Reactions of Copper Group Cluster Anions with Oxygen and Carbon

Oct 1, 1994 - Masahiko Ichihashi, Charlotte A. Corbett, Tetsu Hanmura, James M. Lisy, and Tamotsu .... Vincent F. DeTuri, Paul A. Hintz, and Kent M. E...
0 downloads 0 Views 1MB Size
10023

J. Phys. Chem. 1994,98, 10023-10031

Reactions of Copper Group Cluster Anions with Oxygen and Carbon Monoxide Taeck Hong Lee and Kent M. Enin* Department of Chemistry and Chemical Physics Program, Universiv of Nevada, Reno, Nevada 89557 Received: June I I , 1994; In Final Form: July 28, 1994@

Reactions of copper group cluster anions, Cu,- (n 5 13), Ag,- (n I13), and Au,- (n I7), with oxygen and carbon monoxide are investigated using a flow tube reactor instrument. The metal cluster anions are prepared in a flowing afterglow ion source with a cold cathode dc discharge. Thermal effective bimolecular reaction rate coefficients are reported along with determinations of the product ions. The primary reaction channel is addition of one CO or 0 2 molecule to the metal cluster anion. A dramatic evedodd altemation with cluster size n is found for the reaction rates with oxygen; even n clusters (odd valence electron) anions are more reactive than odd n clusters. In the reactions of Cu,- and Au,- with carbon monoxide, larger clusters (n > 4) are more reactive than small clusters, and there is no evedodd altemation in rates. Silver cluster anions are unreacted toward CO. The effective bimolecular rate coefficients are pressure dependent for the reactions CO ( n = 7-11), Au,CO (n = 5-7), and Au,02 (n = 2, 4, and 6). Lifetimes of the Cu,cluster/adsorbate complexes are estimated from temolecular rate coefficients.

+

+

I. Introduction Clusters of the copper group metals have drawn the interest of both experimental and theoretical researchers, in part because the "alkali-metal-like" dlOsl electronic configuration of the atoms leads to relatively tractable electronic structure compared to clusters of transition metal with open d shells. For example, electron affinities,'-9 ionization potential^,'^-'^ and bond energiesI5 of copper group clusters have been measured. These physical properties of copper group clusters have been related to the jellium shell model of electronic structure.16 Clusters possessing a number of s valence electrons corresponding to a closed shell in the jellium model are found to be particularly stable with respect to electron removal. Features in the electronic band structure of the copper group clusters observed in photoelectron of the anions can also be related to the shell model. The electronic structure of the copper group clusters also plays a determining role in their chemical reactivity, the focus of this study. For example, Riley and c o - ~ o r k e r s 'have ~ found that copper cluster neutrals with closed electron shells according to the jellium model are particularly unreactive toward 0 2 . Cox et studied reactions of gold cluster cations, neutrals, and anions with hydrogen, methane, and oxygen. They found a dependence of reactivity on both size and charge state of the gold clusters; for Au,- only even n sizes reacted with oxygen. Nygren et 01.19,20 have implicated shell closing effects in the binding energies of CO to copper cluster neutral and cations. They find that the eight-electron species cU&o and cu7Co' are especially strongly bound. The reaction rates of transition metal clusters with CO are found to be relatively insensitive to either size (for n > 4), charge, or elemental composition of the cluster^;^^-^^ depending on reaction conditions, the reactivity may or may not be sensitive to the metal-CO bond strengths. Most of the available information about the electronic and geometric structures of copper group clusters comes from theory,20*25-35 Theoretical and experimental evidence indicates that the ground states of the trimer anions are linearFg The 19,21922

* To whom correspondence should be addressed. Internet: ervin@ chem.unr.edu. Abstract published in Advance ACS Abstracts, September 1, 1994. @

0022-365419412098-10023$04.5010

+

most stable structures calculated for the tetramer and pentamer anions are the rhombus and planar trapezoid, r e s p e ~ t i v e l y . ~ ~ , ~ ~ BonanEiC-Kouteckf et have recently calculated vertical excitation energies for partially optimized geometries of silver cluster anions up to n = 9 and have found good agreement with experimental photodetachment spectra. The silver clusters anions larger than five atoms assume three-dimensional structure~.~~ The kinetics of reactions of copper group metal clusters, mostly neutrals and cations, with gaseous molecules has been studied by several groups.17-22,36-43In accord with the behavior of the bulk copper group metals, copper group clusters are relatively inert compared to other transition metals. Small copper group cluster cations are typically either unreactive with water, ammonia, benzene, alkanes, alkenes, and alcohols or only undergo slow addition reactions, occasionally with metal atom loss.37~39,41,42 Morse et al.36found no evidence for reaction of neutral copper clusters ( n = 1-14) with H2. In mass spectral surveys of the reactivity of Cu,- ( n 5 10) and Au,- (n 5 7) cluster anions in our laboratory,44no reactions were observed with H2, N H 3 , C€&, and N2, but reactions were found for these clusters with CO and 0 2 . In this work, we present a detailed study of the kinetics of thermal reactions of copper, silver, and gold cluster anions with oxygen and carbon monoxide. Comparisons are made between the anionic cluster reactions and previous s t ~ d i e s ' ~ , of ~ ~reactions , ~ ~ , ~of~neutral ,~~ and cationic copper group clusters with 0 2 and CO, providing information about electronic structure effects. Investigation of all three elements in the group under identical experimental conditions makes evident periodic trends in the reactivity.

II. Experimental Models A flow tube reactor a p ~ a r a t u sis~ used ~ , ~ ~to investigate the kinetics of metal cluster anion reactions at thermal energies. The experimental and data analysis procedures are described in detail elsewherez3and will be only outlined here. Reactions are carried out in the flow tube (7.3 cm i.d. x 170 cm long), through which a buffer gas mixture (helium with 1-5% argon, 99.995% or greater purity) flows at a typical total pressure of 0.45 Torr and a velocity of about 3 x lo5cm s-l. Metal cluster ions are created in the upstream portion of the flow tube by a 0 1994 American Chemical Society

10024 J. Phys. Chem., Vol. 98, No. 40, 1994

Lee and Ervin

dc discharge ion source. The ion source has a high-voltage cathode (about -2.5 kV with respect to the grounded flow tube) fabricated from copper rod, silver foil, or gold foil (Johnson Matthey, >99.9% purity) to produce the corresponding metal clusters. Clusters are produced by sputtering by Ar+ or other cations in the discharge plasma that are accelerated toward the cathode; anionic clusters are either formed directly or by electron attachment to neutral clusters. The ions are carried downstream by the heliudargon buffer gas, undergoing lo5 collisions with the buffer gas before they reach the reaction zone. These collisions serve to cool the cluster anions to near room temperature. In the reaction zone, the neutral reagent gas, CO (Matheson, 99.99%) or 0 2 (Sierra Airgas, 99.5%),is introduced via a movable ring inlet in the flow tube. Ion-molecule reactions occur in the reaction zone while the pressure, temperature, reaction distance, and reagent flow rates are carefully controlled or measured. At the end of the flow tube, a small fraction of the reactant and product ions is sampled through a 1 mm diameter orifice. The ions are gently extracted and focused by a series of aperture lenses into a quadrupole mass spectrometer with a capability of unit mass resolution over its range of 2- 1500 amu. Reactant and product ion intensities are measured using a collision dynode/particle multiplier detector with pulse counting electronics. The kinetic rate law for the bimolecular reaction M,L products, where M,- is the metal cluster anion and L is the reactant gas, can be expressed as

250

1

1

200

150

100 50

0 200

0

400

800

800

1000

'E: 260 4

a

g

-2

200

D

150

3

; 100

-

4

E

-0

50

0

2

+

-

d [M,-] ldt = -k, [M, -] [L]

(1)

The experiment is operated under pseudo-first-order conditions, [MJ > [L] >> [Mn-]), the termolecular rate coefficient is related to the effective bimolecular rate coefficient by

n

B

v

k, = k,,[HeI

(3)

To check for termolecular behavior, the kinetic measurements are repeated as a function of buffer gas pressure over the experimentally accessible range 0.2-0.6 Torr. o

111. Results A. Mass Spectra. Mass spectra for copper, silver, and gold cluster anions produced by the cathode discharge source are presented in Figure 1. The most intense peaks in the mass distributions are Cug-, Ag3-, and AUZ-for the corresponding metals. Intensities are sufficient for kinetics measurements up to n = 13 for copper and silver; the quadrupole mass range limits measurements to n 5 7 for gold. While the gold cluster anion mass spectrum is fairly clean, some copper and silver oxide impurity ions are observed. A few small copper oxide anions (e.g., CuO-, CuOz-, and Cu302-) are about as intense as the corresponding bare copper species. Oxide formation for the larger copper and silver cluster anions is less pronounced. Since the measurement of rate constants involves the decay of the bare cluster ion intensities as a function of reagent gas concentration, oxide contaminants do not impede

2

1

1

4

e

I

,

1

e i o i z i r l

Cluiter S i z e (n)

Figure 2. Measured effective bimolecular rate coefficients for the reactions of copper cluster anions (open circles) and gold cluster anions (closed circles) with carbon monoxide at a buffer gas pressure of approximately 0.45 Torr. No reaction was observed for silver cluster anions with CO.

most of the kinetic rate measurements. In a few cases, however, the contaminants can interfere with product identifications and products of bare clusters or contaminants can overlap with the mass peak of the next larger bare cluster. B. Carbon Monoxide Reactions. Effective bimolecular rate coefficients for the reactions of copper and gold cluster anions with carbon monoxide at a buffer gas pressure of 0.45 Torr are presented in Figure 2 and Table 1. These values represent the total initial reaction rates of the bare clusters including all product channels. The given uncertainties represent the 95%

Reaction of Copper Group Cluster Anions with

0 2

TABLE 1: Effective Bimolecular Rate Coefficients for M.CO at 0.45 Torr of He

+

kn (lo-"

cluster

cm3 s-l) 1.3 f 0.2 1.6 f 0.2 1.1 f 0.2 1.4 f 0.2 2.4 f 0.3 3.1 f 0.3 5.2 f 0.5 6.1 f 1.0 4.9 f 1.0 4.5 f 1.3

cuq-

cu5cu6cu7cus-

cus-

twocull-

cu12cw3-

k, (10-lo cm3s-l)a

kdkb

6.65 6.58 6.54 6.51 6.48 6.46 6.45 6.44 6.42 6.42

0.020 f 0.002 0.030 f 0.002 0.016 f 0.002 0.020 f 0.003 0.037 f 0.004 0.048 f 0.005 0.081 f 0.007 0.09 f 0.02 0.08 f 0.02 0.07 f 0.02

6.74 6.53 6.46 6.42 6.40 6.38 6.38

0.003 f 0.001 0.003 f 0.001 0.003 f 0.001 0.004 f 0.001 0.015 f 0.001 0.020 f 0.001 0.038 f 0.002

0.20 f 0.09 0.22 f 0.09 0.18 f 0.08 0.23 f 0.11 0.96 f 0.09 1.26 f 0.08 2.43 f 0.13

AuAu2Au3Au4Au5AU.5-

Au~-

Calculated collision rate [ref 451. Reaction efficiency.

F '

-- I m

I

I

0.3

0.4

'

2

"El

-

3

I

0 Y r (

1

v-

0.0

0.1

0.2

0.6

0.8

Total Pressure (Torr)

Figure 3. Buffer gas pressure dependence of the effective bimolecular CO (n = 5-7). rate coefficients for the reactions Au,-

+

confidence level based on the standard deviation of the mean of replicate measurements taken over a period of several months. Table 1 also provides reaction efficiencies, knlk,, relative to the ~ a l c u l a t e dcollision ~~ rate based on the ion-dipole potential of the cluster anions with CO. The rates are a strong function of the size of the cluster, increasing from k x 2 x lo-" cm3 s-l for Au,- ( n = 1-4) to cm3 s-l for Cu,- ( n = 10-13). The fastest about 8 x reaction rates are still only 7-9% of the calculated45collision rates for the ion-dipole potential. The reaction rates for copper and gold roughly overlap and follow the same trend in the region of size overlap, n = 4-7. The rates for small copper clusters (n < 4) could not be measured reliably because of their inertness toward CO and because of mass overlaps from products of reactions of copper oxide impurities. In contrast to copper and gold, Ag,- (n = 1-13) cluster anions exhibit no evidence of reaction for CO flow rates up to 130 STP cm3 min-'. The reaction rates for silver cluster anions are less than our detection limit, kn I5 x cm3 s-'. The effective bimolecular rates were found to be pressure dependent for Cu,- -t CO (n = 7- 11) and Au,CO ( n = 5-7) but pressure independent for Cu,CO ( n = 4-6) and Au,CO ( n = 1-4). Figure 3 shows the pressure dependence measurements for the example of Au,- ( n = 5-7). The termolecular rate coefficients, k n ~in eq 3, are obtained by a linear regression fit to the slopes of kll versus pressure and

+

+

J. Phys. Chem., Vol. 98, No. 40, 1994 10025

and CO

+

TABLE 2: Termolecular Rate Coefficients and Complex CO He Lifetimes for M,cluskm" klZb (10-lo kzxc (10-lo ter cm3s-l) cm3s-I) cm3s-l) td s) Cu70.85 f 0.09 6.59" 5.32 2.4 f0.3 6.57" 5.32 6.4 f 0.5 cus2.2 f 0.2 6.55" 5.32 7.1 f 0.7 CUS- 2.5 f 0.3 12.4 f 1.2 6.53" 5.31 Cuio- 4.3 f 0.4 5.31 16.5 f 1.1 6.52" Cull5.7 f 0.4

+

Au5AUg-

Au~-

0.49 f 0.04 0.68 f 0.04 0.91 f 0.06

+

6.48" 6.47" 6.46"

5.31 5.31 5.31

1.42 f 0.12 1.98 f 0.12 2.7 f 0.2

"Experimental, this work. bCollision rate [ref 451, Ma- + CO. Collision rate [ref 461, M,- He. Estimated M,CO- unimolecular dissociation lifetime (see text).

+

are given in Table 2. For Au5- and Aug-, the kn intercept in Figure 3 is zero within experimental uncertainty, which indicates that the true bimolecular reaction rate is zero or very small. For Au7-, the intercept is nonzero. If real, this would indicate that a true bimolecular process is occurring, either radiative association or a product channel involving metal atom or electron loss. However, it is likely that Au7- falls in the transition region between the low-pressure and high-pressure limits in our pressure range, which would give a false intercept. Because the available pressure range is limited (0.2-0.6 Torr), any curvature in the plots would be difficult to detect. Therefore, our termolecular rate coefficients should be considered as lower limits. The effective bimolecular rate coefficients at a fixed pressure provide a guide to the relative reactivities of different clusters. For this purpose, we have chosen the convenient pressure of 0.45 Torr for reporting kII values (Table 2 and Figure 2). The primary product channel responsible for the initial reaction rates of the copper and gold cluster anions is the addition of CO to the clusters without loss of metal atoms. Both copper and gold cluster anions exhibit further reaction involving sequential addition of CO. Figure 4 displays a representative plot of the product intensities as a function of CO flow rates for the case of Aus-. The largest observed addition products at our highest CO flow rate (130 STP cm3 min-l) are cu4(co)4-, cu5(co)4-, cu6(co)5-, cu7(co)6-* cuS(c0)5-, Cug(CO)5-, CU~O(CO)S-, cU11(co)6-, and Cu12(CO)6- for the copper systems and Au3(CO)-, Auq(C0)2-, Au5(C0)2-, Aug(C0)3-, and Au7(C0)3- for gold. The sequential addition reactions are at least as slow as the initial addition of CO, and in some cases the largest product is still growing in at our highest accessible CO concentrations. Therefore, we do not claim that these largest observed products necessarily represent saturated species. Figure 4 shows that the sum of the ion intensities for the Au5- parent and all Aus(CO),- products is constant as a function of CO concentration. Thus, parent cluster ions are converted quantitatively into sequential addition products; i.e, there are no competing metal atom loss or electron loss product channels. This behavior generally holds for the copper and gold cluster anion reactions with carbon monoxide. C. Oxygen Reactions. The effective bimolecular rate coefficients for the reactions of Cun-, Ag,-, and Au,- with 0 2 at a buffer gas pressure of 0.45 Torr are presented in Table 3 and Figure 5 . Table 3 also gives reaction efficiencies, knlk,, relative to the Langevin& collision rate based on the ioninduced dipole potential. Kinetic rate data for some small copper cluster anions, CU,- ( n = 2, 3, 4, and 6), could not be obtained because of fragmentation of larger clusters or mass overlaps with Cun-104- secondary products.

10026 J. Phys. Chem., Vol. 98, No. 40, 1994

Lee and Ervin

3

I

n

-

lt!l?Ll 0

20

40

80

80

100 120

140

C O Flow Rate (STP em3 min-')

+

Figure 4. Product ion intensities for the reaction AusCO as a function of the carbon monoxide flow rate. Sequential addition of CO molecules to the cluster is shown. The solid line represents the sum of the parent cluster ion and product ion intensities. TABLE 3: Effective Bimolecular Rate Coefficients for M,,0 2 at 0.45 Torr kn (lo-" k, (10-lo cluster cm3s-l) cm3s-l)O kdkcb cu56.1 f 1.7 5.46 0.11 f 0.03 cu74.7 f 0.3 5.39 0.088 f 0.006 cu81Of3 5.36 0.19 zt 0.06 cus4.2 f 1.0 5.34 0.08 f 0.02 10f3 5.33 0.19 f 0.06 Cu105.32 0.12 f 0.03 cull6.3 f 1.8

+

Ag2-

Ag3Ag4-

AgsAg6Ag7Ag8AgsAgioAgiiAg12Agl3AuAu~Au3Au4AUSAU.5-

Au7-

6

v

IO*

Ag-

m

0.24 f 0.3 0.27 f 0.16 0.16 f 0.09 2.4 f 0.4 0.25 f 0.19 3.1 f 0.6 0.34 f 0.20 12.9f 2.5 0.32 f 0.25 8.6 f 1.1 0.5 f 0.3 7.6 f 1.1 0.6 f 0.3

5.92 5.57 5.45 5.39 5.36 5.33 5.31 5.30 5.29 5.28 5.27 5.21 5.26

0.004 f 0.005 0.005 f 0.003 0.003 f 0.001 0.044 f 0.007 0.005 f 0.003 0.059 i0.010 0.006 f 0.004 0.24 i 0.05 0.006 f 0.005 0.160 i0.020 0.009 f 0.006 0.14 f 0.02 0.012 f 0.006

50.11 f 0.03 0.19 f 0.12 50.06 f 0.03 0.34 f 0.09 50.06 f 0.04 3.2 i0.5 50.11 f 0.09

5.61 5.41 5.34 5.31 5.29 5.27 5.26

50.002 f0.0005 0.004 f 0.002 50.001 f 0.0005 0.006 f 0.002 50.001 f 0.0007 0.060 f 0.009 50.002 f 0.001

Collision rate [ref 461. Reaction efficiency. The reactions are slow in general; reaction efficiencies (Table 3) range from a high of about 25% for the fastest reaction to a low of 0.1% which is near our detection limit. The rates of reactions of the M,- cluster anions for all of the copper group metals with oxygen show a dramatic evedodd altemation with cluster size n. Even n (odd valence electron) cluster species react faster than neighboring odd n (even valence electron) species. The altemation is most pronounced and obvious for the silver system, with even n rates a factor of 10-40 larger than neighboring odd n rates for n = 3-13. Among the gold clusters, Aug- reacts particularly rapidly, at rates 30-60 times faster than Au5- and A u ~ - . Because the experimental mass range prevents rate measurements for gold clusters larger than n = 7 , we cannot determine whether this trend continues for larger gold clusters. The intensity altemation is somewhat less pronounced for small silver and gold anions (n 4). The even/

0

2

4

8

8

1

0

1

2

1

4

Cluster S i z e (n)

Figure 5. Measured effective bimolecular rate coefficients for the reactions of copper cluster anions (circles), silver cluster anions (triangles),and gold cluster anions (squares) with oxygen at a buffer gas pressure of approximately 0.45 Torr. Error bars are not plotted because of congestion, but uncertainty limits are given in Table 1. TABLE 4: Termolecular Rate Coefficients and Complex 0 2 He Lifetimes for Au.knI" klb (lo-'" k2c (10-lo cluster cm3s-l) cm3s-l) cm3s-') td s) AUI0.39 f 0.26 5.41 5.32 1.4 f 0.9 0.43 f 0.11 5.31 Am5.31 1.5 f 0.4 AU60.90 f 0.17 5.26 5.31 3.2 f 0.6 "Experimental. this work. bLangevin rate [ref 461, Au,- + 02. Langevin rate [ref 461, Au,- + He. Estimated Au,Oz- unimolecular dissociation lifetime (see text).

+ +

odd altemation is also observed for copper cluster anions, but even n clusters react faster than odd n clusters by only a factor of about 2 for the observable range of Cu,- ( n = 7- 11). In measurements of the dependence of these rates on the buffer gas pressure, we observed that Cu,- ( n = 5, 7 - 11) and Ag,- ( n = 2, 4, 6, 8-12) show no pressure dependence, but Au,- ( n = 2, 4, and 6 ) do show a pressure dependence. The pressure dependence of the bimolecular rates for those cluster anions not listed here could not be measured reliably because of their slow reaction rates. Apparent termolecular reaction rate coefficients, ~ I I I ,calculated from the slopes of k~ versus [He] according to eq 3 for the even n gold cluster anions are given in Table 4. The primary product channel observed in the oxygen reactions is addition of 0 2 units. Since we detect only the product ion masses, we cannot directly determine whether the addition is molecular adsorption or dissociative adsorption. Product channels are described in detail below for each metal system. Copper. Addition of molecular oxygen, Cu,0 2 Cu,02-, is the primary reaction channel for most copper cluster anions. The larger cluster anions ( n = 6-12) exhibit only addition of one 0 2 unit in our range of oxygen concentrations, but further reactions are observed for Cu,- ( n = 2-5) with the formation of secondary products such as Cu,O4-. The larger copper clusters also exhibit substantial fragmentation (loss of metal atoms to form smaller bare clusters or smaller cluster oxides), either upon the initial reaction with 0 2 or upon secondary reactions of Cu,Oz- products. For example, only 35-40% of the initial intensities of Cu,- ( n = 8 and 9) are converted to Cu,Oz-. Because of this fragmentation, determination of product branching ratios is difficult. In the product ion scans for the copper cluster anions, a few product species are observed with odd numbers of oxygen atoms: Cu03-, Cu2O3-, and Cu303-. However, in examining

+

-

Reaction of Copper Group Cluster Anions with

and CO

0 2

1""'"""""'l

t

I

0

, . . .

h .

...

5 0 , Flow Rete

1

1

.

.

.

.

1

1

15

10

(STP cm'min-')

+

Figure 6. Product ion intensities for the reaction Ag402 as a function of oxygen flow rate. The solid line represents the sum of the parent cluster ion and product ion intensities.

the absolute ion intensities, it is found in each of these cases that the products could arise from copper oxide impurity ions from the ion source, Le., Cu,O0 2 Cu,O3-. Thus, there is no unambiguous evidence for production of atomic oxygen in the reactions of the copper cluster anions. Silver. The major primary reaction channel for silver cluster anions is addition of oxygen molecules, Ag,0 2 Ag,Ozfor n = 2 , 4 , 6 , and 8-12. Compared with copper, little cluster fragmentation is observed. For these reactions, the decay in the reactant cluster ion intensity is accounted for by the corresponding increase in the 0 2 addition product. An example for n = 4 is shown in Figure 6. The reactions of Ag,- species with odd n are very slow and no products of the form Ag,Ozfor n = 3,5,and 7 are observed. We speculate that the products of the odd n clusters may be of the form Agn-102-, but these would be overwhelmed by the products from the much faster 0 2 addition reactions of the even n clusters. At high oxygen concentrations, large silver cluster anions exhibit secondary addition of oxygen, Ag,020 2 Ag,04for n = 9-12. For n = 10 and 11, it is observed that the loss of Ag,O2- cannot be completely accounted for by production of Ag,04-. For these two cases, the reaction Ag,O?;- 0 2 Agn-104Ag is the only other product channel that can explain the observed product intensities. Gold. The only products observed in the reaction of gold cluster anions with oxygen are Au,Oz- ( n = 2,4, and 6). The intensities of these products are completely accounted for by the decays of the corresponding even n clusters, Au,0 2 Au,02-. Odd atom gold cluster anions ( n = 1, 3, 5 , and 7) exhibit extremely slow depletion rates near our detection limit. Therefore, we report these rates as upper limits in Table 3. As with silver, no products of the form Au,Oz- with odd n are observed. Cox et ul.18 reported that odd n gold cluster anions are unreactive with oxygen, in agreement with our results. For the clusters where identification of product channels is ambiguous, reactive detachment (electron loss) is another possibility. While reactive detachment cannot be ruled out as a minor process, the ion intensity summed over all masses does not support a loss of total ion current as the 0 2 flow rate is increased.

+

+

-

+

-

+

-

+

-

+

-

IV. Discussion The copper group cluster anions are less reactive in general than clusters of the other transition metal elements. The largest effective bimolecular reaction rate observed for any of the clusters is 25% of the collision rate with 0 2 and 9% with CO.

J. Phys. Chem., Vol. 98, No. 40, 1994 10027

Furthermore, our previous survey of reactivity44 showed that copper and gold cluster anions are inert toward hydrogen, ammonia, methane, and nitrogen. This low reactivity is generally consistent with other work on copper group cluster neutrals and cations, summarized in the Introduction. The reaction efficiencies (Tables 1 and 3) may be compared to the sticking coefficients for scattering from bulk metal surfaces. For the coinage metals, sticking coefficients vary greatly depending on the system, crystal face, and conditions. The sticking coefficients are 10-3-0.5 for CO and 0 2 on c ~ p p e F ~ - ~ ~ but are as low as for silver and gold.72,73The underlying reason for the inertness of the copper group bulk metals and the copper group clusters is the same, namely, that the filledshell d orbital electrons in the copper group lie at low energy and are not easily available for interaction with reactants. In contrast, transition metals with open d shells are generally much more reactive. For example, most nickel group cluster anions with n 2 4 react on nearly every collision with CO, 0 2 , COz, and N20,23324347 and the sticking coefficients for CO on nickel group metal surfaces are around 90%.74 Striking differences are observed in the reactivity of copper cluster anions with 0 2 versus CO. While the reaction rates with oxygen display a dramatic evedodd alternation with cluster size, the predominant trend for carbon monoxide is simply an increase in rates with increasing cluster size. In the following sections, we will attempt to explain these trends in terms of the electronic structure of the clusters, the different reactive properties of 0 2 and CO, and the thermochemistry of the reactions. A. Carbon Monoxide Reactions. Copper and gold cluster anions react with CO, but no reaction is detected between silver cluster anions and CO. Castleman and co-workersZ2 have previously observed association reactions of copper cluster anions with CO. Cox et aL21 reported that neutral copper clusters react with CO, but no products were detected. The lack of detectable products was explained by unimolecular dissociation of weakly bound species before arrival at the photoionization detector. Recently, Lian et ~ 1 observed . ~ ~ association reactions of Cu2 and A u with ~ CO but reported that no reaction was found for Agz. The inertness of both neutral silver dimers and silver cluster anions suggests that this behavior may be independent of size or charge state of the clusters. To gain insight into this behavior, we will first consider what is known about the bonding of CO to copper group metal centers in organometallic complexes. Stable binary carbonyl compounds of the copper group metals as solids or in solutions are unknown, and complexes including carbonyl ligands bound to a copper group metal center are rare.48-51 AuCl(C0) can be synthesized; its carbonyl stretching frequencies are similar to free CO, and the carbonyl is easily displaced, indicating weak binding.51 M(CO),(M = Cu, Ag, Au, n = 1-2) and M(CO)3 and Mz(CO)6 (M = Cu, Ag) have been identified in rare gas matrices.53 In complexes with transition metals, carbonyl ligands are typically u electron donors and x* electron acceptor^.^^ Optimal bonding occurs for elements with partially occupied d shells, i.e., with an empty d o acceptor orbital and a filled drc donor orbital. Carbonyls are weakly bound to copper group metal centers because the d shell is full. Carbonyl u donation is possible only into an su (for MI oxidation states) or a pa metal orbital, and dx(M)x*(CO) back-bonding is unfavorable because of the low energy of the d orbitals in the coinage group metal^?^,^^ Spectroscopic characterization of the matrix-isolated carbonyls suggests that there is x back-bonding, although it is weak and partially involves px(M) molecular orbitals in addition to &(M) orbitais.53

10028 J. Phys. Chem., Vol. 98, No. 40, 1994

These ideas may be incorporated into a model of CO bonding to copper group clusters. The cluster LUMOs (lowest unoccupied molecular orbitals) are constructed from metal s orbitals. Calculated LUMO energies for Ag,- clusters are several electronvolts above the HOMO (highest occupied molecular orbital);35this should hold for copper and gold also. Therefore, electron donation from the 5a orbital of CO to the cluster LUMO will provide at best a slightly attractive interaction. Back-bonding from a metal dn orbital, or from a cluster molecular orbital of the appropriate symmetry, to the CO 2n* antibonding orbital could lead to an attractive interaction. An approximate measure of the relative d electron energies is given by the lowest dl0s1 d9s2atomic promotion energies,7511 200, 30 200, and 9160 cm-' for Cu, Ag, and Au, respectively. As pointed out by Lian et for the dimers, the low energy of the silver d orbitals compared to copper and gold means that n back-bonding is especially unfavorable for silver, preventing bonding of the silver clusters with CO. It is worth noting that this argument implies that &(M)-n*(CO) back-bonding is indeed important in the binding of CO to copper group clusters. The differences in CO interactions with copper, silver, and gold are also apparent in adsorption of CO on the bulk metal surfaces. Heats of adsorption for CO at low coverage are 5070 kJ/mol for ~ o p p e r , 27-42 ~ ~ - ~kJ/mol ~ for silver,56,60and 5558 M/mol for gold61,62(ranges refer to various crystal surfaces for which values have been measured). Thus, the binding energy of CO is significantly less on silver surfaces than on copper and gold surfaces. From these and other CO is believed to be only physisorbed on most silver crystal faces, while CO chemisorbs to copper and gold. Under the conditions of the present experiment, if CO is weakly bound (physisorbed) to Ag,- clusters, the intermediate complexes may not be sufficiently stable to survive collisions with the buffer gas. The weaker binding of CO on silver surfaces compared to copper and gold may be attributed to d band energy level differences of the metals; the electron binding energy of the d band in silver is 2 eV higher than for copper and gold.67 This is confirmed by photoelectron spectros~opy,6~ which shows that the 2n* level of CO adsorbed on copper lies close to the Fermi level, while the corresponding level of CO on silver has a large energy gap above the Fermi level. The observed CO association rates increase with cluster size (Figure 2) for copper and gold. This may be attributed either to greater metal-CO interaction energies for larger clusters or simply to increasing statistical lifetimes with more degrees of freedom (discussed below). Assuming the metal d electrons are mainly localized on atoms, the dn(M)-n*(CO) backbonding interaction will not be affected much by cluster size for a given element. The HOMO-LUMO band gap decreases to zero in the bulk limit.20 Therefore, the LUMO(M)-a(C0) interaction may be more favorable for larger clusters, although the variation may be small for the cluster size range under investigation. Following the initial association reaction, additional CO molecules are observed to adsorb sequentially onto copper and gold cluster anions. Castleman and co-workersZ2also observed sequential addition for CO on Cu,+. Given the experimental conditions for both experiments, the observed maximum numbers of CO molecules added are not necessarily indicative of true saturation limits. For small sizes (n 7), Castleman observed one or two more carbonyls on Cu,+ than the maximum number we observed on Cu,-. Because each CO donates two electrons, we would expect to add one less carbonyl to anion clusters compared to cations if the saturation limit is controlled by the electronic structure (number of available acceptor

-

Lee and Ervin orbitals). For our larger copper cluster anions (n = 7-12), the maximum number added is five or six carbonyls, which is probably an artificial maximum due to the limited experimental reaction time and CO concentrations and the slow sequential addition rates. For gold cluster anions, only two carbonyls add for n = 4 and 5 , while three carbonyls add for n = 6 and 7. The total valence electron counts for these species are roughly comparable with those found in stable organometallic complexes containing multiple gold atom^.^^,^^ Organometallic gold compounds have much lower electron counts than expected for 18-electron metals because the high-energy gold p orbitals are not involved in ligand bonding.76 B. Oxygen Reactions. The M,- copper group clusters with even n are reactive with oxygen, while those with odd n are less reactive or unreactive. The primary reaction channel for silver and gold cluster anions is addition of an 0 2 molecule to the cluster; for copper, addition is sometimes accompanied by cluster fragmentation. Our results for the gold cluster anions agree with those of Cox et u1.,l8who found that the odd n cluster anions were unreactive with oxygen. Irion and Selinger38 reported that copper cluster cations CU,+ (n 5 41) are unreactive with oxygen at thermal energies but undergo fragmentation mainly by loss of CUZwhen collisionally excited. The absence of thermal oxygen addition reactions in that work might be due to the low-pressure reaction conditions; in our experiments, collisions by the helium buffer gas stabilize the intermediate complexes. Riley and co-workers17 observed that neutral copper clusters reacted with 0 2 but that clusters with closed shells of electrons according to the jellium model were particularly unreactive. The strong evedodd reaction tendency for the reactions with oxygen is in striking contrast to the behavior for CO addition reactions. The evedodd alternations may be understood in terms of the frontier orbitals of the clusters and ~ x y g e n . ' ~ , * ~ The dengerate 2pn* antibonding orbitals of oxygen each have an unpaired electron. Oxidative addition to the cluster requires interaction of an 0 2 n* orbital with the HOMO of the cluster. In the case of even n clusters, the HOMO has an unpaired electron. Pairing this electron with the unpaired electron in an 0 2 n* orbital in the bonding combination of the two orbitals results in an attractive interaction. For odd n clusters, the HOMO is doubly occupied and the interaction with 0 2 n* is much less attractive because the additional electron must be placed in the antibonding orbital of the complex. As a result, the odd n clusters are either completely unreactive or else require a costly electron promotion in the cluster before reaction can occur. Riley and co-workers17proposed that the reactions of neutral copper clusters with oxygen occur by dissociative chemisorption promoted by electron transfer from the metal cluster to the 0 2 n* orbital, which weakens the 0-0 bond. Reactions between oxygen and transition metal surfaces78 also involve electron density transfer from the metal to the n* orbital of 0 2 , leading to 0-0 bond cleavage. The electron transfer mechanism is promoted by low cluster ionization potentials for neutral clusters17 and by low electron affinities for anionic clusters. Ionization energies1°-14 and electron affinitie~l-~ for the copper group clusters exhibit strong evedodd alternations, which reinforce the frontier orbital explanations given above for the evedodd alternation in reactivity with oxygen. That is, the clusters with an unpaired electron in the HOMO also have low electron binding energies, and both of these factors promote interaction with the unpaired electron in the 0 2 x* orbital. Our experiments cannot distinguish between molecular and dissociative adsorption reactions on the clusters. Observation

Reaction of Copper Group Cluster Anions with

0 2

and CO

J. Phys. Chem., Vol. 98, No. 40, 1994 10029

of product clusters with a single 0 atom is unexpected because buffer gas collisions either in the formation region or in the photoionization detection region, which makes the product the formation of both metal-0 atom bonds is required energetidistributions more sensitive to the stability of the ions. cally to drive the breaking of the 0 2 bond. Oxygen adsorbs dissociatively on copper surfaces, both dissociatively and The evedodd alteration in reaction rates for the reactions of molecularly on silver, but only molecularly on gold surfaces.78 the metal cluster anions with oxygen can be explained by Heats of dissociative adsorption of oxygen on the metal surfaces molecular orbital interactions for one or two electrons in the are -326 kJ/mol for copper, -62 kJ/mol for silver, and +54 HOMO, as described above. However, we observe only weak indications of special effects in the chemical kinetics due to kJ/mol for gold.78 Assuming these thermochemical values hold jellium shell closings. The eight-electron M7- clusters have approximately for the cluster anions, we would expect oxygen low reactivity with oxygen, but no more so than neighboring to be dissociatively adsorbed on copper clusters and possibly odd n cluster anions. The reaction rate constant for Aggalso on silver clusters, but molecularly on the gold clusters. The 0 2 is the largest of all of the measured oxygen reactions. This large exothermicity for copper explains the extensive fragmentahigher reactivity could be attributed to the single electron in a tion channels observed in reactions of copper cluster anions with new shell, which leads to a larger HOMO and a lower electron oxygen. Collisions with the buffer gas may be insufficient to binding energy, both of which promote interaction with Oz(n*). remove that much excess energy before fragmentation occurs, Unfortunately, rate data could not be obtained for c U 6 - and especially for the second addition of 0 2 . The exothermicity Au,- (n > 7), preventing verification of this effect for the copper may also explain the relatively fast reactivity of Cu,- with odd and gold systems. Winter et al.17reported that copper cluster n (only half the rate of the even n copper cluster anions) cations with closed electron shells according to the jellium model compared to that of odd silver and gold cluster anions. The show particular inertness in the cluster reactivity toward 0 2 . higher exothermicity could drive electronic rearrangement Nygren et al.19have shown experimentally that Cu7CO+ and needed for an attractive interaction with the 0 2 n* orbital. We Au7COf possess greater stability relative to other M,CO+ observe that gold cluster anion add oxygen only in a termospecies, which they attribute to the eight-electron closed shell lecular process, while addition of 0 2 to copper and silver clusters counting the copper s valence electrons and two electrons is in the high-pressure limit. This behavior is consistent with donated by the CO ligand. Nygren and SiegbahnZ0calculated a lower M,O2- complex binding energy for gold and therefore CO binding energies for partially optimized neutral and cationic with molecular rather than dissociative adsorption. copper cluster geometries and found that the eight-electron C. Shell Effects. Many properties of the copper group species CU&o and Cu7CO+ have higher binding energies than clusters follow trends that have been explained by the jellium neighboring sizes. The corresponding closed-shell species shell model,16which treats the clusters as spheres or ellipsoids among the anionic clusters would be MsCO-. Figure 2 shows of uniform positive charge that form an electron well with that the rate constant for CugCO Cu5CO- is a local quantized energy levels. The jellium model predicts shell maximum and that the rate constants for Au,CO have a closings for spherical clusters at total valence electron counts sharp step between n = 4 and n = 5. In the copper cluster of 2, 8, 18, 20, 34,40, 58, and others.16 For example, electron cation reactions with CO investigated by Castleman and coaffinitie~l-~ and ionization potentials1°-14 indicate that copper workers,22a step in the rates between CU6’ and CUT+is evident. group clusters exhibit particularly high electron binding energies These features could be residual effects from the special for M7- and Mg, which have closed shells of eight valence stabilitiesof the eight-electron cluster carbonyl species; however, electrons. This behavior is analogous to the electron shell they are not distinct enough to be convincing as magic numbers. closings responsible for the high ionization potentials of rare As suggested by Nygren et al.,19 the absence of distinct shellgas atoms. An electron in the doubly occupied HOMO of a closing effects in the reactivity data is likely due to a lack of closed-shell cluster feels a stronger effect core potential, and sensitivity of the kinetics of the reaction to the stability of the thus has a larger electron binding energy, than an electron in a CO adduct. To the extent that the rates of the reactions reflect singly occupied HOMO because of less effective screening by the activation energy for addition of CO, these appear not to same-shell electrons. In this section, we discuss some of our be especially sensitive to the exothermicity of the overall observations from the viewpoint of the jellium model. reaction. That further implies that the activation energies Strong evedodd intensity altemation in mass spectra has been correspond to an early transition state in the adsorption process observed for fast atom bombardment (FAB) production of Ag,+ that is not greatly affected by final product stabilities. On the (n = 1-16)41 and of copper, silver, and gold cluster anions79 other hand, as discussed in the next section, the termolecular and cationss0for sizes up to n = 20-40. In the latter s t u d i e ~ , 7 ~ - ~ ~association rates can be influenced by the stability of the intensity anomalies (“magic numbers”) were reported for intermediate complex. expected shell closings of 2, 8, 18, 20, ... electrons, including D. Pressure Dependence. The pressure dependence of the M9+ and M7- (eight electrons). Yamada and Castlemansl also bimolecular rate coefficients indicates that there is a competition reported magic numbers for M9+ and M21+ (M = Cu, Ag) from between collisional stabilization of the cluster-adsorbate coma gas aggregation source with photoionization detection. In our plex by the buffer gas and decomposition back to reactants. mass spectra, which are similar to those of Lineberger and coWe can model the lifetime of the intermediate complexes using w o r k e r ~using ~ ~ ~the same type of ‘source, the heptamer cluster the simple Lindemann mechanismg2 for association reactions ions, Cu7- and Ag7-, have particularly strong intensities relative given by eqs 4 and 5 . to neighboring sizes, while the hexamers, Cu6- and Ags-, are weak. These intensity anomalies could be due to the special stability of the eight-electron heptamer anions according to the (4) jellium model; however, the trend is not followed in the gold mass spectra, in which the hexamer anion has a larger intensity k2 M,L-* He M,LHe than its neighbors. In any case, the mass spectral intensities are a function of not only cluster thermochemical stabilities but also the kinetics of the ion source. The ion sources that exhibit Using the steady state approximation for the metastable magic numbers in mass spectra share an absence of stabilizing complexes, the observed termolecular rate is given by km =

+

+

+

-

-

+

+

Lee and Ervin

10030 J. Phys. Chem., Vol. 98, No. 40, 1994 Vibrational degrees of freedom 0

6

12

18

30

24

36

I . ' . ' " - ' ' " I

1

.I

0 cu,(co)-

10-9

0

2

4 6 6 Cluster size (n)

1

0

I

Z

Figure 7. Estimated lifetimes of the cluster-ligand complexes,

calculated from measured termolecular rate coefficients as described in the text. Lifetimes are plotted as a function of cluster size (lower axis) or the total Vibrational degrees of freedom (upper axis).

+

(klkz)/(k-1 k2[He]). In the low-pressure limit, k2[He] > k-1, collision stabilization is much more rapid than the unimolecular decomposition, and the effective bimolecular rate coefficient k11 is independent of the pressure of the helium gas, kn = kn~[He]= k l . This implies that the rate coefficients should equal the collision rate for the reactions in which the effective bimolecular rates are independent of pressure. Actually, the reaction efficiencies in the highpressure limit are less than unity even for the fastest reaction rates measured. In that case (kl < kc), the M,CO- lifetimes calculated according to the simple Lindemann model are too short. Castleman and co-workersZ2incorporated an arbitrary "sticking coefficient", kl = 0. 19kc,into their RRKM model for the Cu,+ CO reactions. Because our data set is limited with regard to the number of different sizes, we do not believe an RRKM analysis or determination of a sticking coefficient is warranted. Instead, we will discuss the lifetime trends while recognizing that the absolute values may be low. The classical RRK model for unimolecular dissociation yields lifetimes given by

+

where Y is a characteristic vibrational frequency, E is the total energy of the complex, EOis dissociation energy of the bond to be broken, and s is the number of vibrational degrees of freedom. Cox et have developed this model of cluster chemisorption in greater detail. While the simple RRK model is certainly inadequate for a quantitative description of the lifetimes, it

provides an appropriate physical picture. In Figure 7, we have plotted the logarithm of the experimental lifetimes as a function of clusters size or, equivalently, the total number of vibrational degrees of freedom, 3n - 6. If the ratio ( E - &)/E is approximately the same for all of the clusters, this plot should be linear according to the RRK model. In this case, EO is the binding energy of the adsorbate molecule to the cluster, and E is equal to EOplus the initial thermal vibrational energy of the cluster.21 The lifetimes for the carbon monoxide adducts do follow an approximately linear trend, consistent with this model and roughly equal CO binding energies for all the clusters. The oxygen adducts for gold cluster anions do not lie on the same line, as expected because the binding for oxygen and carbon monoxide is quite different and because the number of active degrees of freedom may differ. Our measured termolecular rates for association of Cu,- (n = 7-1 1) with CO are 2-10 times smaller than those found by Castleman and co-workers22for Gun+ clusters of the same size. This could partly be an artifact since our termolecular rates represent lower limits, as discussed above. However, over this size range, ow termolecular rates are increasing with increasing cluster size, while the cations are already in the high-pressure limit for n > 7. This implies that anionic clusters bind CO less strongly and therefore have shorter complex lifetimes than cationic clusters of the same size. The calculations of Nygren and SiegbahnZ0indicate that CO is more strongly bound to cationic copper clusters than to neutral copper clusters, for the same size or the same number of valence electrons. In the context of the molecular orbital description of the CO bonding (discussed above), u(C0) electron donation into the cluster LUMO would be favored for electron-deficient clusters, giving the following order for the bond strength with CO: cations > neutrals > anions. It would be interesting to test this conclusion with desorption energy measurements or with theoretical calculations for the anions.

V. Conclusions From flow tube reactor kinetic measurements, we have investigated the reactivity of coinage metal cluster anions with oxygen and carbon monoxide. The copper group cluster anions are relatively unreactive, with reaction efficiencies of 25% or lower for 0 2 and 9% or lower for CO. The low reactivity compared to other transition metals can generally be attributed to the stability of the full d electron shell in the copper group atoms. In the oxygen reactions, dramatic even/odd alternations in reactivity are observed. Even atom (odd valence electron) cluster anions react faster than odd atom (even valence electron) clusters. Copper and gold cluster anions react with CO, but silver cluster anions do not. The general trend for copper and gold is that larger size cluster anions show faster reaction rates with CO. These variations in reactivity can be understood by examining the molecular orbitals involved in the reactions. CO binds to the clusters by u donation into the cluster LUMO and ~d back-bonding from metal d electrons, both of which are fairly weak interactions that vary smoothly as a function of cluster size. The inertness of silver clusters toward CO can be ascribed to the much lower-lying d orbitals in silver compared to copper and gold. In contrast to carbon monoxide, oxygen interacts with the cluster HOMO through the unpaired electron in an 0 2 n* antibonding orbital. This interaction is attractive for odd electron clusters with an unpaired electron in the HOMO but is unfavorable for even electron clusters.

Reaction of Copper Group Cluster Anions with

0 2

and CO

The primary product channel for both the oxygen and carbon monoxide reactions is addition of the molecule to the cluster. From the pressure dependence of the effective bimolecular rates, we have estimated lifetimes of the cluster-adsorbate complexes. The lifetimes exhibit an exponential dependence on cluster size, as expected from simple statistical models.

Acknowledgment is made to the donors of the Petroleum Research Fund, administered by the American Chemical Society, for support of this research. We thank Dr. Paul A. Hintz for helpful discussions. References and Notes (1) Zheng, L. S.; Karner, C. M.; Brucat, P. J.; Yang, S. H.; Pettiette, C. L.; Craycraft, M. J.; Smalley, R. E. J. Chem. Phys. 1986, 85, 1681. (2) Pettiette, C. L.; Yang, S. H.; Craycraft, M. J.; Conceicao, J.; Laaksonen, R. T.; Cheshnovsky, 0.; Smalley, R. E. J. Chem. Phys. 1988, 88, 5377. (3) Taylor, K. J.; Pettiette-Hall, C. L.; Cheshnovsky, 0.;Smalley, R. E. J. Chem. Phys. 1992, 96, 3319. (4) Leopold, D. G.; Ho, J.; Lineberger, W. C. J . Chem. Phys. 1986, 86, 1715. ( 5 ) Ho, J.; Ervin, K. M.; Lineberger, W. C. J . Chem. Phys. 1990, 93, 6987. (6) Gantefor, G.; Gausa, M.; Meiwes-Brcer, K.-H.; Lutz, H. 0. Faraday Discuss. Chem. Soc. 1988, 86, 1. (7) Begemann, W.; Hector, R.; Liu, Y. Y.; Tiggesbaumker, J.; MeiwesBroer, K.-H.; Lutz, H. 0. Z. Phys. D 1989, 12, 229. (8) Gantefor, G.; Gausa, M.; Meiwes-Brcer, K.-H.; Lutz, H. 0.J . Chem. Soc., Faraday Trans. 1990, 86, 2483. (9) Gantefor, G. F.; Cox, D. M.; Kaldor, A. J. Chem. Phys. 1992, 98, 4102. (10) Powers, D. E.; Hansen, S. G.; Geusic, M. E.; Michalopoulos, D. L.; Smalley, R. E. J . Chem. Phys. 1983, 78, 2866. (11) Jackschath, C.; Rabin, I.; Schulze, W. Z. Phys. D 1992, 22, 517; Ber. Bunsen-Ges. Phys. Chem. 1992, 96, 1200. (12) Zhao, J.; Chen, X.; Wang, G. Phys. Left. A 1994, 189, 223. (13) Knickelbein, M. B. Chem. Phys. Left. 1992, 192, 129. (14) Alameddin, G.; Hunter, J.; Cameron, D.; Kappes, M. M. Chem. Phys. Lett. 1992, 192, 122. (15) Jarrold, M. F.; Creegan, K. M. Int. J. Mass Specfrom. Ion Processes 1990, 102, 161. (16) Cohen, M. L.; Chou, M. Y.; Knight, W. D.; de Heer, W. A. J . Phys. Chem. 1987, 91, 3141. (17) Winter, B. J.; Parks, E. K.; Riley, S. J. J . Chem. Phys. 1991, 94, 8618. (18) Cox, D. M.; Brick”, R.; Creegan, K.; Kaldor, A. Z. Phys. D 1991, 19, 353. (19) Nvaen. M. A.: Sieebahn.’ E. M.: Jin., C.:, Guo. T.: Smallev. ,. R. E. J. Chem. kiys. 1991, 95, 6T81. (20) Nvmen. M. A.: Sieabahn. E. M. J. Phvs. Chem. 1992. 96. 7579. (21) Cox, D. M.; Reich-ann, K. C.; Trevor,-D. J.; Kaldor, A. J . Chem. Phys. 1988, 88, 11 1. (22) Leuchtner. R. E.: Harms. A. C.: Castleman. A. W.. Jr. J . Chem. Phys. i990, 92, 6527. (23) Ren. X.: Hintz. P. A.: Ervin. K. M. J. Chem. Phvs. 1993.99.3575. (24) Hintz, P. A.; Ervin, K. M. J . Chem. Phys., 1994, 100, 5715. (25) Kouteckf, J.; Fantucci, P. Chem. Rev. 1986,86539 and references therein. (26) Baetzold, R. C. J . Chem. Phys. 1978, 68, 555. (27) Post, D.; Baerends, E. J. Chem. Phys. Lett. 1982, 86, 176. (28) Had, J.; Igel-Mann, G.; Preuss, H.; Stoll, H. Chem. Phys. Len. 1984, 90, 257. (29) Bauschlicher, C. W., Jr.; Langhoff, S. R.; Partridge, H. J . Chem. Phys. 1989, 91, 2412. (30) Bauschlicher, C. W., Jr.; Langhoff, S. R.; Partridge, H. J . Chem. Phys. 1990, 93, 8133. (31) Erkoc, S. Chem. Phys. Lett. 1990, 173, 57. (32) Balasubramanian, K.; Liao, M. Z . Chem. Phys. 1988, 127, 313. (33) Balusubramanian, K.; Feng, P. Y. Chem. Phys. Left. 1989, 159, 452. (34) Liao, D. W.; BalasubramMan, K. J . Chem. Phys. 1992,97,2548. (35) BonanEiC-Kouteckf, V.; CeSpiva, L.; Fantucci, P.; Pittner, J.; Kouteckf, J. J. Chem. Phys. 1994, 100, 490. I

J. Phys. Chem., Vol. 98, No. 40, I994 10031 (36) Morse. M. D.: Geusic. M. E.: Heath. J. R.: Smallev. R. E. J . Chem. Phys. 1985, 83, 2293. (37) Buckner. S. W.: Gord. J. R.: Freiser. B. S. J. Chem. Phvs. 1988. 88,‘3678. (38) Irion, M. P.; Selinger, A. Chem. Phys. Lett. 1989, 158, 145. (39) Irion, M. P.; Schnable, P.; Selinger, A. Ber. Bunsen-Ges. Phys. Chem. 1990.94, 1291. (40) Gord, J. R.; Bemish, R. J.; Freiser, B. S. Int. J. Mass Specfrom. Ion Processes 1990, 102, 115. (41) Shatpe, P.; Cassady, C. J. Chem. Phys. Left. 1992,191, 111; 1992, 197, 338. (42) Cheeseman, M. A,; Eyler, J. R. J . Phys. Chem. 1992, 96, 1082. (43) Lian, L.; Hackett, P. A.; Rayner, D. M. J. Chem. Phys. 1993, 99, 2583. (44) Ren, Xiaoli; M. S. Thesis, University of Nevada, Reno, 1992. (45) Su, T.; Chesnavich, W. J. J . Chem. Phys. 1982, 76, 5183. (46) Gioumousis, G.; Stevenson, D. P. J. Chem. Phys. 1958, 29, 294. (47) Hintz, P. A,; Ervin, K. M. To be published. (48) Bruce, M. I. J . Organomet. Chem. 1972, 44, 209. (49) Mingos, D. M. P. In Dictionary of Organometallic Compounds; Buckingham, J., Ed.; Chapman and Hall: London, 1984; Vol. 1, pp 1-10, 188-203, 569-582. (50) Van Koten, G.; Noltes, J. G. In Comprehensive Organometallic Chemistry; Wilkinson, G., Stone, F. G., Abel, E. W., Eds.; Pergamon: Oxford, 1982; Vol. 2, p 709. (5 1) Puddephatt, R. J. In Comprehensive Organometallic Chemistry; Wilkinson, G., Stone, F. G., Abel, E. W., Eds.; Pergamon: Oxford, 1982; Vol. 2, p 765. (52) Mingos, D. M. P. In Comprehensive Organometallic Chemistry; Wilkinson, G., Stone, F. G., Abel, E. W., Eds.; Pergamon: Oxford, 1982; Vol. 2, p 1. (53) (a) Huber, H.; Kundig, P.; Moskovits, M.; Ozin, G. A. J . Am. Chem. SOC. 1975, 97, 2097. (b) McIntosh, D.; Ozin, G. A. J . Am. Chem. SOC. 1976, 98, 3167. (c) McIntosh, D.; Ozin, G. A. Inorg. Chem. 1977, 16, 51. (54) Horn, K.; Hussain, M.; Pritchard, J. Surf. Sci. 1977, 63, 244. (55) Papp, H.; Pritchard, J. Sug. Sci. 1975, 53, 371. (56) McElhiney, G.; Papp, H.; Pritchard, J. Surf. Sci. 1976, 54, 617. (57) Tracy, J. C. J . Chem. Phys. 1972, 52, 2748. (58) Truong, C. M.; Rodriguez, J. A.; Goodman, D. W. Surf. Sci. Lett. 1992, 27, L385. (59) Hollins, P.; Pritchard, J. Surf. Sci. 1979, 89, 486. (60) Doyen, G.; Ertl, G. Surf. Sci. 1977, 60, 397. (61) McElhiney, G.; Pritchard, J. Surf. Sci. 1976, 60, 397. (62) Kottke, M. L.; Greenler, R. G.; Tompkins, H. G. Surf. Sci. 1972, 32, 231. (63) Lindgren, S.A.; Paul, J.; Wallden, L. Chem. Phys. Lett. 1981, 84, 487. (64)Watanabe, M.; Wissmann, P. Surf. Sci. 1984, 138, 95. (65) Trapnell, B. M. W. Proc. R. SOC. London 1953, A218, 566. (66) Bradshaw, A. M.; Pritchard, J. Proc. R. SOC. London 1970, A316, 169. (67) Duckers, K.; Bonze], H. P. Surf. Sci. 1989, 213, 25. (68) Krause, S.; Mariani, C.; Prince, K. C.; Horn, C. Surf. Sci. 1984, 138, 305. (69) Nesbitt, A.; Lewin, A. K.; Hodgson, A. J . Phys. Condens. Matfer 1991, 3, s71. (70) Ge, Q.; Wang, L.; Billing, G. D. Surf. Sci. 1992, 277, 237. (71) Harendt, C.; Goschnick, J.; Hirschwald, W. Surf. Sci. 1985, 152/ 153, 453. (72) Elliott, G. S.; Miller, D. R. Gas Dyn., Proc. Int. Symp. 14th 1984, 1, 349. (73) Spmit, M. E. M.; Kleyn, A. W. Chem. Phys. Lett. 1989,159, 342. (74) D’Evelyn, M. P.; Steinriick, H. P.; Madix, R. J. Surf.Sci. 1987, 180, 47. Wang, Y. W.; Lin, J. C.; Engel, T. Sug. Sci. 1993, 289, 267. Harris, J.; Luntz, A. C. J . Chem. Phys. 1989, 91, 6421. (75) Moore, C. E. Atomic Energy Levels; National Bureau of Standards: Washington, DC, 1952. (76) Owen, S. M. Polyhedron 1988, 7, 253. (77) For a discussion of the application of electron counting rules to clusters, see refs 23 and 24 and references therein. (78) Rao, C. N. R.; Kamath, P. V.; Yashonath, S. Chem. Phys. Left. 1982, 88, 13. (79) Katakuse, I.; Ichihara, T.; Fujita, Y.; Matsuo, T.; Sakurai, T.; Matsuda, H. Inf. J . Mass. Spectrom. Ion Processes 1986, 74, 33. (80) Katakuse, I.; Ichihara, T.; Fujita, Y.; Matsuo, T.; Sakurai, T.; Matsuda, H. Inf. J. Mass Spectrom. Ion Processes 1985, 67, 229. (81) Yamada, Y.; Castleman, A. W., Jr. J . Chem. Phys. 1992,97,4543. (82) Steinfeld, J. I.; Francisco, J. S.; Hase, W. L. Chemical Kinetics and Dynamics; Prentice Hall: Englewood Cliffs, NJ, 1989.