Reaction of β-SiC with thermal atomic hydrogen by modulated molecular beam mass spectrometry

..,:“.‘::

....:..... ..:::,.> :,.,.,, ,,,, ,, ,.(..,,,

.‘:‘:’ :.j::: .:.:. ,,::-.‘: ..,. :... ..... ,.Y.

c:: 7::.

..(. ,::::.:.: . . . :: ,.,. >:. ,y~;>: : :.:.. ;,y

surface

: .’

..

.:.: ,... :.:., ,. >y (:. ., : . . ..,.,.. .,:,, :‘_::::

,,:::.,,:.,.:

::

,. . . . . ,::.:j ::

science

.:..,:.:,._,. . :;:::,z;: :,, .::.: ,..,...) ...,::,. .v::...: “./,(. :.:’.:.i:.~:,~:i_j:::::.~.~:,,~.:~,:~.: .. .‘.‘V ...i::...._. :.:.y.:i:;_:_ ~‘:;, :‘._:,~_ “‘%+ %::::ix::$:; ,._..: )‘,_;.,:~ :.:..........,.,. ,.,.... .._i L‘.:c.. ELSEVIER

Surface

Science

313 (1994) 399-416

Reaction of P-Sic with thermal atomic hydrogen by modulated molecular beam mass spectrometry Yongsoo Kim a,*, D.R. Olander b a Department of Nuclear Engineering, Hanyang Vnirwsity, Seoul, South Korea b Department of Nuclear Engineering, University of California, Berkeley, CA 94720, USA (Received

2 August

1993; accepted

for publication

17 February

1994)

Abstract The reaction of polycrystalline @-Sic with thermat atomic hydrogen is investigated by modulated molecular beam mass spectrometry. The temperature and equivalent pressure on the surface are 300-1100 K and 6 x lOme- X IOv5 Torr, respectively. The reaction probability of atomic hydrogen is 9 X lo-‘-5 X 10m4. Grain boundary effects are negligible and no bulk or surface diffusion is observed. The major reaction products are SiH,, CH,, and C,H,. SiH, and CH, formation reactions are linear with respect to the H-atom pressure while C,H, formation is nonlinear. A precursor model is proposed for SiH, and CH,. C,H, appears to be produced in the defect site where SIH, forms and desorbs.

1. Introduction Silicon carbide is a high-temperature semiconductor with unique properties. Various solid-state devices using SiC have been investigated to exploit its excellent electrical properties, such as wide bandgap, good thermal conductivity, superior high-breakdown electric field properties, and a low dielectric constant [l-41. Sic devices, especially using p-Sic, would be particularly valuable in high temperature and/or corrosive environments. Dry-etching by reactive gases at low temperature is a common process in the fabrication of silicon integrated circuits. Although singlecrystal Sic devices have been prepared, there is

* Corresponding

author.

~39-6028/94/$07.00 0 1994 Elsevier SSDZ 0039-6028~94~00121-0

Science

no efficient technique of dry-etching this material because of its extremely high resistance to volatilization by molecular gases. If Sic could be selectively etched by atomic hydrogen, fabrication of integrated circuit devices might become easier and faster. Sic is also a prime candidate first-wall material for a thermonuclear fusion reactor. It was introduced because neither stainless steel nor refractory metals could fulfil the performance criteria for the first-wall material as well as silicon carbide [S-S]. Plasma impurity control in the the~onuclear fusion reactor is a particularly critical problem. Plasma-facing components experience high heat fluxes and are bombarded by hydrogen isotopes, ionic or neutral. The volatile products, desorbed or sputtered physically or chemically, contribute to contamination of the plasma.

B.V. All rights reserved

Y. Kim, D.R. Obnder/Swface

400

Many experiments have investigated the sputtering yield (mostly physical) of Sic with H+ and D+ with energies above 1 keV [9-141. However, no information is available on the reaction of atomic hydrogen with Sic. This paper deals with the kinetics and mechanism of the reaction of polycrystalline P-Sic with thermal atomic hydrogen.

2. Experimental The experiments were performed in an ultrahigh vacuum (UHV) modulated molecular beam system with mass spectrometric detection. The schematic of the apparatus is shown in Fig. 1. The details of the apparatus and experimental approach are discussed elsewhere [15]. A mixture of atomic and molecular hydrogen ( - 13% atomic hydrogen) generated in a tungsten oven at 2350 K

Science 313 (1994) 399-416

by resistive heating effuses from a 1 mm diameter orifice in the oven source and is chopped by a rotating disk. It is collimated into a collision-free beam through another 1 mm diameter hole between source and target chamber and strikes the SIC target which is inclined at 45” to the incident molecular beam. Reflected H and H, and any reaction products emitted from the target surface are sampled by a 2 mm diameter collimator separating the target and the detector chamber housing the mass spectrometer. The mass spectrometer signal, combined with the reference signal derived from the chopper motor, is analyzed by a lock-in amplifier to provide the amplitude and phase of reaction products relative to the incident atomic hydrogen beam. The apparent reaction probability, E, which is the ratio of the amplitudes of the reaction product to reactant signals, and the phase lag, 4, between the two signals are measured as functions of incident atomic hydro-

Turbomolecular Pump

Fig. 1. Schematic

of the modulated

molecular

beam apparatus

with electron

beam target

heater

(plan view).

Y Kim, D.R. Olander / Surface Science 313 (1994) 399-416

gen beam intensity, I,, Sic surface temperature, T, and beam modulation frequency, f. The surface temperature was varied from 300-1100 K by a combination of resistive and electron-beam heating. For this purpose a thin tantalum plate was tightly attached to the back surface of the sample to attract the electrons from the heated tungsten coil. The temperature of the Sic target surface was measured by an infrared pyrometer through the viewport (Fig. 1). The atomic hydrogen beam intensity on the target surface ranged between 2 X 1015 and 9 X 1015 H-atoms/cm’ * s, corresponding to surface equivalent pressures of 6 X 10e6 to 2 x lo-’ Torr. The beam intensity calculation is described in detail in Ref. [15]. The beam modulation frequency was varied from 40 to 500 Hz. The base pressure is 8 X 10e9 Torr in the target chamber and 4 x 10-l’ Torr in the detector chamber. To improve the signal-to-noise ratio and the mass spectrometer resolution, deuterium is used as the reactant gas instead of protium; however, the generic name, hydrogen, and the symbol, H, are used throughout this paper except when the mass difference between H and D is essential to the data interpretation. The sample is a 1 mm thick disk of polycrystalline P-Sic supplied by the Lawrence Livermore National Laboratory (LLNL). Prior to mounting, the sample was polished with 1 km diamond paste to a mirror finish. The native surface oxide was removed by vacuum annealing the sample for two hours at 1000°C. This procedure minimizes (but does not eliminate) the surface structure and stoichiometry changes due to carbon segregation and silicon vaporization which becomes significant above 1300 K [16-191.

3. Experimental

results

3.1. Surface topology and stoichiometry with reaction

The Sic surface topology reaction with thermal atomic by atomic force microscopy Fig. 2. While the unreacted

changes

change due to the hydrogen examined (AFM) is shown in area of the sample

401

surface remains flat (Fig. 2a), the grain boundaries in the reacted area appear clearly (Fig. 2b) after 80 h of exposure to the atomic hydrogen beam (- 2 x 1021 H-atoms/cm2) at 500 K. The depth profile in Fig. 2b shows that the grain surface remains relatively flat with narrow depressions at the grain boundaries. This evidence suggests that although the reaction takes place both on the grain surface and in the grain boundaries, the former is dominant. The boundary between the reacted and unreacted surface on the same specimen was examined with SEM and scanning Auger-electron microscopy (SAM). The SEM micrograph shows a clear contrast between the reacted and unreacted area (Fig. 3a) and the silicon-to-carbon ratio in the SAM spectrum changes abruptly at the boundary (Fig. 3b). This demonstrates that longrange surface diffusion of adsorbed hydrogen is insignificant. The Si/C ratio in Fig. 3b is obtained when the sensitivities of silicon and carbon to Auger-electron emission [20] are taken into consideration. The ratio is 0.43 in the unreacted area whereas it increases to 0.65 after reaction with thermal atomic hydrogen at 500 K. The former is because of the prior heating of the sample at lOOo”C, which renders the surface carbon-rich [16-191 and the latter is due to the nature of the reactions of atomic hydrogen with the two elements on the surface. The reaction with atomic hydrogen eliminates the strong thermal segregation effect that existed prior to the reaction but does not produce a stoichiometric surface because of the higher intrinsic reactivity of surface Si. 3.2. Molecular beam mass spectrometer results Reaction product signals measured by the mass spectrometer as functions of three experimental parameters, surface temperature, atomic hydrogen beam intensity and beam modulation frequency, are converted to the apparent reaction probability, E, and phase lag, 4, of each distinct reaction product [15]. It was found that silicon carbide is not etched by molecular hydrogen but interacts with atomic hydrogen to produce three major volatile species,

402

I’. Kim, D.R. Olander/Surface

silane WH,), methane (CH,), and acetylene (C,H,). Apparent reaction probabilities for these three products range from 9 x 10m5 to 5 x 10m4, of which silane has the highest and acetylene the lowest. To confirm the identification of the products, the signal amplitudes of cracked species of each product (e.g., Si+, SiH+, SiHl, SiHI, and SiHi from SiH,) were compared to published fragmentation patterns [21]. Although some ions were not detectable because of the signal levels below the overall system sensitivity, all detected lower-mass ions had the same phase lags as the parent ions, which is a characteristic of cracked species. The possibility of other intermediate reaction products such as CH, [22] cannot be discounted. However, reaction products with signal amplitudes below the overall system sensitivity of 10P5-10-4 or small compared to that of one of the cracked species from a major product could not be detected.

Science 313 (1994) 399-416

3.2.1. Effect of surface temperature The surface temperature was varied from 300 to 1100 K at a beam chopping frequency of 95 Hz and H-atom beam intensity of 7 x 1015 atoms/ cm2. s. This procedure was repeated several times to ascertain data reproducibility. As seen in Fig. 4 the apparent reaction probability of SiH, production increases slowly with temperature and then decreases slightly above 800 K. On the other hand, CH, formation continues to increase even above 800 K. Apparent reaction probabilities for SiH, and CH, range from 3 X 10m4 to 5 x 10e4 and from 2 X lop4 to 3 x 10 P4, respectively. However, decreasing phase lags indicate that the reactions become faster with increasing temperature. Acetylene formation is unexpectedly detected at temperatures as low as 500 K. Its temperature dependence follows that of SiH,. Between 500 and 1100 K, the apparent reaction probability of

Fig. 2. AFM micrographs of the etched surface of SIC. (a) Top: unreacted, (b) bottom: 80 h exposure (2 T= 500 K. Vertical and horizontal scales are in nanometers.

X

10zl H-atoms/cm’)

at

403

Y. Kim, D. R. Olander / Surface Science 313 (1994) 399-416

this species is one-fourth that of silane. According to standard thermodynamic calculations [22] and observations in the H-graphite reaction 2231, significant acetylene formation was expected only above 1800 K. Thus, in order to verify the identification of this product, two potential experimen-

tal artifacts were examined. Both are related to CO, which is the most abundant background molecule in a UHV system and has the same mass number as C,D,. First, when the D beam and the tungsten furnace were turned off, only DC noise re-

I I

reacted area

0

108

unreactd area

+----d---w I

300

2w) Distance,

400

MO

microns

Fig. 3. (a) SEM micrograph of the etching boundary after 5 h exposure (1.3 X 10’” H-atoms/cm21 at T= 500 K, (b) Si/C ratio from SAM analysis along the line in (a) after correction for the sensitivities of silicon and carbon to Auger-electron emission [ZO].

404

Y: Kim, D.R. i?lander/Surface

Science 313 (1994) 399-416

TO 800

1100

1 o-3

660

600

300 60

SiH, 50

40

30

20

10 60

50

40

_ 30

23

20

iz

f

10

50

40

30

20

A

8 I 5

10

15

20

25

30

f 10

35

1 O4I T(K) Fig. 4. Apparent reaction probabilities, E, and phase lags, I$, of reaction products, (a) SiH,, (b) CHg, and (c) C,H,, temperature at f = 95 Hz and I, = 7 X 10 I5 H-atoms/cm’. s (solid curves: model).

versus surface

Y. Kim, D.R. Olander/Surface

mained: its average mass 28 signal amplitude was zero without any phase lag. If the unmodulated CO background contributed to the mass 28 signal, it should have behaved like DC noise. A leak of CO after the hydrogen filter in the reactant feed line would have been modulated and interpreted as C,D,. However, the scattered CO from this source would also show zero phase lag whereas the mass 28 signal showed a definite phase lag. A leak of 0, or H,O after the hydrogen filter would produce a modulated source of 0 and OH radicals in the oven source. These radicals form CO on carbon-containing surfaces and the CO so produced would have a non-zero phase lag. This potential source of CO is ruled out for two reasons. First, the mass 28 signals were quite reproducible from one run to the next, which would not be expected from an uncontrolled, adventitious leak in the source tube. Second, no 0, or H,O was detectable in the scattered D, flux from the cold oven source. With the hot oven source, therefore, the incident flux of 0 or OD on the surface would have been many orders of magnitude smaller than the Datom flux. It is highly unlikely that the reaction probability of 0 or OD with C on the surface to produce CO would be sufficiently higher than that of D with C to produce C,D, so that the former would mask the latter. Second, the possibility of D-atom reaction with CO adsorbed on the specimen surface to produce acetylene was checked by testing a non-carboncontaining target surface (Al,O,). CO was led into the target chamber and its pressure was maintained at 2 x lop6 Torr while the atomic deuterium beam was modulated before it impinged on the alumina surface. No evidence of C,D, formation was observed in the mass 28 modulated output. Based on the negative results of these tests the signal at the mass of 28 in the mass filter is believed to arise from a true acetylene product of the reaction. 3.2.2. Effect of H-atom beam intensity The atomic hydrogen beam intensity was varied from 2 X 1015 to 9 X 1015 H-atoms/cm2. s, which corresponds to equivalent pressure on the

Science 313 (1994) 399-416

405

SiC surface of 6 x 10e6-2 X lop5 Torr. These measurements provide information on the order of the reaction. The measurements were done at two surface temperatures (1100 and 800 K) at the beam modulation frequency of 95 Hz. The reaction probability and phase lag dependence on the beam intensity at 1100 K surface temperature are shown in Fig. 5 and at 800 K in Fig. 6. These results show that SiH, and CH, production is not influenced by the beam intensity, implying that formation of these species occurs by a firstorder process. The departure of the first two data points from constancy of the reaction probability of CH, in Fig. 5b is due to extraction of the median signal values from large fluctuating amplitudes at low H-atom beam intensities. The constancy at high beam intensity is confirmed from another set of reaction probability measurements of CH, (Fig. 6b). The reaction probability and phase lag dependence of C,H, on the beam intensity are not reported here because of experimental uncertainties due to signal amplitudes close to the sensitivity limit of the detection system.

3.2.3. Effect of H-atom beam modulation frequency The measured phase lag contains information on the “residence time” of incident reactant atoms for the interaction with the surface atoms. In the present experiment the atomic hydrogen beam was modulated at frequencies from 40 to 500 Hz at two surface temperatures, 800 and 1100 K. During the scan of modulation frequencies, the H-atom beam intensity was fixed at 7 X 1015 atoms/cm2. s. Only SiH, and CH, production reactions are reported here for the same reason as in the previous section. Effects of beam modulation frequency on the reaction probability and phase lag measurements are shown in Figs. 7 and 8. At high beam modulation frequency the phase lag from the beam transit effect [24] is considerable. The phase lags shown in Figs. 7 and 8 have been corrected for this and other mechanical components in the phase measurements through the data interpretation process [15]. The error bars in Fig. 8 represent the range of fluctu-

Y Kim, D. R. Olander / Surface Science 313 (1994) 399-416

406

ations of the signal amplitudes, which are larger at 800 than at 1100 K. With increasing modulation frequency, the phase lags increase rapidly while the apparent reaction probabilities decrease slowly. At high beam modulation frequencies, the amplitudes of product signals approach the overall system sensitivity limit, thereby increasing the error in the data points, especially at low temperature.

4. Reaction mechanism

analysis

4.1. Overall reaction mechanism Recent temperature programmed desorption (TPD) experimental results [25] demonstrate that molecular hydrogen does not adsorb on the Sic surface, whereas atomic hydrogen does. According to the TPD spectra, high coverage (greater

50

SiH, 40 0

O-

0

8 30

20 A---c A

%

a

* 10 50

CH4 40 0

0

to 0

O

8

30

20 -

A

h i

/3

Beam

intensity,

10 10.0

7.0

4.0

1.0

I, (atoms/cm%)

Fig. 5. Apparent reaction probabilities, E, and phase lags, 4, of reaction intensity at f= 95 Hz and T, = 1100 K (solid curves: model).

x 10m”

products,

(a) SiH,,

and (b) CH,,

versus

H-atom

beam

Y Kim, D.R. Olander/Surface

407

Science 313 (1994) 399-416

single-step, first-order reaction behavior. That is, the rate-controlling surface process is the attachment of H atoms to SiH, and CH, complexes and the remaining hydrogen-collection steps occur rapidly and terminate with volatile tetrahydrides. There is considerable experimental evidence that hydrogen forms clusters, often called complexes, on single-crystal silicon and diamond surfaces [26-341. The SiH complex was observed on the Si(111) surface with a low H-atom coverage. SiH2 becomes observable with H coverages of about 50%, and even SiH, was detected on

than 1 monolayer) was maintained up to 800 K, which is ascribed to saturation of dangling bonds. In the current experiment the incident atomic hydrogen flux (I, = 7 x 1015 H-atoms/ cm2 * s) is high enough to maintain a high coverage on the SiC surface at all temperatures tested. The high H-atom coverage and the experimentally-observed linearity in the silane and methane formation reactions suggest a precursor model for both reactions. As silane and methane molecules are tetrahydrides, it would be extremely improbable that without precursor formation on the surface, adsorption of four H atoms would give a

50

IO-3 ....*

SiH, c)

40

0

-

Q 30 w

t

0

rodi

\

..,__

10

50

40

0 0

e-0

30

8

0

20

2.0

3.0

4.0

Beam

5.0

Intensity,

6.0

I, (atoms/cnJ”o)

7.0

8.0

9.0

10

x 10”

Fig. 6. Apparent reaction probabilities, 6, and phase lags, c$, of reaction products, (a) SiH,, and (b) CII,, versus intensity at f= 95 Hz and T, = 800 K (solid curves: model).

H-atom

beam

408

Y Kim, D.R. Olander / Surface Science 313 (1994) 399-416

SXlll) at high H coverage [Z&-29]. CH and CH, complexes were observed on the diamond (100) surface at high coverages of atomic hydrogen [30]. Sic has a ZnS structure which would become a diamond or a silicon structure if silicon atoms are replaced by carbon atoms or vice versa. Therefore it is reasonable to suppose that SiH, SiH,, CH, and CH, complexes form on the Sic surface at the high H-atom coverage sustained by the incident atomic hydrogen beam. Thus, it is proposed that the Sic surface is covered with an over-layer of hydrogen atoms adsorbed on tightlybound surface monohydrides (SiH and CH) and dihydrides (SiH, and CH,). The dihydride com-

plexes might be in the form of C(SiH,)C or Si(CH,Bi, whose two dangling bonds terminate with adsorbed H atoms while the other two bond to neighboring surface atoms. The dihydrides are believed to be the precursors of final reaction products, SiH, and CH,. During the surface reaction, diffusion of H atoms into the bulk or on the surface may occur. However, SEM micrographs and the SAM spectrum in Section 3.1 demonstrate that surface diffusion of adsorbed hydrogen atoms is negligible. With significant hydrogen diffusion, the apparent reaction probability would decrease and the phase lag increase with increasing temperature, which is

100

70

SiH,

60

lo4

60

Beam Modulation

froqwney

(Hz)

Fig. 7. Apparent reaction probabilities, e, and phase lags, (6, of reaction products, (a) SiH,, and Cb) CH,, versus H-atom beam modulation frequency at T, = 1100 K and I, = 7 x 10” H-atoms/cm’. s (solid curves: model).

Y. Kim, D.R. Olander / Surface Science 313 (1994) 399-416

opposite to the experimental results. Therefore, no diffusion term is included in the kinetic model for SiH, and CH, productions. The grain boundary influence on the reaction mechanism is not taken into consideration. The chief effect of the slight preferential reaction in the grain boundaries appears to be surfaceroughening. The schematic diagram of the overall reaction mechanism of the Sic-H reaction showing reaction paths for SiH,, CH,, and C,H, production is shown in Fig. 9. The C,H, production path is drawn as thick dotted lines because the detailed kinetic mechanism is not given. Kinetic mechanisms for the two major reaction products, SiH,

409

and CH,, are given in the following section. Plausible detailed mechanistic models are discussed in Section 5. 4.2. Silane and methane production reaction mechanism

Experimental results in Section 3.2 reveal that the dependencies of SM, and CH, formation on the three experimental variables are very similar. In addition, the two reaction processes are dependent on each other because both draw H atoms from the same source (i.e., the adsorbed H layer). Thus the reaction mechanisms of both

lo-3

70

60

50

40

30

20 3 C! 10

P E

l@ 109 -

0

a

70

9

^

60

f

a’

Boom

Modulation

Frequmcy

(Hz)

Fig. 8. Apparent reaction probabilities, E, and phase lags, 4, of reaction products, (a) SiH,, modulation frequency at T, = 800 K and I,, = 7 X 1Ol5 H-atoms/cm’. s (solid curves: model).

and (b) CH,,

versus

H-atom

beam

410

Y. Kim, D.R. Olander /Surface

Science 313 (I 994) 399-416

silane and methane formation are simultaneously dealt with in this section. The effect of increasing temperature on the reaction probability and phase lag measurements for silane and methane production are shown in Figs. 4a and 4b. The decrease of silane formation and the steady increase of methane formation above 800 K are attributed to the silicon depletion on the Sic surface [16-191. When the carbon segregation effects on the high temperature reaction probability measurements are excluded [35], the responses of E and $ of both silane and methane formation to the temperature scan show a simple adsorption-desorption reaction behavior. Figs. 5 and 6 show silane and methane formation reaction probabilities and their phase lags to be independent of atomic hydrogen beam intensity, which is indicative of a first-order surface reaction. The beam modulation frequency dependencies in Figs. 7 and 8 are also consistent with the linear reactions. Based on these experimental results a precursor reaction model is proposed for both the silane and methane production reaction. In the model the Sic surface is covered with an overlayer of adsorbed hydrogen atoms, Hcadj, surface monohydride, MH, and surface dihydride, MH,, where

dkCtiOtl

H(adj

+

M is Si or C. H atoms in the incoming beam adsorb on the Sic surface with an effective sticking probability, rleff. MH, interacts with an adsorbed H atom and forms an MH, complex. The MH, complex instantaneously picks up an additional adsorbed hydrogen to form stable MH, that desorbs from the surface. In this desorption process the reaction of MH, with an adsorbed H atom is the rate-limiting step. With n(t) denoting the surface concentration of adsorbed hydrogen atoms on the Sic surface covered with MH and MH, complexes, the balance equation for Hcadj is as follows: dn = qeffZOg( t) - 4kzLf4n - 4kzet4n, dt

where I, is the amplitude of the incident atomic hydrogen beam intensity, g(t) is the gating function of the square-modulated beam, and, qeff, kziH4, and k,‘*4 are the effective sticking probabilly and the$ffective desorption rate constants of SiH, and CH,, respectively. The term “effective” implies that the coefficients result from the overall effect of more than one physical process, which will be discussed in Section 5. The first term on the right hand side of Eq. (1) represents the effective rate of adsorption of H atoms per

CH2

2

kd,

t CH3

fast

CH4w

fast

s*4cP)

HW

1-q

H(ad)

11

:--I I

kd,

1 eff

t siH3

SIC

no

I

‘____‘.l!df_

(ER mechanism)

C/SiC(defect)

_____________----(detailed mechanism not given)

Fig. 9. Schematic

diagram

(1)

of the overall SIC-H

reaction

mechanism.

b

CzHxg)

Y Kim, D.R. Olander /Surface

unit area.

The second and third term account for Head) atom loss by first-order silane and methane production. The typical truncated Fourier expansion technique is used in the analysis of this model to solve Eq. (1). The gating function is written as g(t) = i
The reaction product vectors for silane and methane production are defined so that the maximum reaction probability is unity: EM” 4 e-i$“b

=

4k”%i, d,, iZ”&

Science313 (1994) 399-416

411

of a pre-exponential factor, k,, and an activation energy, E, so that five constants are to be determined by the fitting technique. Fitting of the reaction model to the experimental results was performed by a Monte Carlo method in which random generation of the five constant values is repeated until the deviation of the theoretical predictions from the measured values is minimized. The advantage of the Monte Carlo method is its ability to simultaneously fit the entire experimental database, which is especially important in this case of a coupled-parameter model. The best-fitting parameters obtained by this method 5 4 x 102e(- i.b/RT) are: Teff = 9 x 10P4, k;ib‘1 (= 08,RTj se1 where s-l, and kzet4 = 4.6 x 10 e the activation energies are in kcal/mol. Theoretical curves of E and 4 from these parameters are shown as solid lines in Figs. 4-8. In general there is good agreement between the model and the experimental results. 4.3. Acetylene production reaction mechanism

’

where M = Si or C. Substitution of Eq. (2) into Eq. (3) and simple vector and complex manipulation yield:

-l/2

As the temperature increases, the C,H, phase lag decreases almost linearly while the apparent reaction probability increases, reaches a maximum and then decreases slightly again above 800 K, despite carbon segregation on the surface, which becomes kinetically feasible when the surface temperature is above 500 K [36,37]. Thus above 800 K carbon becomes readily available on the Sic surface to react with Head) so the methane production increases over that without carbon segregation. At the same time, silane formation decreases because of the depletion of Si on the surface. If the acetylene formation took place on random carbon sites, the apparent reaction probability above 800 K should have increased steadily with increasing temperature, just as that of methane does. However, the C,H, apparent reaction probability has exactly the same temperature dependence as that of silane formation. This observation implies that acetylene forms by reaction with the carbon atoms in the defected site exposed by SiH, desorption (Fig. 10) and explains why C,H, formation is possible even at low temperature.

) (4)

Eqs. (4) and (5) predict that the phase lags in the silane and methane formation reactions are identical. The apparent reaction probabilities of the silane and methane formation, however, are different but coupled to each other. Eqs. (4) and (51 are used to compute the theoretical values of as functions of the three experiEMH, and 4MH, mental variables, T, I,, and w for comparison with the experimental data. The model contains three fitting parameters: kzt$4, and k2Ct4. The last two are made up 77eff,

412

Y. Kim, D.R. Olander/Surface

Science 313 (1994) 399-416

0~~~~~ ’ Si

.:.y

SiH, complex

OC .

H

1

formation of SiH4 and a defect site left by its desorption

saturation of carbon dangling bond in defect site

formation of C2H2 and its desorption

Fig. 10. Defect-site

mechanism

There are several experimental observations of chemical sputtering of graphite or carbides at temperatures below 1000 K supporting the claim that C,H, forms in the defect site [38-441. Heavier hydrocarbons (C,H,, C,H, including acetylene) were detected during chemical sputtering of graphite due to bombardment by H+ ions 138-401 and combined H+ and Ho bombardment 141,421. In an investigation of TiC coating behavior under H+ bombardment, acetylene formation was observed [43]. In the erosion of H-covered amor-

for C,H,

production.

phous graphite by He and Ne ions, acetylene was the dominant hydrocarbon [44].

5. Discussion 5.1, Hydrogen recombination Production of SiH,, CH,, and C,H, can be influenced by H, recombination on the Sic surface, which is thermodynamically favorable. The

413

Y. Kim, DR. Olander/Surface Science 313 (1994) 399-416

very small value of the effective sticking probability (neff = 9 x 10w4) suggests that most of the incident H atoms from the incident beam that stick to the surface reappear as H,,. However, H, recombination cannot be measured because the reactant beam is predominantly H, and the mass spectrometer cannot distinguish the small signal of recombined H, on the surface from the large signal arising from the H, molecules reflected from the surface. Both Eley-Rideal (ER) recombination between H,, and Hfadf and Langmuir-Hinshelwood (LH) recombination between two Hcadj species or between Hcadj and H in the surface hydrides are possible. The TPD experimental results on the Sic-H system reveal two broad first-order H, desorption peaks at approximately 800 and 1200 K corresponding to the activation energies of 63 and 72 kcal/mol, respectively. These two distinct recombinations are attributed to surface heterogeneity from the two types of surface atoms [25]. When a linear H, recombination by the LH mechanism is included in the kinetic model for SiH, and CH, formation, the reaction product vectors for MH, formation and H, recombination can be derived by the same procedure used in Section 4.2. This anaIysis yields the following reaction product vector for H, recombination:

be extremely small because of the large difference in the activation energies, which are about 1 kcal/mol for MH, production and 60-70 kcal/ mol for H, production. That is, H, recombination by an LH mechanism cannot affect the kinetics of MH, production. Therefore, only H, recombination by an ER mechanism remains to explain the expected large hydrogen recombination process. The incident beam delivers about 5 monolayers of H atoms per second to the hydrogen-saturated surface, of which a fraction, 1 - n, scatters without any interaction, where n is the physical sticking probability of H atoms on the surface. Of the remaining portion, 7, that interacts with the surface, recombination occurs by the ER process: H(p) + MH, a

Hz --E Eh/l

kH2 bff.

M kM% d eff

MH, ’

(6)

where kdyF4 and kriM are the effective desorption rate constants fear MH, and recombined H, in the M site (M = Si or C), respectively. Eq. (6) demonstrates that LH recombined H, has the same phase lag as the corresponding tetrahydride reaction product and the apparent reaction probability differs from that of MH, by the ratio of the two rate constants. Since H, recombination both in the present model and in the TPD experiments are linear processes, estimation of the rate constant ratio in Eq. (6) is possible using the known activation energies. This ratio turns out to

(7)

which is governed by the cross section, (T, of the interaction. This is followed by rapid replenishment of the original surface hydride by pick-up of an H atom from the adsorbed population: MH,-,

+ Hcadj -

MH,,

(8)

where M is Si or C and y1= 1 (monohydride) or II = 2 (dihydride). Reactions (7) and (8) are equivalent to the overall ER recombination: H(gf +

and

Hzcgj + MH,_,

Il@+df ““,

H w

(9)

Thus, the effective sticking probability in Eq. (1) is defined as the difference between the probability of H-atom adsorption by physical sticking and the probability of loss by ER recombination:

(10)

where NMH and NMH are the concentration of surface monohydrides’ and dihydrides, respectively. From the fitting of the kinetic model to the experimental results q,rf = 9 x 1O-4 was obtained. Even though H-atom cross section data are not available, the geometric cross section, rTTs2,can be estimated from the H-atom diameter, 5 (0.75 A). The summation of surface hydride densities in Eq. (10) must be close to the surface

414

Y. Kim, D.R. Olander /Surface Science 313 (I 994) 399-416

atom density (- 1 X lo’* cmP2>, thus, the product a[(NsiH + Ncu) + 2(Ns,uz + NcHT)] is about 10W2.If the physical sticking probabihty, n, is just about 10% larger than this figure, only 10% of the incident H atoms survive ER recombination and participate in the tetrahydride-producing reactions.

is combined with the small positive activation energy of the subsequent rate controlling step, Hfadf + MH, --+MH,, the total activation energy for the product desorption is barely positive. The activation energies in the current Sic-H reaction are comparable to 1.8 kcal/mol for the Si-H reaction 1451 and 1.0 kcal/moI for the Sic-C1 reaction [46].

5.2. Rate constants The precursor model for SiH, and CH, production entails the sticking of H atoms from the molecular beam to an Head) overlayer on the Sic surface consisting of MH and MH, compiexes (where M = Si or Cl which are in equilibrium with each other. Addition of an adsorbed H atom to the MH, complex forms MH, and rapid addition of another H atom forms the final volatile tetrahydride. The effective rate constant, kz,“4, consists of two factors, the surface dihydride concentration, NMHZ, and the true rate constant, kyHz for the rate-limiting reaction: Head) + MH, -+ MI-I,. In the experiment it is impossible to obtain kyH2 and NrCIH separately. Thus, the 2MH,= k,MhN,, z, is aseffective rate constant, kdcm sessed. Direct comparison of the kinetic constants derived in Section 4.2 with other data is impossible because no comparable kinetic constants, theoretical or experimenta1, have been reported for the Sic-H system. The pre-exponential factors derived from the current model, 5.4 X lo2 se ’ for SiH, and 4.6 x lo2 s-l for CH,, can be compared to the value of 2.7 X lo3 s-’ for silane formation in the Si-H system 1451.The activation energies obtained in the current model are small compared to those of most chemical reactions. This is believed to be due to precursor formation in the SiH, and CH, production reactions. Surface monohydrides and dihydrides are thermally and chemically equilibrated on the Sic surface in both the SiH, and CH, production reaction, i.e., MH + MH ++ MH, + M. There is evidence of a negative reaction enthalpy in the equilibrium between SiH and SiH, [27,45]. The H atom is a free radical whose reactions in general require relatively small activation energies. Thus, when a negative enthalpy of the MH/MH, equilibrium

5.3. Grain boundary reaction The AFM micrographs in Fig. 2 reveal that the Sic reaction with thermal atomic hydrogen takes place not only on the grain surface but also in the grain boundaries (Section 3.1). It is also seen from the micrographs that etching in the grain boundaries produces small narrow pits. However, the grain surface remains essentially flat and the roughness due to the pits does not influence the experimental measurements. A rough estimate of the ratio of the area of the pits to the total etched area is about 0.1, based on the assumption that the original elevation of the Sic surface before etching is the top of the highest spot in the profile of the micrograph in Fig. 2b. This is in good agreement with Hallum and Herbell’s observation that the SiC surface underwent only slight grain boundary corrosion after reaction with hydrogen at 1000°C [22]. Also Fischman et al. reported that H, attacks Lu-Sic grains on the grain surface as well as into the grain boundaries [47]. 5.4. hydrogen bulk or sur$ace diffusion Figs. 3a and 3b demonstrate that long-range diffusion of adsorbed hydrogen atoms on the Sic surface is insignificant. This is understandable because the chemical bonding between adsorbate (H atom) and adsorbent (Si or C atom) is strong (Si-H: 70.4 kcal/mol, C-H: 98.8 kcal/mol) [48]. The apparent reaction probabili~ and phase lag measurements exhibit the characteristic of the surface reaction without bulk diffusion. It is well-known that hydrogen diffusion and solubility in silicon carbide are much lower than those in metals [491.

E Kim, D.R. Olander/Surface

5.5. Material balance on product fluxes

During the reaction the Si atom removal rate on the surface must be equal to the C atom removal rate: Rsiu, = RCH, + 2R~~r.r~’

(11)

where Ri is the desorption rate of each reaction product from the surface. As the apparent reaction probability, E, is defined as the product of the number of H atoms in the product molecule and the ratio of product molecule flux to reactant molecule flux, Eq. (11) can be rewritten as: ESiH4 = ECH,

+ %,Hf

(12)

Experimentally the ratio (ecu, + 4EC2H2)/EsiH, is 1.4 + 0.5 over the temperature range investigated (300-1100 K). The deviation from the unity is believed to be primarily due to the uncertainty in the combined detection efficiency calibrations. In any case, the experimental results are consistent with congruent gasification, in which the Si/C ratio of the reaction product gases is equal to that of bulk Sic. 6. Conclusions Silicon carbide is not etched by molecular hydrogen but interacts with atomic hydrogen to produce the volatile species, silane (SiH,), methane (CH,), and acetylene (C,H,). Below 1100 K, the apparent reaction probabilities for SiH,, CH,, and C,H, range from 3 X lop4 to 5 X 10F4, from 2 X low4 to 3 X 10e4, and from 9 x 1o-5 to 1.2 x 10-4, respectively. These reaction probabilities translate to an etching rate of about 2 monolayers/ hour at an equivalent atomic hydrogen pressure of 2 x 1O-5 Torr. Due to the carbon segregation effects the reaction probability for SiH, formation increases slowly with temperature and then decreases slightly above 800 K, whereas CH, formation keeps increasing even above 800 K. C,H, formation is detectable above 500 K and its temperature dependence follows that of SiH,. No diffusion of H atoms into the Sic matrix or on the surface is observed and surface topological changes examined by AFM reveal that the reac-

Science 313 (1994) 399-416

415

tion takes place on the grain surface as well as in the grain boundaries. The grain boundary reaction produces small narrow pits along the grain boundaries, but their effect on the reaction mechanism seems to be insignificant. A mechanistic interpretation of the modulated molecular beam data, coupled with TPD results, suggests a precursor reaction model for SiH, and CH, formation and a defect-site model for C,H, production. The mechanism involves adsorbed atomic hydrogen overlaying monohydrides, SiH and CH, and dihydrides in the form of C(SiH,)C and Si(CH,)Si clusters on the surface. The dihydride clusters act as precursors in the reactions that produce SiH, and CH,. C,H, forms from the defected site consisting of two loosely-bound carbons left in the C(SiH,)C cluster by SiH, desorption. Tetrahydride production is found to be satisfactorily modeled by a first-order reaction of the overlayer surface hydrogen with MH2 clusters. Rate constants characterizing the elementary steps of the mechanism were determined by fitting this model to the experimental data. The small difference between the physical sticking probability and the H 2 recombination probability explains why the effective sticking probability of the H atom on the Sic surface is low. Activation energies of the rate constants for tetrahydride production are small because of the low activation energy of the free-radical-like adsorbed hydrogen reactant and the negative enthalpy of the equilibrium between surface monohydrides and dihydrides.

The equipment used in this research was provided by the US DOE, Office of Basic Energy Science, Material Science Division. Conduct of the research was supported by the National Science Foundation under grant CBT-891240. References tll W. von Miinch, P. Hoeck and E. Pettenpaul, Int. Electron Devices Meeting, Washington, DC, 1977, p. 337.

416

Y. Kim, D.R. Olander/Sutface

[2] W. VOII Miinch and P. Hoeck, Solid State Electron. 21 (1978) 479. [3] J.D. Parsons, R.F. Bunshah and O.M. Stafsudd, Solid State Technol. 28 (1985) 133. [4] W. von Miinch, J. Electron. Mater. 6 (1977) 449. [S] L.H. Rovner and G.R. Hopkins, Nucl. Technol. 29 (1976) 274. [6] G.G. Trantina, Nucl. Eng. Des. 54 (1979) 67. [7] G.G. Trantina, J. Nucl. Mater. 85/86 (1979) 415. [S] G.R. Hopkins, J. Nucl. Mater. 85/86 (1979) 409. [9] J. Bohdansky and J. Roth, J. Nucl. Mater. 122/123 (1984) 1417. [lo] K. Sone, M. Saidoh, K. Nakamura, R. Yamada, Y. Muragami, T. Shikama, M. Fukutomi, M. Kitajima and M. Okada, J. Nucl. Mater. 98 (1981) 270. [ll] M. Mohri, K. Watanabe and T. Yamashina, J. Nucl. Mater. 75 (1978) 7. [12] T. Yamashina, M. Mohri, K. Watanabe, H. Doi and K. Hayakawa, J. Nucl. Mater. 76/77 (1978) 202. [13] J. Bohdansky, H.L. Bay and W. Ottenberger, J. Nucl. Mater. 76/77 (1978) 163. 1141 C.M. Braganza, G.M. McCracken and S.K. Erents, in: Proc. Int. Symp. on Plasma Wall Interaction, Jiilich, 1976 (Pergamon, New York, 1977) p. 257. [15] Y. Kim, The Investigation of Silicon Carbide Reaction with Thermal Atomic Hydrogen by Modulated Molecular Beam Mass Spectrometry, PhD Thesis, University of California, Berkeley, 1992. [16] R. Kaplan, Surf. Sci. 215 (1989) 111. [17] S. Adachi, M. Mohri and T. Yamashina, Surf. Sci. 161 (19851 479. [18] S. Adachi, S. Fukuda, M. Mohri and T. Yamashina, J. Nucl. Mater. 133/134 (198.5) 263. [19] S. Fukuda, S. Kato, M. Mohri and T. Yamashina, J. Nucl. Mater. 111/112 (1985) 839. [20] L.E. Davis, N.C. MacDonald, P.W. Palmberg, G.E. Riach and R.E. Weber, Handbook of Auger Electron Spectroscopy, 2nd ed., Physical Electronics Division (PerkinElmer, Eden Prairie, MN, 1978). [21] UT1 Quadrupole Mass Spectrometer Manual, lOOC-02 (1977). [22] G.W. Hallum and T.P. Herbell, High Temperature Effect of Hydrogen on Sintered Alpha-Silicon Carbide, NASA TM 88819 (1986). [23] M. Balooch and D.R. Olander, J. Chem. Phys. 63 (1975) 4772. [24] H. Harrison, D.G. Hummer and W.L. Fite, J. Chem. Phys. 41 (1964) 2567.

Science 313 (1994) 399-416 1251 M.D. Allendorf and D.A. Outka, Surf. Sci. 258 (1991) 177. 1261 G. Schulze and M. Henzler, Surf. Sci. 124 (1983) 336. 1271 H. Wagner, R. Butz, U. Backes and D. Bruchmann, Solid State Commun. 38 (1981) 1155. [28] H. Froitzheim, H. Lammering and H.L. Giinter, Phys. Rev. B 27 (1983) 2278. [29] J.A. Schaefer, F. Stucki, D.J. Frankel, W. Giipel and G.J. Lapeyre, J. Vat. Sci. Technol. B 2 (1984) 359. [30] A.V. Hamza, G.D. Kubiak and R.H. Stulen, Surf. Sci. 237 (1990) 35. [31] C.J. Wu and E.A. Carter, Chem. Phys. Lett. 185 (1991) 172. [32] L.S.O. Johansson, R.I.G. Uhrberg and G.V. Hansson, Phys. Rev. B 38 (1988) 13490. [33] T. Sakurai and H.D. Hagstrum, Phys. Rev. B 14 (1976) 1593. [34] S. Ciraci, R. Butz, E.M. Oellig and H. Wagner, Phys. Rev. B 30 (1984) 711. [35] Y. Kim and D.R. Olander, J. Nucl. Mater., to be published. [36] R.A. Swalin, Thermodynamics of Solids, 2nd ed. (Wiley, New York, 1972). [37] R.J. Good, L.A. Girifalco and G. Kraus, J. Phys. Chem. 62 (1958) 1418. [38] R. Yamada, J. Nucl. Mater. 174 (1990) 118. 1391 R. Yamada, J. Vat. Sci. Technol. A 5 (19871 305. [40] R. Yamada, J. Nucl. Mater. 145-147 (1987) 359. [41] J.W. Davis, A.A. Haasz and P.C. Stangeby, J. Nucl. Mater. 155-157 (1988) 234. [42] A.A. Haasz, J.W. Davis, 0. Auciello, P.C. Stangeby, E. Vietzke, K. Flaskamp and V. Philipps, J. Nucl. Mater. 145-147 (1987) 412. [43] C. Ferro, E. Franconi, A. Neri and F. Brossa, J. Nucl. Mater. 101 (1981) 224. [44] E. de la Cal, D. Tafalla and F.L. Tabares, J. Nucl. Mater. 196-198 (1992) 1041. [45] J. Abrefah and D.R. Olander, Surf. Sci. 209 (1989) 291. [46] M. Balooch and D.R. Olander, Surf. Sci. 261 (1992) 321. [47] G.S. Fischman, A. Zangwil and S.D. Brown, Adv. Ceram. Mater. 3 (19881 66. [48] L. Pauling, The Nature of the Chemical Bond, 3rd ed. (Cornell University Press, Ithaca, 1960). [49] R.A. Causey, J.D. Fowler, C. Ravanbakht, T.S. Elleman and K. Verghese, J. Am. Ceram. Sot. 61 (1978) 221.