Ionization potentials of oxidized copper clusters

Ionization potentials of oxidized copper clusters

6 June 1997 CHEMICAL PHYSICS LETTERS ELSEVIER Chemical Physics Letters 271 (1997) 61-66 Ionization potentials of oxidized copper clusters O John ...

418KB Sizes 0 Downloads 35 Views

6 June 1997

CHEMICAL PHYSICS LETTERS

ELSEVIER

Chemical Physics Letters 271 (1997) 61-66

Ionization potentials of oxidized copper clusters O

John L. Persson, Mats Andersson, Lotta Holrngren, Thorbjtirn Aklint, Arne RosEn Department of Physics, Chalmers Unit,ersi~ of Technology and Ggteborg University, S-412 96 Gi~teborg, Sweden Received 3 June 1996; in final form 7 April 1997

Abstract

The ionization potential (IP) of CunO 2 clusters (n = 15-46) has been measured and the shift in IP induced by the adsorption of the 02 molecule evaluated. There are no large discontinuities in the IP of Cu,O 2, in contrast with the electronic shell closings observed for pure Cu n, but the even-odd alternation in the IP persists upon oxidation. On average, the IP increases after oxidation, with larger shifts for clusters where the pure Cu, have low IP (n = 21, 41) and small or negative shifts for clusters originally having high IP (n = 30, 34, 40).

1. Introduction The reactivity of metal clusters is one of the most important properties for any application based on metal clusters, as it governs the stability as well as possible catalytic properties. The reactivity is also of fundamental interest in helping us to understand such diverse subjects as the electronic structure of pure metal clusters, the geometric structure of pure and reacted species, reaction and dissociation dynamics, etc. The study of copper oxide clusters is part of our broader programme, aimed at understanding the reactivity and catalytic potential of small metal clusters. Cu was chosen because of its interesting electronic structure, with the clusters adhering well to the spherical jellium model for the electronic levels [ 1 5]. We have previously measured the sticking probability of oxygen molecules on the Cu clusters and found a direct dependence on the shell structure [6,7], in accordance with previous findings [8]. Knowing that the shell structure influences the initial adsorption process, we are now interested in finding

out whether the electron counting rules of the shell model and the spheroidal geometries of pure Cu clusters survive the addition of adsorbed oxygen. Neither the geometry nor the electronic structure can be directly observed, but we hope to draw some indirect conclusions through measurements of the ionization potentials of copper oxide clusters.

2. Experimental The experimental apparatus has been described in previous papers [6,7]; here we will only give a brief description. A pulsed beam of pure Cu clusters (carrier gas He) is produced in a laser vaporization source, cooled with liquid nitrogen. After expanding into vacuum and passing through a skimmer, the cluster beam enters a 50 mm long reaction cell, filled with oxygen gas at 1.5 X 10 - 3 o r 2.5 x 10 - 3 mbar. The dominant reaction products are of type CunO 2. Few C u , O 4 clusters are produced at this oxygen pressure, and no odd-numbered oxide products are

0009-2614/97/$17.00 Copyright © 1997 Elsevier Science B.V. All rights reserved. PH S 0 0 0 9 - 2 6 1 4 ( 9 7 ) 0 0 4 2 7 - 2

62

J.L. Persson et al. / Chemical Physics Letters 271 (1997) 61-66

ever detected. The reaction products and the remaining pure clusters are detected by laser ionization and time-of-flight mass spectrometry. For the determination of ionization cross sections, we used two lasers: a 6.4 eV ArF excimer laser as reference, and a tunable, frequency-doubled dye laser for the measurements. The two lasers were fired on every other pulse from the cluster source (for normalization purposes), and the signal from the detector was switched accordingly into two separate digitizers, one accumulating the 6.4 eV reference mass spectrum, and one accumulating the dye-laser mass spectrum. Mass spectra were recorded at integer or half-integer nm intervals in the 240-206 nm wavelength range, i.e. in the 5.2-6.0 eV photon-energy range. By normalizing each mass peak in the dye laser spectrum to its corresponding 6.4 eV spectrum, we obtained ionization cross sections normalized to the 6.4 eV cross section, for each Cu,O 2 cluster and each Cu,, cluster.

3. Results and discussion Some examples of photoionization cross section curves are shown in Fig. 1. In these curves the ionization potential, marked with a vertical bar, is determined as the linear extrapolation to the baseline. The linear extrapolation method is the best choice when information crucial to the modelling of the threshold (such as the internal temperature and the density of states) is not known. Disadvantages are that it contains a subjective element and that the threshold values determined with this method are sensitive to the internal energy of the cluster. The latter is of special concern in this experiment. The time between reaction and ionization, 100 I~s, is long enough for the clusters to reach vibrational equilibrium, but with only a few (2-3) collisions in the reaction cell, most (or all) of the chemisorption energy of the reaction will still remain in the Cu~O 2 clusters. Previous IP determinations using the linear extrapolation method have shown that a higher internal energy results in an apparently lower IP [9]. Thus, the IP values determined here will describe the properties of Cu~O 2 as formed by a single-collision reaction and will most likely be lower than the zero K vertical IP. However, the variations in IP with

cluster size will still reflect the size evolution of the electronic properties. The magnitude of the chemisorption energy depends on whether or not the oxygen molecule dissociates on the cluster surface. In the size range studied here the clusters are large enough that we assume bulk behavior in this respect, that the adsorbed oxygen molecules will dissociate. The chemisorption energy of oxygen molecules dissociating on bulk Cu surfaces is not precisely known, one experimental estimate is 2.9 eV (from the chemisorption energy of oxygen atoms on Cu, 90 kcal/mol (theory), 97 kcal/mol (exp.) [10] and the 02 dissociation energy: 5.11 eV). The temperature T of the CunO 2 clusters can then be estimated from A E = 3nkaAT (with A E = 2.9 eV, k B = Boltzmann constant). Assuming that the pure clusters are equilibrated at the cluster source temperature (80 K), we get a temperature of 500 K for Cu3002. Smaller clusters with fewer vibrational modes will be hotter and larger clusters colder. This is only a rough estimate since it is entirely based on copper bulk properties. As discussed in conjunction with the reactivity measurements [7], the exothermic oxidation could lead to fragmentation of the clusters. Since the Cun-O 2 reaction probability and possibly also the chemisorption energy vary with cluster size, one would expect a preferential fragmentation of highly reactive clusters, but in the mass spectra we do not observe any size variations in the cluster abundances that could indicate significant fragmentation. The ionization potentials of Cu~O 2 (n = 15-46) and Cu n (n = 15-41) are shown in Figs. 2 and 3, respectively. The error bars indicate the accuracy with which the linear extrapolations could be done. Some of the pure Cu n IPs are at the lower edge of the photon energy range studied here and, thus, the determinations are less accurate. The size range feasible for study is at small sizes limited by a low product formation rate combined with a high deflection probability in the reaction cell, ionization potentials expected at or above the upper edge of the available photon energy range and an even more serious influence from the chemisorption energy. IP determinations for larger clusters become less accurate because of limited mass resolution, due to the isotope distribution of Cu, and possible interference from products with several 02 molecules adsorbed.

63

J.L. Persson et a l . / Chemical Physics Letters 271 (1997) 6 1 - 6 6

o~;,5o~

oO 'Cu2~02

!

I

° °°

Cu27O2

°

Cu22;2~° oooO~.OoSo

[

CuI7 02

r.,'3

-~u~

=

° ]

Cu340 2

1oooooo /. °

1o. o:)

Cu2402

~°-, o~

Cu3502 .=°

Cu410a

° °

_ ;.o.--:°°°°

--Cu3002



Cu3602

~o

% oo *g

o I oooo° °

°

Cu4202

1o]

> ¢Z

%°=0°

I o,o~o°°°°

cu2~2 °

&oo°o°,-oooo

o.5- Cu1802

Cu4002 oo ~,°

°o °

°

c-, 0

° 2o

Cu2802

* %**°"~° °

o~ Cu3902

.... ~,~oo°°°~°:°°°°:

0.5~ CUl602

0

Cu3302

oo o°]

° *° °°o.~,e*° = °%°° °. 0.5-



=

CUl9 02

_.~1°** °o~.°oOo:::~ - -

CU2502

0.5-~

CU20 0 2

°

_. 5 ooo,o,',°°°'°

CH3102

°°°"°

/

° ° °,°

_._.~a~** °°°'~'~°°~'el~

-Cuz6O2

e. ] .

. ,~o**: Cu43 02

Cu3"~O 2

=*'*~'°*"

~

~.~

•,~

,

°~*

•~ °°°°"%

Is:2g:~~s:6=~:88:o 0"5 1 CU|9 /

Cu24

_5oo~O~o..~° s2

512

s4

s.~

s:a

,=~ Cu29

5.4

o.

°~°°°°

Cu380 2

Cu32Oz

5.oooo.

516

, Cu34 ° /

* Cu4402 I

,e ,oO~°

°°°°°

~ ^

518 610 5.2 ~

o~

-

514

516

518

oO C1.139 o i

oo° =2 i

,9°°°°°° I o;o :i ~ I ..... ~oo°° t - . l...--o° o , l~.oo ~°°°~°° ~......._--o~oo.2_~ _ _ ~ ; ° L

6.os.2

s.4

s:6

s:a

6:os.a

s.4

s.6

s.a

sos2

s.4

s.6

sa

sos2

s4

610

s6

sa

6.o

N ,i

Photon Energy (eV) Fig. 1. P h o t o i o n i z a t i o n cross sections for c o p p e r o x i d e clusters in the photon e n e r g y range 5.2-6.0 eV, n o r m a l i z e d to the photoionization cross section at 6.42 eV photon e n e r g y . T h e ionization potentials are d e t e r m i n e d as a linear extrapolation to the baseline and the arrows indicate the a s s i g n e d values. D a t a for a f e w pure c o p p e r clusters are included for c o m p a r i s o n .

In Fig. 4, the shift (or difference) in ionization potential of the CunO 2 compared to the Cu~ clusters is plotted. Ionization potentials of pure Cu, clusters have also been measured by Knickelbein [4], and our data for the pure clusters compare well with those results, with only a few data points not within the combined error bars of the two experiments. The differences are most likely due to the subjective process of extrapolating the thresholds to the baseline and to the fact that the experimental conditions, although similar, were not identical.

The electronic shell closings, according to the spherical-jellium model, can be observed in the IP of the pure clusters. The closed-shell clusters, Cul~, Cu2o, Cu34 and Cu40 in Fig. 3, exhibit a high IP, whereas the IP of the open-shell clusters is lower and especially at Cu20-Cu2~ and Cu40-Cu4~ there are large drops in the IP. In the measurement of the 0 2 sticking probability on Cu, [6,7] we also found a strong correlation to the shell closings of the spherical jellium model. A low sticking probability was measured for the closed-shell clusters, whereas the

J.L Persson et al. / Chemical Physics Letters 271 (1997) 61-66

64 I

I

I

I

I

I

I

0.6

5,8-

"~

5.6-

~

5.4-

"

I

I

I

I

I

I

40

45

0.4-

0.2-

1ri ll- I !

5.2-

"]

-0.2 -

-0.4

,'5

20 '

25'

3'0

;5

40 '

45'

Cluster size - n

open-shell clusters exhibited a high reactivity. These reactivity characteristics are believed to be caused by the lower ability of clusters with closed shells and high IP to donate charge to the adsorbing molecule and form strong bonds. One might have expected to find shell closings (possibly shifted as a function of size) in the ionization potentials of the oxides as well, but as can be seen in Fig. 2 there are no striking discontinuities. The dominating feature is rather the persistent e v e n odd variation. There is, however, a separation into two regions: for n < 21 the absolute IP values are higher and also the e v e n - o d d variations are larger I

I

I

I

i

i

I

Cun

5.8-

r.~ m

20'

25 '

;0

35 '

C l u s t e r size - n

Fig. 2. The ionization potentials of CunO2 as a function of n.

6

15 '

5.6-

e~ ~ 5.4-

] 5.2-

i 15

Cluster size - n

Fig. 3. The ionizationpotentials of Cu, as a function of n.

Fig. 4. The shift in ionization potential upon adsorption of an 02 molecule, i.e. IP(CunO2)-IP(Cu,).

whereas the IP of larger C u , O 2 varies within a relatively narrow range, 5.35-5.60 eV. If the oxygen atoms reside on the surface of the cluster, the disappearance of shell structure can be easily understood in terms of the strong perturbation on the spherical degeneracy that this would cause. The closed-shell pure Cu clusters gain their high stability and high IP from a spherical geometry, resulting in highly degenerate electronic energy levels and large separations between these levels. In the C u , O z clusters, electrons are localized at the oxygen atoms. This negative charge breaks the spherical symmetry of the cluster, and the distinct shell structure is removed. The amount of charge transfer from the clusters to the oxygen would also affect the shell structure, since the bound electrons cannot contribute to the shell filling. With no clear structure in Fig. 2, it is not possible to use electron counting to determine the number of delocalized 4s valence electrons compared to the number of 4s electrons in the copp e r - o x y g e n bonds, i.e. the charge transfer can not be determined. The stepwise decrease in IP of odd-numbered oxide clusters around the expected shell closings, n = 21, 23, 25 and n = 41, 43, 45 does not give a clear indication• In contrast, shell structure in the ionization cross sections has been observed for C s , O x, for several values of x [11,12], where each O atom binds two Cs 6s electrons and the shell-closing numbers appear

J.L. Persson et al. / Chemical Physics Letters 271 (1997) 6•-66

at z = n - 2x = 8, 18, 20, 34, 40, 58 and 92. In that case, it is understood that the oxygen acts as a nucleating site for the cluster, and the Cs atoms may adopt a spherical shape around a Cs2xO x core. Abundance spectra of Na,O~- ions ( x - 1, 2) have also shown enhanced stability for clusters with closed electronic shells, counting z = n - 2x + 1 delocalized electrons [13]. The reason for the differences between the alkali-oxide clusters and the Cu,O 2 is not clear, i.e. preserved shell structure and well-defined charge transfer is observed for the alkali oxides, but not for the Cu oxides, This can be an effect of the different chemistry of the alkali and the coinage metals. The metal-oxygen bond is expected to be ionic for alkali metals, whereas the C u - O bond is assumed to have a more covalent character. However, the difference can also be an effect of how the clusters are produced. In experiments with alkali oxides [11-13], the oxides were grown in the cluster source, whereas, in our experiment, the cluster formation and reaction were separated. The even-odd variation seen for the pure clusters is preserved in the oxides, but with a smaller amplitude. This variation is interpreted as an effect of electron pairing in non-degenerate orbitals, and is a significant characteristic for alkali and coinage metal clusters where each atom contributes one s electron to the filling of delocalized valence orbitals. Also clusters that are to first approximation expected to exhibit degenerate levels are often reduced to nondegenerate levels through Jahn-Teller distortions. The fact that the even-odd variation persists uninterrupted over the whole size range of Cu,O 2 clusters indicates that the character of the oxygen-copper bond is similar (involves the same number of electrons) for all the clusters, though the bond strength may vary. Fig. 4 shows the shifts in ionization potential, IP(CunO 2) - I P ( C u , ) and, as expected, the shifts are, on average, positive. The shifts presented here are probably an underestimation of the true shifts in the adiabatic ionization potential, since the oxidized clusters will have a higher internal energy, as discussed above. The variations in Fig. 4 are mainly due to discontinuities in the IPs of pure Cun. High positive shifts are observed at n = 21, 27, 31 and 41, and negative shifts at n = 30, 34 and 40. Considering only the IP, oxidation appears to render the

65

open-shell Cun more stable, while it does not improve the stability of the closed shell species. In fact, CU340 2 appears to become less stable. One reason for this might be that the open-shell clusters interact more strongly with the oxygen atoms, as indicated by their higher reactivity. Thus, these clusters experience a larger reconstruction of the electronic structure and a larger shift of the IP than the less reactive closed-shell clusters do. This interpretation is consistent with recent calculations of molecularly chemisorbed 02 on Cu-jellium clusters [14], where the chemisorption energy, charge transfer and shift in IP were calculated for clusters around the electronic shell closings at 8, 20 and 40 atoms. For clusters with closed or almost closed shells, a low chemisorption energy and charge transfer were calculated. Conversely, the adsorption of 0 2 on the open-shell clusters Cu 9, Cu2~ and Cu41 resulted in a strong hybridization between oxygen and jellium orbitals and a relatively large charge transfer and binding energy. The calculated shifts in IP are in good qualitative agreement with our experimental results. 4. Conclusions The main characteristics of the IP scheme for CunO 2, the disappearance of large drops in IP at shell openings and the persistence of even-odd variations, might be understood as the breaking and lowering of the symmetry of pure Cu n clusters. The closed-shell pure Cu clusters are assumed to gain their high stability and IP from a spherical geometry, resulting in highly degenerate electronic energy levels and large separations between these levels. If oxygen is adsorbed onto the cluster surface, electrons will be transferred to the oxygen atoms and this localized negative charge will break the spherical symmetry and the distinct shell structure will be perturbed. The remaining even-odd variation shows that the pairing of electrons, two electrons to each delocalized orbital, is stabilizing the even-numbered CunO 2, just as for the pure Cu,. The large increase in IP upon oxidation for the most reactive open-shell clusters is likely an effect of these clusters interacting strongly with the oxygen atoms, leading to a large modification of their electronic structure, whereas closed-shell clusters with a weaker interac-

66

J.L. Persson et al. / Chemical Physics Letters 271 (1997) 61-66

tion with o x y g e n show only a small shift in IP, or even a negative shift due to the breaking of spherical symmetry.

Acknowledgements W e would like to thank Henrik Gr/Snbeck for stimulating discussions and Dr, M a r k K n i c k e l b e i n for providing information about his experiment. Financial support from the N U T E K / N F R Materials Research C o n s o r t i u m 'Clusters and Ultrafine Particles' and the Swedish Research Council for Engineering Sciences ( T F R ) is gratefully acknowledged.

References [1] I. Katakuse, T. Ichihara, Y. Fujita, T. Matsuo, T. Sakurai, H. Matsuda, Int. J. Mass. Spectrom. Ion Processes 67 (1985) 229; 74 (1986) 33.

[2] C.L. Pettiette, S.H. Yang, M.J. Craycraft, J. Conceicao, R.T. Laaksonen, O. Cheshnovsky, R.E. Smalley, J. Chem. Phys. 88 (1988) 5377. [3] C.-Y. Cha, G. Gantef'rr, W. Eberhardt, J. Chem. Phys. 99 (1993) 6308. [4] M.B. Knickelbein, Chem. Phys. Lett. 192 (1992) 129. [5] W.A. deHeer, Rev. Mod. Phys. 65 (1993) 611. [6] J.L. Persson, M. Andersson, A. Ros~n, Z. Phys. D 26 (1993) 334. [7] M. Andersson, J.L. Persson, A. Rosrn, J. Phys. Chem. 100 (1996) 12222. [8] B.J. Winter, E.K. Parks, S.J. Riley, J. Chem. Phys. 94 (1991) 8618. [9] G. Alameddin, J. Hunter, D. Cameron, M.M. Kappes, Chem. Phys. Lett. 192 (1992) 122. [10] A. Mattsson, I. Panas, P. Siegbahn, U. Wahlgren, H. ,~keby, Phys. Rev. B 36 (1987) 7389, and references therein. [11] T. Bergmann, H. Limberger, T.P. Martin, Phys. Rev. Lett. 60 (1988) 1767. [12] T. Bergmann, T.P. Martin, J. Chem. Phys 90 (1989) 2848. [13] A. Nakajima, T. Sugioka, K. Hoshino, K. Kaya, Chem. Phys. Lett. 189 (1992) 455. [14] H. GriSnbeck,A. Ros~n, Chem. Phys. Lett. 227 (1994) 149.