- Email: [email protected]
A comparative study of oxidation ofTi and CoTi by HREELS and XPS
Journul of Electron Spectroscopy and Related Ph~rwmena, 54155 (1990) 1065-1074 Elsevier Science Publishers B.V., Amsterdam
1065
A comparative studyof oxidationof Ti and CoTi by HREELS and XPS. S.J. Garretta, R.G. Egdella and J.C. Rivibreb
aDepartment 2Az, UK. bMaterials Oxfordshire
of Chemistry,
Imperial
Development Division, OX11 ORA, UK.
College,
Harwell
South Kensington,
Laboratories,
London
Harwell,
SW7
Didcot,
Abstract. High resolution electron energy loss spectroscopy (HREELS) has been used in conjunction with X-ray photoelectron spectroscopy (XPS) to study the oxidation of Ti(0001) and CoTi at room temperature. Oxygen exposure of Ti(OOO1) produces a HREELS loss peak at 65meV for exposures up to 0.5L. Thereafter a shoulder appears at 87meV and becomes the dominant loss peak after 25L exposure. The high energy peak continues to grow in intensity for exposures up to 17,000L and shifts upward in energy to 95meV. Similar changes in HREEL spectra accompany oxidation of CoTi, although the intensity of loss peaks grows more slowly than for Ti. The HREELS results are in broad agreement with XPS data, pointing to TiOa as the limiting oxide phase on both Ti and CoTi.
1. INTRODUCTION. High resolution electron energy loss spectroscopy (HREELS) has been widely used to study the surface chemistry of transition metals. However there have been surprisingly few studies of titanium with this technique, despite the technological importance of this metal and its alloys. In particular HREELS work on oxidation of single crystal titanium has been restricted to a range of exposures extending up to only 2OL [ll. However it is known from XPS that a surface oxide phase continues to grow on polycrystalline Ti surfaces up to much
0363~2043/90/$03.50
0 1990 Elsevier Science Publishers B.V.
1066
higher exposures [2]. In the present communication we use HREELS in conjunction with XPS to study oxidation of Ti(OOO1) over a wide range of oxygen exposures from O.lL up to 10,OOOL. The oxidation of Ti is compared with that of CoTi. Interest in oxidation of the alloy derives from its possible application as a hydrogen storage medium [3] and the inhibition of hydrogen uptake by oxide overlayers [4]. Comparison is also made with HREELS data from single crystal TiO#lO) [5,63.
2.EXPERlMENTAL. HREEL
spectra
were measured
in a turbomolecular
and ion pumped UHV
system (base pressure 5x10-11 torr) equipped with a Leybold-Heraeus ELS 22 HREELS module, a 1OOmm mean radius spherical sector analyser used in conjunction with a VSW twin anode X-ray source for measurement of XI’S and VG four grid LEED optics. A polished single crystal of (0001) oriented Ti 4mm in diameter was laser welded to a lmm thick Ti bar and mounted between Ta support rods at the end of the long travel manipulator of the HREEL spectrometer. The crystal was cleaned in situ in the spectrometer by alternate cycles of argon ion bombardment and annealing at 600°C over a period of several days. This removed oxygen and carbon contamination seen in XI’S but led to appearance of weak S related core peaks. Segregation of S was encouraged by a prolonged 3 day anneal at 600°C and S was finally removed by a final brief etch. After flashing to 6OO”C, the XPS was free of contaminant structure and the crystal gave a sharp 1x1 LEED pattern. However, a strong loss peak remained at 120meV in HREELS. This has been attributed previously to Si contamination Ill. However, the difficulty in removing hydrogen from Ti(0001) is well documented [7,81 and we believe the peak relates to a Ti-H stretch associated with residual hydrogen on the crystal surface. To confirm this hypothesis, the crystal was exposed to Da. The 12OmeV loss was found to decrease in intensity and the expected Ti-D stretch grew up at 85meV (figure 1). Further cycles of oxygen exposure and annealing reduced the intensity of the Ti-H related losses, although it proved impossible to achieve zero loss intensity around 120meV. As the hydrogen contamination level was reduced, the rate of H-D interchange or further Hz uptake decreased as compared with the data of figure 1. Thus we believe the reactivity of Ti(0001) toward H2 or D2 is strongly dependent on the perfection of the surface. Limiting surfaces with the lowest level of H contamination (e.g. figure 2a) showed no evidence of additional H2 adsorption even for exposures up to 10,OOOL. Oxygen exposures were studied in the range up to 17,000L in several series of experiments. Oxygen free surfaces were easily recovered after oxygen exposure
1067
by simply annealing at 600°C in UHV. Nominally single crystal CoTi(ll0) was subject to a cleaning procedure similar to that for Ti(OOOl), as described in detail elsewhere [9]. Ion bombardment alone generated a Co rich surface, but subsequent annealing led to segregation of Ti and associated bulk dissolved oxygen [lo]. By careful manipulation of annealing and bombardment conditions it was possible to obtain nominal 1:l Co:Ti stoichiometry, but the 01s signal in XPS could not be reduced below a Ols/Ti2p intensity ratio of 0.15 despite prolonged cleaning cycles over periods of many weeks. The cleaning procedure also led to macroscopic surface roughening of the crystal and it was not possible to observe well defined LEED patterns. For these reasons our data must be regarded as characteristic of polycrystalline CoTi.
Energy
loss I meV
Figure l.HREEL spectra of H-Ti(ooO1) 1x1 as a function of exposure to D2. (a) OL (b) 13L (c) 5OL (d) 400L. Ep=2eV Bi=8s=60’ relative to surface normal.
3. RFXJT.,‘lS AND DISCUSSION. Energy loss spectra of Ti(0001) as a function of exposure to oxygen are shown
1068
in figure
2. At very low exposure a sharp loss peak develops at 65meV, but even
by 1L exposure the peak has broadened considerably. After 4L exposure a shoulder at 87meV has become apparent and dominates the HREEL spectra after 30L exposure. The high energy loss continues to grow in intensity even up to the highest exposures and both broadens and shifts up to about 95meV after 17,000L exposure (figure 3).
(h)
(e) (d)
(cl (b)
b
x 300
a)
.:: I
0
x300
I
I
1
I
50
loo
150
200
Energy
loss / meV
Figure 2. HREEL spectra of Ti(OOO1)as a function of exposure to 02. (a) OL (b) 0.25L (c) 1L (d) 4L (e) 7L (f) 30L (g) 3000L (h) 17,OOOL.Ep=2eV Bi=6,=60’ relative to surface normal.
k loo. t.$ 90. :
0.4,
d 8
.
l.
00
0 0
.
l
**
. . ..
.
0 ooo
=0.2.
5 4
90.
5
70.
0 00 0
Q
0 o” 0.0,
: OQeO 0
1
10 ‘102 103 Oxygen Exposure /L
104
15= 10
E ._ ::
oo” 60.
:
0
1
10 102 103 Oxygen Exposure /L
104
105
Figure 3. Left hand panel: total loss intensity I relative to elastic peak in HREELS of Ti(0001) as a function of oxygen exposure (Ep = 2eV). Right hand panel: energy of principal loss features as function of oxygen exposure.
(b)
I
0
0% Binding
I
660
465
1 470
energy /eV
Figure 4. MgKa excited XPS of Ti(OOO1). (a) clean surface showing 2p spin orbit doublet with 2~3~ peak at 453.8eV. (b) after exposure to 17,000L 02, showing new doublet with 2p3/2 peak at 459&V.
1070
Ti2p core level XPS for clean Ti(0001) and for the surface after exposure to 17,000L of oxygen are shown in figure 4. These data are similar to those found previously for polycrystalline Ti foil [21. The 2~312 component of the 2p doublet of titanium metal is found at 453.8eV. After oxygen exposure a new doublet appears whose 2~312 component is at 459.0eV. This corresponds to the binding energy
of TiO2 [2,61. For this reason it is argued that the limiting
oxide after
exposure of Ti metal to oxygen is essentially TiO2. In this context it is interesting to compare the energy loss spectrum of TiO2(110) in figure 5 with the spectra 2f2g. The limiting energy of 95meV found for the broad oxygen induced loss on Ti(0001) corresponds exactly with the energy of the highest surface optical
Energy
Figure 5. HREELS of TiOz(llO),
Loss I meV
taken from reference 6. Ep=7eV. 0i=8,=60” relative to
surface normal. phonon mode of the TiO2 crystal. Moreoever
the upward energy shift of this peak
with increasing
downward
oxygen exposure finds parallel behaviour in the progressive shift in the surface optical phonon frequency of TiO2 with increasing
oxygen vacancy defect concentration induced by ion or electron bombardment [5,6]. The change is related to an increase in the background dielectric constant in the vibrational region associated with defect electronic excitations and thus the HREELS indicates that the TiOa-like layer formed on Ti(0001) is initially oxygen deficient. This conclusion is supported by detailed analysis of the XPS lineshapes which indicate that there is a third spin orbit doublet with the 2p3/2 peak at 457.5eV [2]. This is probably associated
with the nominal Ti3+ found in
oxygen deficient x-utile Ti0z.x. Despite the close correspondence between frequencies of the most intense loss peaks, comparison between figures 2 and 5 highlights the fact that the oxide overlayer on Ti(OOO1) does not support the same sharply defined surface phonon structure as found on TiOz(110) or indeed on bulk polycrystalline TiO,, which yields HREEL spectra similar to those of figure 4 [ll]. Thus the 95meV peak is very much broader on O-exposed Ti(OOO1) and the surface phonon loss at 55meV seen in HREELS of TiOa is not resolved, although there is strong loss intensity in this region. The most obvious explanation of the differences lies in terms of an amorphous structure for the oxide overlayer with local sixfold oxygen coordination around titanium as found in rutile, but without the long range order characteristic of the rutile structure. Turning now to the low exposure regime, the spectra of figures 2a-c demonstrate the remarkable sensitivity of HREELS to small oxygen uptake, Growth of the loss peak at 65meV is accompanied by rapid quenching of the Ti-H related structure at 120meV. It is not obvious whether this results from chemical displacement of H or from screening of Ti-H losses. The 65meV loss itself has been previously associated with a so-called 0: Ti-0 phase [l]. This may simply correspond to oxygen in threefold hollow sites of Ti(OOOl), although it has been argued that the decrease in work function for oxygen exposures below 1L and the low 0+ yield in photon stimulated desorption may be associated with subsurface oxygen [1,123. Clearly, however, HREELS supports the hypothesis of two different adsorption states for oxygen on Ti(OOO1) [12,131. We consider next the HREELS data from O-exposed CoTi (figure 6). In contrast to Ti(OOOl), a significant loss peak is found at 65meV even before oxygen exposure. This is associated with the irreducible 01s intensity seen in XPS (section 2). The problems in cleaning CoTi appear to result from the greater mobility of oxygen in this alloy as compared with Ti and the uptake of around 1OOOppm dissolved oxygen impurity during crystal growth. The initial intensity of the 65meV loss is in fact much greater than can be seen on Ti(OOOl), despite the lower beam energy used in the study of the pure metal. This suggests that HREELS is probing a surface layer with significant concentration of dissolved oxygen impurity. The evolution of HREEL spectra with increasing oxygen exposure is in some way similar to that for Ti(OOOl), with increasing spectral weight appearing to the high binding energy side of the initial 65meV peak. However, there is no clear emergence of a dominant loss peak above 9OmeV and the spectra remain dissappointingly ill-structured. In fact XPS proves to be much more informative than HREELS in probing the oxidation of CoTi. The initial Ti2p spin orbit doublet of CoTi is shifted by about 0.6eV to higher binding energy as compared with Ti(OOOl), so that the 2~312 peak appears at 454.4eV (figure 7). This shift is due both to charge transfer from Ti to
1072
Co in the alloy and to the effects of the dissolved oxygen. With increasing oxygen exposure a high binding energy spin orbit doublet grows in intensity, with the 2~312component at 458.9eV as for Ti itself. After 15,OOOLoxygen exposure this doublet dominates the XPS. In contrast to Ti(OOOl),the 2~312oxide peak is clearly distinguishable from the initial 2~~2 peak because of the high binding energy shift of the latter.
:e)
W
a
0
,
I
50
100
Energy
I
I
150
200
loss lmeV
Figure6. HREELspectraof CoTi as a functionof oxygen exposure.(a) OL(b) IL (c) 4L (d) 400L (e) 4000L.Ep=SeV.Bi=8,=60’relativeto surfacenormal.
1073
In contrast to pronounced changes in the Ti2p region of the spectra, there is little change in the Co2p region with oxygen exposure up to 15,OOOL (figure 7). In particular, the changes are much more muted than for Co itself [141 where oxidation is accompanied both by emergence of chemically shifted Co2p peaks to 2.OeV higher binding energy of the original metal peaks and appearance of strong satellites associated with Co0 in the oxide overlayer. The picture to emerge from figures 7 and 8 is thus that oxidation of CoTi leads to formation of an esentially TiOa-like overlayer, which protects the Co component of the alloy from oxidation. This conclusion is supported by more detailed analysis of XPS data [9], which reveals that oxygen exposure is accompanied both by an decrease in the CopTi intensity ratio and an increase in background intensity in the Co2p region.
I
1’
(a)
M
lb)
(b) .
+J I
455 Binding
460
465
energylev
Figure 7. (Left hand panel) MgKa
4‘7(1
: 1
L I
I
760
790
Binding
800
I
energy/eV
excited XPS of CoTi in Ti 2p region. (a) clean surface
showing 2p spin orbit doublet with 2~3~ peak at 454.4eV. (b) after exposure to 15,OOOL02, showing new doublet with 2~3~ peak at 458.9eV. Figure 8. (Right hand panel) MgKa excited XPS of CoTi in Co2p region. (a) clean surface (b) after exposure to 15,OOOL 02 .The changes accompanying oxygen exposure are very small compared with those in figure 7.
1074 Whilst the HREELS data of figure 6 are not at variance with these conclusions, clearly the technique is less specific than XPS in distinguishing between Co and Ti oxides. However, HREELS does allow the possibility of monitor-ring H or D adsorption on CoTi. Somewhat surprisingly we have so far failed to observe hydrogen-associated vibrational modes on CoTi. It is unclear at present whether this is due to rapid diffusion of hydrogen into the bulk of CoTi or to inhibition of Hz dissociation on the crystal surface by the residual oxide.
4. ACKNOWIEDG~~. We are grateful to Harwell studentship to SJG.
Laboratories
and SERC for award
of a CASE
5. REFERENCES. 1. R.L. Strong and J.L. Erskine, J. Vat. Sci. Technol. A3 (1985) 1428. 2. A.F. Carley, P.R. Chalker, J.C. Riviere and M.W. Roberts, J. Chem. Sot. Faraday Trans. I, 83 (1987) 351. 3. Y. Osumi, H. Suzuki, A. Kato, K. Oguro and M. Nakane, J. Less Common Metals, 74 (1980) 271. 4. M.C. Burrell and N.R. Armstrong, Surf. Sci., 160 (1985) 235. 5. G. Rocker, J.A. Schaefer and W. Gape& Phys. Rev. B, 30 (1984) 3604 6. S. Eriksen and R.G. Egdell, Surf. Sci., 180 (1987) 263. 7. P.J. Feibelmann, D.R. Hamann and F.J. Himpsel, Phys. Rev. B, 22 (1980) 1734. 8. Y. Fukuda, F. Honda and J.W. Rabalais, Surf. Sci., 91 (1980) 165. 9. S. Garrett, R.G. Egdell and J.C. Rivibre, Surf. Sci., submitted. 10.S. Garrett, R.G. Egdell and J.C. Rivi&re, J. Phys. Condens. Matter,
1 (1989)
SB229. 11. W.R. Flavell, D. Phil. Thesis, Oxford 1986. 12.D.M. Hanson, R. Stockbauer and T.E. Madey, Phys. Rev. B, 24 (1981) 5513. 13.B.T. Jonker, J.F. Morar and R.L. Park, Phys. Rev. B, 6 (1971) 2951. 14.C.R. Brundle, T.J. Chuang and D.W. Rice, Surf. Sci., 60 (1976) 286.
1065
A comparative studyof oxidationof Ti and CoTi by HREELS and XPS. S.J. Garretta, R.G. Egdella and J.C. Rivibreb
aDepartment 2Az, UK. bMaterials Oxfordshire
of Chemistry,
Imperial
Development Division, OX11 ORA, UK.
College,
Harwell
South Kensington,
Laboratories,
London
Harwell,
SW7
Didcot,
Abstract. High resolution electron energy loss spectroscopy (HREELS) has been used in conjunction with X-ray photoelectron spectroscopy (XPS) to study the oxidation of Ti(0001) and CoTi at room temperature. Oxygen exposure of Ti(OOO1) produces a HREELS loss peak at 65meV for exposures up to 0.5L. Thereafter a shoulder appears at 87meV and becomes the dominant loss peak after 25L exposure. The high energy peak continues to grow in intensity for exposures up to 17,000L and shifts upward in energy to 95meV. Similar changes in HREEL spectra accompany oxidation of CoTi, although the intensity of loss peaks grows more slowly than for Ti. The HREELS results are in broad agreement with XPS data, pointing to TiOa as the limiting oxide phase on both Ti and CoTi.
1. INTRODUCTION. High resolution electron energy loss spectroscopy (HREELS) has been widely used to study the surface chemistry of transition metals. However there have been surprisingly few studies of titanium with this technique, despite the technological importance of this metal and its alloys. In particular HREELS work on oxidation of single crystal titanium has been restricted to a range of exposures extending up to only 2OL [ll. However it is known from XPS that a surface oxide phase continues to grow on polycrystalline Ti surfaces up to much
0363~2043/90/$03.50
0 1990 Elsevier Science Publishers B.V.
1066
higher exposures [2]. In the present communication we use HREELS in conjunction with XPS to study oxidation of Ti(OOO1) over a wide range of oxygen exposures from O.lL up to 10,OOOL. The oxidation of Ti is compared with that of CoTi. Interest in oxidation of the alloy derives from its possible application as a hydrogen storage medium [3] and the inhibition of hydrogen uptake by oxide overlayers [4]. Comparison is also made with HREELS data from single crystal TiO#lO) [5,63.
2.EXPERlMENTAL. HREEL
spectra
were measured
in a turbomolecular
and ion pumped UHV
system (base pressure 5x10-11 torr) equipped with a Leybold-Heraeus ELS 22 HREELS module, a 1OOmm mean radius spherical sector analyser used in conjunction with a VSW twin anode X-ray source for measurement of XI’S and VG four grid LEED optics. A polished single crystal of (0001) oriented Ti 4mm in diameter was laser welded to a lmm thick Ti bar and mounted between Ta support rods at the end of the long travel manipulator of the HREEL spectrometer. The crystal was cleaned in situ in the spectrometer by alternate cycles of argon ion bombardment and annealing at 600°C over a period of several days. This removed oxygen and carbon contamination seen in XI’S but led to appearance of weak S related core peaks. Segregation of S was encouraged by a prolonged 3 day anneal at 600°C and S was finally removed by a final brief etch. After flashing to 6OO”C, the XPS was free of contaminant structure and the crystal gave a sharp 1x1 LEED pattern. However, a strong loss peak remained at 120meV in HREELS. This has been attributed previously to Si contamination Ill. However, the difficulty in removing hydrogen from Ti(0001) is well documented [7,81 and we believe the peak relates to a Ti-H stretch associated with residual hydrogen on the crystal surface. To confirm this hypothesis, the crystal was exposed to Da. The 12OmeV loss was found to decrease in intensity and the expected Ti-D stretch grew up at 85meV (figure 1). Further cycles of oxygen exposure and annealing reduced the intensity of the Ti-H related losses, although it proved impossible to achieve zero loss intensity around 120meV. As the hydrogen contamination level was reduced, the rate of H-D interchange or further Hz uptake decreased as compared with the data of figure 1. Thus we believe the reactivity of Ti(0001) toward H2 or D2 is strongly dependent on the perfection of the surface. Limiting surfaces with the lowest level of H contamination (e.g. figure 2a) showed no evidence of additional H2 adsorption even for exposures up to 10,OOOL. Oxygen exposures were studied in the range up to 17,000L in several series of experiments. Oxygen free surfaces were easily recovered after oxygen exposure
1067
by simply annealing at 600°C in UHV. Nominally single crystal CoTi(ll0) was subject to a cleaning procedure similar to that for Ti(OOOl), as described in detail elsewhere [9]. Ion bombardment alone generated a Co rich surface, but subsequent annealing led to segregation of Ti and associated bulk dissolved oxygen [lo]. By careful manipulation of annealing and bombardment conditions it was possible to obtain nominal 1:l Co:Ti stoichiometry, but the 01s signal in XPS could not be reduced below a Ols/Ti2p intensity ratio of 0.15 despite prolonged cleaning cycles over periods of many weeks. The cleaning procedure also led to macroscopic surface roughening of the crystal and it was not possible to observe well defined LEED patterns. For these reasons our data must be regarded as characteristic of polycrystalline CoTi.
Energy
loss I meV
Figure l.HREEL spectra of H-Ti(ooO1) 1x1 as a function of exposure to D2. (a) OL (b) 13L (c) 5OL (d) 400L. Ep=2eV Bi=8s=60’ relative to surface normal.
3. RFXJT.,‘lS AND DISCUSSION. Energy loss spectra of Ti(0001) as a function of exposure to oxygen are shown
1068
in figure
2. At very low exposure a sharp loss peak develops at 65meV, but even
by 1L exposure the peak has broadened considerably. After 4L exposure a shoulder at 87meV has become apparent and dominates the HREEL spectra after 30L exposure. The high energy loss continues to grow in intensity even up to the highest exposures and both broadens and shifts up to about 95meV after 17,000L exposure (figure 3).
(h)
(e) (d)
(cl (b)
b
x 300
a)
.:: I
0
x300
I
I
1
I
50
loo
150
200
Energy
loss / meV
Figure 2. HREEL spectra of Ti(OOO1)as a function of exposure to 02. (a) OL (b) 0.25L (c) 1L (d) 4L (e) 7L (f) 30L (g) 3000L (h) 17,OOOL.Ep=2eV Bi=6,=60’ relative to surface normal.
k loo. t.$ 90. :
0.4,
d 8
.
l.
00
0 0
.
l
**
. . ..
.
0 ooo
=0.2.
5 4
90.
5
70.
0 00 0
Q
0 o” 0.0,
: OQeO 0
1
10 ‘102 103 Oxygen Exposure /L
104
15= 10
E ._ ::
oo” 60.
:
0
1
10 102 103 Oxygen Exposure /L
104
105
Figure 3. Left hand panel: total loss intensity I relative to elastic peak in HREELS of Ti(0001) as a function of oxygen exposure (Ep = 2eV). Right hand panel: energy of principal loss features as function of oxygen exposure.
(b)
I
0
0% Binding
I
660
465
1 470
energy /eV
Figure 4. MgKa excited XPS of Ti(OOO1). (a) clean surface showing 2p spin orbit doublet with 2~3~ peak at 453.8eV. (b) after exposure to 17,000L 02, showing new doublet with 2p3/2 peak at 459&V.
1070
Ti2p core level XPS for clean Ti(0001) and for the surface after exposure to 17,000L of oxygen are shown in figure 4. These data are similar to those found previously for polycrystalline Ti foil [21. The 2~312 component of the 2p doublet of titanium metal is found at 453.8eV. After oxygen exposure a new doublet appears whose 2~312 component is at 459.0eV. This corresponds to the binding energy
of TiO2 [2,61. For this reason it is argued that the limiting
oxide after
exposure of Ti metal to oxygen is essentially TiO2. In this context it is interesting to compare the energy loss spectrum of TiO2(110) in figure 5 with the spectra 2f2g. The limiting energy of 95meV found for the broad oxygen induced loss on Ti(0001) corresponds exactly with the energy of the highest surface optical
Energy
Figure 5. HREELS of TiOz(llO),
Loss I meV
taken from reference 6. Ep=7eV. 0i=8,=60” relative to
surface normal. phonon mode of the TiO2 crystal. Moreoever
the upward energy shift of this peak
with increasing
downward
oxygen exposure finds parallel behaviour in the progressive shift in the surface optical phonon frequency of TiO2 with increasing
oxygen vacancy defect concentration induced by ion or electron bombardment [5,6]. The change is related to an increase in the background dielectric constant in the vibrational region associated with defect electronic excitations and thus the HREELS indicates that the TiOa-like layer formed on Ti(0001) is initially oxygen deficient. This conclusion is supported by detailed analysis of the XPS lineshapes which indicate that there is a third spin orbit doublet with the 2p3/2 peak at 457.5eV [2]. This is probably associated
with the nominal Ti3+ found in
oxygen deficient x-utile Ti0z.x. Despite the close correspondence between frequencies of the most intense loss peaks, comparison between figures 2 and 5 highlights the fact that the oxide overlayer on Ti(OOO1) does not support the same sharply defined surface phonon structure as found on TiOz(110) or indeed on bulk polycrystalline TiO,, which yields HREEL spectra similar to those of figure 4 [ll]. Thus the 95meV peak is very much broader on O-exposed Ti(OOO1) and the surface phonon loss at 55meV seen in HREELS of TiOa is not resolved, although there is strong loss intensity in this region. The most obvious explanation of the differences lies in terms of an amorphous structure for the oxide overlayer with local sixfold oxygen coordination around titanium as found in rutile, but without the long range order characteristic of the rutile structure. Turning now to the low exposure regime, the spectra of figures 2a-c demonstrate the remarkable sensitivity of HREELS to small oxygen uptake, Growth of the loss peak at 65meV is accompanied by rapid quenching of the Ti-H related structure at 120meV. It is not obvious whether this results from chemical displacement of H or from screening of Ti-H losses. The 65meV loss itself has been previously associated with a so-called 0: Ti-0 phase [l]. This may simply correspond to oxygen in threefold hollow sites of Ti(OOOl), although it has been argued that the decrease in work function for oxygen exposures below 1L and the low 0+ yield in photon stimulated desorption may be associated with subsurface oxygen [1,123. Clearly, however, HREELS supports the hypothesis of two different adsorption states for oxygen on Ti(OOO1) [12,131. We consider next the HREELS data from O-exposed CoTi (figure 6). In contrast to Ti(OOOl), a significant loss peak is found at 65meV even before oxygen exposure. This is associated with the irreducible 01s intensity seen in XPS (section 2). The problems in cleaning CoTi appear to result from the greater mobility of oxygen in this alloy as compared with Ti and the uptake of around 1OOOppm dissolved oxygen impurity during crystal growth. The initial intensity of the 65meV loss is in fact much greater than can be seen on Ti(OOOl), despite the lower beam energy used in the study of the pure metal. This suggests that HREELS is probing a surface layer with significant concentration of dissolved oxygen impurity. The evolution of HREEL spectra with increasing oxygen exposure is in some way similar to that for Ti(OOOl), with increasing spectral weight appearing to the high binding energy side of the initial 65meV peak. However, there is no clear emergence of a dominant loss peak above 9OmeV and the spectra remain dissappointingly ill-structured. In fact XPS proves to be much more informative than HREELS in probing the oxidation of CoTi. The initial Ti2p spin orbit doublet of CoTi is shifted by about 0.6eV to higher binding energy as compared with Ti(OOOl), so that the 2~312 peak appears at 454.4eV (figure 7). This shift is due both to charge transfer from Ti to
1072
Co in the alloy and to the effects of the dissolved oxygen. With increasing oxygen exposure a high binding energy spin orbit doublet grows in intensity, with the 2~312component at 458.9eV as for Ti itself. After 15,OOOLoxygen exposure this doublet dominates the XPS. In contrast to Ti(OOOl),the 2~312oxide peak is clearly distinguishable from the initial 2~~2 peak because of the high binding energy shift of the latter.
:e)
W
a
0
,
I
50
100
Energy
I
I
150
200
loss lmeV
Figure6. HREELspectraof CoTi as a functionof oxygen exposure.(a) OL(b) IL (c) 4L (d) 400L (e) 4000L.Ep=SeV.Bi=8,=60’relativeto surfacenormal.
1073
In contrast to pronounced changes in the Ti2p region of the spectra, there is little change in the Co2p region with oxygen exposure up to 15,OOOL (figure 7). In particular, the changes are much more muted than for Co itself [141 where oxidation is accompanied both by emergence of chemically shifted Co2p peaks to 2.OeV higher binding energy of the original metal peaks and appearance of strong satellites associated with Co0 in the oxide overlayer. The picture to emerge from figures 7 and 8 is thus that oxidation of CoTi leads to formation of an esentially TiOa-like overlayer, which protects the Co component of the alloy from oxidation. This conclusion is supported by more detailed analysis of XPS data [9], which reveals that oxygen exposure is accompanied both by an decrease in the CopTi intensity ratio and an increase in background intensity in the Co2p region.
I
1’
(a)
M
lb)
(b) .
+J I
455 Binding
460
465
energylev
Figure 7. (Left hand panel) MgKa
4‘7(1
: 1
L I
I
760
790
Binding
800
I
energy/eV
excited XPS of CoTi in Ti 2p region. (a) clean surface
showing 2p spin orbit doublet with 2~3~ peak at 454.4eV. (b) after exposure to 15,OOOL02, showing new doublet with 2~3~ peak at 458.9eV. Figure 8. (Right hand panel) MgKa excited XPS of CoTi in Co2p region. (a) clean surface (b) after exposure to 15,OOOL 02 .The changes accompanying oxygen exposure are very small compared with those in figure 7.
1074 Whilst the HREELS data of figure 6 are not at variance with these conclusions, clearly the technique is less specific than XPS in distinguishing between Co and Ti oxides. However, HREELS does allow the possibility of monitor-ring H or D adsorption on CoTi. Somewhat surprisingly we have so far failed to observe hydrogen-associated vibrational modes on CoTi. It is unclear at present whether this is due to rapid diffusion of hydrogen into the bulk of CoTi or to inhibition of Hz dissociation on the crystal surface by the residual oxide.
4. ACKNOWIEDG~~. We are grateful to Harwell studentship to SJG.
Laboratories
and SERC for award
of a CASE
5. REFERENCES. 1. R.L. Strong and J.L. Erskine, J. Vat. Sci. Technol. A3 (1985) 1428. 2. A.F. Carley, P.R. Chalker, J.C. Riviere and M.W. Roberts, J. Chem. Sot. Faraday Trans. I, 83 (1987) 351. 3. Y. Osumi, H. Suzuki, A. Kato, K. Oguro and M. Nakane, J. Less Common Metals, 74 (1980) 271. 4. M.C. Burrell and N.R. Armstrong, Surf. Sci., 160 (1985) 235. 5. G. Rocker, J.A. Schaefer and W. Gape& Phys. Rev. B, 30 (1984) 3604 6. S. Eriksen and R.G. Egdell, Surf. Sci., 180 (1987) 263. 7. P.J. Feibelmann, D.R. Hamann and F.J. Himpsel, Phys. Rev. B, 22 (1980) 1734. 8. Y. Fukuda, F. Honda and J.W. Rabalais, Surf. Sci., 91 (1980) 165. 9. S. Garrett, R.G. Egdell and J.C. Rivibre, Surf. Sci., submitted. 10.S. Garrett, R.G. Egdell and J.C. Rivi&re, J. Phys. Condens. Matter,
1 (1989)
SB229. 11. W.R. Flavell, D. Phil. Thesis, Oxford 1986. 12.D.M. Hanson, R. Stockbauer and T.E. Madey, Phys. Rev. B, 24 (1981) 5513. 13.B.T. Jonker, J.F. Morar and R.L. Park, Phys. Rev. B, 6 (1971) 2951. 14.C.R. Brundle, T.J. Chuang and D.W. Rice, Surf. Sci., 60 (1976) 286.