Optical second-harmonic generation studies of molecular adsorption

Lifting the Mirror Symmetry of Metal Surfaces: Decoupling the Electronic and Physical Manifestations of Surface Chirality. Andrew Mulligan, Ian Lane, ...
0 downloads 0 Views 909KB Size
J. Phys. Chem. 1988,92, 1419-1425

1419

(ii) Local biaxial ordering of uniaxial molecules giving a macroscopically uniaxial phase: = S:tK$)

where S$ = (1/2(3n,nb- tiab))is the ordering matrix of the local director axis with respect to the molecular frame. In order to write down the measured susceptibility anisotropy with respect to a laboratory fixed axis system, eq A2 has to be averaged over the orientations of the local director, giving

!!.(3RldRld - 6

1 a8ya),(3R$?R:km - a7&))KP

(

d.7 Ba ('46) 9 2 Assuming the random phase approximation, averaging over the director and molecular axes may be done separately so eq A6 becomes

(Ktp)$kple

= (4/9)st;d,,yas$6&")

(iii) Biaxial molecules giving a locally uniaxial nematic, which is macroscopically biaxial-this could represent the orientational ordering in a tilted smectic phase: (Kap)kpIe

(A7)

( AK)

Id m d am)

(2/3)Sd.gS7d Ky6

(A 10)

kkPle= S > % ; ~ K $ ~ )

('41 1)

In the above the second-rank ordering matrices can be written as simple averages,over appropriate rotation matrices; for example, S$? is given by Sd.m =

A number of special cases can now be identified: (i) Uniaxial molecules with local biaxial order which generates a macroscopic biaxiality:

( ~ t ~ ) =z (4/9)st;d,,y,s$'A~(m) k~~~

('49)

(iv) Biaxial molecules which order locally and macroscopically to give a uniaxial phase:

=

(KtS)&pIe

( b ~ ) k k= ~ l(2/3)s$$,d,.gmA~(~) ~

645)

e cos' m - 1 c o s 0 sin cos e cos @

sin' e c o s sin @ 3 sin' e sin' m - 1 sin e c o s e sin 0

('48)

sin e c o s e c o s sin e c o s e sin @ 3 cos' e - 1

I)

(-412)

ARTICLES Optical Second Harmonic Generation Studies of Molecular Adsorption on Pt( 111) and Ni(111) S. G.Grubb,*t A. M. DeSantolo,t and R. B. Hall* Exxon Research and Engineering Company, Annandale, New Jersey 08801 (Received: December 2, 1986; In Final Form: October 13, 1987) The nonlinear optical process of second harmonic generation (SHG) is, from symmetry considerations, inherently surface specific and has the sensitivity to detect submonolayer coverages of molecules. We have used SHG to monitor the adsorption of CO, 02,and H2 on Ni( 11 1) and Pt( 111) surfaces. In the case of Ni( 11l), hydrogen adsorption follows Langmuir kinetics while the adsorption of both CO and O2 is characteristic of adsorption from a precursor state. The SHG response from these surfaces is dependent on the relative orientation of the crystalline axes and input polarization vector. Under specific wavelength and polarization conditions the SHG response is seen to correlate with work function changes upon adsorption. In the case of Pt( 11l), a striking frequency dependence of the SHG response is observed, suggesting the participation of surface states or resonant effects associated with interband transitions.

Introduction The utility of optical second harmonic generation (SHG) as a surface probe. has now been well demonstrated in a wide variety of Second-order nonlinear optical processes such as S H G are forbidden, within the electric dipole approximation, in media with inversion symmetry. Thus, the SHG process exhibits an intrinsic sensitivity to a symmetry-breaking surface or interface. In addition, S H G is sufficiently sensitive to detect submonolayer coverages of adsorbates on surfaces. The SHG technique has been used to monitor the orientation and structural symmetry of monolayers of adsorbates on a variety of dielectric surfaces.*' In these cases, the nonlinear susceptibility t Present address: Ammo Research Center, Physical Technology Division, P.O. Box 400, Naperville, IL 60566. *Present address: AT&T Bell Laboratories, 600 Mountain Ave, Murray Hill. NJ 07974.

of the adsorbates and, hence, the S H G signal are greater than or comparable to that of the substrate. However, in the case of r ~ ~ ~ ~ the adsorption on smooth metal8 and s e m i c o n d u c t ~ surfaces, ( 1 ) Shen, Y. R. In Chemistry and Structure at Interfaces, New Laser and Optical Investigations; Hall, R. B., Ellis, A. B., Eds.; VCH: Publishers Deerfield Beach, FL, 1986; and references therein. (2) Shen, Y.R. J . Vac. Sci. Technol., B 1985, 3, 1464. (3) Heinz, T. F.; Chen, C. K.; Ricard, D.; Shen, Y. R. Phys. Rev. Lett. 1982, 48, 478. (4) Heinz, T. F.;Tom, H. W. K.; Shen, Y. R. Phys. Rev. A 1983,28, 1883. (5) Tom, H. W. K.; Heinz, T. F.; Shen, Y. R. Phys. Rev. Lett. 1983,51, 1983. ( 6 ) Rasing, Th.; Shen, Y. R.; Kim, M. W.; Valint, P., Jr.; Bock, J. Phys. Rev. A 1985, 31, 537. (7) Rasing, Th.; Shen, Y. R.; Kim, M. W.; Grubb, S . G. Phys. Rev. Lett. 1985, 55, 2963. (8) Tom, H. W.K.; Mate, C. M.; Zhu, X. D.; Crowell, J. E.; Heinz, T. F.;Somorjai, G.; Shen, Y. R. Phys. Rev. Lett. 1984, 52, 348.

0022-3654/88/2092-1419$01 .50/0 0 1988 American Chemical Society

1420 The Journal of Physical Chemistry, Vol. 92, No. 6, 1988

substrates exhibit large surface nonlinear susceptibilities, and changes in the S H G signal upon adsorption have been shown to be dominated by moleculesubstrate interactions. Early studies demonstrated the sensitivity of S H G from metal and semiconductor surfaces to the presence of adsorbates."J2 However, the lack of detailed knowledge of surface condition prevented conclusive interpretation of these results. Recent SHG experiments in ultrahigh vacuum (UHV) have yielded information concerning the nonlinear susceptibility of a single-crystal metal surface and its change upon adsorption.* In the case of molecular adsorption of the electron acceptors C O and O2on a Rh( 1 11) surface, the SH signal from the surface was found to decrease monotonically with increasing adsorbate coverage. This result suggests that the nonlinear optical response of the metal surface is dominated by the surface free-electron density. Strong chemisorption bonds, which partially localize the surface electrons, are thus expected to reduce the surface nonlinear optical response. On the other hand, adsorption of alkali metal atoms which donate electrons to the Rh surface was found to result in an increase of the SH response.13 SHG has the potential to be a useful probe of surfaces in UHV environments especially when used on conjunction with more conventional surface science techniques. However, complete understanding of the origin of surface nonlinearities is needed to make S H G a generally useful surface probe. The fast time response, the ability to probe small (>wavelength of light) sample areas as well as the nondestructive nature of the technique are several of the advantages of SHG. S H G would be especially applicable as an in situ probe for catalytic reactions at high pressure. As outlined below, SHG, using tunable sources, also has the potential to provide information regarding surface electronic states and their change upon adsorption. We have used time-resolved S H G to monitor the kinetics of adsorption of CO, 02,and H2 on Ni( 111) and Pt( 11 1). In the case of Ni( 1 1l ) , the S H G signal varies monotonically with adsorbate coverage and can be modeled to obtain information on the sticking coefficient as a function of coverage. Adsorption of H2 is found to follow Langmuir adsorption kinetics while C O and O2exhibit characteristics of adsorption via a precursor state. The sticking coefficient for C O remains invariant with coverage up to 0.4 monolayer while for O2 the sticking coefficient remains invariant nearly to saturation of the first layer. In the case of CO adsorption on Pt( 11l), we find that the SHG signal does not vary monotonically with coverage because the CO binding site varies with coverage. Under specific excitation conditions, we find that there is detailed correspondence between the S H G signal and the change in work function as a function of coverage. Finally, we observe a striking wavelength dependence of the SHG response from Pt, suggesting a resonant enhancement of the nonlinear susceptibility involving surface states or interband transitions. Experimental Section The experiments were conducted in a UHV chamber with a base pressure of 6 X lo-" mbar. The ion gauge was calibrated for each gas by means of a spinning rotor gauge manufactured by MKS Inc. The Ni( 11 1) surface was cleaned by cycles of Ar+ sputtering and annealing. The P t ( l l 1 ) surface was cleaned by cycles of high-temperature oxygen treatment, Ar+ sputtering, and annealing. Surface cleanliness was monitored by Auger electron spectroscopy (AES), ion scattering spectrometry (ISS), and low-energy electron diffraction (LEED). Samples were flash heated to 700 K prior to each experiment. The adsorption ex(9) Heinz, T. F.; Loy, M. M. T.; Thompson, W. A. Phys. Rev. Lett. 1985, 54, 63.

(10) Tom, H. W. K.; Zhu, X.D.; Shen, Y.R.; Somorjai, G. A. Surf: Sci. 1986, 167, 167. (11) Brown, F.; Parks, R. E.; Sleeper, A. M. Phys. Rev. Lett. 1965, 14, 1029. (12) Chen, J. M.; Bower, J. R.; Wang, C. S.; Lee, C. H. Opt. Commun. 1973, 9, 132. (13) Tom, H. W. K.; Mate, C. M.; Zhu, X. D.; Crowell, J. E.; Shen, Y. R.; Somorjai, G. A,, submitted for publication in Surf: Sci.

Grubb et al. 1

I

I

CO/Pt (111)

I

I

0

1

2

3

4

CO Exposure (Langmuirs)

Figure 1. Second-harmonic signal from Pt(ll1) at 100 K as a function of exposure to CO. The excitation wavelength is 532 nm. Units of exposure are langmuirs, 1 langmuir = 1 X lo6 Torr s.

periments were carried out with the surface at 100 K (except as noted). Coverages were calibrated by thermal desorption spectroscopy (TDS) and LEED. A Nd:YAG laser manufactured by Quantel International provided 7-11s laser pulses at either 532 nm at 1.06 pm at a repetition rate of 30 Hz. The 3-4 mJ per pulse energy in an area of 10 mm2 was well below the threshold for either laser-induced desorption or surface damage. The excitation light was ppolarized and incident on the sample at an angle of 67' from the surface normal. The plane of incidence was approximately 30' off of the projection of the metal crystal (271) axis onto the surface. (The (217) axis is along the C, mirror plane of symmetry of the first two atomic layers of the (1 11) surface.) The SH signal was monitored by a gated detection system. In the case of Ni( 11l), a polarization analyzer was used so that only the p-polarized radiation was detected. In the case of Pt(l1 l ) , the SH signal from the clean surface was a factor of 4 less intense than the signal from the N i surface. Because of this, after determining that more than 80%of the SH signal from the Pt surface at any coverage was inherently p-polarized, the polarization analyzer was removed for the experiments on Pt except as noted. ReSultS The SH signal from a Pt( 11 1) surface versus C O exposure at a pressure of 7.6 X lo4 Torr is shown in Figure 1. The SH signal is normalized to the signal from the clean (8, = 0) metal surface, where 8, is the adsorbate surface coverage relative to the density of substrate surface atoms. The SH signal chapges with CO coverage in a complex way. The SH intensity at first increases, reaching a maximum a t 8, = 0.33 then decreases to just below the bare surface value at 8, = 0.5. The SH intensity then monotonically increases to a limiting value at high coverages. At 100 K we observed a somewhat diffuse ( d 3 X d 3 ) R 3 0 ° LEED pattern at 8, = 0.33 which transformed into a sharp ~ ( 4 x 2 ) structure at e,, = 0.5. The normalized SH signal from a Ni( 11 1) surface during exposure to CO at a pressure of 7.6 X lo4 Torr is shown in Figure 2. The SH response for the adsorption of CO on Ni( 111) is seen to be qualitatively different from that for CO on the P t ( l l 1 ) surface. The SH signal shows a monotonic decrease to a limiting value of 0.14 upon adsorption of a full monolayer of CO. For both metals the SH signal is observed to fully recover to the bare metal value upon desorption of CO. The solid line shows a fit to a model assuming Langmuir adsorption kinetics.* A value of 1.0 f 0.05 is obtained for the initial sticking coefficient, So,of CO on Ni( 11 1). This agrees well with previous values of 1.OI4 (14) Christmann, K.; Schober, 0.;Ertl, G . J . Chem. Phys. 1974,60,4719.

The Journal of Physical Chemistry, Vol. 92, No. 6, 1988 1421 1 CO/Ni (111)

02/Ni (111)

.E

e

-.-w

.6 0

5.

.-r L

u)

E .4

+

I v)

.2

1 1

O O

0

3

2

0

1

CO Exposure (Langmuirs)

Figure 2. Second-harmonic signal from Ni( 111) as a function of exposure to CO at 100 K. The excitation wavelength is 532 nm. The smooth

curve is derived from a Langmuir adsorption model used to fit the experimental data. 0

I

2 3 0, Exposure (Langmuin)

4

Figure 4. Second-harmonic signal as a function of O2dose on Ni( 111) at 100 K. The excitation wavelength is 532 nm. The smooth line is

derived from an adsorption model that assumes the sticking probability is independent of coverage (see text).

H, Exposure (Langmuirs) 20 40 60 I

I

I

HJPt (111)

I HJNi (111)

0

0

.2

.4

Coverage

.6

.8

1

(em.)

F’igure 5. Relative sticking probability (S(e)/S,) from the SH adsorption isotherms for CO (w), O2(e) and H2( 0 )versus coverage on Ni(ll1)

at 100 K.

.2

I0

I

I

5 10 H, Exposure (Langmuirs)

Figure 3. Second-harmonic signals as a function of H2dose on Pt( 111) and Ni(l11) at 100 K. The excitation wavelength in both cases is 532

nm. The smooth curves are fits to a Langmuir adsorption model. and 0.91 l5 obtained by monitoring the change in work function upon adsorption of CO on a Ni( 111) surface. In this calculation, we take for e,, the saturation coverage of CO relative to the (15)

Campuzano, J. C.; Dus, R.;Greenler, R.G . Surf. Sci. 1981,102, 172.

substrate atom density, a value of 0.53,14corresponding to 1.0 X lOIs adsorption sites per cm2. Distinctly different SH responses from the Ni( 111) and Pt(111) surfaces are also observed for adsorption of H2as shown in Figure 3. The SH response in both cases is monotonic but differs in sign. The Langmuir adsorption model gives values of 0.06 and 0.1 5 for the initial sticking coefficient of H2 and Pt( 11 1) and Ni( 11l), respectively. The SH response of Ni( 1 11) as a function of O2exposure is shown in Figure 4. In this case, the change in the SH signal cannot be modeled by assuming Langmuir adsorption kinetics. The smooth line in Figure 4 is derived from a model assuming uniform probability of sticking, independent of adsorption site occupancy. Figure 5 shows the values of the relative sticking probabilities (S(e)/S,) versus coverage of CO, H2, and O2 on Ni( 111) obtained from the slope of the SH intensity as a function of dose as described below. The sticking probability for H2 on N i ( l l 1 ) decreases linearly with coverage, characteristic of Langmuir kinetics. The sticking probability for CO, however, remains constant up to e/€),approaching 0.4. For O2adsorption

Grubb et al.

1422 The Journal of Physical Chemistry, Vol. 92, No. 6, 1988 l"O

A

I fY

I

1.4

I

I

CO/Pt (111)

I

I

CO/R (111)

1.2

Le, = 532 nm

0.90

-

I

B

.-I

I

I

I

1.0

E 0

t

*

-

L 'B

0.8

C

0

-I v)

0.6 -0.025

I

0

I

I

I

1 2 3 CO Exposure (Langmuirs)

I

4

Figure 6. In A, the square root of the second-harmonicsignal from Pt(l11) as a function of exposure to CO at two different incident polarization conditions. Both curves are for p-polarized input and p-polarized SH. The upper curve has the plane of incidence oriented +30° from the projection of the crystal (211) axis on the surface. The lower curve has the plane of incidence oriented +210° from this reference axis (see text). In B, the difference of the two curves is plotted to indicate the anisotropic SH response as a function of CO exposure.

0.4

0.2

0

1

2

3

4

CO Exposure (Langmuirs)

on Ni(l1 l ) , the sticking probability remains at its initial value until near saturation coverage where it decreases rapidly. These results suggest the existence of an extrinsic precursor state for adsorption of CO and 02. The SH response for CO, 02,and H2 adsorption on the Ni( 111) surface shows little or no dependence on the polarization of the incident radiation or on whether the frequency of the incident radiation is 532 nm or 1.06 pm. This is not true for P t ( l l 1 ) . Figure 6A shows the variation of intensity of the SH signal as a function of exposure of the platinum surface to CO at two different input polarizations. The upper curve was obtained at the same incident angle and incident p-polarization as the upper curve in Figure 1, but in this case only the ppolarized SH response was detected. The lower trace was obtained by rotating the input polarization by 180°, as described below. The SH response to CO adsorption on Pt( 111) with 1.06-pm incident radiation is markedly different from the response for 532-nm incident radiation. This is shown in Figure 7. In addition, the SH signal for H2 adsorption on P t ( l l 1 ) with 1.06-pm excitation decreases monotonically in contrast to the increase shown in Figure 3 for 532-nm excitation. As described below, we believe the qualitatively different resposne of Pt at 1.06 pm is due to resonant enhancement of the SH signal by electronic states in Pt. Ni does not have states in the same region and hence the SH response at the two frequencies investigated is the same. Most of our results can be understood in terms of the variation of the free-electron density of the metal upon formation of a surface adsorbate bond in much the same way as the change in work function is understood. The polarization- and frequencydependent results cannot be described by such a simple picture. These points are discussed in the following section. Discussion The adsorption of CO on Pt( 11 1) exhibits several interesting complexities which have made this system the focus of many experimental For example, the data of Ertl (16) Poelsema, B.; Verheiu, L. K.; Comsa, G . Surf.Sci. 1985, 153, 496. (17) Halsey, G. D. J. Chem. Phys. 1976, 65, 2029.

Figure 7. The second-harmonicresponse from Pt( 111) as a function of

exposure to CO for two different incident wavelengths. Both curves are normalized to the clean surface SH response for each input wavelength. et and Norton et demonstrate that the variation of work function, A%, with CO exposure is a complex function of coverage. The change in work function versus CO exposure at first decreases, passes through a pronounced minimum, reaches a maximum, and then decreases again, passing through a second shallow minimum. The variation of A@ versus CO exposure exhibits a striking inverse correlation with the change in SH response shown in Figure 1. Ertl et al. reports that the work function minimum of -1 50 mV occurs at a CO coverage of 8 = 0.33 as evidenced by the appearance of the (v'3Xd3)R30° LEED pattern.21a The ap) pattern at 8 = 0.5 coincides pearance of the sharp ~ ( 4 x 2LEED with the maximum of the work function change. Using TDS and LEED, we have determined that the coverages corresponding to the minimum and maximum of the change in work function correspond to the maximum and minimum of the SH response, respectively. A correspondence between AO and ASH would be expected if the surface nonlinear susceptibility, x ( ~ )is, dominated by the surface free-electron density. The initial decrease in work function upon CO absorption corresponds to a net transfer of electronic charge from CO to the P t ( l l 1 ) surface. The increased surface electron density creates a greater surface nonlinear polarizability and thus an increase in the SH response. The increase in work function between 8 = 0.33 and 8 = 0.5 similarly parallels the decrease in the SH response. The observed variation of work function over this coverage regime has been explained by various groups to be due to a change from linear to bridge bonded CO species as well as changes in ordering of the adsorbed layer which effects the dipole moment by through-metal interaction^.'^ ~

(18) Steininger, H.;Lehwald, S.; Ibach, H. Surf. Sci. 1982, 123, 264. (19) Norton, P. R.; Goodale, J. W.; Selkirk, E.B. Surf. Sci. 1979.83, 189. (20) Hayden, B. E.; Bradshaw, A. M. Surf. Sci. 1983, 125, 787. (21) (a) Ertl, G.; Neumann, M.; Streit, K. M. Surf.Sci. 1977, 64, 393. (b) Norton, P. R.; Goodale, J. W.; Selkirk, E. B. Surf.Sci. 1979, 83, 189.

The Journal of Physical Chemistry, Vol. 92, No. 6, 1988 1423

S H G Studies of Molecular Adsorption on Pt and Ni

As has been well established p r e v i o ~ s l ythe ~ ~SH ~ *generated ~~~~ from a crystalline surface will consist of isotropic and anisotropic terms from both the surface and bulk. The behavior of the anisotropic component of the SH signal has previously been well characterized for semiconductor surfaces and recently for a Cu(1 1 1) surface under UHV conditions.22 The magnitude of the anisotropic contribution to the SH response depends on the relative orientation between the crystalline axes and the laser polarization. The isotropic components are most associated with the surface free electrons, although the bonding electrons can also contribute. Hence, the observed inverse correspondence between the work function and the SH intensity suggests that under these specific conditions the SH signal appears to be dominated by the freeelectron response. As shown below, this correspondence is no longer valid under different polarization and wavelength conditions. The SH response to CO exposure on the Ni( 111) surface differs dramatically from the case of C O adsorption on Pt( 111). There is a monotonic decrease in the SH signal, as shown in Figure 2, and a similarly monotonic increase in the work f ~ n c t i o n . ' ~ *The '~ fact that nearly 90% of the bare Ni( 111) SH signal is extinguished upon adsorption of a monolayer of CO demonstrates both the sensitivity of the S H G technique and its inherent surface specificity. Some of the residual Ni( 1l l ) SH signal after adsorption of a monolayer of C O may arise from the quadrupole allowed signal from the bulk. The smooth line in Figure 2 is the fit expected if the adsorption of C O on Ni( 111) follows Langmuir adsorption kinetics. Unlike the case of C O on Pt( 11 l), CO on Ni(l11) adsorbs only at twofold bridge sites until the coverage is very near s a t u r a t i ~ n . ~Assuming ~ - ~ ~ all adsorption sites are equivalent and noninteracting, we adopt the same simple model as Tom et aL8 and write the surface nonlinear susceptibility as

xS(') = A

+ B 0/0,

(1)

where 0 is the fractional coverage of adsorbate relative to the metal surface atoms, 0, is the saturation value of 0, A is the bare metal contribution to the nonlinear susceptibility, and B is a constant. The Langmuir adsorption model dictates that the rate of change of adsorbate coverage with time varies as d 0 / d t = Kp(1 - e/e,)

(2)

where p is the pressure of adsorbate and the constant K is related to the initial sticking coefficient So by K = ( S o / N , ) ( R T / N, is the density of adsorption sites, Mgis the molecular weight of the gas, and Ngis the gas density. Integration of (2) yields e(t) = e,[l

- exp(-~pt/0,)]

(3)

The intensity of the second-harmonic signal, ISH, is proportional , to the square of the surface nonlinear susceptibility, I X S ( ~ ) I ~and therefore has the following form for Langmuir adsorption (4)

The fit to a Langmuir isotherm for CO adsorption on Ni( 1 11) a t 100 K,shown in Figure 2, gives the following values: B / A = 1.1 exp(i(163')) and ~ / 0=, 0.84. With 0, = 0.53, the latter term leads to a value for the initial sticking coefficient of So = 1.O f 0.05. Upon close inspection of the Langmuir model fit to the SH data of Figure 2, it is seen that this simple model does not adequately describe the adsorption kinetics of C O on N i ( l l 1 ) . A more general form than that of eq 2 for the kinetics of adsorption may be written as d0/dt = K P S ( ~ )

(5)

where S ( 0 ) is the coverage-dependent sticking coefficient. With

the assumption of eq 1 that coverage, it follows that

x ( ~changes ) linearly with absorbate

Thus S(0)may be obtained from the slope of (ZsH)1/2 versus time. The relative sticking probability S(0)/Sofor CO adsorption on N i ( l l 1 ) obtained in this manner from the data in Figure 2 is shown in Figure 5. It can be seen that (S(0)/So)remains constant until 0/0, approaches 0.4, suggesting that adsorption occurs via precursor-state adsorption kinetic^^-^^ rather than the Langmuirian model. These results for the dependence of the sticking probability on coverage agree well with those derived from work function measurements of Campuzano et al.ls and qualitatively with those of Christmann et al.14 for C O on Ni(ll1). Both our SH results and the work function measurements differ from recent SH measurements of Zhu et a1.28for this same system at 300 K. The adsorption of C O on Ni( 111) in this case showed an excellent fit to Langmuir kinetics. We have observed adsorption isotherms for CO on Ni( 1 11) that were better described by Langmuir kinetics under high laser flux conditions, although the quality of the fit was poor. However, under the flux conditions reported here, we consistently observe significant deviations from Langmuir kinetics at crystal temperatures of both 100 and 300 K. Recent molecular beam studies26of the dynamics of C O adsorption on Ni( 111) indicate the existence of a long-lived molecular precursor state. It is possible that a significant number of C O molecules bound in a precursor state could be desorbed by the laser before being trapped in the chemisorbed state. It seems likely that the higher laser fluxes used in the SH experiments of Zhu et al. were sufficient to desorb the precursor molecules but not the chemisorbed species, leading to an apparent Langmuir adsorption kinetics. The SH responses to H2adsorption on the Ni( 111) and Pt( 111) surfaces shown in Figure 3, although qualitatively opposite in sign, both follow an inverse correlation with the change in the work function. Hydrogen adsorption on Ni( 11 1) leads to a work function increase of approximately 0.2 V,29 along with the decrease of the surface nonlinear susceptibility. Adsorption of hydrogen on the Pt(111) surface, on the other hand, lowers the work function by approximately 0.23 V30 and increases the SH response. As shown in Figure 3, hydrogen adsorption on both the Ni( 111) and P t ( l l 1 ) surfaces follows Langmuir kinetics. The low initial sticking probabilities of 0.15 and 0.06 for the Ni(ll1) and Pt(ll1) surfaces agree well with literature values of 0.1517 and 0.0516 for both systems. The sticking probability as a function of coverage on Ni( 111) shown in Figure 5 is that expected for Langmuirian adsorption. The adsorption of oxygen on the Ni( 11 1) surface cannot be fit to Langmuir adsorption kinetics. However, we are able to model the data of Figure 4 by assuming that adsorption occurs through a mobile precursor state with the time dependence of the coverage given by de/dt = ~ p [ -l 6 ( t , - t ) ]

(7)

where t, is the time to reach saturation (6 = 0,) and 6 is the Kronecher A function, which assumes a value of unity at saturation. With this, the expression for the SH intensity as a function of adsorbate dose is

where L is the adsorbate dose in langmuirs. A nonlinear least(26) Tang, S. L.; Beckerle, J. D.; Lee, M. D.; Ceyer, S. T. J. Chem. Phys.

-1986. - - -, 84. - ., 6488. - . - -. (22) (23) (24) (25)

Tom, H. W. K.; Aumiller, G. D. Phys. Rev. E 1986, 33, 8818. S i p , J. E.; Moss, D. J.; van Driel, H. M. Phys. Rev. E 1987.35, 1129. Trenary, M.; Uram, K. J.; Yates, J. T., Jr. Surf. Sci. 1985, 163, 513. Netzer, F. P.; Madey, T. E. J . Chem. Phys. 1982, 76, 710.

(27) Kisliuk, P. J. Phys. Chem. Solids 1957, 3, 95. (28) Zhu, X. D.; Shen, Y. R.; Carr, R. Surf.Sci. 1985, 163, 114. (29) Christmann, K.; Schober, 0.;Ertl, G.; Neumann, M. J . Chem. Phys. 1974,60,4528. (30) Christmann, K.; Ertl, G.; Pignet, T. SurJ Sci. 1976, 54, 365.

1424 The Journal of Physical Chemistry, Vol. 92, No. 6, 1988

squares fit of the data of Figure 4 leads to the determination of the parameters: BIA = -0.40 and K/e, = 0.87. Measurements of the work function change upon oxygen exposure of a Ni( 111) surface at 5 K have provided direct evidence for the existence of precursor adsorption preceding dissociative chemisorption.” The sticking probability at 100 K as a function of coverage shown in Figure 5 also suggests the existence of long-range interactions that allow O2to remain near the surface long enough to adsorb even at high coverage. For the polarization conditions used here, the change in the second harmonic signal follows more than just the qualitative direction (sign) of the change in the work function. There is at least a semiquantitative correlation between the two quantities. The ordering of the magnitudes of the change in the work function follows the ordering of the magnitudes of the change in the SH signal for all of the systems investigated. For example, the magnitude of the change of ISHfrom Ni( 111) increases in the order H2 < O2 < CO. This is also the ordering of the increase in the magnitude of the change of the work function. Currently, we cannot put this correlation on a more quantitative basis because the various literature values for the magnitude of the change in the work function vary significantly. We intend to repeat the work function measurements in our own system in order to have the measurements made under comparable conditions. We do not expect this correlation to hold for all input and output polarizations. Specific tensor elements of the nonlinear susceptibility may exhibit strong correlations to work function changes while others may be. more sensitive to adsorbate-induced changes in interband transitions, or to phase transition^.^^ The correlation with changes in the work function is most likely to hold in those cases where the SH response originates predominately from the metal free electrons. The SH response in this case should be isotropic with respect to the angle between the input polarization vector and some reference plane in the metal crystal. In contrast to this, recent work has demonstrated the importance of the anisotropic components of the susceptibility and of interband transitions to the SH response from a metal surface.22 Each polarization geometry contains a linear combination of susceptibility tensor elements that contribute to the SH response. The case of both p-polarized excitation and p-polarized SH response from a (1 11) surface is the most complex, with all five distinct susceptibility components contributing to the SH signal.34 Ideally, each nonlinear susceptibility tensor element should be isolated by the appropriate experimental conditions and followed as a function of adsorbate coverage. It may then be possible to determine the relative contribution of the d-band electrons and the s-p bands close in energy to the Fermi level and how they are affected by adsorption. Recently this has been accomplished for the adsorption of O2 on Si( 11 l ) . 3 5 We expect that additional data on the polarization dependencies of the SH response for several different crystal orientations (e.g. Ni( 100) and Ni( 1 10)) will allow us to resolve the various tensor components responsible for the correlation between the SH response and the change in the work function in the systems studied here. Some measurements of the dependence of the SH response on the input polarization have already been made. For Ni(l1 l), there is little or no polarization dependence detected for the fractional change in the SH intensity upon adsorption of CO, H2, or 02. The same is not true for Pt( 11 1). In Figure 6a, the SH response from P t ( l l 1 ) at two different input polarization orientations is shown as a function of CO adsorption. Some anisotropy might be expected in light of previous work.5,9*22*23 Electrons associated with band states of the metal that have some average directionality (31) Shayegan, M.; Cavallo, J. M; Glover, R. E., 111; Park, R. L. Phys. Reo. Lett. 1984, 53, 1578. (32) Miragliotta, J.; Furtak, T. E. Phys. Rev. E 1988,37, 1028. Koos, D. A.; Shannon, U. L.; Richmond, G. L., to be published. (33) Heskett, D.; Song, K.J.; Burns, A.; Plummer, E. W.; Dai, H. L. J . Chem. Phys. 1986,85,7490. (34) Tom, H. W. K.Ph.D. Thesis, University of California at Berkeley, 1984.

(35) Tom. H. W. K.;Aumiller, G . D. Phys. Reu. B, in press.

Grubb et al. with respect to the crystal lattice or electrons associated with molecular orbitals of the adsorbates will result in an anisotropic contribution to the nonlinear susceptibility. The intensity of the SH signal is a sum of the isotropic and anisotropic contributions from both the surface dipole allowed terms and the bulk quadrupole allowed terms. The general form for ZsH is ISH cc

Icl + cfl+)12

(9)

where C1 is a linear combination of isotropic surface and bulk nonlinear susceptibilities and C2is a combination of the anisotropic susceptibilities.22 For a 3m symmetric crystal, f(+) is a linear combination of cos (3+) and sin (3+) with defined as the angle between the input polarization vector and the projection of the (255) axis on the surface plane. In Figure 6A, the square root of ISH,which is proportional to the nonlinear susceptibility, is plotted as a function of exposure to C O for each of two input polarization conditions. The upper curve is the square root of the SH signal from the ”top” surface of the Pt( 11 1) crystal. This is the same surface measured in Figure 1. The input radiation is p-polarized but, unlike Figure 1, a polarization analyzer is inserted in the reflected beam so that only the ppolarized component of the SH signal is detected. The lower curve is also the p-polarized component of the S H signal, but in this case the plane of incidence has been rotated 180’. This is equivalent to a change of 180’ in and is achieved either by flipping the crystal over so that the SH signal from the “bottom” surface is detected or by reversing the propagation direction of the laser beam. (A similar experiment with the Ni( 11 1) crystal did not result in a change of the SH signal.) For p-polarized incident radiation and p-polarized SHG, both C1and C2 are nonzero andf(+) = cos (3+). Under these conditions, a 180’ change in changes the sign of the second term in eq 9. Figure 6B shows the difference function obtained by subtracting the two curves in Figure 6A. The isotropic components cancel in this subtraction, and the difference curve gives an indication as to the behavior of the anisotropic term, Czf(+), as a function of C O exposure. The difference curve remains approximately zero up to an exposure of 0.3 langmuir, which we have calibrated to be a coverage of e,, = 0.17. It is at this coverage that we observe a rather diffuse (v‘3Xv‘3)R30° LEED pattern as reported by Ibach et a].’* The difference curve term then increases rapidly between 8, = 0.17 and 8, = 0.33. It is apparently in this coverage region that the anisotropic components are most strongly perturbed. We suspect that the Pt d-bands, which have a relatively directed spatial distribution, contribute most to the anisotropic terms and hence that it is the effects on these bands that we are observing. The subtraction of the two curves in figure 6A to isolate the anisotropic term is not strictly valid as there is an undetermined difference in phase between C1and C2,but it does serve to show the trend of the change of the anisotropic component as a function of CO coverage. The pure anisotropic contribution to the S H response can be isolated by the proper choice of excitation geme try.^^^^^ Further work is now in progress to rigorously isolate the orientation-dependent and orientation-independent contributions to the S H response as a function of adsorbate coverage. The increase in the intensity of the SH responses from 532-nm excitation of the Pt( 11 1) surface upon exposure to C O and H 2 and the decrease in the work functions suggest an increase in the electron density at the surface. Significantly, the SH response for 1.06-pm excitation is dramatically different, showing a decrease instead of an increase in SH intensity upon adsorption. A comparison of the SH response at these two wavelengths for C O adsorption on Pt(l11) is shown in Figure 7 . This wavelength dependence may involve resonance enhancement due to surface states in the vicinity of the 1.06-pm (1.17-eV) excitation. Alternatively, it may be due to resonant enhancement associated with interband transitions. Recent calculations of the local density of states (LDOS) of transition-metal surfaces show an enhancement of state density in the region from 0 to 2 eV below the Fermi level.36 Both calculation^^^ and photoemission ex-

+

+

+

S H G Studies of Molecular Adsorption on Pt and Ni p e r i m e n t ~show ~ ~ , a~ reduction ~ of state density in this low-energy region upon adsorption. A photoemission study of CO on Pt foil'@ (treated by a method shown to give a preferentially (1 11) oriented surface) also shows a decrease in state density in the region 0-2 eV below the Fermi level.'@ If transitions between these low-lying surface states to unoccupied states above the Fermi level are optically allowed, they could dominate the SH signal by resonance enhancement of the nonlinear susceptibility. The decrease of the SH signal in the case of 1.06-pm excitation for the adsorption of O2,Hz, and CO on Pt( 111) could be explained by a quenching of these surface states upon adsorption. A different temperature dependence of the SH signal was also observed for the two frequencies used in these experiments. For 532-nm excitation, the bare Pt(l11) SH signal was constant over the 100-1000 K temperature range. With 1.06-pm excitation however, the SH signal increased by 70% upon heating to 1000 K. The SH signal recovered to its original value upon cooling to 100 K. This effect must arise from a temperature-dependent change in the surface electronic properties of Pt(ll1). The signal rises sharply in the 250-500 K temperature range. It is unlikely that impurity segregation to the surface layer is a major factor at these modest temperatures. The reversibility of the SH signal upon cooling further argues against impurity effects. We found no evidence of surface impurities in the Auger spectrum after repeated temperature cycling of the Pt(ll1) crystal in the 100-800 K range. A decrease of the SH response with increasing temperature has also been observed for the Si(ll1) surface.1° It was suggested that this temperature dependence may be correlated with the temperature dependence of an EELS loss peak at 1.7 eV. Part of the temperature of the Si( 11 1) SH signal was also believed to be due to segregation of bulk phosphorus impurities to the surface layer. The temperature dependence of the S H response as function of wavelength and polarization should provide additional insight into the nature of the surface electronic properties of the metal. Results of these studies will be presented elsewhere. Conclusions The second-harmonic response of a metal surface with adsorbed molecules can provide fundamental information on the nature of the metal electron distribution in the near surface region and on the nature of the surface chemical bond. In most, but not all, of the cases studied here, the intensity of SH radiation provides a quantitative measure of the coverage of adsorbed species. In these cases, the S H G technique is especially useful for the study of the kinetics of molecular adsorption, desorption, and reaction. We observe for example, distinctly different kinetics for the adsorption of H2, 02,and CO on N i ( l l 1 ) . For Hz, adsorption is seen to follow Langmuir kinetics. For CO and 02,the existence of an extrinsic precursor state to adsorption is identified and the coverage dependence of the sticking probability measured. The change in the SH response of the surface upon adsorption is shown to be sensitive to the nature of the surface bonding. A correlation is noted between the change in the SH response and the change in the work function induced by adsorption. This is most clearly evident in the case of CO adsorption on a Pt( 111) surface. Here, the SH intensity as a function of the coverage is nonsingle valued because of a change in the bonding position as a function of coverage. Because of this, it is not possible to determine uniquely the surface coverage from the intensity of the (36) Louie, S. G. In Electronic Structure, Dynamics, and Quantum Structural Properties of Condensed Matter, Devreese, J . T., Van Camp, P., Eds.; Plenum: New York, 1985. (37) Louie, S. G. Phys. Rev. Lett. 1979, 42, 476. (38) Louie, S. G. Phys. Rev. Lett. 1978, 40, 1525. (39) Eberhardt, W.; Louie, S. G.; Plummer, E. W. Phys. Rev. B 1983,28, 465. (40) Collins, D. M.; Lee, J. B.; Spicer, W. E. Surf. Sci. 1976, 55, 389.

The Journal of Physical Chemistry, Vol. 92, No. 6,1988 1425

SH signal. Even with this, the change in the work function upon adsorption is similarly nonmonotonic and varies inversely with the change in the SH intensity at each coverage. This inverse correlation holds only for certain polarization and wavelength conditions. For polarization conditions that project out the isotropic part of the nonlinear susceptibility tensor, the correlation seems to be at least semiquantitative. By appropriate manipulation of the input and output polarizations, the anisotropic part of the SH response can be separated from the isotropic part. In the current experiments, the tensor elements were not determined individually and several tensor elements could contribute to the anisotropic part of the SH response. Nevertheless, we find that the CO coverage dependence of the anisotropic part of the SH response from Pt(111) is different than that of the isotropic part. The anisotropic part increases sharply between Oc0 = 0.17 and ec0= 0.33 and varies very little at coverages above or below this range. The anisotropic response should provide information on the orientation of the surface chemical bonds, possibly on how the metal d-electron bands are perturbed upon adsorption as a function of the orientation with respect to the lattice atoms.35 Isolation of the individual anisotropic tensor elements will be necessary to test this further. The wavelength dependence of the SH response should provide new information on the surface electronic properties and their change with adsorption. For P t ( l 1 l ) , there is a dramatic difference in the SH response depending on whether 532-nm or 1.06-pm excitation is used. This difference may be due to resonant enhancement of the SH signal by surface states or interband transitions around 1.2 eV. Significantly, it is only the SH response for 1.O6-pm excitation that exhibits a temperature dependence. A more detailed investigation of the wavelength and temperature dependencies of the SH response should provide additional insight into the nature of the metal electronic properties in the near surface region. The S H G technique has the potential to be applied to many problems in surface and interfacial sciences. The specificity to the interface, sensitivity, and time response of the technique make it a valuable complement to existing surface spectroscopiesin UHV environments. In addition, and probably more important in the long run, it provides a means to study interfacial processes in high-pressure or liquid environments. There have been several encouraging recent applications of this technique for the in situ study of electrochemical interface^.^',^^ In addition, a generalization of the SHG process, that of infrared-visible sum frequency generation, SFG, has recently been dem0nstrated.4~ By combining a tunable infrared laser with a visible laser, the surface vibrational spectrum can be measured because of the resonant enhancement of the SFG signal when the infrared laser frequency is near a vibrational frequency of the system. Although a great deal more work is required before the fundamental materials and optical physics underlying S H G and S F G are understood, these two techniques offer considerable promise for providing important new information in the future. Acknowledgment. We especially thank Dr. H. W. K. Tom for many helpful discussions and preprints of his work. We also thank D. R. Herschbach, J. L. Gland, F. M. Hoffmann, C. Friend, T. F. Heinz, and M. W. Kim for their suggestions. We also thank the reviewers for many helpful comments. Registry No. Ni, 7440-02-0; Pt, 7440-06-4; H2, 1333-74-0; 0 2 , 7782-44-7; CO, 630-08-0. (41) Richmond, G. L. Chem. Phys. Lett. 1984, 110, 571. Shannon, V. L.; Koos, D. A.; Richmond, G. L. J. Phys. Chem. 1987, 91, 5548; Appl. Opt. 1987, 26, 3579. (42) Miragliotta, J.; Furtak, T. E. submitted for publication in Surf. Sci. (43) Hunt, J. H.; Guyot-Sinnest, P.; Shen, Y. R. Chem. Phys. Lett. 1987, 133, 189. Zhu, X. D.;.Suhr Hajo; Shen, Y. R. Phys. Rev. B 1987, 35, 3047.