Study of the Oxidation of Molybdenum Surfaces by Energy Loss Spectroscopy Combined with Auger Electron Spectroscopy B Y TOMOJI KAWAI, KIMIO KUNIMORI, TAMOTSU KONDOW, TAKAHARU ONISHI AND KENZI TAMARU" Department of Chemistry, The University of Tokyo, Hongo, Bunkyo-Ku, Tokyo, Japan Received 3rd April, 1973 The oxidation of molybdenum as an evaporated film and of its annealed surface was investigated by Energy Loss Spectroscopy (ELS) combined with Auger Electron Spectroscopy (AES). The 10 eV and the lower energy loss peaks towards 5 eV were identified as the clean surface and the oxidized layer surface plasma loss peaks, respectively. The behaviours of these surface plasma loss spectra revealed that the oxidation of the evaporated film surface proceeded uniformly, rapidly and without any induction period under vacuum of lo-' Torr, whereas that of the annealed sample proceeded more slowly through " oxidized patches ", exhibiting a considerable induction period under the same vacuum conditions.These phenomena were interpreted with reference to the surface conditions and structures studied by the AES and ELS ; the fresh evaporated surface is clean and rough, whereas the annealed surface is smoothed by the heat-treatment and contaminated by sulphur and carbon which partially cover the active surfaces and give rise to the patchy surface oxidation. Energy analysis of scattered electrons has recently been shown to be an extremely useful tool for the study of solid ~urfaces,l-~ especially when various techniques are combined, for instance, low energy electron diffraction and Auger electron spectro- scopy (AES).5* Energy loss spectroscopy (ELS) is of particular interest since it is suitable for revealing the electronic structures of the surfaces of solids and adsorb- a t e ~ .~ ~ AES, on the other hand, reveals the atomic identities existing on the sur- faces. Therefore, simultaneous measurements of ELS and AES may provide a deeper understanding of the surface properties and reactions than is possible with either one alone. On a metal surface, a surface mode of plasma excitation (surface plasmon) exists and can be observed by ELS, preferably in a reflection geometry. The surface mode appears at a lower frequency than the corresponding bulk mode, and shifts and/or reduces its intensity as the surface becomes contaminated by foreign ~pecies.~ In one case, the energy loss peak due to the surface plasmon on the clean surface shifts in the direction of lower energy, approaching a definite position of energy loss as the contamination proceeds. In another case, this initial peak gradually vanishes while another peak appears at the final energy position of the previous case.7 In the oxidation of aluminium, the surface plasma loss peak adopted the former behaviour under high vacuum condition ( Torr), it behaved as in the latter case.2 These phenomena cannot be explained by the different collision number of the gaseous molecules on the surface, but could be closely connected with the mechanism of the surface oxidation processes itself, so 137 T ~ r r ) , ~ whereas under lower vacuum138 OXIDATION OF MO that it can be investigated through the analysis of the behaviours of the surface plasmon loss peaks with reference to the surface conditions.In this paper, the oxidation process of a molybdenum film has been studied by varying its surface properties, taking advantage of the combined ELS and AES techniques. H 60 40 20 I mass number FIG. 1 . 4 0 ) The block diagram of the electron spectrometer for the ELS and the AES. This is a modified Simpson-Kuyatt type with 180" hemispherical energy selector whose optical circle has 25 mm radius. (b) Mass spectrum of the residual gas during the measurements.T . KAWAI, K . KUNIMORI, T. KONDOW, T . ONISHI, K . TAMARU 139 EXPERIMENTAL A modified Simpson-Kuyatt type electron spectrometer was constructed for the measure- ments of ELS and AES (fig.l(a)). The monochromator, as well as the analyzer, was a 180" hemispherical electrostatic lens pair, average radius of 2.5 cm, equipped with retarding electrostatic lenses. The energy resolution (AEG 500 meV) was high and constant enough, throughout the energy range of the measurements, to obtain the true shapes of the energy loss spectra and the Auger electron spectra. The scattered electrons, after being velocity selected, were detected by a single electron counting method. The collision chamber and the main chamber of the spectrometer were differentially pumped. This permitted one to introduce foreign gases into the collision chamber without disturbing the spectrometer. An evaporated Mo film was prepared and annealed in the collision chamber without exposure to air.Molybdenum (H. Cross Co. Ltd. 99.96 % purity) was evaporated onto a Mo foil which was fixed on a sample holder in the centre of the chamber. The Mo foil was chemically etched, heated to 1400°C in uacuo, then alternatively exposed to oxygen and hydrogen and finally checked for surface impurities by AES examination. Both chambers were evacuated by oil diffusion and rotary pumps. In order to prevent these chambers from being contam- inated, a foreline alumina trap was used as well as an ordinary liquid N2 trap. The pressure was measured by Bayard-Alpert gauge and was of the order 1 x lo-' Torr during the measure- ments. A NEVA quadrupole mass spectrometer was used to mass-analyze the residual gases (H20>90 %) (fig.l(b)). The operating parameters were as follows : energy of the primary electron beam, 380 eV for the ELS and 1.5 keV for the AES, primary beam current, 2x lo-* A for the ELS and 3 FA for the AES ; the incident and scattering angles, 27" each. RESULTS AND DISCUSSION 1. OXIDATION OF A FRESHLY EVAPORATED M O FILM When Mo was instantaneously evaporated on a Mo foil under a pressure of 1.2 x Torr, the energy loss spectra exhibited distinct time-dependent changes as is demonstrated in fig. 2. The Mo foil itself had two loss peaks at 23 eV and 5 eV before the evaporation. In terms of a free electron model based on the assumption of 6 free electrons (4a5(5s)l per atom, the 23 eV peak is assigned to be a bulk plasmon e~citation.~. l o After evaporation two peaks appeared at 10 eV and 23 eV the 5 eV loss peak having disappeared.This 23 eV peak did not change during the measure- ments. On the other hand, the 10 eV peak shifted with time towards low energies and finally reached a position of 5 eV energy loss. It took about 30 min to accomplish the complete shift. The rate of the shift was rather faster at the beginning, without an induction period, decreasing gradually with time. The Auger spectra taken immediately after the evaporation exhibited peaks due to molybdenum and oxygen atoms together with very weak carbon peaks which became evident in the latter half of the measurements. Just after the evaporation the amount of surface carbon was less than a few percent of monolayer coverage, judged from the relative intensities of carbon and molybdenum and the change in their intensity.The exact amount of the surface oxygen at the initial stage of the oxidation was difficult to estimate because of the rapid oxidation rate, but the extrapolated value for the oxygen was estimated to be nearly the same as that for carbon. Considerable growth of a 5 15 eV oxygen peak with time indicates that " swface oxidation " proceeds, as shown in fig. 3(a). It is highly likely that the M o surface is oxidized by water, which was one of the major residual gases, as indicated by mass analysis of background gases in the vacuum chambers. Stern and Ferrell have theoretically predicted that the loss peak of surface plasmon undergoes a red-shift depending on the thickness of the oxidized layer formed on the140 OXIDATION OF M O 6) 38min 5) 30min 4 ) Z m i n 3) 12min 2) 6 min 1 ) 3 m i n after e va po rat i D n 8cfore evaporation 1 1 I 1 0 Ib 'io 3b 9 energy losslev FIG.2.-The variation of the characteristic energy loss spectra with time, for the freshly evaporated molybdenum surface. S ( b ) 0 1 v 160 ZbO 240 3b0 500 550 kinetic energy/eV FIG. 3.-Auger electron spectra from a molybdenum surface. (a) Freshly evaporated surface. (6) After anneaiing in hydrogen at 1.5 x Sulphur and carbon are ob- Torr, looO°C, for 20 min. served on this annealed surface.T . KAWAI, K . KUNIMORI, T . KONDOW, T . ONISHI, K . TAMARU 141 metal, provided that the layer is formed uniformly.7* l1 * It can, accordingly, be concluded that the time-dependent peak is attributable to surface plasmon excitation.The thickness of the layer was estimated from the equation of Stern and Ferrell, assuming its validity, and was plotted as a function of reaction time (see fig. 5). In conclusion, the thickness of the oxidized layer increased with time in a uniform manner over the whole surface on this evaporated sample, similar to the oxidation of an A1 film under high vacuum c~nditions.~ If the assignment of the 23 eV peak is correct, the surface plasma loss on the clean surface should be 23/,/2 = 16.5 eV on the basis of the free electron model. This theoretical prediction does not agree with our experimental results. If one or more interband transitions occur with the energy close to that of the surface plasmon excitation expected from the free electron model, the observed surface plasmon loss may deviate from the theoretical prediction of the free electron model.2. OXIDATION OF THE FILM ANNEALED IN HYDROGEN ATMOSPHERE The evaporated film, whose surface plasmon loss was 5eV, was annealed in hydrogen at 1.5 x Torr pressure and a temperature of 1000°C for 20 min. The loss spectra of the annealed specimen were examined at different times after the annealing (see fig. 4). The other experimental conditions were identical to the case of the evaporated film. The behaviours and the shapes of the surface plasmon were quite different from those of the evaporated sample. Immediately after the treat- ment, the 10 eV loss peak was found as well as that at 23 eV. As the surface oxidation proceeded, a new peak appeared at 5 eV and coexisted with the peak at 10 eV whose intensity was gradually decreasing.These two peaks did not shift with time but their intensities changed. The new peak at 5 eV corresponds to the 5 eV peak observed in the experiment on the freshly evaporated film, i.e., the loss peak by the surface plasma excitation of the Mo surfzce covered by a sufficient amount of the oxidized layer. This phenomenon shows that both the oxidized part and non-oxidized part coexist during the course of the oxidation, and suggests that the oxide nuclei grow at the surface layer, the oxidation proceeding at the boundary between the metal and its oxide. The AES of this specimen demonstrated the appearance of sulphur and carbon atoms immediately after the annealing as is shown in fig.3(6). Seemingly, sulphur and carbon had diffused from the inside onto the surface of the specimen during the annealing procedure. Considerable parts of this annealed surface are covered by carbon and sulphur judging from the reduction of the intensity of the molybdenum. The ratio of C and S peak areas to Mo AES peak area were 2.1 and 3.2 respectively. In addition, the relative intensity of the surface to the bulk loss on the fresh evaporated film was approximately 25 times larger than on the annealed one. This means that the surface area of the evaporated film may be larger than that of the annealed one, that is, the surface of the evaporated one may be rougher, though the contaminants may reduce their relative intensity to some extent.12* l 3 3" E+tanhkD 2&+(1 +cZ) tanh kD us = where us is the frequency of the surface plasmon when a surface oxidized layer of dielectric constant E is formed to thickness D ; up is the frequency of the bulk plasmon (fiup = 23 eV for Mo) ; k repre- sents the wave number of the surface wave excited in the solid by the incident electron and is obtained from the momentum and the energy conservation law ; k = mus cos Bltiko where 8 is the incident and scattering angle (27") and ko is the wave number of the incident electron.142 OXIDATION OF M O In fig.6, graph (a) and (b), the percentage ratio of the 5 eV peak to the sum of the 5 and the 10 eV one was plotted against time (t) and t 2 . There is an induction period of about 10 min. In the initial stage of the reaction, the ratio was proportional to t 2 but above a ratio of 50 :d it was expressed in terms of aft2 +/I, where a' and are I t 10) After one day 9) 58min 8) 52min 7) 46min 6 ) 38min 5) 32min 4) 25 min 3) 19 min 2 ) 13 min I ) 4 min a f t e r l e a t i n g up in H2, 1 .5 ~ mmHg 20 min 0 10 20 30 energy losslev FIG. 4.-The variation of the characteristic energy loss spectra with time, for the molybdenum surface annealed in hydrogen at 1.5 x Torr, 1000°C, for 20 min. FIG. 5.-The /O 10 20 30 40 50 elapsed timelniin thickness of the oxidized layer as a function of elapsed time after evayora tion.T. KAWAI, K . KUNIMORI, T . KONDOW, T . ONISHI, K . TAMARU 143 time-independent constants. These phenomena can be explained qualitatively as follows. Assuming that the diameter of the circular patches increases with a constant velocity and the size of every patch is approximately the same, the ratio may be roughly equal to ((n.nE2)/So)t2 before these patches overlap.Here n is the number of the active points, So is the total surface area and E represents the average growth velocity of the diameter. After the overlapping, the effective n may decrease. There- fore the slope is smaller than that at the initial stage. t P) elapsed timelmin 5 + square of the elapsed timelmin FIG. 6.-The percentage ratio of the 5 eV peak to the sum of the 5 and 10 eV ones as a function of (a) elapsed time after annealing in hydrogen in hydrogen and (b) square of the elapsed time. The different behaviours in the ELS and the AES on these different surfaces is of interest in the view of the correlation between the surface reaction and the surface properties.Considerations on the rates of the oxidation at the early stage (fig. 5(a), fig. 6(a)) revealed that the uniform oxidation of the Mo surface starts rapidly just after the evaporation, which suggests a large collisional cross section for the reaction between gases and the evaporated surface. The induction period of about 10 min on the annealed surface, on the other hand, implies that the oxidation scarcely proceeded at the beginning and after about 10min began to develop rapidly. The oxidation cross section is very small at the beginning. As to the surface properties, AES demonstrated that the fresh evaporated surface is clean and is scarcely contaminated by other elements, whereas the annealed surface is contaminated by considerable amounts of carbon and sulphur.Another difference between the films would be the geometric structure of the surfaces, i.e., the roughness factor of the evaporated surface is larger than that of the annealed one. Taking these facts into consideration, the different behaviours in Mo oxidation are interpreted as follows ; on the fresh evaporated surface which is rough and active,144 OXIDATION OF M O the H20 in the ambient gas attacks the surface, colliding with the whole surface area, and the oxidation proceeds uniformly into the inside of the film. On the annealed surface, which is partially covered by carbon or sulphur and not so rough as the evaporated film, on the other hand, the oxidation can start only at a limited number of reactive points, and the oxidized part develops its area through the boundaries of the two phases, forming the " oxidized patches ".The authors are grateful to Prof. Kozo Kuchitsu of the University of Tokyo and Dr. Katsuya Nakayama of the Electrotechnical Laboratory for valuable discussions. L. A. Harris, J. Appl. Phys., 1968, 39, 1419. C. J. Powell and J. B. Swan, Phys. Rev., 1960, 118, 640. E. J. Sheibner and L. N . Tharp, Surface Sci., 1967, 8, 427. D. Edwards, Jr., and F. M. Propst, J. Chem. Phys., 1971,55, 5175. R. E. Weber and W. T. Peria, J. Appl. Phys., 1967, 38,4355. P. W. Palmberg and T. N. Rhodin, J. Appl. Phys., 1968, 39,2425. H . Raether, Springer Tracts in Modern Physics, 1965, 38, 84. H. Iback, J. Vac.Sci. Technol., 1972, 9, 713. C. Kunz, Z. Phys., 1966, 196, 311. lo G. J. Dooley and T. W. Haas, .I. Chem. Phys., 1970,52,993. E. A. Stern and R. A. Ferrell, Phys. Rev., 1960, 120, 130. l2 C. J. Powell, Phys. Rev., 1968, 175, 972. l 3 J. W. Swaine and R. C. Plumb, J. Appl. Phys., 1962, 33, 2378. Study of the Oxidation of Molybdenum Surfaces by Energy Loss Spectroscopy Combined with Auger Electron Spectroscopy B Y TOMOJI KAWAI, KIMIO KUNIMORI, TAMOTSU KONDOW, TAKAHARU ONISHI AND KENZI TAMARU" Department of Chemistry, The University of Tokyo, Hongo, Bunkyo-Ku, Tokyo, Japan Received 3rd April, 1973 The oxidation of molybdenum as an evaporated film and of its annealed surface was investigated by Energy Loss Spectroscopy (ELS) combined with Auger Electron Spectroscopy (AES).The 10 eV and the lower energy loss peaks towards 5 eV were identified as the clean surface and the oxidized layer surface plasma loss peaks, respectively. The behaviours of these surface plasma loss spectra revealed that the oxidation of the evaporated film surface proceeded uniformly, rapidly and without any induction period under vacuum of lo-' Torr, whereas that of the annealed sample proceeded more slowly through " oxidized patches ", exhibiting a considerable induction period under the same vacuum conditions. These phenomena were interpreted with reference to the surface conditions and structures studied by the AES and ELS ; the fresh evaporated surface is clean and rough, whereas the annealed surface is smoothed by the heat-treatment and contaminated by sulphur and carbon which partially cover the active surfaces and give rise to the patchy surface oxidation.Energy analysis of scattered electrons has recently been shown to be an extremely useful tool for the study of solid ~urfaces,l-~ especially when various techniques are combined, for instance, low energy electron diffraction and Auger electron spectro- scopy (AES).5* Energy loss spectroscopy (ELS) is of particular interest since it is suitable for revealing the electronic structures of the surfaces of solids and adsorb- a t e ~ . ~ ~ AES, on the other hand, reveals the atomic identities existing on the sur- faces. Therefore, simultaneous measurements of ELS and AES may provide a deeper understanding of the surface properties and reactions than is possible with either one alone.On a metal surface, a surface mode of plasma excitation (surface plasmon) exists and can be observed by ELS, preferably in a reflection geometry. The surface mode appears at a lower frequency than the corresponding bulk mode, and shifts and/or reduces its intensity as the surface becomes contaminated by foreign ~pecies.~ In one case, the energy loss peak due to the surface plasmon on the clean surface shifts in the direction of lower energy, approaching a definite position of energy loss as the contamination proceeds. In another case, this initial peak gradually vanishes while another peak appears at the final energy position of the previous case.7 In the oxidation of aluminium, the surface plasma loss peak adopted the former behaviour under high vacuum condition ( Torr), it behaved as in the latter case.2 These phenomena cannot be explained by the different collision number of the gaseous molecules on the surface, but could be closely connected with the mechanism of the surface oxidation processes itself, so 137 T ~ r r ) , ~ whereas under lower vacuum138 OXIDATION OF MO that it can be investigated through the analysis of the behaviours of the surface plasmon loss peaks with reference to the surface conditions.In this paper, the oxidation process of a molybdenum film has been studied by varying its surface properties, taking advantage of the combined ELS and AES techniques. H 60 40 20 I mass number FIG. 1 . 4 0 ) The block diagram of the electron spectrometer for the ELS and the AES.This is a modified Simpson-Kuyatt type with 180" hemispherical energy selector whose optical circle has 25 mm radius. (b) Mass spectrum of the residual gas during the measurements.T . KAWAI, K . KUNIMORI, T. KONDOW, T . ONISHI, K . TAMARU 139 EXPERIMENTAL A modified Simpson-Kuyatt type electron spectrometer was constructed for the measure- ments of ELS and AES (fig. l(a)). The monochromator, as well as the analyzer, was a 180" hemispherical electrostatic lens pair, average radius of 2.5 cm, equipped with retarding electrostatic lenses. The energy resolution (AEG 500 meV) was high and constant enough, throughout the energy range of the measurements, to obtain the true shapes of the energy loss spectra and the Auger electron spectra. The scattered electrons, after being velocity selected, were detected by a single electron counting method.The collision chamber and the main chamber of the spectrometer were differentially pumped. This permitted one to introduce foreign gases into the collision chamber without disturbing the spectrometer. An evaporated Mo film was prepared and annealed in the collision chamber without exposure to air. Molybdenum (H. Cross Co. Ltd. 99.96 % purity) was evaporated onto a Mo foil which was fixed on a sample holder in the centre of the chamber. The Mo foil was chemically etched, heated to 1400°C in uacuo, then alternatively exposed to oxygen and hydrogen and finally checked for surface impurities by AES examination. Both chambers were evacuated by oil diffusion and rotary pumps. In order to prevent these chambers from being contam- inated, a foreline alumina trap was used as well as an ordinary liquid N2 trap.The pressure was measured by Bayard-Alpert gauge and was of the order 1 x lo-' Torr during the measure- ments. A NEVA quadrupole mass spectrometer was used to mass-analyze the residual gases (H20>90 %) (fig. l(b)). The operating parameters were as follows : energy of the primary electron beam, 380 eV for the ELS and 1.5 keV for the AES, primary beam current, 2x lo-* A for the ELS and 3 FA for the AES ; the incident and scattering angles, 27" each. RESULTS AND DISCUSSION 1. OXIDATION OF A FRESHLY EVAPORATED M O FILM When Mo was instantaneously evaporated on a Mo foil under a pressure of 1.2 x Torr, the energy loss spectra exhibited distinct time-dependent changes as is demonstrated in fig.2. The Mo foil itself had two loss peaks at 23 eV and 5 eV before the evaporation. In terms of a free electron model based on the assumption of 6 free electrons (4a5(5s)l per atom, the 23 eV peak is assigned to be a bulk plasmon e~citation.~. l o After evaporation two peaks appeared at 10 eV and 23 eV the 5 eV loss peak having disappeared. This 23 eV peak did not change during the measure- ments. On the other hand, the 10 eV peak shifted with time towards low energies and finally reached a position of 5 eV energy loss. It took about 30 min to accomplish the complete shift. The rate of the shift was rather faster at the beginning, without an induction period, decreasing gradually with time. The Auger spectra taken immediately after the evaporation exhibited peaks due to molybdenum and oxygen atoms together with very weak carbon peaks which became evident in the latter half of the measurements.Just after the evaporation the amount of surface carbon was less than a few percent of monolayer coverage, judged from the relative intensities of carbon and molybdenum and the change in their intensity. The exact amount of the surface oxygen at the initial stage of the oxidation was difficult to estimate because of the rapid oxidation rate, but the extrapolated value for the oxygen was estimated to be nearly the same as that for carbon. Considerable growth of a 5 15 eV oxygen peak with time indicates that " swface oxidation " proceeds, as shown in fig. 3(a). It is highly likely that the M o surface is oxidized by water, which was one of the major residual gases, as indicated by mass analysis of background gases in the vacuum chambers.Stern and Ferrell have theoretically predicted that the loss peak of surface plasmon undergoes a red-shift depending on the thickness of the oxidized layer formed on the140 OXIDATION OF M O 6) 38min 5) 30min 4 ) Z m i n 3) 12min 2) 6 min 1 ) 3 m i n after e va po rat i D n 8cfore evaporation 1 1 I 1 0 Ib 'io 3b 9 energy losslev FIG. 2.-The variation of the characteristic energy loss spectra with time, for the freshly evaporated molybdenum surface. S ( b ) 0 1 v 160 ZbO 240 3b0 500 550 kinetic energy/eV FIG. 3.-Auger electron spectra from a molybdenum surface. (a) Freshly evaporated surface. (6) After anneaiing in hydrogen at 1.5 x Sulphur and carbon are ob- Torr, looO°C, for 20 min.served on this annealed surface.T . KAWAI, K . KUNIMORI, T . KONDOW, T . ONISHI, K . TAMARU 141 metal, provided that the layer is formed uniformly.7* l1 * It can, accordingly, be concluded that the time-dependent peak is attributable to surface plasmon excitation. The thickness of the layer was estimated from the equation of Stern and Ferrell, assuming its validity, and was plotted as a function of reaction time (see fig. 5). In conclusion, the thickness of the oxidized layer increased with time in a uniform manner over the whole surface on this evaporated sample, similar to the oxidation of an A1 film under high vacuum c~nditions.~ If the assignment of the 23 eV peak is correct, the surface plasma loss on the clean surface should be 23/,/2 = 16.5 eV on the basis of the free electron model.This theoretical prediction does not agree with our experimental results. If one or more interband transitions occur with the energy close to that of the surface plasmon excitation expected from the free electron model, the observed surface plasmon loss may deviate from the theoretical prediction of the free electron model. 2. OXIDATION OF THE FILM ANNEALED IN HYDROGEN ATMOSPHERE The evaporated film, whose surface plasmon loss was 5eV, was annealed in hydrogen at 1.5 x Torr pressure and a temperature of 1000°C for 20 min. The loss spectra of the annealed specimen were examined at different times after the annealing (see fig. 4). The other experimental conditions were identical to the case of the evaporated film. The behaviours and the shapes of the surface plasmon were quite different from those of the evaporated sample.Immediately after the treat- ment, the 10 eV loss peak was found as well as that at 23 eV. As the surface oxidation proceeded, a new peak appeared at 5 eV and coexisted with the peak at 10 eV whose intensity was gradually decreasing. These two peaks did not shift with time but their intensities changed. The new peak at 5 eV corresponds to the 5 eV peak observed in the experiment on the freshly evaporated film, i.e., the loss peak by the surface plasma excitation of the Mo surfzce covered by a sufficient amount of the oxidized layer. This phenomenon shows that both the oxidized part and non-oxidized part coexist during the course of the oxidation, and suggests that the oxide nuclei grow at the surface layer, the oxidation proceeding at the boundary between the metal and its oxide.The AES of this specimen demonstrated the appearance of sulphur and carbon atoms immediately after the annealing as is shown in fig. 3(6). Seemingly, sulphur and carbon had diffused from the inside onto the surface of the specimen during the annealing procedure. Considerable parts of this annealed surface are covered by carbon and sulphur judging from the reduction of the intensity of the molybdenum. The ratio of C and S peak areas to Mo AES peak area were 2.1 and 3.2 respectively. In addition, the relative intensity of the surface to the bulk loss on the fresh evaporated film was approximately 25 times larger than on the annealed one.This means that the surface area of the evaporated film may be larger than that of the annealed one, that is, the surface of the evaporated one may be rougher, though the contaminants may reduce their relative intensity to some extent.12* l 3 3" E+tanhkD 2&+(1 +cZ) tanh kD us = where us is the frequency of the surface plasmon when a surface oxidized layer of dielectric constant E is formed to thickness D ; up is the frequency of the bulk plasmon (fiup = 23 eV for Mo) ; k repre- sents the wave number of the surface wave excited in the solid by the incident electron and is obtained from the momentum and the energy conservation law ; k = mus cos Bltiko where 8 is the incident and scattering angle (27") and ko is the wave number of the incident electron.142 OXIDATION OF M O In fig.6, graph (a) and (b), the percentage ratio of the 5 eV peak to the sum of the 5 and the 10 eV one was plotted against time (t) and t 2 . There is an induction period of about 10 min. In the initial stage of the reaction, the ratio was proportional to t 2 but above a ratio of 50 :d it was expressed in terms of aft2 +/I, where a' and are I t 10) After one day 9) 58min 8) 52min 7) 46min 6 ) 38min 5) 32min 4) 25 min 3) 19 min 2 ) 13 min I ) 4 min a f t e r l e a t i n g up in H2, 1 . 5 ~ mmHg 20 min 0 10 20 30 energy losslev FIG. 4.-The variation of the characteristic energy loss spectra with time, for the molybdenum surface annealed in hydrogen at 1.5 x Torr, 1000°C, for 20 min.FIG. 5.-The /O 10 20 30 40 50 elapsed timelniin thickness of the oxidized layer as a function of elapsed time after evayora tion.T. KAWAI, K . KUNIMORI, T . KONDOW, T . ONISHI, K . TAMARU 143 time-independent constants. These phenomena can be explained qualitatively as follows. Assuming that the diameter of the circular patches increases with a constant velocity and the size of every patch is approximately the same, the ratio may be roughly equal to ((n.nE2)/So)t2 before these patches overlap. Here n is the number of the active points, So is the total surface area and E represents the average growth velocity of the diameter. After the overlapping, the effective n may decrease. There- fore the slope is smaller than that at the initial stage. t P) elapsed timelmin 5 + square of the elapsed timelmin FIG.6.-The percentage ratio of the 5 eV peak to the sum of the 5 and 10 eV ones as a function of (a) elapsed time after annealing in hydrogen in hydrogen and (b) square of the elapsed time. The different behaviours in the ELS and the AES on these different surfaces is of interest in the view of the correlation between the surface reaction and the surface properties. Considerations on the rates of the oxidation at the early stage (fig. 5(a), fig. 6(a)) revealed that the uniform oxidation of the Mo surface starts rapidly just after the evaporation, which suggests a large collisional cross section for the reaction between gases and the evaporated surface. The induction period of about 10 min on the annealed surface, on the other hand, implies that the oxidation scarcely proceeded at the beginning and after about 10min began to develop rapidly.The oxidation cross section is very small at the beginning. As to the surface properties, AES demonstrated that the fresh evaporated surface is clean and is scarcely contaminated by other elements, whereas the annealed surface is contaminated by considerable amounts of carbon and sulphur. Another difference between the films would be the geometric structure of the surfaces, i.e., the roughness factor of the evaporated surface is larger than that of the annealed one. Taking these facts into consideration, the different behaviours in Mo oxidation are interpreted as follows ; on the fresh evaporated surface which is rough and active,144 OXIDATION OF M O the H20 in the ambient gas attacks the surface, colliding with the whole surface area, and the oxidation proceeds uniformly into the inside of the film. On the annealed surface, which is partially covered by carbon or sulphur and not so rough as the evaporated film, on the other hand, the oxidation can start only at a limited number of reactive points, and the oxidized part develops its area through the boundaries of the two phases, forming the " oxidized patches ". The authors are grateful to Prof. Kozo Kuchitsu of the University of Tokyo and Dr. Katsuya Nakayama of the Electrotechnical Laboratory for valuable discussions. L. A. Harris, J. Appl. Phys., 1968, 39, 1419. C. J. Powell and J. B. Swan, Phys. Rev., 1960, 118, 640. E. J. Sheibner and L. N . Tharp, Surface Sci., 1967, 8, 427. D. Edwards, Jr., and F. M. Propst, J. Chem. Phys., 1971,55, 5175. R. E. Weber and W. T. Peria, J. Appl. Phys., 1967, 38,4355. P. W. Palmberg and T. N. Rhodin, J. Appl. Phys., 1968, 39,2425. H . Raether, Springer Tracts in Modern Physics, 1965, 38, 84. H. Iback, J. Vac. Sci. Technol., 1972, 9, 713. C. Kunz, Z. Phys., 1966, 196, 311. lo G. J. Dooley and T. W. Haas, .I. Chem. Phys., 1970,52,993. E. A. Stern and R. A. Ferrell, Phys. Rev., 1960, 120, 130. l2 C. J. Powell, Phys. Rev., 1968, 175, 972. l 3 J. W. Swaine and R. C. Plumb, J. Appl. Phys., 1962, 33, 2378.