- Email: [email protected]
Si interface by XPS
Vacuum/volume
46/number
Pergamon 0042-207X(OO)E0022-0
An experimental study of an interface practical Pd/Si interface by XPS Dao-xuan PR China
Dai, T D Lee Physics Laboratory,
Department
Z/pages 139 to 142/1995 Elsevier Science Ltd Printed in Great Britain 0042-207x/95 $9.50+.00
reaction at the
of Physics, Fudan University,
200433 Shanghai,
and I Davoli, Department received for publication
of Mathematics
and Physics,
Camerino
University,
62032 Camerino,
Italy
1 March 1994
X-ray photoelectron spectroscopy (XPS) measurements of core-level and valence band have been used to study the practical Pd/Si interface. Evidence of silicide, Pd,Si, formed by chemical reaction at this interface has been obtained. A monotonic increase of the branching ratio for two Pd 3d spin-orbit split peaks of silicide, Pd,Si, with decreasing photoemission angles was measured. Some possible interpretations for this change are also presented.
1. Introduction Transition metal/semiconductor interfaces were widely studied because of the great interest stimulated by the microscopic chemical reaction at these interfaces’-“. A monotonic decrease of the branching ratio of two spin-orbit Pt 4fpeaks with a decreasing photoemission angle has been measured at the practical Pt/InP, Pt/GaAs and PtjSi interfaces ‘J by the authors. Similar research at the practical PdjSi interface has not yet been systematically carried out. As a natural extension of the above work, in this paper we report on some experimental evidence of silicide, Pd,Si, formed at the practical PdjSi interface using XPS measurements. Contrary to the monotonic decrease of the branching ratio of Pt 4fpeaks at the practical Pt/InP, Pt/GaAs and Pt/Si interfaces, it is noted that a monotonic increase of the branching ratio of two spin-orbit Pd 3d peaks for silicide, PdSi, with a decreasing photoemission angle is seen at the practical PdjSi interface and some possible reasons for this are presented. 2. Experimental The experiments were performed on a VG ESCALAE-5 electron spectrometer equipped with X-ray photoelectron spectroscopy (XPS) and a defocused sputtering ion gun. This apparatus has been previously described elsewhere”. All XPS spectra were taken using a Mg K, X-ray source (ho = 1253.6 eV) with a 50 eV pass energy of the hemispherical analyser. The sample used in the present experiment was a lightly n-doped mirror-polished single crystal Si(ll1) wafer with a carrier concentration of 10” cmm3. In order to obtain the results ofthe practical PdjSi interface we adapted the etching and cleaning of the sample to standard procedure. After chemical etching in diluted HF, rinsing with deionized water and blow drying with Nz, the substrate surface was
checked by XPS. Several 8, of silicon oxide and small amounts of C and 0 contamination existed on this surface. The sample was then immediately transferred to the vacuum chamber and a Pd film of about 50 8, was deposited on the surface of Si at room temperature by the dc sputtering method for studying the PdjSi reaction at the interface. Also, as a standard sample, the single phase film of silicide, Pd$i, was obtained by heating another sample with 300 A Pd film at 300°C for 20 min in a nitrogen atmosphere, according to the standard technique”.
3. Results and Discussions 3.1. Interface reaction. 3.1.1. Core-level spectra. First, the contamination of the sample without sputtering was checked by XPS. We observed a strong Pd peak, other small 0 and C peaks, but no Si peak. It shows that small 0 and C contamination exists on this surface. In order to eliminate surface contamination the sample surface was sputtered for 10 min using an argon gun with an energy of 1 KeV and current density of 8 pA cm-*. XPS measurements showed that 0 and C peaks visibly decreased, but the Si peak was still not observed. After continuous sputtering, the intensity of the Pd peak decreased, but Pd and Si peaks coexisted, the intensities of Si 2p and Si 2s peaks were comparable to the intensity of Pd 3d peak. In this condition we take the XPS spectra as the results of the practical PdjSi interface. In Figure 1 the Pd 3d core-level spectra for (a) the clean metal Pd curve; (b) the practical PdjSi interface and (c) an uncontaminated Pd,Si silicide curve are shown. All the above XPS measurements were carried out under the same experimental conditions. The broadened curve (b) is due to the composition 139
Dao-xuan
Dai and I Davoli: Pd/Si interface
reaction
r
I
*
3%
I
.
333
cc
342
(
BE (eV) Figure 1. The Pd 3d core-level spectrum at the practical Pd/Si interface.
1
d - (( a
of the Pd 3d peak for the metal Pd, with the binding energy at 335.3 eV and for the bulk silicide, Pd&, with a binding energy at 336.7 eV, ref (45). The coexistence, in curve (b), of both these peaks is evidence of the interface reaction. Both chemical shift and broadening for the Pd 3d peak show that some chemical reaction between the metal Pd film and the surface of Si substrate took place at the practical PdjSi interface. Conversely, the binding energy of a Si 2p peak at this PdjSi interface, measured by XPS, is 99.5 eV (Figure 2) and is located between the value of the Si substrate at 99.3 eV, measured by XPS, and the value of the bulk silicide, Pd,Si, at 100.5 eV (Figure 2), further indicating that a more stable silicide, Pd$i, has been formed. 3.1.2. Valence band spectra. The above results are supported by the valence band spectrum at this PdjSi interface, as shown in Figure 3(e) which is very different from that of Pd metal shown in Figure 3(a), but nearer to that of silicide, Pd,Si, shown in Figure 3(c). The main peak with a binding energy of 3 eV shifts by about 2 eV to a higher binding energy (with respect to that of Pd metal) and the rise of the spectrum near EF is slower than that of Pd metal, demonstrating a redistribution of the DOS (density of states) for the valence band due to d-p hybridization.
3.2. Branching ratio. The Si 2p and Pd 3d core-level spectra, as a function of photo-emission angle 0 at the practical PdjSi interface, are shown in Figure 2 (a-f) and Figure 4, respectively. Angle B is the electron exit angle relative to the sample surface, providing more detailed information about the interface reaction. We assume that the modification of the emission lines of Pd 3d with 0 could be synthesized by the Lorentzian curve, Figure 1 (a) and (c). In this way, four peaks are found by the least squares fitting of a group of peaks obtained from Figure 1 to the experimental data, including the background for the 332-346 eV energy range. The results of such procedures are reported in the curves labelled 0 from 90” to 20” in Figure 4 (dotted lines). In order to minimize the photoelectron diffraction effect and the effect of the probable presence of other chemical states and/or the contamination at this interface under very low emission angles, such as 20”, our discussion is limited to the range of the emission angle 90’-30’. Now we will give some evidence of a Si rich outermost layer of the practical PdjSi interface. The normalized intensity ratio r(0) between Pd 3d5,* of the Pd$i silicide phase (Figure 4) and the 140
I yr
(
1 96
98
101
102
104
B.E (ev) Figure 2. The Si 2p core-level spectra:
(a-f) at PdjSi interface under different photoemission angles and (A) bulk Pd,Si.
Si 2p of the PdzSi silicide phase (Figure 2(a-f)), at this interface, decreases monotonically from 1.O to 0.68 with a decreasing emission angle from 90” to 30” [Figure 5(a)], indicating a Si rich (in silicide phase) outermost layer of this interface. Here we suppose r(0) = 1 at f3 = 90”. This result is also supported by the valence band spectra at this interface (Figure 3). Curve (e) at emission angle 90” has more features of the curve (c) for Pd,Si silicide; curve (d) at emission angle 30” has fewer features of curve (c) for Pd,Si silicide. A shoulder of curve (d) at emission angle 30” near 4 eV lower than that of curve (b) at emission angle 30”, might be due to a Si rich outermost layer. Up to now, the whole analysis has been made in a very traditional way without invoking a change in the branching ratio R = (Z,~,,‘I,:,). In fact, for both the bulk metal Pd and bulk silicide Pd$i, their branching ratios, of 1.3 and 1.46 respectively, remain unchanged with 0. According to the traditional explanation, the branching ratio R(0) for the silicide Pd,Si at this practical PdjSi interface should stay fixed ; on the contrary, it is very interesting that in Figure 5(b), R(B) increases monotonically from 1.39 to 1.93 when the emission angle decreases from 90” to 30”. Now we attempt to look for some reasons for interpreting the monotonic change of the branching ratio R(O) at the practical PdjSi the interface according to the above information from XPS and ARXPS (angular resolved XPS) measurements though it is difficult to give a unique explanation. 3.2.1.
Change of co-ordination
number. Starenberg
et al’”
Dao-xuan
Daiand
I Davoli:
Pd/Si interface
reaction
1 . Pd
cT 1.2
I
0.61
ia)
.
d
L
3.0”
5d
70
90’
J
8
Figure 5. R(B) and r(Q) vs emission angle 8 at this Pd/Si interface
0
2
4 6 8 ?O B.E (eV) Figure 3. Valence band spectra : (a) Pd metal, (b) bulk Pd,Si at 1) = 30’, (c) bulk Pd$i at 8 = 90’, (d) PdiSi interface interface at 19= 90”.
at 6’ = 30’ and (e) PdSi
observed the splitting spectrum of Cs 4s at low Cs coverage on a InP(ll0) surface and indicated that it is due to the existence of inequivalent Cs adsorption sites. Soukiassian et a/l4 measured that oxygen exposure changes the peak shape of Cs 4d and its branching ratio on an O,/Cs/Si(lOO) system. Parmigiani et alI5 reported the change of (AE) and the intensity ratio Z(Si 2p,,/Z(Si 2~+*) in an As-ion-implanted Si( 100) surface, even after laser annealing attributed it to the loss of co-ordination and the disordering of the surface Si atoms. As we know, the angular scan basically proves the concentration gradient at the interface, since the trend is that of reducing the probing depth when going to low emission angles.
WiSi
w 3%/2
332
336
B
340
E (eV)
344
I 332
336
340 BE
Figure 4. The change of Pd 3dcore-level interface.
with emission
344
I
I
(eV)
angle at this Pd/Si
According to the r(0) data of above ARXPS measurements, our complex practical PdjSi interface cannot be characterized by a single compound phase, though different silicide phases are difficult to identify. We know that the number of Pd neighbours around Si is three, ref(7), in bulk Pd$i phase (hexagonal Fe,P structure) and the number changes to six, ref(4, 11), in bulk PdSi phase (orthorhombic MnP structure), so we can deduce that the change of the co-ordination number in the normal direction to the interface is likely to be of importance for the monotonic changes of R(0). 3.2.2. Presence of metallic Pd environment. At the practical PdjSi interface we have observed two types of Pd co-ordination with comparable intensities (Figure 1), one at the energy of the Pd$i phase silicide, the other corresponding to a metallic environment. We suggest that the presence of metallic Pd may be another factor, because the asymmetic many-body tails connected with metallic environments would affect the height of a 3dx/2peak when measured with standard XPS energy resolution. Because the many-body electronic relaxation is complex and the relative changes of the photoionization cross-sections and asymmetry parameters for two core-levels, as a function of final state energy (chemical shifts), should be small, we do not involve them here. In short, we observed two kinds of changes of the branching ratio with decreasing emission angle from 90^ to 30”, one was a monotonic decrease at the practical Pt/Si, PtjInP and Pt/GaAs interfaces; another was a monotonic increase at the practical Pd/Si interface. The former was due to rich Pt in the outermost layer at these interfaces, the latter was due to rich Si in the outermost layer at this interface. Lastly, we would like to point out two technical problems. One is the ion-sputtering effect which may have had some influence on the coexistence of a metal Pd and a silicide, Pd,Si, and a rich Si on the outermost layer at this Pd/Si interface. For example, a sputtered sample is not flat, it can destroy the expected intensity variation. Another is the oxide effect; the thin surface oxide on Si can be partly decomposed during dc sputtering, so we assume that this influence on the main results was not serious, at least the results are a reflection of the PdjSi interface at the practical device.
Acknowledgements The authors would like to thank Professor Xide Xie for her guidance and encouragement. This work was supported by the National Natural Science Foundation of China. 141
Dao-xuan Dai and
I Davoli: Pd/Si interface reaction
References
‘H Roux, N Boutaoui and M Tholomier, Surface Sci, 260, 113 (1992). 2D A Evans, T P Chen, Th Chasse, K Horn, M von der Emde and DRT Zhan, Surface Sci, 2691270, 979 (1992). ‘A L Musatov, S Yu Smirnov, 0 V Rakhovskaya, S S Elovikov and E M Dubinina, Surface Sci, 269/270, 1048 (1992). 4P J Grunthaner, F J Grunthaner and A Madhukar, J Vuc Sci. Technol, 20, 680 (1982). ‘P J Grunthaner, F J Grunthaner, A Madhukar and J W Mayer, J Vat Sci Technol, 19,649 (1981). 6G W Rubloff and P S Ho, Thin Solid Films, 93, 21 (1982). ‘W D Bucklly and S C Mos, Solid St Electron, 15, 1331 (1972).
142
“Daoxuan Dai, Houshun Tang, Guobao Jiang, I Davoli and S Stizza, Applied Surface Sci, 62,249 ( 1992). ‘Daoxuan Dai and I Davoli, Vacuum, 44, 1189 (1993). loDaoxuan Dai and Furong Zhu, Phys Rev, B43,4803 (199 1). “J M Jaote, K N Tu and J W Mayer (Eds), Thin Film, Interdiffusion and Reaction, pp 376-377. Wiley, New York (1978). US Doniach and M Sunijic, J Ph_vs, C3,285 (1970). “H I Starenberg, P Soukiassian, M H Bakshi and Z Hurych, Surface Sci, 224, 13 (1989). 14P Soukiassian, M H Bahghi, Z Hurych and T M Gentle, Phy.7 Rev, B35,4176 (1987). 15F Parmigiani, P S Bagus, G Pacchioni and A Stella, Phys Rev, B41, 3728 (1990).
46/number
Pergamon 0042-207X(OO)E0022-0
An experimental study of an interface practical Pd/Si interface by XPS Dao-xuan PR China
Dai, T D Lee Physics Laboratory,
Department
Z/pages 139 to 142/1995 Elsevier Science Ltd Printed in Great Britain 0042-207x/95 $9.50+.00
reaction at the
of Physics, Fudan University,
200433 Shanghai,
and I Davoli, Department received for publication
of Mathematics
and Physics,
Camerino
University,
62032 Camerino,
Italy
1 March 1994
X-ray photoelectron spectroscopy (XPS) measurements of core-level and valence band have been used to study the practical Pd/Si interface. Evidence of silicide, Pd,Si, formed by chemical reaction at this interface has been obtained. A monotonic increase of the branching ratio for two Pd 3d spin-orbit split peaks of silicide, Pd,Si, with decreasing photoemission angles was measured. Some possible interpretations for this change are also presented.
1. Introduction Transition metal/semiconductor interfaces were widely studied because of the great interest stimulated by the microscopic chemical reaction at these interfaces’-“. A monotonic decrease of the branching ratio of two spin-orbit Pt 4fpeaks with a decreasing photoemission angle has been measured at the practical Pt/InP, Pt/GaAs and PtjSi interfaces ‘J by the authors. Similar research at the practical PdjSi interface has not yet been systematically carried out. As a natural extension of the above work, in this paper we report on some experimental evidence of silicide, Pd,Si, formed at the practical PdjSi interface using XPS measurements. Contrary to the monotonic decrease of the branching ratio of Pt 4fpeaks at the practical Pt/InP, Pt/GaAs and Pt/Si interfaces, it is noted that a monotonic increase of the branching ratio of two spin-orbit Pd 3d peaks for silicide, PdSi, with a decreasing photoemission angle is seen at the practical PdjSi interface and some possible reasons for this are presented. 2. Experimental The experiments were performed on a VG ESCALAE-5 electron spectrometer equipped with X-ray photoelectron spectroscopy (XPS) and a defocused sputtering ion gun. This apparatus has been previously described elsewhere”. All XPS spectra were taken using a Mg K, X-ray source (ho = 1253.6 eV) with a 50 eV pass energy of the hemispherical analyser. The sample used in the present experiment was a lightly n-doped mirror-polished single crystal Si(ll1) wafer with a carrier concentration of 10” cmm3. In order to obtain the results ofthe practical PdjSi interface we adapted the etching and cleaning of the sample to standard procedure. After chemical etching in diluted HF, rinsing with deionized water and blow drying with Nz, the substrate surface was
checked by XPS. Several 8, of silicon oxide and small amounts of C and 0 contamination existed on this surface. The sample was then immediately transferred to the vacuum chamber and a Pd film of about 50 8, was deposited on the surface of Si at room temperature by the dc sputtering method for studying the PdjSi reaction at the interface. Also, as a standard sample, the single phase film of silicide, Pd$i, was obtained by heating another sample with 300 A Pd film at 300°C for 20 min in a nitrogen atmosphere, according to the standard technique”.
3. Results and Discussions 3.1. Interface reaction. 3.1.1. Core-level spectra. First, the contamination of the sample without sputtering was checked by XPS. We observed a strong Pd peak, other small 0 and C peaks, but no Si peak. It shows that small 0 and C contamination exists on this surface. In order to eliminate surface contamination the sample surface was sputtered for 10 min using an argon gun with an energy of 1 KeV and current density of 8 pA cm-*. XPS measurements showed that 0 and C peaks visibly decreased, but the Si peak was still not observed. After continuous sputtering, the intensity of the Pd peak decreased, but Pd and Si peaks coexisted, the intensities of Si 2p and Si 2s peaks were comparable to the intensity of Pd 3d peak. In this condition we take the XPS spectra as the results of the practical PdjSi interface. In Figure 1 the Pd 3d core-level spectra for (a) the clean metal Pd curve; (b) the practical PdjSi interface and (c) an uncontaminated Pd,Si silicide curve are shown. All the above XPS measurements were carried out under the same experimental conditions. The broadened curve (b) is due to the composition 139
Dao-xuan
Dai and I Davoli: Pd/Si interface
reaction
r
I
*
3%
I
.
333
cc
342
(
BE (eV) Figure 1. The Pd 3d core-level spectrum at the practical Pd/Si interface.
1
d - (( a
of the Pd 3d peak for the metal Pd, with the binding energy at 335.3 eV and for the bulk silicide, Pd&, with a binding energy at 336.7 eV, ref (45). The coexistence, in curve (b), of both these peaks is evidence of the interface reaction. Both chemical shift and broadening for the Pd 3d peak show that some chemical reaction between the metal Pd film and the surface of Si substrate took place at the practical PdjSi interface. Conversely, the binding energy of a Si 2p peak at this PdjSi interface, measured by XPS, is 99.5 eV (Figure 2) and is located between the value of the Si substrate at 99.3 eV, measured by XPS, and the value of the bulk silicide, Pd,Si, at 100.5 eV (Figure 2), further indicating that a more stable silicide, Pd$i, has been formed. 3.1.2. Valence band spectra. The above results are supported by the valence band spectrum at this PdjSi interface, as shown in Figure 3(e) which is very different from that of Pd metal shown in Figure 3(a), but nearer to that of silicide, Pd,Si, shown in Figure 3(c). The main peak with a binding energy of 3 eV shifts by about 2 eV to a higher binding energy (with respect to that of Pd metal) and the rise of the spectrum near EF is slower than that of Pd metal, demonstrating a redistribution of the DOS (density of states) for the valence band due to d-p hybridization.
3.2. Branching ratio. The Si 2p and Pd 3d core-level spectra, as a function of photo-emission angle 0 at the practical PdjSi interface, are shown in Figure 2 (a-f) and Figure 4, respectively. Angle B is the electron exit angle relative to the sample surface, providing more detailed information about the interface reaction. We assume that the modification of the emission lines of Pd 3d with 0 could be synthesized by the Lorentzian curve, Figure 1 (a) and (c). In this way, four peaks are found by the least squares fitting of a group of peaks obtained from Figure 1 to the experimental data, including the background for the 332-346 eV energy range. The results of such procedures are reported in the curves labelled 0 from 90” to 20” in Figure 4 (dotted lines). In order to minimize the photoelectron diffraction effect and the effect of the probable presence of other chemical states and/or the contamination at this interface under very low emission angles, such as 20”, our discussion is limited to the range of the emission angle 90’-30’. Now we will give some evidence of a Si rich outermost layer of the practical PdjSi interface. The normalized intensity ratio r(0) between Pd 3d5,* of the Pd$i silicide phase (Figure 4) and the 140
I yr
(
1 96
98
101
102
104
B.E (ev) Figure 2. The Si 2p core-level spectra:
(a-f) at PdjSi interface under different photoemission angles and (A) bulk Pd,Si.
Si 2p of the PdzSi silicide phase (Figure 2(a-f)), at this interface, decreases monotonically from 1.O to 0.68 with a decreasing emission angle from 90” to 30” [Figure 5(a)], indicating a Si rich (in silicide phase) outermost layer of this interface. Here we suppose r(0) = 1 at f3 = 90”. This result is also supported by the valence band spectra at this interface (Figure 3). Curve (e) at emission angle 90” has more features of the curve (c) for Pd,Si silicide; curve (d) at emission angle 30” has fewer features of curve (c) for Pd,Si silicide. A shoulder of curve (d) at emission angle 30” near 4 eV lower than that of curve (b) at emission angle 30”, might be due to a Si rich outermost layer. Up to now, the whole analysis has been made in a very traditional way without invoking a change in the branching ratio R = (Z,~,,‘I,:,). In fact, for both the bulk metal Pd and bulk silicide Pd$i, their branching ratios, of 1.3 and 1.46 respectively, remain unchanged with 0. According to the traditional explanation, the branching ratio R(0) for the silicide Pd,Si at this practical PdjSi interface should stay fixed ; on the contrary, it is very interesting that in Figure 5(b), R(B) increases monotonically from 1.39 to 1.93 when the emission angle decreases from 90” to 30”. Now we attempt to look for some reasons for interpreting the monotonic change of the branching ratio R(O) at the practical PdjSi the interface according to the above information from XPS and ARXPS (angular resolved XPS) measurements though it is difficult to give a unique explanation. 3.2.1.
Change of co-ordination
number. Starenberg
et al’”
Dao-xuan
Daiand
I Davoli:
Pd/Si interface
reaction
1 . Pd
cT 1.2
I
0.61
ia)
.
d
L
3.0”
5d
70
90’
J
8
Figure 5. R(B) and r(Q) vs emission angle 8 at this Pd/Si interface
0
2
4 6 8 ?O B.E (eV) Figure 3. Valence band spectra : (a) Pd metal, (b) bulk Pd,Si at 1) = 30’, (c) bulk Pd$i at 8 = 90’, (d) PdiSi interface interface at 19= 90”.
at 6’ = 30’ and (e) PdSi
observed the splitting spectrum of Cs 4s at low Cs coverage on a InP(ll0) surface and indicated that it is due to the existence of inequivalent Cs adsorption sites. Soukiassian et a/l4 measured that oxygen exposure changes the peak shape of Cs 4d and its branching ratio on an O,/Cs/Si(lOO) system. Parmigiani et alI5 reported the change of (AE) and the intensity ratio Z(Si 2p,,/Z(Si 2~+*) in an As-ion-implanted Si( 100) surface, even after laser annealing attributed it to the loss of co-ordination and the disordering of the surface Si atoms. As we know, the angular scan basically proves the concentration gradient at the interface, since the trend is that of reducing the probing depth when going to low emission angles.
WiSi
w 3%/2
332
336
B
340
E (eV)
344
I 332
336
340 BE
Figure 4. The change of Pd 3dcore-level interface.
with emission
344
I
I
(eV)
angle at this Pd/Si
According to the r(0) data of above ARXPS measurements, our complex practical PdjSi interface cannot be characterized by a single compound phase, though different silicide phases are difficult to identify. We know that the number of Pd neighbours around Si is three, ref(7), in bulk Pd$i phase (hexagonal Fe,P structure) and the number changes to six, ref(4, 11), in bulk PdSi phase (orthorhombic MnP structure), so we can deduce that the change of the co-ordination number in the normal direction to the interface is likely to be of importance for the monotonic changes of R(0). 3.2.2. Presence of metallic Pd environment. At the practical PdjSi interface we have observed two types of Pd co-ordination with comparable intensities (Figure 1), one at the energy of the Pd$i phase silicide, the other corresponding to a metallic environment. We suggest that the presence of metallic Pd may be another factor, because the asymmetic many-body tails connected with metallic environments would affect the height of a 3dx/2peak when measured with standard XPS energy resolution. Because the many-body electronic relaxation is complex and the relative changes of the photoionization cross-sections and asymmetry parameters for two core-levels, as a function of final state energy (chemical shifts), should be small, we do not involve them here. In short, we observed two kinds of changes of the branching ratio with decreasing emission angle from 90^ to 30”, one was a monotonic decrease at the practical Pt/Si, PtjInP and Pt/GaAs interfaces; another was a monotonic increase at the practical Pd/Si interface. The former was due to rich Pt in the outermost layer at these interfaces, the latter was due to rich Si in the outermost layer at this interface. Lastly, we would like to point out two technical problems. One is the ion-sputtering effect which may have had some influence on the coexistence of a metal Pd and a silicide, Pd,Si, and a rich Si on the outermost layer at this Pd/Si interface. For example, a sputtered sample is not flat, it can destroy the expected intensity variation. Another is the oxide effect; the thin surface oxide on Si can be partly decomposed during dc sputtering, so we assume that this influence on the main results was not serious, at least the results are a reflection of the PdjSi interface at the practical device.
Acknowledgements The authors would like to thank Professor Xide Xie for her guidance and encouragement. This work was supported by the National Natural Science Foundation of China. 141
Dao-xuan Dai and
I Davoli: Pd/Si interface reaction
References
‘H Roux, N Boutaoui and M Tholomier, Surface Sci, 260, 113 (1992). 2D A Evans, T P Chen, Th Chasse, K Horn, M von der Emde and DRT Zhan, Surface Sci, 2691270, 979 (1992). ‘A L Musatov, S Yu Smirnov, 0 V Rakhovskaya, S S Elovikov and E M Dubinina, Surface Sci, 269/270, 1048 (1992). 4P J Grunthaner, F J Grunthaner and A Madhukar, J Vuc Sci. Technol, 20, 680 (1982). ‘P J Grunthaner, F J Grunthaner, A Madhukar and J W Mayer, J Vat Sci Technol, 19,649 (1981). 6G W Rubloff and P S Ho, Thin Solid Films, 93, 21 (1982). ‘W D Bucklly and S C Mos, Solid St Electron, 15, 1331 (1972).
142
“Daoxuan Dai, Houshun Tang, Guobao Jiang, I Davoli and S Stizza, Applied Surface Sci, 62,249 ( 1992). ‘Daoxuan Dai and I Davoli, Vacuum, 44, 1189 (1993). loDaoxuan Dai and Furong Zhu, Phys Rev, B43,4803 (199 1). “J M Jaote, K N Tu and J W Mayer (Eds), Thin Film, Interdiffusion and Reaction, pp 376-377. Wiley, New York (1978). US Doniach and M Sunijic, J Ph_vs, C3,285 (1970). “H I Starenberg, P Soukiassian, M H Bakshi and Z Hurych, Surface Sci, 224, 13 (1989). 14P Soukiassian, M H Bahghi, Z Hurych and T M Gentle, Phy.7 Rev, B35,4176 (1987). 15F Parmigiani, P S Bagus, G Pacchioni and A Stella, Phys Rev, B41, 3728 (1990).