Al(110) interface studied by electron spectroscopy

148

Surface Science 138 (1984) I\;orrh-ltolland.

THE Yb/AI( 110) INTERFACE SPECTROSCOPY J. ONSGAARD, 0. SIZIRENSEN

STUDIED BY ELECTRON

1. CHORKENDORFF.

Received 3 May 1983: accepred for publication

0. ELLEGAARD

3 Kwemher

and

1983

Thin films of Yb overlayers on an Al(1 IO) surface have been s1udied with different electron

spectroscopy.

The applicability

electron energy-loss spectroscopy (ELS) ing mixed-vale”1

Yb in the Yb/AI

methods of

of using Auger elec1ron spec1roscopy (AES). and X-ray photoelectron

spectroscopy (XPS)

interface is discussed. Comparison

shows that core hole interactions

play a strong role in the ionization

ionization

Diffusion

of the 4d shell in Yb.

14X-15X

Amsterdam

of Yh 1n1o Al(l10)

reflection

for recognir-

be1ween Yh t Oz and Yh/AI process which involves an

can he described as a two s1ep

process.

1. Introduction The surface can modify the valence of a rare-earth metal or a rare-earth compound. It was shown (I] that the first atomic layer of trivalent metallic samarium has a large divalent component. Wertheim et al. (21 concluded that the ytterbium surface atoms in YbAu,, which according to the lattice volume is trivalent, exhibit an initial mixed valence state with a 3’ to 2’ ratio of about 0.4. The valence state of rare-earth metals were discussed by Johansson [3]. Surface shifts on various rare-earth metals have been reported [4,5] recently and theoretical models have been developed [6-81 in order to explain the shifts. X-ray photoemission XPS data (91 of the intermetallic compound YbAl, showed that Yb is in an intermediate valence state. Single crystal studies of YbAl z by Kaindl et al. [lo] showed that bulk YbAlz has a mean valence of 2.4 and that two surface layers are divalent. Mixed-valent ytterbium,-aluminium thin films were studied by Tibbets and Egelhoff [11,12), who found that pure ytterbium surface layers show only the valence two, whereas a mixed-valent state was observed at 150°C where interdiffusion takes place. The 4p and 4d Auger electron spectra of Yb in the gas phase and the solid state were studied both experimentally and theoretically by Chorkendorff et al. [13]. It was concluded that the N,,N,,N,, Auger electron spectrum of solid Yb has a quasiatomic nature. 0039-6028/84/$03.00 8 Elsevier Science Publishers (North-Holland Physics Publishing Division)

B.V.

J. Onsgaard et al. / Yb/AI(IIO) interface

149

Electron energy loss spectra of Yb and oxidized Yb were measured by Bertel et al. [14], who found that the 4d to conduction band transitions were absent in clean Yb. However, a pronounced peak corresponding to a 4d to 4f transition was seen in oxidized ytterbium reflecting a shift in valence from 2+ to 3+. The purpose of this paper is to characterize the ytterbium-aluminium interface at room temperature and at elevated temperatures. The conditions for creation of the mixed-valent state are studied. The applicability of using different methods of electron spectroscopy like ELS, AES and XPS for recognizing the valence shift in the present metal-metal, Yb/Al, interface is discussed. It is demonstrated that electron energy-loss spectroscopy is a sensitive tool for registrating changes in the electronic structure in the interface region. Diffusion of Yb into Al(llO), as followed by AES, can be characterized as a two-step process. A slow diffusion takes place at room temperature and a strongly enhanced diffusion occurs at a temperature increase to 300°C as observed by XPS and AES. We observe changes in the Auger electron spectra from clean ytterbium and mixed ytterbium-aluminium thin films in contrast to the findings of Tibbets and Egelhoff [ll]. The differences are discussed from the point of view of different valence state configurations with the possibility of autoionization processes, or the possible influence of changes in the energy-loss function of the Auger electrons. Core hole interactions influence strongly the spectra which involve an ionization of the 4d shell in ytterbium.

2. Experimental The Yb/Al interface measurements were performed in an UHV chamber with a base pressure of 1.1 x 10e8 Pa. High purity ytterbium was evaporated in the UHV vessel from a tungsten wire shaped like a helix. The evaporation rate was controlled by a quartz crystal microbalance and a rate of one monolayer (ML) per minute was obtained with a pressure equal to 0.7 x lo-’ Pa. The Al(110) crystal had the dimensions 4 X 4 X 1 mm3 and could be heated by electron bombardment of the rear side of the crystal. The temperature was followed by a thermocouple in direct contact with the front side of the crystal. A clean (110) surface was obtained by a 2 keV Art ion bombardment at 45O’C for a period of 30 min and a subsequent annealing at 500°C resulted in the face centred cubic diffraction pattern with no impurities as judged from the Auger electron spectra. The main facility in the chamber for performing the analyses of the surface by AES, ELS and LEED was a four-grid retarding field analyzer and a 3 keV electron gun. A 3 keV ArC ion gun was used for the purpose of cleaning the

150

J. Onsganrd

et al. /

Yb/AI(lIO)

interface

sample. In the energy analyzing set-up, the analyzer was run both in the N(E) mode and in the first derivative mode. Most of the Auger measurements were carried out in the dN/dE mode with an incoming electron beam energy of 2 keV, a current density of 7 x 10’ PA/cm’ and a modulating voltage in the range 2.5 to 5 V peak-to-peak. In order to follow quantitative evaluations in the electron energy-loss spectra as a function of coverage, the ELS spectra were recorded in direct mode. Primary electron beam energies were varied between 60 and 1200 eV. The XPS measurements were carried out at Haldor Topsere, Copenhagen, with an Al I&Y X-ray source. Cleaning, evaporation and temperature reading were analogous to the above mentioned procedure.

3. Results 3.1. The b~iI~ up of Yb rnonol~~e~~ on the AI~ll~) surface The evolution of the electron energy-loss spectra as a function of increasing number of monolayers of Yb is shown in fig. 1. A low primary electron beam

A: 1110) I Yb Loss Ep=

Spectrum

126 eV

0

ZOML

x8

35ML

---“\

x4

_i--_--30

20

Energy

10

(eV)

Fig. 1. Electron energy-loss spectra of the Yb/AI(IlO) monolayers.

interface for increasing numbers of

J. Onsgaard et d

/ Yb/AI{I IO) imerface

151

energy, 126 eV, was chosen in order to obtain maximum surface sensitivity. Collective excitations, the bulk plasmon at 15.1 eV and the surface plasmon at 20.6 eV dominate the clean Al(ll0) loss spectrum. The other extreme in fig. 1 is the loss spectrum corresponding to twenty mono~ayers of Yb evaporated on the aluminium surface. A band transition is observed at 4.2 eV and a broad collective excitation occurs at 8.6 eV. In between the electron energy-loss spectrum of a surface with a coverage of two monolayers shows an interface plasmon at 12.4 eV. A “depth profile” of the interface for a given coverage, 3.5 monolayers, can be obtained, fig. 2, by variation of the primary electron beam energy. For E, = 90 eV the overwhelming part of the loss spectrum consists of the ytterbium contribution, whereas the loss spectrum is more aluminium like for E,, - 1200 eV. 3.2. Mixing effect.9fallowed by AES, ELS and XI’S A very slow diffusion AlIllO) I Yb Loss e -35Mi

of Yb into the substrate

is observed

at room

Spectrum

ELI

800

eV

x

400eV

x15

126 eV

-

15

x 125

I

I

I

30

20

10

Energy Fig. 2. The Yb(3.5 ML)/Al(llO)

(eV) interface studied by electron energy-loss spectroscopy.

J. Onsgaard el al. / Yb/AI(IIO)

N(E)

loss

interface

spectra

f -

0

T - 45°C

-AI-LVV

------rd Auger

peaks

w

,*--.._,% 25 ,’

JJ

(2‘1 ,‘,__jg________-------__

I

40

Energy

I

I

/

30

20

10

- loss

(eV)

Fig. 3. Diffusion of Yb into Al followed by AES and ELS. The Auger electron spectra, left part of the figure, are recorded in the first derivative mode, whereas the energy-loss spectra, right part of the figure, are recorded in the direct mode with Er, = 126 eV.

XPS

of YblAl

XPS

of YblA;

al

--..__ .-i....___i 24

16

E,;eW

EF

236

216

196

176

Ebb+)

Fig. 4. (a) The XPS spectra of the YbJAI(l10) interface at T = 0°C and T = 265’C. (b) The XPS spectra of Yb 4d recorded at T = 1OO’C and T = 265°C.

J.

Onsgaard er al. / ~~/Al(~iO~

interface

153

temperature, probably due to electron beam effects. However, slight raise of the temperature causes a strong acceleration of the diffusion process. The AES to a coverage of 3.5 and ELS spectra with Ep = 126 eV, corresponding monolayers of Yb on Al(110) and measured at a temperature T = 4S°C, are shown in fig. 3 as the upper curves. A temperature increase to T = 70°Cresults in the spectra shown as the lower curves in fig. 3. There are only minor changes in the line shapes of the Auger electron spectra of Yb, recorded with a 5 V peak-to-peak modulation voltage, as the diffusion takes place, whereas strong changes are observed in the electron energy-loss spectra, where the valence band to conduction band transitions in Yb nearly disappear. The XPS spectra of the Yb/Al(llO) interface at different temperatures, T = 0°C and 265’C, are shown in fig. 4a. The lower curve represents a thick

dNW/dE,

01

Auger spectrum Ea=2000 eV

150

of Yb

170

160

Energy

180

(eV)

Yb in Al Ep = 2000 eV 20 scans, 3 Vpp

I

/ 150

/ 160

/ 170

/ 180

/ 190

Energy (eV) Fig. 5. (a) The Auger ytterbium in aiuminium.

ektron

spectrum of ytterbium, (b) The Auger

electron

spectrum

of

154

layer of ytterbium at 0°C measured in the energy range from the Fermi level, E,, to 25 eV below. This spectrum characterizes Yb in the valence state 2+ with the 4f lines 2.2 eV below E, and some weak structures at higher binding energies. In agreement with ref. [ll] the XPS spectrum corresponding to the high temperature shows strong changes with a multiplet structure 7712 ev below E,. The XPS spectra of Yb 4d recorded at different temperatures are shown in fig. 4b. In analogy to the evolution of the ELS and XPS results a pronounced change could be expected in the AES data. However, as demonstrated in the Auger electron spectra of figs. 5a and 5b, recorded with a higher energy resolution than the spectra in fig. 3, only moderate changes in the intensities are observed. The variations in the first derivative spectra are identified at the positive excursion of the 166.2 eV peak and in the relative heights of the negative excursions at 168.2 and 179.4 eV. Reasons for these changes are discussed in the next section. The diffusion of Yb into Al(110) is displayed in fig. 6, where the Auger signal intensity of Yb is shown as a function of temperature and time. A fast diffusion takes place as the temperature rises to 240°C as indicated by a strong reduction in the signal intensity. After relatively longer time the intensity again falls off, indicating that a second and slower diffusion process occurs.

4. Discussion 4.1. Mixed valency in the Yb/Al

interface

Two aspects of the valence transition in the Yb/Al interface will be discussed in the following. Under which conditions does the transition occur and which methods, based upon electron spectroscopy, are suitable for recognizing the change in valence? 4.2. XPS A strong indicator of the valence shift in the Yb/Al interface is found in the development of the XPS 4f line spectrum, fig. 4a, where the XPS spectrum recorded at the high temperature represents a mixed-valent state. A new observation is that the XPS spectrum of Yb 4d reflects the Yb valency in the Yb/Al interface. The divalent phase is shown in fig. 4b with the 4d electrons causes a high intensity in split into d3,2 and d5,2. A very intense background state is clearly demonstrated in the 4d 3,2 peak. The effect of the mixed-valent the spectrum recorded at 265°C. A hole in the 4f shell couples with the 4d hole from the photo-ionization process and a multiplet structure is created. Some of the ytterbium will still be in the divalent state and the complex structure in the

J. Onsgaard et al. / Yb/AI(IIO) Time

5000

(set

10000

Time

interface

155

1

15000

20000

(set)

Fig. 6. The temperature and the intensity of the main Auger transition in Yb are shown as a function of time on two time scales. The graphs which decrease with time represent the Auger signals, and the graph which increases with time refers to the temperature.

spectrum at 265’C represents a superposition of the spectra from the two valence states. A shake-up process of the type hv + 4d”4f I4 + 4d94f l3 + 2ecan occur in the XPS spectrum, but it will be indistinguishable from the photoemission process from trivalent Yb. A comparison of the present 4d and 4f XPS spectra with the corresponding spectra of clean [15] and oxidized Yb [15] (valence 3+) shows that a superposition of the last mentioned spectra can explain the Yb/Al spectra. Thus the Yb/AI system is characterized by a mixed-valent state. The observed 4d-4f coupling does not support the 4f refilling mechanism proposed in ref. [12]. 4.3. ELS Electron energy loss spectroscopy can be used as a tool for probing an interface due to the possibility of varying the incoming primary beam energy and thereby obtaining information from different depths. In this respect ELS offers an advantage to other spectroscopies like XPS and AES where the photon energy and the Auger transition energy are fixed. This is illustrated in fig. 7 where the energy of the bulk plasmon loss is shown as a function of Ep for an Yb coverage of 3.5 monolayers. A minimum value occurs for Ep - 100 eV where the maximum surface sensitivity is expected. From fig. 1 it is deduced that clean Yb has a bulk value at 8.6 eV in agreement with the findings of Bertel et al. [14]. The fact that the minimum of the graph in fig. 7 lies at 11.3 eV reflects that the collective excitation involves an effective free electron density higher than two electrons per atom. It should also be taken into consideration that the electrons penetrate normal to the

surface. A glancing incidence electron beam would result in a lower lying energy minimum. At E, = 1200 eV the energy position of the bulk plasmon indicates that the substantial part of the free electron density originates from aluminium. It is an empirical fact that the energetic positions of the bulk and the surface plasmon of a metal do not shift significantly with the primary electron beam energy. In terms of creations of new surfaces with deposition of overlayers. the 3.5 monolayers of Yb define a new surface where the free electron density of Yb determines the surface plasmon energy and the energy position of the bulk plasmon is derived of the free electron density of aluminium. The intensities of the respective collective excitations would then be determined by the path lengths of the electrons and the effective free electron densities in the respective media. In this picture the interface collective excitation is thus a superposition of the two plasmons. An argument for the superposition point of view is also found in the considerable broadening of the excitation which is larger than the half widths of bulk plasmons in Al and Yb. This interpretation means that the plasmon dispersion in fig. 7 represents a centroide shift of two partly overlapping plasmon peaks with increasing primary beam energy. In analogy. a correlation between bulk plasmon energy shift and the number of ytterbium monolayers can be established. Thus the energy dispersion of the plasmon can be used for a rough estimate of the overlayer thickness in the low coverage regime. In the above given model, where the loss spectrum is characterized as a linear combination of the two pure metals, changes in the free electron density in the interface region are not taken into consideration. In an alternative model an interface plasmon is excited and its energy will be proportional to the square root of the mean free density of the two metals. ho, a [( nA, + ,1Yh)/2]‘/‘2. This results in an interface plasmon energy around 12.3 eV for the Yb/AI interface. Two observations can be made from fig. 1. First, the upcome of the 4.2 eV energy loss peak with increasing numbers of monolayers represents a valence band to conduction band transition. This peak disappears when ytterbium diffuses into the aluminium substrate at 70°C as shown in fig. 3. Second, a bulk plasmon position at 8.6 eV in the energy loss spectrum representative for an overlayer thickness of twenty monolayers points at a valence state of 2’. This excitation moves gradually to higher energies with decreasing coverage, reflecting a higher free electron density - and indicating a mixed valance. A similar effect was observed by Bertel et al. [16] in Sm. where a divalent surface character was identified by an intense plasmon feature at 7 eV observed at low primary energies.

J. Onsgaard Ed al. /

Yb/AI(I

10) mterjace

157

7Or Energy

Loss

AI/Vb

----.vb+cb

8-35ML

I

- 110

/ , '._I 40

0

200

LOO

Em

800

1000

- 100 1ZCIl

EpW

Fig. 7. The position

of the plasmon

excitation

(right

scale) in the energy-loss

function of primary electron beam energy for an Yb coverage of 3.5 monolayers. scale shows the energy position of the valence band to conduction

spectrum

Similarly.

as a

the left

band transition.

4.4. AES Only minor differences in intensities, see figs. 5a and 5b, are observed betwen the NqgN6,N6, Auger transitions characteristic of the clean Yb and the mixed phase compared to the pronounced differences in the Auger profiles of clean Yb and oxidized Yb [14]. In an earlier report [13] it was shown that the could be reproduced in a intense 4d4f4f transition in Yb2+ theoretically quasiatomic description where in the solid state the electron energy-loss spectrum characteristic of Yb for a primary electron beam energy of 170 eV was used as a suitable energy-loss function for the Auger electrons. In the present case the energy-loss function of the Yb/Al interface corresponding to the mixed-valent state of Yb modifies the profile of the Auger spectrum. Furthermore with a valence of two in the uppermost two layers, as found by Kaindl et al, [lo] in YbAl, and recently by Nyholm et al. (171 for the present system, a valence change in the next layers is hardly observable in the 160-170 eV Auger spectrum of Yb. It is concluded that the modifications of the Auger spectrum are mainly due to changes in the electron energy-loss function with minor contributions from autoionization processes or direct Auger transitions from trivalent Yb. 4.5. Inrerdiffusion The diffusion behaviour of Yb in Al consists of a first step, see fig. 6, where Yb overcomes the first potential barrier at the selvedge, and a second step with a much lower diffusion constant. It is proposed that the mixed Yb/Al phase is

characterized by a higher potential barrier. For temperatures higher than 300°C the diffusion rate increases so fast that the intensity of the ytterbium signal is strongly reduced as observed with both AES and XPS. The explanation of the plateau in the diffusion graph is found in the measuring method. With an Auger transition energy of 170 eV, the information originates from the first few layers. When the mixed phase has been established for several layers in the crystal, the Auger signal scans the taii of a distribution. A relatively slow diffusion is therefore first registered in the rear front of the profile after a certain period. Support for this interpretation is found in the time derivative of the signal, which is lower in the second stage compared to the first step. 5. Conciusion

The valence state of Yb in thin films of Yb on the Al(110) surface and of Yb diffused into the uppermost surface layer of Al has been studied by electron spectroscopy. Electron energy-loss spectroscopy is a sensitive tool, especially in the l-15 eV low energy region, for following changes in the electronic structure of the interface. The mixed valence state is clearly demonstrated in the 4d and 4f XPS spectra. In the Auger electron spectra the modifications can be explained by changes in the electron energy-loss spectra influencing the lineshapes in the M,,KN region. Diffusion of Yb into Al is characterized by a two-step process.

References [I] 121 [3] [4]

G.K. Wertheim and G. Crecelius. Phys. Rev. Letters 40 (1978) 813. G.K. Wertheim, J.H. Wernich and G. Crecelius, Phys. Rev. RI8 (1978) 875. B. Johansson. Phys. Rev. El9 (1979) 6615. F. Gerken, J. Barth, R. Kammerer, L.f. Johansson and A. Flodstrom, Surface Sci. 117 (19X2) 468. 151 L.I. Johansson, A. Flodstrom. S.-E. Hornstrom, B. Johansson. J. Barth and F. Gerkcn. Surface Sci. 117 (1982) 475. [6] B. Johansson and N. Martensson. Phys. Rev. B21 (1980) 4427. [7] A. Rosengren and B. Johansson. Phys. Rev. B22 (1980) 3706. [8] A. Rosengren and B. Johansson. Phys. Rev. 823 (1981) 3852. 191 K.H.J. Buschow, M. Campagna and G.K. Wertheim, Solid State Commun. 24 (1977) 253. [IO] G. Kaindl, B. Reihl. D.E. Eastman. R.A. Pollak. N. Martensson, B. Barbara, T. Penncey and T.S. Plaskett, Solid State Commun. 41 (1982) 157. Ill] G.G. Tibhetts and W.F. Egelhoff. J. Vacuum Sci. Technol.. 17(l) (1980) 458. 1121 W.F. Egelhoff and G.G. Tibbetts, Phys. Rev. Letters 44 (1980) 482. [13] 1. Chorkendorff. J. Onsgaard. H. Aksela and S. Aksela. Phys. Rev. B27 (1983) 945. [ 141 E. Bertel, G. Strasser, F.P. Netzer and J.A.D. Matthew, Surface Sci. 11X (1982) 387. [15) B.D. Padalia, W.C. Lang. P.R. Norris, L.M. Watson and D.J. Fabian. Proc. Roy. Sot. (London) A354 (1977) 269. [I61 E. Bertel, G. Strasser. F.P. Netzer and J.A.P. Matthew. Phys. Rev. 825 (1982) 3374. [17] R. Nyholm. 1. Chorkendorff and J. Schmidt-May, to be published.