Kiss, J. In Proc. 8th Infern.Congr. on Catalysis, Berlin, July 1984; Verlag Chemie: Weinheim, 1984; Vol. 3, p 111-359. (30) Levis, R. J.; Jiang, Z. C.; Winograd, N. J . Am. Chem.Soc. 1988,110, 443 I : 1989. I I I . 4605. (31) Levis, R’. J.; Jiang, Z. C.; Winograd, N.; Akhter, S.;White, J. M. Catal. Lett. 1988, I , 385. (32) Russell, J. N., Jr.; Chorkendorff, I.; Yates, J. T., Jr. Surf. Sci. 1987, I83 316 ---. (33) Steinbach, F.; Krall, R. Surf. Sci. 1987, 183. 331.
(34) Davis, J . L.; Barteau, M. A. Surf. Sci. 1988. 197, 123. (35) Stuve, E. M.; Jorgensen, S. W.; Madix, R. J. Surf. Sci. 1984, 146, 179. (36) Thiel, P. A.; Madcy, T. E. Surf. Sci. Rep. 1987, 7,21 I .
9462 The Journal of Physical Chemistry, Vol. 95, No. 23, 195’1 1.2,
z
I
1
I
1
+ +
0
P
,
+ T f
0.61
+ ++
P0
t,
m cu
0.2 0.0
++++
100
150
200
+
250
300
TEMPERATURE (K)
Figure 5. Comparison of 1(32) for Binitial = 0.09 ML with model calculations b a d on eq 1. For curve a, vI = lo9 s-I and E, = 10.3 kcal/mol. For curve b. uI = 100 s-I and E , = 3.8 kcal/mol. For curve c, uI = IO9 s-I and E, = 9.0 32BCHIO kcal/mol. The model curves have been normalized to unity. The experimental data have been rescaled from that shown in Figure 2 so that the average initial signal is unity and so that for T > 250 K the points scatter around zero.
+
reasonable value, the calculated methanol coverage decreases much more rapidly than the experimental data. Even for vI = lo2 s-I, which we consider an unphysically low value, the calculations slightly overestimate the rate of decrease of 1(32) at low coverages. An alternative explanation to consider is that the I ( 3 2 ) signal does indeed represent a simple unimolecular process with coverage-independent kinetic parameters, but with the ion-stimulated desorption signal either being a nonlinear function of surface coverage or being perhaps temperature dependent. While it is difficult to determine the dependence of I ( 3 2 ) on the methanol coverage during TPR, adsorption measurements at constant temperature are consistent with an approximately linear dependence of I(32) on methanol coverage.13 In addition, as discussed below, I(32)’s for different initial coverages during TPR are very similar. This would not be expected for a unimolecular reaction with coverage-independent kinetic parameters if the ISD signals were nonlinearly dependent on surface coverage. Also, purely temperature-dependent effects on ISD signals have not been observed for this system under conditions where no surface reaction is expected. For example, following decomposition of CH30(a) (or absorption of CO), 1(28) is constant with increasing temperature until C O desorbs from the surface. For CH30H(a), which is more weakly bound than CO(a), the stimulated desorption (sputter) yield may be expected to change during TPR if the methanol binding energy is significantly affected by the accumulation of H(a) and/or CH30(a) on the surface. However, it is difticult to construct a physically reasonable coverage-dependent yield function that would bring a calculated methanol coverage during TPR, e.g., curve a of Figure 5, into agreement with the experimental data. For example, an inverse dependence of the sputter yield on binding energy, which is used to describe preferential sputtering in alloys,37 and a linear dependence of the binding energy on the extent of reaction cannot account for the differences between the calculated and experimental data. In addition, TPR measurements in which 4 X IO” Torr s of deuterium was coadsorbed with 0.09 ML of methanol showed no effect of the deuterium on 1(32). This observation indicates that H(a), which forms as the reaction given by eq 1 proceeds, does not significantly affect 1(32) during TPR. Further evidence against significant temperature-dependent effects comes from TPR measurements in which the temperature ramp was reversed at temperatures between 180 and 220 K. For these measurements, as the temperature decreased, the ion-stimulated desorption signals remained constant. This result precludes a large effect of temperature on the ISD signals in the absence of reaction. We also considered the possibility that 1(32) represents a reversible process with eq 1 at (or near) equilibrium as the temperature is increased. However, the observation that 1 ( 3 2 ) (37) Kelley, R. Nucl. Instrum. Methods 1986, B14, 421.
Vajo et al. remains constant as the temperature ramp is reversed indicates that eq 1 is not reversible under the conditions of our measurements. Previously, we suggested that coverage-dependent kinetic parameters of the rate constant For the reaction in eq l could account for the observed kinetics.” Indeed, a coverage-dependent activation energy was measured for O-H bond cleavage in methanol on Ni( 100) by using LITD.*’ To demonstrate that CH30H(a) decomposition via eq 1 may be described with coverage-dependent kinetic parameters, a model calculation was performed by using a constant preexponential and an activation energy given by E l = El(0)+ aOCHIO, where El(0)is the initial activation energy and a is the coefficient representing the increase in E, per monolayer of CH30(a). Curve c in Figure 5 shows that a reasonable fit to the data is obtained with uI = lo9 s-I, El(0)= 9 kcal/mol, and a = 32 kcal/(mol ML). Holding u, and E,(o)constant, acceptable values of a range from 28 to 40 kcal/(mol ML). When compared using a common definition of a monolayer (1.14 X 1015cm-2 is used in the present study and 4 X lOI4 cm-2 is used in ref 21), the coefficient a is determined here as approximately twice that determined for low initial coverages on Ni(100).2’ This difference may indicate a greater susceptibility of the more open Ni( 1 10) surface to poisoning by electron-withdrawing species such as CH30(a). Although for einit= 0.09 M L a reasonable fit to the data is obtained with a = 32 kcal/(mol ML), for Oidt = 0.29 ML a value of approximately 2 0 kcal/(mol ML) is required to describe the experimental data. This value is approximate because desorption of -0.08 ML of methanol occurs during TPR for e, = 0.29 ML, which complicates the modeling process. Thus, a appears to decrease with increasing initial coverage. This trend in a with einit has also been determined for methanol decomposition via eq 1 on Ni( As suggested previously,2’ the formation of islands of methanol could account for the observed kinetics. Specifically, if methoxy formation occurred from within islands of a common local coverage for both overall initial coverages of 0.09 and 0.29 ML, then a single value of 8, which is equal to the coverage within the islands, should be used to model both initial coverages. In that case, a single value of a would describe both sets of data. Although direct evidence for formation of islands following methanol adsorption on Ni( 110) is not available, islanding of methanol due to hydrogen bonding on Fe(ll0) has been suggested based on EELS.38 However, before a decomposition mechanism based on island formation could be fully substantiated or refuted, additional questions must be addressed. For example, how does decomposition from the edges of the islands contribute to the observed kinetics and how does the desorption of methanol compete with decomposition. Another possibility, also discussed previously,2’ is that the increase in El with &.-Hlo is not linear and may, for example, begin to saturate as eCHIO increases. 4.4. Decomposition of Adsorbed Methoxy. Previously, the kinetics of CH30(a) decomposition on Ni( 110) to CO(a) and H(a), via eq 2, were studied by following the electron energy loss intensities for CH30(a) vibrations during TPR by using TREELS.” By measuring the peak temperature, Tp, corresponding to the maximum rate of decrease of the vibrational spectral intensities as a function of heating rate, the activation energy and preexponential factor for CH30(a) decomposition were independently determined to be 16 f 2 kcal/mol and PI, respectively.” In the present study, the heating rate was not varied. However, by using the Redhead equation,16 by using an experimental value for T from Figure 2, and assuming a preexponential factor of 109-1013\-1, an activation energy for CH30(a) decomposition of 12.5-17.2 kcal/mol is found. To explore further the validity of these kinetic parameters, both the parameters measured previously with TREELS and those obtained from the Redhead equation from our ISD data were used to calculate 8CH10during TPR. A curve calculated assuming E2 = 16.2 kcal/mol and v2 = 10l2s-’ is shown as curve a in Figure 6 together with I ( 1 3 ) for Oinit = 0.09 ML. The calculated curves (38) McBreen, P. H.; Erley, W.; Ibach, H. SurJ Sci. 1983, 133, L469.
The Journal of Physical Chemistry, Vol. 95, No. 23, 1991 9463
CHjOH Adsorbed on Ni( 1 10)
together with 2H,CO(a)
-
or H2CO(a) H,CO(a)
I
200
I
250
,
,
,
l
I
I
300
I
I
350
TEMPERATURE (K)
Figure 6. Model calculations for methoxy coverage during TPR and I( 13) for Binithl = 0.09 ML. For curve a, methoxy decomposition was assumed to proceed via eq 2 in a single step with u2 = IO1* s-I and E2 = 16.2 kcal/mol. For curve b, decomposition was modeled by using eqs 6, 8, and 9 with u6 = I O l 3 s-l, E6 = 12.3 kcal/mol, us = 0.1 cm2/s, E8 = 4.0 kcal/mol, u9 = 5 X 1Olo s-I, and E9 = 6.0 kcal/mol. These parameters are not unique, although the range of acceptable parameters was not explored. Both calculations assume that all surface species occupy single and identical adsorption sites. The experimental data have been rescaled so that for T > 300 K the points scatter around zero.
and experimental data agreed near T,,. However, the calculated curves decrease more rapidly than I( 13). Calculations including and omitting possible steric effects of H(a) yield similar results, as expected for dinit= 0.09 ML. We note that for T 5 250 K the detailed shape of the I(13) curve in Figure 2a is nearly identical with that for the intensity of u(CH3) versus temperature and time in Figure 9 of ref 17. Therefore, the calculated curves also decrease more rapidly with temperature than the TREELS data. The similarity between the TREELS data in ref 17 and the SALI data presented here suggests that, for both sets of data over the temperature range in which CH30(a) decomposition occurs, the measured signal is approximately linearly dependent on methoxy coverage. This follows from the argument that it would be unlikely for both measured signals, which originate from different physical processes, to be nonlinearly dependent on methoxy coverage in similar manners. To describe the experimental data over the complete temperature range using coverage-independent kinetic parameters for eq 2, an extremely low value for u2 of approximately IOs s-’ is required. Although preexponential factors for unimolecular processes on surfaces vary widelygw4 and similarly small values for u2 have been determined on Ru(001),7 10s s-’ is an anomalously small value and difficult to explain in terms of transition-state theory. This result suggests that a simple unimolecular interpretation of eq 2 may not be adequate, and consequently, the kinetics for CH30(a) decomposition are complicated perhaps by either coverage-dependenteffects and/or additional reactions, for example, reactions involving H2CO(a). For dinit = 0.09 as well as 0.29 ML, the decomposition of CH30(a) is complete by -290 K. Therefore, any coverage-dependent effects appear small, although no model calculations were performed to quantify these effects. Islanding effects, such as those discussed for eq 1, may also be present. Here we briefly consider the effects of additional reactions during CH30(a) decomposition on the observed kinetics. Two reaction schemes that could affect the overall observed kinetics for CH30(a) decomposition can be written as CH30(a)
-
H2CO(a) + H(a)
(6)
(39) Baetzold, R. C.; Somorjai, G. A. J. Caral. 1976, 45, 94. (40) Zhou, X. L.; White, J. M. Chem. Phys. Lett. 1987, 142, 376. (41) Greenlief, C. M.; Radloff. P. L.; Zhou, X. L.; White, J. M. Surf. Sci. 1987, 191, 93. (42) Gurney, B. A.; Ho, W . J . Chem. Phys. 1987,87, 1376. (43) Whitman, L. J.; Ho, W. J . Chem. Phys. 1988, 89, 7621. (44) Seebauer, E. G.; Kong, A. C. F.; Schmidt, L. D. J. Chem. Phys. 1988, 88, 6597.
+ CO(a) + H(a)
(7)
-
(8)
CH30(a)
+ H(a)
CH30(a)
-,CO(a) + 2H(a)
(9)
where eq 8 is the reverse of eq 6. Because H2CO(a) is written as a distinct intermediate in eqs 6-9, although it has not been detected during methanol decomposition by using T R E E S , ’ ’ H2CO(a) must be short-lived and consequently be present in only minor quantities if present at all. The TREELS measurements indicate that the transient concentration of any intermediate species implicit in eq 2 must be less than 10% of the initial coverage.” A disproportion of H 2 C 0 on Ni( 1 10) (eq 7) has been observed previously following adsorption of H 2 C 0 at 96-300 K.45*46 The reaction given by eq 8 has not been observed directly. However, the temperatures at which which is considevidence for eq 8 was sought were 1200 K,45346 erably lower than 240-290 K, the temperature range in which CH30(a) decomposition, via eq 6, would produce H2CO(a) on the Ni( 110) surface. Thus, eq 8 may occur during TPR of methanol on Ni( 110). If neither reaction 7 or reaction 8 occurs to a significant extent, then insofar as the observed kinetics are concerned, eqs 6 and 9 may be combined to recover eq 2. Since there is insufficient data to unambiguously discriminate between or establish the validity of eqs 7 and 8 and other possible reactions, detailed calculations of CH30(a) decomposition for various initial coverages incorporating additional reactions were not performed. However, to establish whether additional reactions could both describe the overall observed kinetics for CH30(a) decomposition and yet maintain the concentration of any intermediate species, such as H2CO(a), below a reasonable detection limit (S10-3 ML), dCHIOduring TPR was modeled by using eqs 6, 8, and 9. The results, an example of which is shown as curve b in Figure 6, illustrate that if k8 = w8nsexp[-E8 RT] >> k6 and k9,where the surface site density n, = 1 X 10’ cm-2, then the detailed shape of the & H 3 0 curve during TPR can be described with preexponential factors from 1OIo to 10“ s-I while is ML. For k9 > k6, the rate-limiting reaction is eq 6, and although not simply unimolecular, both a Redhead and heating rate variation analysis of the calculated data yielded activation energies near E6. In addition, in agreement with isothermal measurements of CH30(a) decomposition on Ni( 1 lO)I9 and Ni(l1 1),26 calculations using eqs 6, 8, and 9 at constant temperatures between 260 and 280 K yield nearly exponential decays of OCH,O with time. Further work is necessary to establish whether a reaction mechanism composed of several elementary steps is consistent with both temperature-programmed and isothermal measurements. However, it seems plausible that additional reactions, which are intermediate steps in the overall reaction given by eq 2, can be occurring without being easily detected during decomposition of methanol on nickel and on other surfaces. Moreover, these additional reactions may explain some of the previously observed anomalous kinetic parameters for reactions comprised of several elementary steps.
/
5. Summary By using SALI, the reaction of methanol adsorbed on Ni( 1 10) was studied during TPR for initial coverages of 0.04,0.09,0.15, and 0.29 ML. Decomposition of adsorbed methanol occurred via C&H bond cleavage from 140 t o 240 K with kinetics that appeared similar for the range of coverages examined. Kinetic modeling of the loss of molecularly adsorbed methanol during TPR suggested that 0-H bond cleavage occurred with coverage-dependent kinetics although the data for initial coverages of 0.09 and 0.29 ML could not be described by a single set of kinetic parameters. Decomposition of adsorbed methoxy, formed during 0-H bond (45) Dickinson, J. T.; Madix, R. J. Int. J. Chem. Kiner. 1918, IO, 871. (46) Richter, L. J.; Ho. W . J . Chem. Phys. 1985.83, 2165.
9464
J . Phys. Chem. 1991, 95, 9464-9469
cleavage in molecularly adsorbed methanol, occurred from 240 to 290 K and formed exclusively adsorbed carbon monoxide and hydrogen. The kinetics of methoxy decomposition also appeared independent of the initial methanol coverage. Modeling of the methoxy decomposition reaction indicated that several elementary reactions may contribute to the observed overall kinetics although no intermediate species were detected. The model calculations
indicate that any intermediate species associated with methoxy decomposition could easily be present at transient concentrations below detection limits. Acknowledgment. This research was supported by the National Science Foundation, Division of Chemistry. Registry No. CH,OH, 67-56-1; Ni, 7440-02-0.
Ion-Pairing Effects on Viscosity/Conductance Relations in Raman-Characterized Polymer Electrolytes: LiCIOl and NaCF3S03in PPG(4000) M. G.McLint and C. A. Angell* Department of Chemistry, Arizona State University, Tempe, Arizona 85287- 1604, and Department of Chemistry, Purdue University, West Lafayette, Indiana 47907 (Received: February 19, 1991)
Precise measurements of viscosity and conductivity have been made of the solutions of sodium triflate and lithium perchlorate in poly(propy1ene oxide) of MW 4000 (PPG(4000))studied spectroscopically by Torell and co-workers for the extent of ion association. Viscosity, but not conductivity, data conform very well to the VTF equation and yield ideal glass transition temperatures that track the measured glass transition temperatures as closely as in the case of glass-forming aqueous solutions. The differences in temperature dependences of the two processes are demonstrated by using Walden rule plots. Using the data of Torell et al. at ambient temperature as a calibration point, we show that the deviations of conductivity from viscosity-dictated behavior predict almost quantitatively the strong temperature dependence of the free ion concentration found by Torell et al. up to the high temperature limit of their measurements.
Introduction Interest in polymer electrolytes is primarily due to their potential application as electrolytes for secondary batteries. These have now been under intensive study since the pioneering work of Wright' and Armand.* However, although much progress has been made in improving electrolyte performance, a detailed microscopic mechanism for ionic conduction in these materials remains elusive. The recognition that ionic conduction occurs in the amorphous phase of PEO electrolytes and that mobility was controlled primarily by the temperature interval above the glass transition temperature, TB,by Armand2J marked significant progress toward this goal. Armand showed that the conductivity could be well accounted for by a modification of the Arrhenius equation generally known as the VTF equation which is applicable to all transport processes, y (e.g., viscosity, diffusion, conductivity)
solutions because of the low solvent dielectric constant, to = 5 . With such a value of eo the electrical work of separating a pair of ions initially at contact to an infinite distance is a factor of -80/5 greater than in aqueous solutions. Hence, in molecular solutions of to = 5 it would be expected that dissociation comparable to that found in aqueous solutions would only be obtained at much higher thermal energies. However, in polymer solutions a further factor associated with the peculiarities of macromolecular solution thermodynamicsi0-" enters into consideration, and instead of increases in temperature causing the expected increases in dissociation, the reverse is observed. The ion-pairing phenomenology can be studied quantitatively by spectroscopic methods that can distinguish between free anions and anions distorted by interactions with near-neighbor cations. Torell and co-workers,i2 using Raman spectral analysis, find the percentage of free ions in NaCF3S03solutions in PPG(4000), at a concentration of 30 ether oxygens (U) per cation (designated hereafter 1:30) to decrease from 85% at 186 K to 39% at the highest temperature of their study, 360 K. In the case of the stronger electrolyte LiC104 the corresponding values are reported as 96% at 200 K and 78% at 340 K. In both cases the free and
where A, and D, are constants specific to the process and To is a constant characteristic of the liquid and common to all processes. The paramettr To is dominant and, as the present work will confirm, is closely related to the experimental glass transition temperature, T,, as in the case of glass-forming nonpolymeric ( I ) Wright, P. V. Br. Polym. J . 1976, 7, 319. (2) Armand, M. B.;Chabagno, J. M.; Duclot, M. J. In Fasr-lon Transport electrolyte solution^.^^ This fact alone establishes that the in Solids; Vashishta, P.,Mundy, J. N., Shenoy, G., Eds.; North-Holland: conduction mechanism is basically a liquidlike mechanism, as Amsterdam, 1979; p 13 1. indeed is now generally agreed. The basic difference between (3) Berthier, C.; Gorecki,W.; Minier, M.; Armand, M. B.;Chabagno, J. polymer and nonpolymer electrolytes, as many workers have M.; Riguad, P.Solid State Ionics 1983,11, 19. (4) Angell, C. A. J . Phys. Chem. 1966, 70,3988. pointed out, is that the friction on the ion motion is determined Angell, C. A.; Bressel, R. D. J. Phys. Chem. 1972, 76, 3244. (b) by the microscopic rather than the macroscopic v i s c o ~ i t y . ~ ~ ~Angell, ~( 5 ) (a) C. A.; Pollard, L. J.; Strauss. W. J. Solution Chem. 1972, I , 516. While the microscopic friction, which is involved in the glass (6) Papke, B. L.; Ratner, M. A.; Shriver, D. F. J . Phys. Chem. Solids transition, is the dominant control on this conductivity, it is 1981, 42, 493. (7) Ratner, M. A. In Polymer Electrolyte Reviews I; MacCallum. J. R., recognized that the conductivity is also strongly dependent on the Vincent, C. A., Eds.; Elsevier: New York, 1988; Chapter 7. proportion of the ions that are not bound up as ion pairs. Ion (8) Cameron, G. G.;Ingram, M. D.; Sorrie, G. A. J. Chem. Soc., Faraday pairing, which is a minor effect in solutions of alkali-metal salts Trans. 1 1987, 83, 3345. in water, is expected to be much more important in polymer (9) Torell, L. M.; Angell, C. A. Br. Polym. J . 1988, 20, 173. (10) Patterson, D. Macromolecules 1969, 2, 672.
'Research contained in this paper was performed in partial fulfillment of the Ph.D. requirements in Physical Chemistry at Purdue University. *To whom correspondence should be addressed at Arizona State University.
( I I ) Nitzan, A.; Ratner, M. A. Symp. Mater. Res. SOC.1991, 210, 203. (12) (a) Kakihana, M.; Schantz, S.;Torell, L. M. J . Chem. Phys. 1990,
92, 6271. (b) Kakihana, M.; Schantz, S.;Mellander, B.-E.; Torell, L. M.In Proceedings of the 2nd International Symposium on Polymer Electrolytes; p 23.
0022-365419112095-9464302.50/0 0 1991 American Chemical Society