Surface-enhanced Raman spectroscopy as a probe of adsorption at

Surface-enhanced Raman spectroscopy as a probe of adsorption at transition metal-high-pressure gas interfaces: nitric oxide, carbon monoxide, and oxyg...
0 downloads 0 Views 2MB Size
Langmuir 1991, 7, 714-721

714

Surface-Enhanced Raman Spectroscopy as a Probe of Adsorption at Transition Metal-High-pressure Gas Interfaces: NO, CO, and Oxygen on Platinum-, Rhodium-, and Ruthenium-Coated Gold Todd Wilke,l Xiaoping Gao,2 Christos G. Takoudis,l and Michael J. W e a v e r * 1 2 School of Chemical Engineering and Department of Chemistry, Purdue University, West Lafayette, Indiana 47907 Received July 18, 1990 Surface-enhanced Raman (SER) spectra are reported for the adsorption of NO, CO, and oxygen in flowing streams at ambient pressures on gold surfaces coated with ultrathin (ca. two to three monolayers) films of platinum, rhodium, and ruthenium. The gold surface provides an inert substrate engendering stable and intense SERS for adsorbates bound to the transition-metal overlayer. A buildup of surface carbonaceous impurities was detected on these surfaces in a flowing argon stream from the appearance of broad Raman bands in the ca 1000-1700-cm-~region. These features, however, were essentially removed in atmospheres containing oxygen or NO. Dosing with oxygen yielded broad bands in the 500-600-cm-1 region due to metal-oxygen stretches while NO exhibited an additional feature at 325 cm-l on Pt and at 310 cm-l on Rh and Ru. This band is ascribed tentatively to a surface-atomic nitrogen stretch from dissociative NO chemisorption. Spectral evidence for molecular NO adsorption was only obtained on ruthenium in the form of a ca 1900-cm-l feature ascribed to a N-0 stretch for NO adsorbed on oxidized sites. Dosing with On/CO mixtures on Pt and Rh yielded CO adsorption as discerned from the characteristic C-0 and surface-C vibrations, although surface oxidation on Ru was sufficiently extensive to prevent the observation of adsorbed CO on this metal surface. Spectra obtained from CO/NO mixtures are largely similar to those from NO adsorption alone, indicating that the latter species exhibits stronger surface binding. The temporal replacement of adsorbed CO by NO, as well as temperature-induced changes in surface composition, could be followed on a seconds time scale by means of a spectrograph-charge-coupled device detector arrangement. These results illustrate the real-time capabilities of the present SERS technique. The findings for gas-phase adsorption are also compared with SER spectra for the same adsorbates at the corresponding metal-aqueous electrochemical interfaces.

Introduction Surface-enhanced Raman scattering (SERS) is now a well-established technique for obtaining vibrational spectra a t metal surfaces, especially in electrochemical and other condensed-phase environments where interference from bulk solution species can constitute a serious problem.3 The choice of SERS-active substrates is restricted in practice primarily to silver, gold, and copper. However, we have demonstrated recently that the technique can be extended to transition-metal and other surfaces by electrodepositing them as thin films on gold ~ u b s t r a t e s . ~Silver J substrates have also been utilized for this purpose.6 Using multichannel detection, the SERS technique can provide a sensitive means of following heterogeneous processes in real time.7 Given these attributes, SERS would appear to provide a viable a n d perhaps valuable means of examining (1)School of Chemical Engineering.

(2) Department of Chemistry. ( 3 ) For recent overviews see: (a) Weitz, D. A.; Moskovits, M.; Creighton, J. A. In Chemistry and Structure at Interfaces-New Laser and Optical Techniques; Hall, R. B., Ellis, A. B., Eds.; VCH Publishers: Deerfield Beach, FL, 1986; Chapter 5. (b) Hester, R. E. In Comprehensiue Chemical Kinetics; Compton, R. G., Hamnett, A., Eds.; Elsevier: Amsterdam, 1989; Vol. 29, Chapter 2. (4) (a) Leung, L.-W. H.; Weaver, M. J. J . Am. Chem. SOC.1987, 109, 5113. (b) Leung, L.-W. H.; Weaver, M. J. Langmuir 1988, 4, 1076. (5) (a) Desilvestro, J.; Corrigan, D. A.; Weaver, M. J. J . Phys. Chem. 1986,90,6408. (b) Desilvestro, J.;Corrigan, D. A.; Weaver, M. J. J . Electrochem. Soc. Interfacial Electrochem. 1988,135, 885. (c) Gosztola, D.; Weaver, M. J. J . Electroanal. Chem. 1989, 271, 141. (d) Gosztola, D.; Weaver, M. J. Langmuir 1989,5, 776. (6) (a) Feilchenfeld, H.; Gao, X.; Weaver, M. J. Chem. Phys. Lett. 1989,161,321. (b)Gao,X.;Weaver,M.J. Tobesubmittedforpublication. (7) For example: Gao, P.; Gosztola, D.; Weaver, M. J. Anal. Chim. Acta 1988, 212, 201.

0743-7463/91/2407-0714$02.50/0

adsorption a t such metal surfaces in gas-phase environments, especially a t high pressures of relevance to heterogeneous catalysis where other vibrational techniques are of limited utility. Surprisingly,however, such reports have only appeared sporadically in the literature. Besides the substantial body of work under high vacuum conditions a t lower temperatures, especially by Moskovits and coworkers,8applications of SERS to metal-gas interfaces a t higher temperatures and/or pressures have been limited chiefly to silver substrates.g-l3 Other than the substrate requirement noted above, one factor that has probably limited such applications is the temporal stability of the SERS activity, especially at elevated temperatures. However, we have found that gold surfaces prepared by electrochemical roughening14 upon which thin (three to four monolayer) films of platinum, rhodium, or ruthenium are subsequently electrodepos(8) For example: (a) Moskovits, M.; DiLella, D. P.; Maynard, K. J. Langmuir 1988,4,67. (b) McBreen, P. H.; Moskovits, M. J.Phys. Chem. 1983, 87, 4843. (c) DiLella, D. P.; Moskovits, M. J . Phys. Chem. 1981, 85, 2042. (9) (a) Dorian, P. B.; Von Raben, K. U.; Chang, R. K. Chem. Phys. Lett. 1981,84, 405. (b) Von Raben, K. U.; Dorian, P. B.; Chen, T. T.; Chang, R. K. Chem. Phys. Lett. 1983, 95, 269. (10) (a) Matsuta, H.; Hirokawa, K. Surf. Sci. 1986, 172, L555. (b) Matsuta, H.; Hirokawa, K. Appl. Surf. Sci. 1987, 27, 482. (c) Matsuta, H.; Hirokawa, K. Appl. Surf. Sci. 1988/89,35, 10. (d) Matsuta, H.; Hirokawa, K. Appl. Spectrosc. 1989, 43, 239. (11)McBreen, P. H.; Moskovits, M. J . Catal. 1987, 103, 188. (12) Nowobilski, P. J.; Wilke, T.; Davies, J. P.; Patterson, M.; Weaver, M. J.; Takoudis, C. G. In Catalysis: Theory to Practice; Phillips, M. J., Ternan, M., Eds.; Chemical Institute of Canada: Ottawa, 1988; Vol. 3, D 1159. (13) Boghosian, S.; Bebelis, S.; Vayenas, C. G.; Papatheodorou, G. N. J . Catal. 1989, 117, 561. (14) Gao, P.; Gosztola, D.; Leung, L.-W. H.; Weaver, M. J. J . Electroanal. Chem. Interfacial Electrochem. 1987, 233, 211.

0 1991 American Chemical Society

Langmuir, Vol. 7, No. 4, 1991 715

Adsorption at Transition Metal-Gas Interfaces

VACUUM

PUMP

-

REACTOR

MANIFOLD

COLLIMATING

LENS

- MIRROR

s4 -

MIRROR LKYS

Figure 1. Schematic illustration of the gas flow system, sample

chamber, and collection optics.

ited4 exhibit excellent stability under such conditions in ambient pressure gas-phase environments. This property makes these materials excellent candidates with which to explore the utility of SERS for probing adsorption a t such catalytically significant interfaces. Described herein are results from a study of the adsorption of NO, CO, and oxygen on these platinum-, rhodium-, and ruthenium-coated gold surfaces a t ambient pressures in the temperature range 25-200 "C. The use of a silver substrate is also explored here, but gold is shown to exhibit markedly superior properties for the present purposes. Besides the practical importance of these systems in catalytic devices for automobile and other emissions, the adsorbates examined here are of widespread fundamental significance. Although most Raman spectra were acquired with a conventional scanning monochromator, some results are included that were obtained with a recently procured charge-coupled device (CCD) multichannel system. The latter exhibits sufficient sensitivity to enable spectral sequences to readily be obtained on a ca. 1-s time scale, allowing the time evolution of the surface speciation to be monitored following gas dosage. Some related results are also presented here for the same adsorbate systems in related aqueous electrochemical environments.

Experimental Section A schematic diagram of the gas-phase flow apparatus used in this study is shown in Figure 1. The sample chamber was a

stainless-steel six-way cross that was operated as a continuous flow reactor at atmosphericpressure. Samples were spot-welded to a tantalum foil holder mounted on an electrical feedthrough and held parallel to an optical glass viewport mounted on the bottom flange of the six-way cross. The tantalum foil holder could be heated resistively with a dc power supply; chromelalumel thermocouples were spot-welded to the sample edge to monitor the temperature. The gas manifold allowed the mixing of up to four gas streams. The inlet to the reactor directed the gas flow across the sample surface so the gas composition at the surface was that of the feed. Both the sample chamber and the manifold could be evacuated to below Torr by a mechanical rough pump. The Raman excitation source was a Spectra-Physics Model 165Kr+laser operated at 647.1 nm and a power of 20 mW on the sample. The Raman-scattered light was collected with a 50 mm diameter f/O.95camera lens (DO Industries, Model DO-5095) and focused into either a SPEX Model 1403monochromator or a SPEX Model 1877Triplemate spectrometer. The latter forms a multichannel spectrograph device,utilizinga Photometri'cs PM

512 CCD detector cooled to -1 10"C, with a Photometrics CC200 camera controller interfaced to a Zenith 386 computer for data acquisition and storage. All spectra were recorded with a bandpass of 5 cm-'. The samples used in this study were 6 mm diameter gold or silver disks which were cut from a 0.1 mm thick foil (99.99%, Johnson Matthey). The surfaces were mechanically polished with 1.0- and 0.3-pm alumina and then placed in a Teflon holder that exposed a region 3 mm diameter at the center of the surface. The gold surface was electrochemically roughened by oxidationreduction cycles in 0.1 M KCl to yield SERS activity as described in ref 14. The silver surfaces were roughened in 0.1 M KCl by performing five successive potential step cycles from -0.6 to 0.15 V and return, holding the potential each time until 30 mC cm-2 anodic charge had passed. Surface Raman measurements in the electrochemical environment were made with another electrode subjected to the same preparation, but consisting of a 4 mm diameter gold or silver disk sheathed in Teflon. Essentially identical electrochemical behavior was obtained with these two types of surface fabrication. Details of the electrochemical SERS procedures are essentially as described earlier! The electrodepoM solutions sition of the metal overlayers utilized ca. 5 X of ruthenium(II1) chloride trihydrate (RuCl3*3H20),rhodium(111)chloride trihydrate (RhCls*3H20),and hydrogen hexachloroplatinate(1V) (H2PtChexH20)(Aldrich). The electrolytes used were 0.5 M H2S04 for Pt and 0.5 M HC104for Rh and Ru. The deposition procedures used were essentially as outlined in ref 4. Optimal SERS intensity on the overlayers was usually obtained with about two equivalent monolayers of Pt and three equivalent monolayers of Rh and Ru. Nitric oxide (99%) and oxygen (99.998%)(Matheson) and 15N0(Icon)were used without further purification. Hydrogen (99.99%)and argon (99.99891)(Matheson) were passed through molecular sieve water traps. Carbon monoxide (99.871)was obtained from Matheson in an aluminum cylinder that was filled at the source to eliminate the presence of iron carbonyls. All electrode potentials are quoted versus the saturated calomel electrode (SCE) and measurements were made at room temperature unless noted otherwise.

Results After the sample was mounted in the reactor, the chamber was evacuated to below 1mTorr and then filled with the desired atmosphere. Argon was used as the carrier gas. The gas flow rates were usually about 100 mL min-l. Exposure of the surface to ostensibly pure argon yielded a gradual continuous rise in the continuum Raman "background" together with an increasing emergence of broad features in the so-called "cathedral peak" region, ca 1000-1700cm-l. Comparable behavior was obtained for all the metal surfaces studied here. These features are very probably due to the build-up of carbonaceous material formed by photolysis of organic impurities. Persuasive evidence for a photochemical mechanism includes the observed dependence of the rate of background rise on the light intensity and wavelength. The use of red (647.1 nm) rather than green Raman excitation was advantageous from this standpoint. The former excitation wavelength was therefore utilized for SERS on silver as well as gold surfaces here. (Note that wavelengths above ca. 600 nm are required for the observation of SERS on gold.) With the exception of measurements performed on sufficiently short time scales ( 5 2 min) so to avoid these difficulties, it was found to be necessary to include in the carrier gas a significant proportion (25%)) of a suitable oxidant, oxygen or nitric oxide, so to remove the carbonaceous material and maintain a suitably "clean" background Raman spectrum. This procedure was therefore typically followed in the present study. Gold and Silver Substrates. In order to evaluate the suitability of gold and silver as substrates for the transitionmetal overlayers in the present work, it is necessary initially

Wilke et al.

716 Langmuir, Vol. 7, No. 4 , 1991

I*

to examine the SER spectra of these unmodified surfaces T 500 325 in the presence of the adsorbates of interest here. Exposing gold to oxygen or 02/CO mixtures yielded no significant Raman features other tlian a band a t 260 cm-' due to residual chlorine from the prior electrochemical roughening.15 Similarly featureless spectra were obtained for nitric oxide or NO/CO mixtures on gold, except for an additional broad weak band a t 2200 cm-I (vide infra). Rather different results were obtained for silver. While no significant Raman bands were observed for dosing with C 470 I 0 2 or with C 0 / 0 2 mixtures, several spectral features were obtained in the presence of NO. Similarly to earlier reports of SERS for NO on silver p ~ w d e r s , ~ bands ~ J ~ ~were ,~ observed a t 825 and ca. 1300 cm-l. These features have been assigned to the NO2 bending mode, 6(N02), and the NO2 symmetric stretch.gb The NO2 species is presumably formed by reaction of NO with the adsorbed oxygen which should usually be present on silver under these conditions. We also observed a third, weaker, band from NO on silver a t 1045cm-'; this feature became much more intense upon dosing with NO/Oz mixtures, so that gas-phase NO2 was 500 300 present. This band has also been reported p r e v i o u ~ l y ~ ~ J ~ ~ ~ ~ A wavenumber and assigned to the symmetric NO3 stretch. On the basis of these findings, then, gold represents a more suitable substrate for the transition-metal overlayers in the presence of NO and related adsorbates. Platinum. Dosage with oxygen/argon mixtures on platinum-coated gold yielded largely featureless spectra. Figures 2 shows typical spectra in the 200-600 and 18002300 cm-' frequency regions. Besides a feature centered a t 250 cm-I, due to the metal-chlorine stretch from residual chloride,15only a weak band is observed in the 450-600 cm-l region where a Pt-0 stretch, upto, might be expected. Dosing with an equivolume mixture of oxygen and carbon monoxide, however, yields several features characteristic of the presence of adsorbed CO on platinum (Figure 2 ) . Thus the high-frequency region contains a band a t 2060 cm-I and a weak feature at ca. 1900 cm-'. These C-0 stretches ( U C O ) are diagnostic of the presence of CO bound to atop ("terminal") and 2-fold bridging sites, respectively (cf. ref 4a). The low-frequency pair of bands a t 475 and 390 cm-' (Fig. 2A) are ascribed to metal-C0 vibrations, u p t c , associated with terminal and 2-fold bridging CO by comparison with related electron energy 2200 2000 1 a00 loss spectroscopy (EELS) data.16 The intensities of these bands were unaffected by variations in the CO/O2 pressure A wavenumber ratio from 0.3 to 3. This indicates that CO is adsorbed F i g u r e 2. SER spectra for a platinum-coated gold surface preferentially on Pt a t room temperatures even in 02-rich exposed to various flowing argon mixtures: 20% 02, 10% 0 2 + atmospheres. As noted above, corresponding experiments 10% CO, 20% NO, and 10% NO + 10% CO. The Pt overlayer performed on unmodified gold surfaces yielded no discoverage was two equivalent monolayers, the pressure was 1atm, and the total gas flow rate was 100 cm3 min-1. The spectrometer cernible SERS features, confirming that the spectra in was a SPEX Model 1403 scanning monochromator. Laser Figure 2 arise entirely from adsorption on the platinum excitation was 20 mV at 647.1 nm; the scan rate was 0.5 cm-* SI. overlayer. This conclusion is similar to that for the coralthough a weak uco responding electrochemical was obtained with a 30-s acquisition time. While no change band a t ca. 2100-2120 cm-l due to CO adsorbed on gold in the uptc (and also the U C O ) bands can be discerned up is usually obtained in the latter case. to ca 100 "C, the band intensities decrease sharply toward Time-resolved Raman experiments were also performed higher temperatures, especially those associated with with CO/Op mixtures by utilizing the CCD Raman system. terminal CO. The bands disappear by 200 "C, but return The addition of CO into the flowing Oz/Ar gas stream fully upon returning the surface temperature to 25 "C. rapidly yields uco and uptc bands (within 1min), indicating This reversible behavior confirms the ability of the SERS the facile nature of the initial chemisorption kinetics, activity to survive such elevated temperatures. Most although up to ca. 1 h is required for the band intensities likely, then, the temperature-induced decreases in the uptC to reach their maximum level. Figure 3 shows spectra in and uco band intensities reflect diminutions in the steadythe low-frequency region obtained during a ramp in temstate CO coverage, BCO, brought about by the acceleration perature from 25 to 200 "C a t 0.4 deg s-l. Each spectrum in the CO oxidation kinetics expected under these conditions. (15)Gao, P.; Weaver, M. J. J. Phys. Chem. 1986, 90, 4057. Dosing the platinum surface with NO/argon mixtures (16)For example, see: Steininger, H.; Lehwald, S.; Ibach, H. Surf.Sci. resulted in spectra as also indicated in Figure 2. Two strong 1982, 123, 264.

fi2060

Langmuir, Vol. 7, No. 4, 1991 717

Adsorption at Transition Metal-Gas Interfaces

550

350

450

A wavenumber

Figure 3. Temperature-dependent SER spectra in the lowfrequency (300-600 cm-l) region for a platinum-coated gold surface in 100 cm3 min-l of 10% 02 10% CO in argon. The data were obtained by use of a multichannel CCD detector. The temperature was ramped from 25 to 200 "C at 0.4 deg s-l and spectra were recorded every 60 s. Laser excitation was 20 mV at 647.1 nm; the integration time was 30 s.

+

bands are consistently observed, a t 325 cm-' and a t about 2200 cm-l. A band similar to, although much weaker than, the latter is also obtained on unmodified gold under these conditions. Careful examination of the 1500-1900-~m-~ region, within which N-0 stretching vibrations ( U N O ) for molecularly adsorbed NO should be located," failed to detect any clearcut U N O bands. The assignments of the 325 and 2200 cm-l features are not immediately self-evident. Both of these bands were downshifted significantly (10 and 60 cm-', respectively) when '5NO was substituted for 14N0, confirming the involvement of nitrogen atom stretching in both bands. Dosing with N2O yielded no discernible bands in the 2200-cm-' region, eliminating this species as being responsible for the high-frequency feature. Spectra obtained for mixtures of CO and NO, rather than CO and 0 2 as considered above, proved to be interesting. Typical results are included in Figure 2. In the low-frequency region, the 325-cm-' NO feature is retained along with the terminal uptc band at ca. 470 cm-', although the bridging uptC feature is now absent (Figure 2A). Consistent with this, the uco region shows only a terminal uco band, now narrower and upshifted to 2090 cm-1 (Figure 2B). In addition, the broad 2200-cm-1 band seen with NO dosage is still present but is now partnered with an additional strong feature a t 2160 cm-l (Figure 2B). The latter feature is assigned tentatively to a C-N stretching vibration ( U C N ) of adsorbed cyanide formed by the reaction of CO (or carbon-containing impurities) and adsorbed nitrogen formed by dissociative NO chemisorption (vide infra). The presence of nitrogen in the adsorbate was confirmed by the observation of the anticipated ca 60-cm-' frequency downshift when 15N0 was substituted for 14N0. An alternative assignment of the 2160-cm-' band is the asymmetric stretch of isocyanate, formed from CO (17) Gland, J. L.; Sexton, B. A. Surf. Sci. 1980, 94, 355.

and adsorbed nitrogen. Such an assignment has been made for infrared bands seen a t similar frequencies during the reduction of NO by CO on supported Pt, Rh, and Ru surfaces.18 The formation of CN- in similar metal-gas systems has also been observed, but appears to commonly require higher temperatures.lg Nevertheless, supporting the UCN assignment is the appearance of an essentially identical feature at 2160 cm-l resulting from dosage with HCN.20 The 2200-cm-l band could also result from NCO or CN formed by reaction between adsorbed nitrogen and carbonaceous impurities. Corresponding measurements were also attempted for platinum overlayers on silver. Even for ultrathin (one to two equivalent monolayer) films, however, the SERS intensities were extremely weak, so that useful spectra could not be obtained. Rhodium. A typical SER spectrum obtained in the 200-900-cm-' region for oxygen dosed onto a rhodium overlayer on gold is shown in Figure 4A. A broad yet clearly discernible band is seen a t ca. 500 cm-l, attributed to the Rh-0 stretch (VRhO) for adsorbed atomic oxygen. Bands for adsorbed oxygen at comparable frequencies have been seen a t monocrystalline rhodium ultrahigh vacuum (UHV) surfaces.21 A band centered a t 520-530 cm-I is also observed for the present Rh/Au surface in acidic aqueous solution a t potentials where electrochemical oxide formation is known to The inclusion of CO in the dosing mixture with 0 2 leads to a diminution of the URhO band intensity and the appearance of two new bands a t 465 and 2030-2040 cm-l (Figure 4), attributable to the Rh-CO and C-0 stretch of terminal adsorbed CO (cf. EELS behavior at Rh UHV interfaces21bc).No clear evidence for a bridging uco band, anticipated at ca 1950-2000 ~ m - l , ~ b wobtained. as Similar spectra were obtained for Oz/CO pressure ratios up to ca. 10. The temporal replacement of adsorbed oxygen by CO was also examined. Typical time-resolved spectra in the low- and high-frequency regions are shown in Figure 5. The bottom spectrum was observed for an O2/Ar gas flow; the upward-going spectral sequence was obtained from 30 to 240 s, as indicated, after initiating the flow of equal CO and 02 partial pressures. Figure 5 shows that there is substantial replacement of adsorbed oxygen by CO even at short times, ca. 1min. Time-resolved spectra for CO/ 0 2 mixtures were also obtained on rhodium during linear temperature increases (cf. Figure 3). A t heating rates of ca. 1deg s-l, a marked diminution in the URhC and uco band intensities occurs a t temperatures above 100 "C. Dosing the rhodium surface with NO yielded an intense band at 310 cm-'and a weaker feature at 2200 cm-' (Figure 4); the latter is weaker than the 2200-cm-' band observed on Pt/Au, and similar to that on unmodified gold. An additional band was also obtained at 840 cm-'. A residual broad band a t 500 cm-', due to atomic oxygen, was usually discernible. Prior to NO dosage, the surface was cleaned of atomic oxygen by heating in hydrogen for 1 min at 100 "C. Similar spectra were also obtained following NO dosage onto surfaces dosed previously with CO/O2 mixtures, the CO features being replaced entirely by the NO ~~

(18)For example, see: Hecker, W. C.; Bell, A. T. J . Catal. 1984,85, 389, and previous references cited therein. (19) (a) Lorimer, D.; Bell, A. T. J . Catal. 1979,59,223. (b) De Louise, L. A.; Winograd, N. Surf. Sci. 1985, 154, 79. (20) The surface was dosed with HCN by evacuating the chamber and then admitting 20 Torr of HCN from a bulb filled by reacting NaCN with H~SOI.The chamber was then brought up to an atmosphere with argon and a spectrum obtained in a flowing mixture of oxygen and argon. (21) (a) Dubois, L. H. J . Chem. Phys. 1982, 77, 5228. (b) Gurney, B. A.;Richter,L. J.; Villarrubia, J. S.;Ho, W.J. Chem. Phys. 1987,87,6710. (c) Dubois, L. H.; Somorjai, G. A. Surf. Sci. 1980, 91, 514.

Wilke et al.

718 Langmuir, Vol. 7, No. 4 , 1991

;

A

2 40 s .

CO+NO

NO

400

600

400

800

A wavenumber

Ti00 cos

I

2040

I

I

I

2200

2000

1800

A wavenumber

Figure 4. SER spectra for a rhodium-coated gold surface exposed to various flowing argon mixtures: 20% 02,10% 02 + 10% CO, 20% NO, and 10% NO + 10% CO. The Rh overlayer coverage was three equivalent monolayers; other details are as in caption to Figure 2.

bands noted above. Not surprisingly, then, dosage of CO/ NO mixtures onto rhodium yielded spectra essentially identical with those obtained with NO alone (Figure 4). Time-resolved spectra were also obtained following NO dosage onto rhodium with preadsorbed CO in order to examine the temporal relationship between the appearance of the 310-cm-' and the URhO bands. Figure 6 shows such a spectral set. The bottom spectrum, obtained in argon, shows the characteristic 465-cm-' YRhC band for adsorbed CO and no evidence of adsorbed oxygen. The upwardgoing spectral sequence was obtained a t the times indicated after initiating the NO dosage. The URhC band is removed rapidly, and the ensuing VRhO and 310-cm-' bands are seen to grow in with time in a roughly similar fashion. This last observation sheds some light on the likely origin of the 310-cm-I feature. The presence of the Rh-0 stretching band indicates that NO dissociative chemisorption is occurring. The commensurate growth of the

2100

1900

A wavenumber

Figure 5. Time-dependent SER spectra for an oxidized rhcdiumcoated gold surface exposed to CO. The surface was initially in a flowing O2 argon, 10 and 80 cm3 min-l, respectively, mixture and then 10cm3min-l of CO was added to the feed and a spectrum recorded at the times indicated. Laser excitation was 20 mW at 647.1 nm; the integration time was 10 s.

+

310-cm-' feature suggests that it arises from another dissociation product; the simplest possibility is adsorbed atomic nitrogen. While this assignment is speculative, supporting evidence is given below. The effect of dosing NO on rhodium that was predosed with oxygen was also investigated. The URhO band gradually diminished with time as the 310-cm-' band grew in, the process taking ca. 1 h. This indicates that the NO dissociative chemisorption products can slowly displace previously chemisorbed oxygen. The substitution process was found to be entirely reversible, the original VRhO band intensity returning and the 310 cm-' being removed upon redosing with oxygen. Sequential oxygen and NO dosing yielded a more intense 840-cm-' feature than observed in the absence of oxygen preadsorption. This evidence,

Adsorption a t Transition Metal-Gas Interfaces

500

Langmuir, Vol. 7, No. 4, 1991 719

300

700

A wovenumber

500

300

A wovenumber

Figure 6. Time-dependentSER spectra for a reduced rhodiumcoated gold surface exposed to NO. The surface was first reduced with CO at room temperature, evacuated, and then held in 80 cm3min-l of argon where the initial, t = 0, spectrum was obtained. NO (20 cm3 mi+) was then added to the feed and a spectrum recorded at the times indicated. Laser excitation was 20 mW at 647.1 nm; the integration time was 10 s.

together with the proximity of the band frequency to the bending mode for nitro complexes, ~ ( N O Z(cf. ) ~SERS ~ on silver, vide supra), leads us to assign tentatively the 840-cm-I feature to 6(N02) for adsorbed nitrogen dioxide. The corresponding symmetric NO2 stretch, anticipated a t ca. 1300cm-l (vide supra),could not be clearly discerned, probably due to interference from features arising from residual carbonaceous impurities. Corresponding experiments performed with rhodium overlayers on silver rather than gold yielded markedly different results. The 825-and 1300-cm-' spectral features seen upon NO dosing on silver are largely removed in the presence of k 1-2 equivalent monolayers of rhodium. In the presence of 02,Oz/CO, and NO/CO mixtures, as well as NO, only a single clearcut band a t ca 510 cm-l could be discerned, attributed to the Rh-0 stretch. Heating the Rh/Ag surface to temperatures above 100 "C in a reducing atmosphere of CO or H2 attenuates irreversibly the 510-cm-I feature and yields a strong band a t 240 cm-', attributed to the Ag-Cl stretch from residual chlorine. Evidently, the rhodium overlayer is sintering under these conditions. Ruthenium. Typical SER spectra obtained upon dosing 02, 02/CO, NO, and NO/CO onto a ruthenium overlayer on gold, obtained similarly to the above data on platinum and rhodium overlayers, are given in Figure 7. Dosing with 0 2 yields a pair of broad features a t ca 500 and 600 cm-' in addition to the Ru-Cl stretch a t 280 cm-' (Figure 7A). The former two bands are attributed to adsorbed atomic oxygen and to more tightly bound ruthenium oxides, respectively. Vibrational bands a t comparable frequencies for oxygen on ruthenium have been observed with EELS.Z3 The present UR"O bands survive unchanged upon heating to 150 "C and subsequent cooling. (22) Nakamoto, K. Infrared and Raman Spectra of Inorganic and Coordination Complexes; Wiley: New York, 1086; p 223. (23) Thomas, G. E., Weinberg, W. H.J. Chem. Phys. 1979, 70,954.

~

COtO,

I

I

2200

2000

1800

A wovenumber

Figure 7. SER spectra for a ruthenium-coated gold surface exposed to various flowing argon mixtures: 20% 02, 10% 02 + 10% CO, 20% NO, and 10% NO + 10% CO. The Ru overlayer coverage was three equivalent monolayers; other details are as in caption to Figure 2.

Exposure of the ruthenium surface to CO/Oz mixtures yields little change in these spectral features, although the 500-cm-' band is sharpened significantly (Figure 7A). In particular, there are no clearly discernible bands attributable to adsorbed CO. Apparently, then, oxygen adsorption is sufficiently extensive to largely preclude competitive CO binding under these conditions. The interaction of NO with the ruthenium surface yields several significant Raman features. In the low-frequency region (Figure 7A), a band at 310 cm-l is obtained, similar to (but more intense than) corresponding bands with NO on rhodium and platinum. A band a t 460 cm-l is also observed, along with a feature a t 600 cm-' which is sharper than that obtained with 02 (Figure 7A). Besides a weak band a t 2200 cm-', similar to that obtained on rhodium,

Wilke et al.

720 Langmuir, Vol. 7, No. 4 , 1991 a more pronounced feature is observed on ruthenium a t 1900 cm-l (Figure 7B). The 1900-cm-l band is tentatively assigned to the N-0 stretch, U N O , of molecularly adsorbed NO and the 460-cm-l band to the corresponding Ru-NO stretch. Surface NO bands a t comparable frequencies have been observed with EELSz4 This latter assignment is supported further by the observed close correlation between the appearance and disappearance of the 1900- and 460-cm-' features, which occurs rapidly upon the initiation and termination, respectively, of NO dosing. The rapid reversible nature of the adsorption, combined with the absence of these features on platinum and rhodium, suggests the presence of NO weakly bound to oxidized ruthenium sites. A comparable feature has been observed a t oxidized ruthenium on silica by infrared s p e c t r o ~ c o p y . ~ ~ Mixtures of NO and CO yielded an additional weak band at 2005 cm-l (Figure 7B), assigned to uco of terminal adsorbed CO. The detection of a corresponding UR"C band is probably thwarted by the presence of stronger overlapping low-frequency features (Figure 7A). Similarly to rhodium, ruthenium overlayers on silver in the presence of 0 2 , NO, Oz/CO, and NO/CO mixtures yielded only Ru-0 bands, similar to those obtained on the Ru/Au surface (Figure 7A). Related Electrochemical Systems. Given the detailed, and in some cases surprising, SERS features obtained for the transition-metal overlayers in the gas phase, it is of interest to compare these findings with spectra obtained for the same adsorbates a t corresponding aqueous electrochemical interfaces. As alluded to above, SER spectra for CO electrosorbed a t these overlayer surfaces have been outlined in detail previously and are compared further with the present results below. We now summarize briefly relevant SERS findings for the electrosorption of NO along with CO on platinum, rhodium, and ruthenium overlayers in aqueous media purged previously with nitrogen. Unlike unmodified gold in the gas phase, evidence for molecular NO adsorption a t this surface in NO-sparged aqueous 0.1 M HC104 was obtained from a clearcut SERS band a t 1600 cm-', attributed to U N O . This feature is attenuated a t potentials positive of 0.6 V vs SCE or negative of 0.3 V. The latter effect is presumably due to NO electroreduction.26 Brief subsequent sparging with CO, so to yield a NO/CO mixture in solution, gave rise to a pronounced band a t about 2190 cm-l and a weaker additional feature a t 2120-2150 cm-l a t more negative potentials. Similar results were also obtained in 0.1 M NaC104 electrolyte (pH 7). The 2190-cm-' band intensity increased with irradiation time of the laser on a given surface region, indicating that it is associated with the product of a photochemical reaction. These spectral features are consistent with the presence of adsorbed cyanide, formed by reaction between adsorbed NO and carbonaceous impurities. Support for this assignment was obtained from the closely similar potentialdependent spectra obtained on gold in 0.1 M NaC104 in the presence of cyanide. Similarly to the gas-phase systems, exposure of the platinum, rhodium, and ruthenium overlayers to NOsparged 0.1 M HC104 yielded bands a t 325 cm-' (Pt) or ~

(24) Lambert, R. M.; Bridge, M. E. In T h e Chemical Physics o f Solid Surfaces and Heterogeneous Catalysis: King, D. A,, Woodruff, D. P., Eds.; Elsevier: New York, 1984; Vol. 3, p 102. (25) Davydov, A. A,; Bell, A. T. J . Catal. 1977,49, 332. (26) (a) Plieth, W. J. In Encyclopedia of Electrochemistry of the Elements; Bard, A. J.,Ed.; Marcel Dekker: New York, 1978;Vol. 8, p 321. (b) Colucci, J. A.; Foral, M. J.; Langer, S. H. Electrochim. Acta 1985,30, 1675.

310 cm-l (Rh, Ru),the latter being especially intense. These features were observed over the potential range 0 to 0.6 V. No marked higher-frequency features were observed on platinum. On rhodium, however, a band a t 830 cm-l was observed. The close similarity in frequency with the corresponding gas-phase feature on rhodium identifies this band as the bending mode for adsorbed N02. This assignment is supported by the additional observation of a band a t 1600 cm-l, consistent with the N-0 symmetric stretch (vide supra). A similar 830-cm-1 band was also observed on ruthenium in the electrochemical enivronment. In addition, a weak band a t 1890 cm-I was observed on the ruthenium electrode a t potentials positive of 0 V. As for the corresponding gas-phase feature (Figure 7B), this band is attributed to U N O arising from NO on oxidized Ru sites. Comparison of the surface-oxygen bands on rhodium and ruthenium in the gas phase and electrochemical environments is of interest. Regarding the latter systems, a single broad band is seen a t ca. 530 cm-I on rhodium and a t 490 cm-' on ruthenium electrodes. Upon the addition of NO, these bands become markedly more intense and survive to more negative potentials, a t least to 0 V and -0.3 V on rhodium and ruthenium, respectively. On ruthenium, an additional UR"O feature a t ca. 600 cm-l is obtained in the presence of NO (cf. Ru-gas phase data, Figure 7A). This effect of NO on the potential-dependent oxide formation is consistent with a NO-induced chemical oxidation of the rhodium and ruthenium electrodes, similarly to that discussed previously for gold in the presence of peroxodisulfate anions.27 Mixtures of NO and CO on the overlayer electrodes yielded intense features between 2100 and 2200 cm-l, comparable to those obtained on unmodified gold. These bands are again closely similar to those obtained a t the same surface in the presence of solution cyanide. Only very weak spectral features in this frequency region were obtained on the Pt, Rh, and Ru overlayers (as well as unmodified gold) in the presence of cyanate anions. Unlike the reports for CO/NO mixtures on these transition-metal surfacesls (vide supra),then, adsorbed NCO-is apparently not responsible for the present 2100-2200-cm-' features, a t least in the electrochemical environment.

-

Discussion Of central interest here is the intercomparison of the adsorbate compositions as inferred from the SER spectra on the three transition-metal overlayers in the gas phase with data for the corresponding electrochemical systems as well as with other vibrational spectral and related data reported previously for similar metal-gas interfaces. The increasing prevalence of surface oxidation with 0 2 or NO as deduced from the metal-oxygen stretching bands in the sequence Pt < Rh < Ru is consistent with chemical expectations. The decreasing CO coverages attained from CO/O2 mixtures in this sequence as gleaned from the vco and metal surface-C0 ( U M - C O ) band intensities can be understood in terms of increasing competition both from oxygen adsorption and from CO surface oxidation to CO2. Examination of the vco frequencies themselves is instructive. The observed uco frequency for CO/Oz mixtures on platinum, 2060 cm-l (Figure 2B), is relatively low for Ptgas phase interfaces.28 A similar finding for CO on (27) Desilvestro, J.; Weaver, M. J. J . Electroanal. Chem. Interfacial Electrochem. 1987, 234, 237. (28) For example: Barth, R.; Pitchai, R.; Anderson, R. L.; Verykios, X. E. J. Catal. 1989, 116, 61.

Adsorption at Transition Metal-Gas Interfaces alumina-supported Pt has been attributed to adsorption on P t microparticles.z8 An alternative interpretation can be couched in terms of the effective surface potential, cp. Comparisons of infrared spectra for CO on monocrystalline P t surfaces have shown that the uco frequency differences can be accommodated largely in terms of the variations in 9;29for terminal CO, typically dvcold4 = 30 to 40 cm-' V-'. Given that 4 for gold is significantly (at least 0.5 V) lower for gold than for platinum30 and that the effective 4 for the present overlayers should be determined partly by the gold substrate, somewhat lower uco frequencies can be anticipated. However, other factors will alsoaffect these vco frequencies, suchas the CO coveragezgb and the presence of coadsorbates. The latter is exemplified by the ca 30 cm-1 higher terminal vco frequency observed on Pt for CO/NO mixtures (Figure 2B). The above assignment of the ca. 300-cm-' band to the surface-atomic nitrogen stretch from dissociative NO chemisorption is admittedly speculative. On the basis of frequency alone, it is difficult to distinguish such vibrations from those involving N-bound NO since the stretching frequencies are usually determined chiefly by the lead-in atom.15 There have been only occasional reports of similar vibrations for NO chemisorption on Pt-group metal surfaces.31 However, this can be attributed in part to the inability of infrared spectroscopy to readily detect such low-frequency vibrations along with the virtual absence of applications of Raman spectroscopy to such systems. The larger intensities of the ca. 300-cm-I band on Rh and Ru compared with Pt are consistent with the greater extent of NO dissociative chemisorption on the former metals (both single-crystal and polycrystalline surfaces) a t nearambient temperatures, as inferred from thermal desorption and other measurements (refs 32-34 are representative for Pt, Rh, and Ru, respectively). Alternatively, the 300-cm-' features may arise partly (or even entirely) from molecular NO absorption. The absence of detectable U N O bands on the platinum and rhodium overlayers is in any case surprising since substantial molecular absorption is commonly observed a t high NO dosages on these metal^.^^-^^ Most likely, their absence in the gas-phase SER spectra results from primarily low Raman-scattering cross sections. The present results for CO/NO gas-phase mixtures indicate that while CO is adsorbed extensively on platinum (Figure 21, the CO coverages are much lower on the rhodium and ruthenium overlayers as evidenced by the absence of uco and Q-CO bands under these conditions on the latter surfaces (Figures 4 and 7 ) . The diminution of the 310-cm-I NO feature on platinum in the presence of CO (Figure 2A) and the relative lack of influence of CO on the corresponding features on rhodium and ruthenium (Figures 4A and (29) (a) Chang, S.-C.; Leung, L.-W. H.; Weaver, M. J. J . Phys. Chem. 1989,93,5341. (b) Chang, S.-C.; Weaver, M. J. J . Chem. Phys. 1990,92, 4582. (30) (a) Eastman, D. E. Phys. Rev. E: Solid State 1970,2, 1. (b) Demuth, J. E. Chem. Phys. Lett. 1977, 45, 12. (31) For example: Pirug, G.; Bonzel, H. P.; Hopster, H.; Ibach, H. J . Chem. Phys. 1979, 71, 593. (32) (a) Ibach, H.; Lehwald, S.Surf. Sci. 1978, 76, 1. (b) DeJong, K. P,; Meima, G. R.; Geas, J. W. Appl. Surf. Sci. 198213, 14, 73. (33) (a) Chin, A. A.; Bell, A. T. J . Phys. Chem. 1983, 87, 3700. (b) Campbell, C. T.; White, J. M. Appl. Surf. Sci. 1978, I , 347. (c) Ha,P.; White, J. M. Surf. Sci. 1984, 137, 103. (d) Root, T. W.; Fisher, G. B.; Schmidt, L. D. J . Chem. Phys. 1986,85,4679. (34).(a)Thiel, P. A.; Weinberg, W. H. J . Chem. Phys. 1980, 73, 4081. (b) Thiel, P. A.; Weinberg, W.H. ACS Symp. Ser. 1980, No. 137, 191.

Langmuir, Vol. 7, No. 4, 1991 721 7A) are also consistent with this interpretation. The temporal replacement of preadsorbed CO on rhodium with NO (Figure 6) certifies the ability of the latter species to oxidize adsorbed CO. The inability of CO to form high coverages in the presence of NO on rhodium and ruthenium can be understood in part from the extensive oxygen (and nitrogen) coadsorption and also the consequent CO oxidation that occurs under these conditions. These observations are broadly in accord with deductions in the literat~re.~~ Concluding Remarks Although the present study constitutes only a preliminary survey, the results suggest that the technique of SERS with appropriately prepared metal overlayers should provide a significant and possibly even valuable means with which to characterize adsorbate vibrational properties in high-pressure gas-phase environments. Infrared spectroscopy has already proved to be invaluable for this purpose, initially for examination of high-area supported catalysts and lately also for single-crystal and other lowarea metal surfaces.36 In some respects, ,Raman spectroscopy provides a complementary vibrational probe in that the factors influencing infrared absorption and Raman scattering cross sections, along with surface selection rules, are sufficiently distinct so that different vibrational modes are often detected with these techniques. The ability of SERS to access sensitively the metal-surface vibrational region at low frequencies represents a major additional virtue of this technique. Admittedly, some major stumbling blocks to the viable application of SERS to such catalytically relevant metalgas-phase systems remain. The strong possibility exists that the ostensibly inert laser probe is altering photochemically the adsorbate system a t hand, even a t low powers (20 mW, as used here). Indeed, such photochemistry may be responsible in part for the apparent dissociative NO chemisorption discerned above. The observed buildup of carbonaceous material observed here in the absence of oxidizing species is another likely manifestation of this effect. Nevertheless, an enticing virtue of SERS is the ability to acquire time-resolved spectra; with the advent of CCD technology, real-time sequences even in the subsecond time regime are now becoming straightforward. The use of such sensitive and ultralow-noise detectors has the added benefit of allowing interfaces with relatively weak surfaceenhancing properties (such as the present bimetallic overlayer systems) to yield detectable Raman spectra. Given these factors, there appears to be good reason to expect that SERS will finally become a valued technique for the vibrational characterization of catalytically relevant systems in gas-phase as well as electrochemical environments. Acknowledgment. This work is supported in part by grants from the National Science Foundation to M.J.W. (CHE-88-18345) and to C.G.T. (CBT-8611176). ( 3 5 ) (a) Root, T. W.; Schmidt, L. D.; Fisher, G. B. Surf. Sci. 1985,150, 173. (b) Schwartz, S. B., Fisher, G. B., Schmidt, L. D. J . Phys. Chem. 1988, 92, 389. (c) Oh, S. H.; Fisher, G. B.; Carpenter, J. E.; Goodman, D. W. J . Catal. 1986, 100, 360. (36) For reviews, see: Bell, A. T. In Vibrational Spectroscopy of Molecules on Surfaces; Yates, J . T., Jr., Madey, T. E., Eds.; Plenum: New York; 1987, Chapter 3. Hayden, B. E. In Vibrational Spectroscopy of Molecules on Surfaces; Yates, J. T., Jr., Madey, T. E., Eds.; Plenum: New York, 1987; Chapter 7.