AES measurements of adsorption isobars for oxygen on tungsten at low pressure and high temperature

SURFACE

SCIENCE 40 (1973) 205-215 o North-Holland

AES MEASUREMENTS OXYGEN

A. E. DABlRItt, Mechanical

OF ADSORPTION

ON TUNGSTEN HIGH

Publishing Co.

AT LOW

PRESSURE

TEMPERATURE

V. S. ARAMATl*

ISOBARS

FOR

AND

+

and R. E. STICKNEY

Engineering Department, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, U.S.A.

Received 7 March 1973; revised manuscript

received 22 May 1973

Auger electron spectroscopy (AES) has been employed to determine the relative coverage of oxygen on polycrystalline tungsten at high temperatures (1200 i T < 2500 K) and low 02 pressures (5 x lO-g < poI < 5 x 1O-6 Torr). We believe that this is the first demonstration that chemical analysis of solid surfaces by AES is possible even at temperatures as high as 2500 K. It is assumed that the relative oxygen coverage is directly proportional to the peak-to-peak amplitude of the first derivative of the 509 eV oxygen Auger peak. The experimental results illustrate the dependence of coverage on temperatureand pressure, and it is shown that the results for low coverages may be described reasonably well by a simple first-order desorption model plus a semi-empirical expression for the equilibration probability (or sticking coefficient). On the basis of this approximate model, the binding energy of oxygen on tungsten is estimated as a function of coverage, giving a value of N 140 kcal/mole in the limit of zero coverage.

1. Introduction

The interaction of oxygen with tungsten has been studied by many investigatorsr) using a variety of different techniques because it is encountered in numerous technological devices ** and it is considered as a classical example of high-temperature oxidation. The chemical reaction of oxygen with tungsten at high temperatures and low 0, pressures results in the formation of volatile tungsten oxides (WO, W02, WO,, W,O,, etc.) in addition to 0 and 0,. This reaction is described reasonably well by the quasi-equilibrium analysis t Work supported in part by the Advanced Research Projects Agency [Contract DAHCl5-67-C-02221 and the Joint Services Electronics Program [Contract DA28043_AMC02536(E)]. tt Present address: Mechanical Engineering Department, Arya-Mehr University of Technology, Teheran, Iran. * Present address: Bell Telephone Laboratories, North Andover, Massachusetts, U.S.A. ** For example, incandescent and arc lamps, ionization gauges, electron tubes, thermionic energy converters. ion propulsion engines, and high-temperature vacuum furnaces. 205

206

A. E.DABIRI,

V. S.ARAMATI

AND R. E. STICKNEY

proposed by Batty and Stickneys*s), but their analysis relies upon a semiquantitative estimate of the equilibriation probability, co2, defined as the probability that an impinging 0, molecule will adsorb and be equilibrated at the tungsten surface rather than being scattered without equilibration. Batty and Stickney suggest that co, depends primarily on the fraction of the tungsten surface that is not covered by oxygen, but existing experimental data are not sufficient to verify this suggestion. Although the results of flash desorption studies415) provide valuable information on the coverage (i.e., number of oxygen atoms adsorbed per unit surface area) of oxygen on tungsten as a function of temperature, the pressure range is not sufficient to determine the dependence of coverage on pressure. In an attempt to obtain a more complete understanding of the dependence (AES) to of Lo, on coverage, we have used Auger electron spectroscopy perform approximate measurements of the coverage of oxygen on a polycrystalline tungsten sample as a function of sample temperature (1200 to 2500 K) and 0, pressure (5 x 10m9 to 5 x 10e6 Torr).

2. Experimental

apparatus and procedures

The apparatus is a conventional ultrahigh vacuum system equipped with an Auger electron spectrometer (cylindrical analyzer, Model 10-243, Physical Electronics Industries, Inc.). In the present study, an electron beam current of 100 uA at 3000 eV is supplied to the sample from a gun mounted for grazing incidence. The electronic circuitry is essentially identical to that described by Palmberg et al.s), with the analyzer frequency being 30 kHz. The tungsten sample is a Ushaped polycrystalline ribbon mounted on a manipulator in such a manner that only the mid-point of the U is at the focal point of the AES unit. The sample is heated resistively by a dc current, and the temperature at the mid-point is measured with an optical pyrometer. The sample surface is cleaned by a combination of heating in 0, and flashing to 2500 K; this procedure successfully reduces the concentrations of surface impurities (predominantly carbon) to a level that is below the limit of detection of AES. Both a nude ionization gauge and a partial pressure analyzer are used to measure the 0, pressure in an attempt to minimize the various possible errors associated with 0, pressure measurements?). The desired steady-state 0, pressure, po2, is attained by adjusting the inlet valve controlling the flow of research grade 0, into the chamber. The experimental procedure includes the following steps: (1) Clean the surface as described above; (2) Set po, at the desired value; (3) Flash the sample to 2500 K; (4) Reduce the sample temperature, T,to the desired value,

AES MEASUREMENTS

and record

the Auger

spectrum

OF ADSORPTION

ISOBARS

after steady-state

FOR 02

207

ON w

conditions

are attained

*;

(5) Reduce (or increase) T to another value and again record the Auger spectrum after attaining steady-state conditions; (6) Repeat step #5 for various temperatures ; (7) Repeat the entire procedure for various values of po, in the range 5 x lo-’ to 5 x 10e6 Torr. It is assumed here that N, the coverage of oxygen on the tungsten surface, is directly proportional to So, the peak-to-peak amplitude of the first derivative of the 509 eV oxygen Auger peak. This assumption has proven to be accurate for a variety of systemss), and we expect that it is a valid approximation for the O-W system if the coverage is sufficiently low (e.g., less than a monolayer) and if the adsorbed 0 atoms do not penetrate into the lattice as a result of surface reconstruction (oxidation). The experimental results of Musket and Ferrantes) show that the shape and width of the oxygen Auger peak at 509 eV do not vary significantly with coverage on the (110) face of tungsten. Since we were unable to evaluate the proportionality constant by a direct calibration test, the resulting data represent relative values of the coverage rather than absolute values. However, by comparison with existing coverage data based on flash desorption measurements, we are able to approximate the scale factor relating the relative and absolute values (see section 3). 3. Experimental

results

Fig. 1 shows the first-derivative Auger spectra we obtained for four values of the sample temperature, T, at one level of 0, pressure, poz = 1 x 10e6 Torr. The spectra have been displaced by an arbitrary amount in the vertical direction to prevent the curves from overlapping. A most interesting feature of these results is the fact that the signal-to-noise ratio does not appear to be degraded when the sample is heated to high temperature. Notice that the peak-to-peak amplitude of the oxygen peak follows the expected trend of decreasing with increasing T. From Auger spectra recorded for various choices of T andpo2, we evaluated the relative change of So, the peak-to-peak amplitude of the 509 eV oxygen peak. The results presented in fig. 2 are restricted to the range Ts 1200 Kbecause it was difficult to obtain reproducible data at lower temperatures where a surface oxide layer forms at a very slow rate49 5). The data obtained at T 2 1200 K were quite reproducible and did not exhibit hysteresis (i.e., the value of So measured for a particular choice of T and pol was essentially the same regardless of whether the preceding values of T and po2 were above * For temperatures above 1200 K, it appeared that the oxygen coverage attained a steadystate value in less than 4 min; at lower temperatures, however, the 02-W reaction rate is so slow that we did not attempt to attain completely steady-state conditions.

208

A. E. DABIRI,

V. S. ARAMATI

AND R. E. STICKNEY

=Ix

IO-= Torr

p%

0 I

0

tl

I

I

I

I

400

200 E.

ELECTRON

I

600 ENERGY

I

I

800 IeV)

Fig. 1. Examples of Auger spectra recorded at various temperatures of the tungsten sample and at a constant 02 pressure, pol = 1 x 1O-6 Torr. Electron beam: 100 uA and 3000 eV. Modulation: 30 kHz and 15 V peak-to-peak. For clarity, the curves have been displaced vertically to prevent overlapping. The arrows indicate the expected positions of tungsten and oxygen peaks based on Chang’s tabulatior@).

20

0 1200

1400

1600

It300

2000

2200

2400

TP’K)

Fig. 2. Dependence of So, the peak-to-peak amplitude of the 509 eV oxygen Auger peak, on the temperature of the tungsten sample and on the 02 pressure. Electron beam and modulation conditions: same as in fig. 1.

AES MEASUREMENTS

OF ADSORPTION

ISOBARS

FOR

02 ON

w

209

or below the chosen values) in the few checks that we performed. In some cases we flashed the sample to 2500 K between each measurement at different T and poz, but the resulting values of So were essentially identical to those obtained when the sample was not flashed between measurements. in fig. 2, notice that at 1200 K there is an anomalously large gap between the data for 1 x lop8 Torr and 1 x lo-’ Torr. Although we do not have a conclusive explanation of this gap, we will mention one possibility. At the lowest pressures, the rate at which 0, molecules impinge upon the surface may be so low that the steady-state coverage may be lowered significantly by the fact that the electron beam used for Auger spectroscopy produces electronsimulated desorption (ESD) of oxygen from tungstenlo). Existing data show that the ESD rate increases markedly with increasing coveragelo), thereby providing a possible explanation of the observation in fig. 2 that the gap between the data for 1 x lo-* and 1 x lo-’ Torr increases with increasing coverage. Since ESD data are not available for the high electron energies and high temperatures employed in this study, we can do no more than provide a rough estimate of the magnitude of the ESD rate, R,, relative to the oxygen impingement rate, 2, expressed in terms of 0 atoms rather than 0, molecules. Assuming an electron beam current of IN 100 uA, a beam cross sectional area of A N I x lo-’ cm’, a coverage of NE 1 x lOI5 atoms/cm’, and an ESD cross-section of Q- 1.5 x 1O-2o cm* (see ref. 1 l), we obtain* R,/Z

II

1.5 x 1O-9/po,,

(1)

where po, is the 0, pressure in Torr. This very approximate relation predicts that R,/Z=0.15 atpoz= 1 x lOmE Torr, and we would therefore expect that ESD may produce a significant reduction in the coverage (especially when one realizes that the sticking probability is significantly less than unity at high coverages, thereby causing the adsorption rate to be considerably smaller than the impingement rate, Z). Because of these uncertainties in the data obtained at the lowest pressures and temperatures, our discussion (section 4) will favor the data for higher pressures and temperatures. To obtain a rough evaluation of the constant of proportionality between So and N, we will use the results of Ptushinskii and Chuikovb) to estimate the coverage at one particular choice of T and po,. At 1600 K and 1 x IO-’ Torr, their data (fig. 7 of ref. 4) indicate that NE 1.1 x IO” oxygen atoms per cm* (counting all 0 atoms on the surface, regardless of their molecular state of adsorption). Since So-9 for the same state in fig. 2, the resulting estimated relation is N 21 1.2 x 1Ol4 So. * As shown in ref. 10, Re=QNI/eA, where E is the electronic charge.

(2)

210

A. E. DABIRI,

V. S. ARAMATI

AND

R. E. STICKNEY

0 A

15d 5 z

lXld8 lxlo-6

IO -

z G 0

b co

5-

0

’ 300

700

1100

1500 T(“Kl

1900

2300

I 2700

Fig. 3. Extension of fig. 2 to lower temperatures. Note: as mentioned in the text, the data below 1200 K do not correspond to steady-state conditions and may be influenced by electron-stimulated desorption.

We emphasize that this relation is very approximate because of various uncertainties; furthermore, a slightly different proportionality constant is obtained if we select a different temperature for fitting our data to those of Ptushinskii and Chuikov. Although the data taken below 1200 K are inaccurate because steady-state conditions were not fully obtained, fig. 3 is included here since the results for po,=l x 1O-6 T orr exhibit an interesting maximum at N 1100 K. This maximum is consistent with results based on flash desorption measurements41 5), and it is suspected to be a consequence of the fact that the surface oxidation rate is highest in the vicinity of 1100 K. The absence of a maximum in the curve for po, = 1 x 10-s Torr is most likely associated with the problems discussed above. 4. Discussion As indicated in the introduction, a principal objective of the study was to determine if co2, the equilibration probability defined by Batty and Stickney 3), depends primarily on coverage, iV,as they suggested. Since Batty and Stickney did not have data on IV, the expression they adopted is in terms of T (K) and po, (Torr) * : co, = 80 (~o,)-‘.~ exp (- 18400/T). (3) [Note: since values of co2 greater than unity are absurd, we will assume co, = 1 when eq. (3) yields co, > I.] We have used this equation to construct * This expression

is obtained from eq. (7) in ref. 3 with the aid of eq. (2) in that paper.

AES MEASUREMENTS

OF ADSORPTION

ISOBARS

FOR 02

211

ON w

1.0 : i

m 2

0.8

8

-

I 2

d

So, OX6yGEN

a

SIGNAL

IO

12

14

Fig. 4. Relationship between the equilibration probability, C+, and the oxygen coverage (assumed to be directly proportional to SO). See text for details.

fig. 4 in the following

manner. Each data point in fig. 2 represents the value of So corresponding to particular choices of T and pOZ. By using eq. (3) to calculate Co, for the same T and po2, we obtain the values of co, corresponding to various values of So. The results are shown in fig. 4. If Co, depends primarily on the coverage, N, and if N is directly proportional to So, then the points plotted in fig. 4 should form a single curve corresponding to the functional relationship co,(N). Therefore, we have attempted to draw a smooth curve through the points, but ignoring the unreliable points (section 3) for 5 x 10e9 and 1 x lo-* Torr in the range SoX5. The scatter is sufficiently large that these results provide only marginal support of the suggestion that co, depends primarily on coverage. Since the results are not conclusive, we do not feel that it is worthwhile to include a detailed description of our attempts to find a model which would yield a co,-N relationship that agrees with the results in fig. 4. We should mention, however, that the simplest forms of the popular models suggested by Roberts 12) and by Kisliuk l3) do not agree closely with the present results over the entire range. Based on experimental results obtained previously in our laboratory 1%9, we would expect that the present data (fig. 2) may be correlated in terms of the parameter T/T,*, where T is the sample temperature and Tz is the saturation temperature corresponding to the oxygen pressure, po2, in our experiments (i.e., T,*is the temperature at which the equilibrium vapor pressure of liquid oxygen is equal to po2). Engelmaier and Stickney14) adopted the following relationship between T,*and po, based on available vapor pressure

212

A. E. DABIRI,

V. S. ARAMATI

AND R. E. STICKNEY

data for 0,: po, = 9.25 - 486.45/T,*.

log,,

(4)

Using this equation, we have transformed the data in fig. 2 to obtain fig. 5. Notice that T/T,* is very successful in correlating the data if we ignore the unreliable points for 5 x 10m9 and 1 x lo-* Torr in the range SoZ5. We have also attempted to describe the data in fig. 2 by means of’kinetic models. Since the kinetics of the O-W interaction appear to be extremely

20

I

15 -

I

0

d 5

LIYI

20

I

I

pop(Torr) 0 0 rJ . 0

*

IO -

0

I

.D

z z

I

5x10-9 1x10-9 1x10-7 1X10-6 5x10-6

_

&

-1

I

I

I

I

30

40

50

60

;4a 70

J

_

80

90

T/T:

Fig. 5.

Correlation

of the oxygen coverage data by means of the parameter (See text for details.)

T/TR*.

complex, we will consider only the low coverage region where the predominant desorption product is atomic oxygenl-5). Under steady-state conditions, the rates of adsorption and desorption may be equated to give 25,, p,,, (2nm,,kT’)-+

= vN ev

(-

EIRT) 3

(5)

where mo2 is the molecular mass of 02, k is the Boltzmann constant, T’ is the temperature of the 0, gas (~300 K), v is the frequency factor in the assumed first-order form of the desorption rate, E is the desorption energy for 0 on W, T is the temperature of the tungsten sample, and the other symbols have been defined before. Assuming v is constant and &,, is a function of coverage alone, partial differentiation of the logarithm of eq. (5) leads to [a(ln

po,)P(l/T)IN

= - Elk.

(6)

AES MEASUREMENTS

OF ADSORPTION

ISOBARS

FOR

02

213

ON w

T(“K) 2500 I I ,

10-S

10-9

3

4

2000 I , I

1500 I

1

5

6 104/T

1250 I

7

8

9

(OK-‘1

Fig. 6. Pressure-temperature plots of data in fig. 2 for various values of coverage. For convenience, the coverage is described in terms of the fractional coverage, 0, based on the assumption that 0 = 1 corresponds to N N 1.2 x 1Ol5 atoms cm-2.

Therefore, by replotting the data in fig. 2 in the form In pol versus l/T for various choices of N (i.e., choices of So), we may evaluate E(N) from the slopes of the lines of constant coverage. This has been done in fig. 6, where for convenience we have expressed the coverage in terms of 8, the fractional coverage, defined by utilizing eq. (2) together with the assumption that a monolayer consists of 1.2 x lOi atoms cm-’ (as suggested in ref. 5). That is, 8 = N/1.2 x 1o15 = 0.1 so.

(7)

From the slopes of the lines in fig. 6, we have evaluated E versus 0 in fig. 7. Notice that extrapolation of E to the limit of zero coverage yields - 140 kcal/ mole (- 6.1 eV), which is consistent with the values obtained by other investigators using different techniques43 5,i4-17). With these data on E versus coverage, and with Batty and Stickney’s approximate expression for lo, [eq. (3)], eq. (5) provides a relationship between N, T, and poz in terms of only one adjustable parameter, v. This relationship is plotted in fig. 8 for v=4 x lOi set-‘, which was selected because it leads to the best overall agreement of the curves with the experimental data. The magnitude of v appears to be reasonable for a frequency factor in a first-order desorption model. In view of the many assumptions

214

A. E. DABIRI,

0

AND

V. S. ARAMATI

0.2 8.

0.4

R. E. STICKNEY

0.6

FRACTIONAL

0.6

1.0

COVERAGE

Fig. 7. Variation of the desorption energy with coverage, as estimated from the slopes of the constant-coverage lines in fig. 6. The bars indicate the estimated uncertainty in determining the slopes.

12

2

0

z ” v,

6-

g

4-

5x10-6

’

5 0 2

2 -

0 1500

I700

1900

T. TEMPCRATURE

2100

2300

2500

(OK)

Fig. 8. Comparison of a simple, semi-empirical model (solid curves) with experimental data (from fig. 2) on the dependence of coverage on Tand po2. The bars on the curve for ~0, = 1 x 1O-7 Torr are included to illustrate the effects of uncertainties associated with the desorption energy (fig. 7).

and uncertainties associated with both the model and the experimental data, the agreement exhibited in fig. 8 is about as good as we should hope for. The curves have not been extended above So 2 6 because the model is invalid at higher coverages where the desorption rates for oxides (WOz, WOs) are no longer negligible.

AES MEASUREMENTS

OF ADSORPTION

ISOBARS

FOR 02

ON w

215

5. Conclusions We believe that the most valuable aspect of the present work is the demonstration of the feasibility of extending the range of AES to extremely high temperatures. We expect that this extension in the range of application of AES will prove to be useful in studies of adsorption and desorption, segregation and solution of impurities or alloying components, surface reactions, and thermal cleaning in controlled atmospheres. The results of the O-W study are generally consistent with existing data obtained by flash desorption techniques 4l 9, and they contribute by providing adsorption isobars for a considerable range of T and po2. At low coverages, a simple, semi-empirical model of the adsorption and desorption processes provides an approximate description of the experimental results, with the desorption energy of 0 atoms being N 140 kcal/mole in the zero-coverage limit. References 1) For example, see the references listed by B. McCarroIl, J. Chem. Phys. 46 (1967) 863, and by R. G. Musket, J. Less-Common Metals 22 (1970) 175. 2) J. C. Batty and R. E. Stickney, J. Chem. Phys. 51 (1969) 4475. 3) J. C. Batty and R. E. Stickney, Oxidation of Metals 3 (1971) 331. 4) Y. G. Ptushinskii and B. A. Chuikov, Surface Sci. 6 (1967) 42; 7 (1967) 90. 5) D. A. King, T. E. Madey and J. T. Yates, Jr., J. Chem. Phys. 55 (1971) 3236. 6) P. W. Palmberg, C. K. Bohn and J. C. Tracy, Appl. Phys. Letters 15 (1969) 254. 7) P. A. Redhead, Vacuum 13 (1963) 253; J. H. Singleton, J. Chem. Phys. 45 (1966) 2819; A. E. Dabiri and R. E. Stickney, J. Vacuum Sci. Technol. 9 (1972) 1032. 8) R. E. Weber and A. L. Johnson, J. Appl. Phys. 40 (1969) 314; M. Perdereau, Surface Sci. 24 (1971) 239. 9) R. G. Musket and J. Ferrante, J. Vacuum Sci. Technol. 7 (1970) 14. 10) T. E. Madey and J. T. Yates, Jr., J. Vacuum Sci. Technol. 8 (1972) 355; T. E. Madey, Surface Sci. 33 (1972) 355. 11) R. G. Musket, Surface Sci. 21 (1970) 440. 12) J. K. Roberts, see description in A. R. Miller, 77re Adsorption of Gases on Solids (Cambridge Univ. Press, Cambridge, 1949). 13) P. Kisliuk, J. Phys. Chem. Solids 5 (1958) 78. 14) W. Engelmaier and R. E. Stickney, Surface Sci. 11 (1968) 370. 15) W. Greaves and R. E. Stickney, Surface Sci. 11 (1968) 395. 16) H. W. Wassmuth, H. Werner and A. K. Mazumdar, J. Vacuum Sci. Technol. 9 (1972). 17) See table 1 in ref. 14 for a survey of data obtained prior to 1968. 18) C. C. Chang, Surface Sci. 25 (1971) 53.