J. Phys. Chem. 1992,96,7695-7703 (47) Gordon, M. S.;Gano, D. R.; Binkley, S.;Frish, M. J. J . Am. Chem. SOC.1986,108,2191-2195. (48) Ho,P.; Coltrin, M. E.; Binkley, J. S.;Melius, C. F. J. Phys. Chem. 1986, 90,3399-3406. (49) Vanderwielen, A. J.; Ring, M. A,; ONeal, H. E. J . Am. Chem. Soc. 1975, 97, 993-998.
76%
(50) McQuarrie, D. A. Statistical Mechanics; Harpor and Row: New
York, 1973. Hill, T. L. An Introduction to Statistical Thermodynamics; Addison-Wesley: Reading, MA, 1960. (51) Kee,R. J.; Rupley, F. M.; Miller, J. A. Sandia NalionaI Laboratwies Report, SAND87-8215, 1987. (52) Gordon, S.;McBride, B. D. NASA Report, NASA-SP-273, 1971.
Estimation of Arrhenius Parameters for the 1,l Elimination of H, from Si2H6and the Role of Chemically Activated Disilane in Silane Pyrolysis Harry K. Motfat,*.+Klavs F. Jensen,t and Robert W. Cam Department of Chemical Engineering and Material Science, University of Minnesota, Minneapolis, Minnesota 55455 (Received: May 12, 1992)
Arrhenius parameters for the 1,l H2 elimination from disilane are estimated. Chemically activated disilane should play a significant role in the high-temperature pyrolysis of silanes, such as the shock tube dissociation of SiH4and the atmospheric are also made that are consistent with those pressure silicon CVD from a SiH4 source. Estimates for AII'O~,~~~(H~S~S~H) Arrhenius parameters.
Introduction In a companion paper,' we estimated fall-off parameters for the 1,2 H shift disilane decomposition channel. In this paper we address and estimate the importance of other decomposition channels for disilane. For clarity, the numbering of the reactions will remain consistent between the two papers. Though the main channel for disilane dissociation is firmly believed to be reaction R1, the possibility of alternate channels has been addressed by several Because of the relative stability of the silylene species to that of the silyl species, it is not thought that the atomic hydrogen loss reaction (R12 would be competitive with the above reaction (even under chemical activation conditions). Moreover, the 75 kcal mol-' Si-Si bond dissociation energy'" precludes reaction R6 from playing much of a role during thermal processes. A classical trajectory analysis suggested that this mode may be the dominant dissociation mechanism at higher internal However, the 1,l loss of molecular hydrogen (R3) may be important in some circumstances for thermal reactions. Si2H6
1
SiH2
+ SiH4
3
SizH6f H2 + H3SiSiH
6
Si2H6 f SiH3 + SiH3 Si2H5+ H Though predicted to be the least endothermic of the disilane dissociation reaction channels, the four centered 1,2 H2 elimination reaction (RS) has been estimated to have a high activation energy by recent ab initio electronic structure These same calculations, however, indicate that the three-centered 1,l H2elimination reaction (R3) has an activation energy that is only 5 kcal mol-' above that of the main channel (Rl).4 Becerra and Walsh's work on the product decomposition pathways of chemically activated Si2H6from the silyl radical recombination reaction2v3has yielded experimental estimates for the reaction rate *Author to whom correspondence should be addressed. Present address: Sandia National Labs, P.O. Box 5800, Albuquerque, NM 87185. 8 Present address: Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139.
0022-365419212096-7695SO3.0010 , I
,
of R3 at one value of the internal energy of disilane. These may be combined to yield estimates for the Arrhenius parameters. Recently, Martin, ONeal, and Ring (MOR)6 have redone their earlier disilane pyrolysis work13and trisilane pyrolysis work14using silylene trapping agents to limit secondary reactions. They also extended their experiments to study tetrasilane pyrolysis. From a careful analysis of the trapping products, Arrhenius parameters for reaction R3 were estimated. The reaction channel R3 should be especially important as a competitor with the bimolecular direction of reaction R1, already shown to be in the pressure fall-off regime under typical CVD conditions. This chemical activation induced reaction will be labeled reaction R4 SiH2 + SiH4 F? Si2H6*2 H2 + H3SiSiH In the bimolecular direction of reaction R1, which will be labeled R-1, vibrationally excited disilane is formed with roughly an average of 59 kcal mol-I of internal energy at 700 K. This chemical activation can cause alternate channels for disilane dissociation that have large pre-exponential factors, which would not be competitive with the unimolecular direction of reaction 1, to be competitive in the bimolecular direction. In addition, these alternate channels will have a different pressure dependence in the fall-off region. In the low-pressure limit, the alternate channel reaction rate coefficient will be pressure independent, while the main reaction channel, stabilizationof vibrationally excited Si2&*, will be linear with respect to the pressure. This has important repercussions for understanding the homogeneous gas phase chemistry of low-pressure CVD processes employing silanes as growth precursors, where the recent emphasis has been to go to lower reactor pressures, below 1 Torr. The species formed from the alternate channel reaction (R4), silylsilylene, H3SiSiH, may therefore be an important intermediate in the formation of the higher silanes and, possibly, in the initiation of homogeneous nucleation of silicon particles. In this paper, an analysis of the 1,l H2 dissociation channel reaction rate is given. Fixed transition state RRKM numerical analysis is used as presented in our earlier work.lJ5 Arrhenius parameters for this reaction may be evaluated by combining Gordon et al.'s potential energy ~ t i m a t e sBecerra ,~ and Walsh's reaction rate data,3 thermodynamics, and Arrhenius parameters for the other silanelike dissociation reactions and silylene insertion reactions. These estimates are reviewed and compared with MORS recent indirect experimental measurement: The reaction rate of the chemically activated channel (R4) is also calculated. 0 - 1992 American Chemical Societv
7696 The Journal of Physical Chemistry, Vol. 96, No. 19, 1992
A discussion of when this channel becomes important compared to the stabilization reaction is presented, and a discussion of the importance of this channel to silane and disilane pyrolysis chemistry and previous experiments that have led to the establishment of the silane and disilane dissociation rate constant are made. Numerical Method Using fixed transition state theory summarized in Robinson and Holbrook16and previously presented in Moffat et al.,'J5 the rate of collisional stabilization of the main channel (R-1) in the presence of a chemically activated channel (R4) becomes
The rate of reaction of the bimolecular chemically activated channel (R4), k4(T,P)can be expressed as
(2) The rate of the unimolecular direction of the main channel is related to its bimolecular direction via microscopic reversibility. K , ( T ) is the equilibrium constant for R1. Equation 3, which represents the usual relationship between the forward and reverse rate constants of a chemical reaction, holds for all pressures despite the presence of chemical activation induced reaction channels. k-,(T,P) = k l ( T , P ) / K , ( T ) (3) In eqs 1 and 2, integration is carried out over E+, the energy above the zero point of the activated complex not including the energy in any adiabatic rotations of the activated complex. W(E+) is the sum of states of the active rotations and vibrations in the transition state whose total energy is less than E+. k,'(Eo E+) is the energy-dependent microscopic rate coefficient. ku[M] is the Lennard-Jones collision rate. flc is the weak collision efficiency. The sum over j in the above integrals is a sum over all reaction channels other than the main reaction channel, i.e., the channel along which the chemically activated species is being formed from the reactants. Fj is the Whitten-Rabinovitch correction factor for thejth channel. k,i(Eo E+) is the internal energy dependent rate of reaction along thejth channel, with an internal free energy for reaction along that channel of Eo E+ - Ed, where Ed is the critical energy for the jth channel. Eo is the zero point energy of the reactant. For the case where the chemically activated channel has a higher critical energy than the main channel (i.e., where Ed is >Eo),the internal free energy available at the transition state of the channel can become negative. In this instance, k,/ is defined to be equal to zero. Further details on the equations, on the construction of the transition state for the main channel, and on the statistical mechanical/thermochemicalconstants used for each molecule are contained in the work of Moffat, Jensen, and Carr.'
+
+
+
Construction of a R R K M Model for the 1,l H2Disilsne Dissociation ExperimentalRate Constant Measurements. There have been two indirect experimental measurements of R3's rate constant.2*6 Olbrich et al.2 estimated Arrhenius parameters for R3 based on a product analysis of the chemically activated dissociation of Si2H6
Moffat et al. created via the silyl recombination reaction ( R d ) SiH3 + SiH3 2 The silyl radicals were formed from the Hg-induced photodecomposition of H2 leading to the formation of silyl radicals from the resulting H abstraction reaction with SiH,. The Si2&* formed from R-6 has an average internal energy of 70-76 kcal It can either dissociate via the SiH2 dissociation channel or the 1,l H2dissociation channel. Collisional stabilization is negligible for such high internal energies. Becerra and Walsh3 have recently reanalyzed this experiment by kinetically modeling the observed primary production of Si2H6,Si3H8,and Si4Hloand have determined that a value of ka1((E*))/k,3((E*)) = 0.18 f 0.5 best fits the product distribution curves. RRKM modeling of the SiH2 dissociation and the 1,l H2dissociation reaction assumed that the two reactions have equal pre-exponential factors. However, this should be questioned considering the difference in the pre-exponential factors and reverse insertion reaction rates between the SiH4 and Si2& dissociation reactions. For example, the rate constant for the silylene insertion reaction into H2 (k2= 1.9 X 10I2cm3 mol-' s-') has been measured to be roughly 70 times slower than the rate constant for the silylene insertion reaction into SiH4(k-' = 1.3 X lOI4 cm3mol-' Using ( E * ) = 73.7 kcal mol-' yielded a difference of 3.9 kcal mol-' in the critical energies of the two dissociation channels. The resulting highpressure Arrhenius parameters for the 1,l H2 dissociation channel were predicted to be log k3(s-') = 15.91 - 56.4 kcal mo1-'/2.3RT at 574 K with a critical energy of 54.2 kcal mol-'. This should be compared to their values for the Arrhenius parameters for R l at 574 K, log kl(s-') = 15.91 - 52.5 kcal mol-'/2.3RT with a critical energy of 50.3 kcal mol-'. SiH,
2
z= H2 + SiH2
In the second indirect experimental measurement, Martin, ONeal, and Ring (MOR)6 have recently measured the energetics of R3 by analyzing product distributions from trisilane pyrolysis in a H2 bath gas. One of the two major decomposition pathways for Si3H8under pyrolysis conditions is R13. The other pathway is R7
13
Si3H8 SiH4 + H3SiSiH Under excess H2 conditions, silylsilylene will react with H2. The resulting SizHs will be vibrationally energized. In can either undergo collisional stabilization (R-3), react via the chemically activated pathway (R14) to form SiH2and SiH4, or react along the original pathway to re-form the reactants. MOR measured H3SiSiH + H2 ~t Si2H6*2
-
H3SiSiH + H2 i=?Si2H6*
14
SiH2 + SiH4
the "extra" production of SiH4 and Si2H6from trisilane pyrolysis under H2 trapping conditions versus under butadiene or trimethylsilane trapping conditions. The ratio of the relative rates of the extra production of SiH4 to Si2H6(0.43 at 569.2 K in 392.7 Torr of H2 and 7.3 Torr of Si3H8)then yields a measure of the relative rates for R-3 to R14. MOR used a simplified form of RRKM modeling that does not require a transition state formulation for R1 (all hot disilane molecules were assumed to have the same internal energy) to calculate an average vibrational energy content of the energized disilane molecule, 63.5 kcal mol-'. Assumptions on &,,(569 K) = 0.24, flcSijH (569 K) = 0.8, and the energetics of R1 had to be made, though. MOR calculated 57.8 kcal mol-' for the activation energy of R3 at 569 K when an activation energy of 52.2 kcal mol-' is assumed for R1. Energetic Information from Electronic Structure Cdcuhtions. Additional methods for estimation of the Arrhenius parameters for reaction R3 exist. The critical energy of the forward and reverse direction of R3 may be estimated from the ab initio
The Journal of Physical Chemistry, Vol. 96, No. 19, 1992 7697
Estimated Arrhenius Parameters
TABLE I: Summary of Arrhenius Parameters for 1.1 H2Disilrne Dissociation and Trisilane Dissociation" Si2H6a H2 + H3SiSiH Si2H6s SiH2 + SiH4 a H2 + Si3H8a SiH2+ method 1 method 2 SiHacase A SiH, fitted 3 Si,H, 49.01 15.91 60.17 0.14 x 10-7 15.42 58.60 14.68 57.21 55.72 57.41 1.5 X 10l2 3.0 X lo1* 0.39 55.79 0.07 76.51
k(925 K) @-I) log A(925 K) E,(925 K)
k(550 K) (s-l) log A(550 K) E,(550 K) log A(298 K) E,(298 K) M 0 ( O K) AHO(298 K) k,(298 K) (cm' mol-l A' E,' E,'
s-I)
EO' AHo1,298( HSiSiHJ or
411.2 16.30 57.92 2.7 x 10-7 15.8 1 56.3 1 15.03 54.86 53.26 54.96 2.0 x 10'2 4.5 x 10'2 0.49 53.26 0.0 74.06
3451. 16.04 52.91 1.2 x 10-5 15.83 52.20 15.48 51.54 52.32 53.46 1.28 x 1014 1.38 x 1 0 1 3 -1.33 50.15 -2.18 64.36
41.8 15.68 59.50 0.14 x 10-7 15.30 58.2 14.62 57.0 55.34 57.29 1.78 X 10l2 3.0 X lo'* 0.30 55.88 0.54 65.49
1931. 16.14 54.4 0.36 x 10-5 15.94 53.8 15.56 53.1 54.59 55.40 4.8 x 1014 2.7 X 10') -1.72 51.72 -2.91 64.80
Si3H8a H3SiSiH + SiH4 3115. 15.95 52.7 1.2 x 10-5 15.74 52.0 15.35 51.3 53.24 53.76 2.38 x 1014 1.02 x 10'3 -1.88 49.97 -3.28 74.06
M0f,298(SiH2) All energies are in kcal mol-'. electronic structure calculations of Gordon et a1.4 and Ho et al." Gordon et al. predict M o ( O K) for the 1,l H2 dissociation and for the SiH2 dissociation reaction to be equal to 53.4 and 50.0 kcal mol-', respectively. No energetic barrier for the reverse recombination direction is predicted for the SiH2 dissociation reaction, but a 2.1 kcal mol-' barrier is calculated for the 1,l Hz dissociation channel. However, the authors predict that the bamer will be reduced by 1 to 2 kcal mol-', if the complexity of the basis set used in the calculations is increased. Ho et al. predict W 3 ( 0 K) and W I ( O K) values (after application of a temperature correction to their reported MO(298 K) values) of 59.61 and 56.66 kcal mol-' for the 1,l H2 dissociation and the SiH2dissociation channels, respectively. The absolute magnitude of the W ( 0 K) values from both of these calculations is not in agreement with current estimates of Mor(SiH2)from thermal data. Also, the W r ( S i H 2 )value consistent with Gordon et ala's prediction of A P l ( O K), 62.04 kcal mol-', is quite different from their estimation of Wf,298(SiHz)= 65.3 kcal mol-', derived from the SiH4 1, H2 + SiHz reaction, using a larger set of basis functions.I8 However, Ho et al.'s predicted difference in W ( 0 K) between the two channels, 2.95 kcal mol-', is close to Gordon et al.'s 3.4 kcal mol-'. Therefore, it seems appropriate to scale Gordon et al.3 W 3 ( 0 K) value to our W I ( O K) value optimized in out fitting study, 52.32 kcal mol-' (from case A in our previous paper'). This W , ( O K) value is consistent with Wf,298(SiH2) = 64.36 kcal mol-'. Thus, a prediction of W 3 ( 0K) = 55.72 kcal mol-', which produces Wf,298(H3SiSiH)= 76.51 kcal mol-l, seems to be a reasonable number based on current electronic structure calculations. Method 1. These estimates of W 3 may be combined with estimates of the pre-exponential factor from the analogous reaction, postulated to have the same kind of transition state structure SiH4 f H2 + SiH2 2
to yield the Arrhenius parameters for 3
Si2H6z= H2 + H3SiSiH Our current estimate for the high-pressure A factor for SiH2 insertion into H2 at 298 K is 3.0 X 10l2 cm3 mol-' s-I.Is The activation energy of the reverse reaction at 298 K is 0.30 kcal moP, based on a value for the critical energy of 0.54 kcal mol-'. This same A factor can be used as a basis for constructing a transition state model. We expect that silyl substitution should have little influence on the pre-exponential factor or activation energy. This is in contrast to the effects of methyl substitution on 1,l H2eliminati~n,'~ which is seen to lower the pre-exponential
factor by roughly s-l and to raise the activation energy by 4-5 kcal mol-' per methyl group. However, this decrease in reactivity of silylene due to methyl substitution has been attributed to methyl being an electronegative substituent.28 The silyl group constitutes an electropositive substituent of silylene. Thus, if a silyl group substituent influence of silylene reactivity toward H2 is to be inferred, it would be toward greater reactivity. However, for the moment, we will assume no substituent influence and will consider the pre-exponential factor for R3 thus derived to be a lower bound. The critical energy for the reverse direction, EC3, may additionally be assumed to be in the 0 4 . 5 kcal mol-' range following ref 15. When combined with the W 3 ( 0K) value from the previous paragraph, a value of E< is arrived at. Therefore, and EC3,we have gathered enough by estimating W 3 ( 0K), L3, information to formulate a transition state. A set of four transition state vibrational frequencies can be adjusted so as to reproduce the needed A factor. The result of the transition state model produced by this procedure is presented in Table I under the column heading, method 1. The transition state vibrational frequencies used in method 1 are 2150 (4), 1400f (2), 900 (3), 800, 1328, 550, 488f (2), 400, 432, and 170, where f (= 0.281) is the factor adjusted to yield the desired pre-exponential factor. The statistical factor for R3, Lt3, is 6. Method 2. Arrhenius parameters for Si3H8dissociation from both types of 3-centered H migration reactions that lead to the formation of silylenes R7 and R13 have been measured by Vanderwielen et al.I4 They observed Arrhenius parameters of log k,(s-I) = (15.69 f 0.18) - (52.99 f 0.43) kcal mo1-'/2.3RT and log k13(~-')= (14.68 f 0.23) - (49.24 f 0.55) kcal mol-'/ 2.3RT. However, given suitable assumptions concerning the rate constant and activation energy for the reverse reaction (R-13), AHDf,z98(H3SiSiH) may be estimated via a third law approach from the Arrhenius parameters for reaction R13. Then, the critical energy for the forward and reverse directions of reaction R3 may be determined by the condition that they be consistent with the Wf,298(H3SiSiH) value determined from reaction R13. Combined with expected values for the pre-exponential factor and activation energy for the reverse silylene insertion reaction, the Arrhenius parameters for R3 may then be determined. We estimate that k-13(300K) = 1.1 X l O I 4 cm3 mol-* s-', based on Jasinski and Chu's measurement of the SiH2insertion rate into SiH420 Combined with Vandenvielen et ale'srate constant, kl3( T ) at 550 K, and the assumption that the k-13(T) is independent of T, this yields an estimate for the equilibrium constant at 550 K and thereby AGoI3(550K) = 35.92 kcal mol-'. Using ASoI3(550 K) = 31.21 kcal mol-I, derived from the assumptions concerning H3SiSiH's molecular parameters presented in the Appendix, produces W I 3 ( 5 5 0K) = 53.09 kcal mol-I. Adjusting for tem= 28.5 kcal mol-', yields perature and using AiYof,z9s(Si3H8) AHof,298(H3SiSiH)= 74.06 kcal mol-'. In turn, this value of
7698 The Journal of Physical Chemistry, Vol. 96, No. 19, 1992
AWf,298(H3SiSiH) implies W 3 ( 2 9 8 K) = 54.96 kcal mol-' and W 3 ( 0K) = 53.26 kcal mol-'. When this is combined with an assumption of 0 kcal mol-' for the critical energy of the reverse reaction, Eo3 = 53.26 kcal mol-' is estimated. Combined with our estimated A factor developed in the previous paragraph, this yields a full transition state model formulation, whose results are presented in Table I under the column heading, method 2. We used a slightly increased estimate for the A factor over that used in method 1 in the spirit of making method 2's Arrhenius parameters an upper bound for the 1,l H2 disilane decomposition channel. Values for the vibrational frequencies are the same as those used in method 1, except f equals 0.230. The resulting Arrhenius parameters for k3(T ) from method 2 are close to Becerra and Walsh's parameters. For example, BW's Arrhenius parameters result in k3"(550K) = 3.1 X lO-'s-', while the k3"(550 K) value from method 2 is 2.7 X 10-7s-1.A lower A factor for the 1,l H2dissociation reaction than BWs is predicted at the expense of a lower relative dierence in the critical energies between the SiH2and 1,l H2 dissociation channels. This is due to our original assumption that the 1,l H2 dissociation channel should have a lower A factor than the 3-centered H migration SM2disilane dissociation channel. Note also that the relative differences in 0 K heats of reaction between reactions R1 and R3 is 0.96 kcal mol-' (despite the 3.09 kcal mol-' difference in critical energies), when method 2's model is used, instead of Gordon et al.'s predicted 3.4 kcal mol-'. We probably could have been equally justified in assuming a 1 kcal mol-' critical energy for E c 3 instead of 0 kcal mol-' (e&, we used a critical energy of 0.54 kcal mol-' for SiH2 H2 SiH,). However again, method 2 serves as a reasonably tight upper bound to the magnitude of the 1,l H2 decomposition channel. Becerra and Walsh's predicted ratio of products from the decomposition of hot disilane created from silyl radical recombination was checked against our two transition state models for R4. We used Arrhenius parameters of log k6(T) = 16.60 - 74.50 kcal mol-'/2.3RT'O in our RRKM calculations to create a distribution of energized disilane molecules. The average internal energy of the distribution (300 K) was roughly ( E * ) = 74.1 kcal mol-'. Using method 2's Arrheniw parameters for the 1,l H2dissociation channel and case A's Arrhenius parameters for the SiH2dissociation channel,' we obtained a product ratio of ( k a ' ) / ( k 2 )= 0.16. &,'((E*))was equal to 2.8 X 1O'O s-I, while k,3((E*))was equal to 4.47 X lo9 s-l. Thus, not surprisingly considering the small difference in Arrhenius parameters between BW's recommended values and ours, method 2's transition state predicts BWs product ratios, 0.18 f 0.05, nearly to within the level of experimental uncertainty. When method 1's Arrhenius parameters for the 1,l H2 dissociation channel are used, a product ratio of {ka')/(k,3)= 0.02 is obtained. This ratio lies outside the bounds of Olbrich et ale's experimental uncertainty.2 Method 1 stays faithful to Gordon et al.'s predicted difference in W ( 0 K) between R1 and R3. Its 550 K activation energy, 58.60 kcal mol-', is 0.8 kcal mol-' more than MORS prediction, 57.8 kcal mol-'. MOR did not predict the pre-exponential factor for R3. Both method 1 and method 2's Arrhenius parameters are consistent with MORS experiments on the product yields from trisilane pyrolysis. MOR concluded that under their experimental conditions, the ratio, k14/k-3,was equal to 0.43. We calculate using the full RRKM model (eqs 1 and 2) that method 2's Arrhenius parameters for the 1,l H2 dissociation channel yield this ratio ifflc,,,(300 K) = 0.33. Method 1 yields this ratio if flc,,2(300 K) = 0.466. Collision efficiencies in the silane system are not that well-known to distinguish the veracity of either set of Arrhenius parameters from MOR'S experiments. The value of the ratio, k14/k-3,is roughly inversely proportional to flc,H2under MORS conditions.
+
-
Mscussion of the 1,l H2Elimination Reaction Thus, we have constructed two sets of Arrhenius parameters for the 1,l H2 elimination reaction, one conservative and the other not so conservative but supported by indirect experimental evidence. The rate constants produced by these two models will be
Moffat et al. compared under several conditions not used in their derivation to evaluate the importance of the 1,l H2 decomposition channel. Comporisoa witb Otber Experimental Data. Dzarnoski, Rickborn, O'Neal, and Ring (DROR) studied the pyrolysis of Si2H6 and Si2D6under shock tube condition^.^ The experimentalconditions were centered around 925 K, 2500 Torr in an argon bath gas. The mechanism for the decomposition of disilane was evaluated by analyzing the yield data from the Si2D6shock tube results. Si2D6 Z D2 + D3SiSiD I'
Then, the overall rate of reaction in terms of the observed rate of disilane loss was determined from the Si2H6shock tube experiments. Relative yields of SiD4and D2 from Si2Dsshocks were assumed to correspond directly to the relative values of kloto kll. It was also assumed that the ratio of kloto kll was equal to the ratio of kl to k,, thereby neglecting any possible isotope effects, in order to derive rate constants for kl and k3 from the Si2& shock tube data on the overall rate of Si2H6destruction. In analyzing DROR's results, we also assume that there is no isotope effect. Thus, we will discuss whether the entire yield of SiD4 and D2 observed by DROR can be attributed to R10 and R11 only by evaluating the rate constants for R1 and R3. What we find is that even assuming method 2's Arrhenius parameters, the 1,l H2 decomposition channel for disilane cannot account for all primary D2 production observed by DROR. We used Arrhenius parameters for the 1,2 H shift disilane decomposition channel of log kl"(T = 500 K) = 15.78 - 52.07 kcal mol-'/2.3RT. These were determined by optimizing the Arrhenius parameters against experiments on both the unimolecular and bimolecular directions (case A'). Case A's thermochemistry is summarized in Table I. From Table I, kl"(925 K)/k3"(925 K) = 8.39 is predicted. However, when fall-off effects are taken into account, kl(925 K, 2500 Torr)/k3(925 K,2500 Torr) = 28.1 is calculated. DRORs experimentally observed primary product formation ratios of SiD4/D2= 4.26 i 1.0 were attributed solely to the ratio of the rates for the primary pyrolysis channels for disilane under their experimental conditions, k3(T,P)/k,(T,P). However, in order for this to be true, k1"(925 K)/k3"(925 K) would have to equal 0.44,' resulting in a difference in critical energies between the two channels of 0.68 kcal mol-'. This small difference is not supported by recent electronic structure calculations~Moreover, if this small difference in critical energies between the two channels were true and physically reasonably estimates for the preexponential factor for the 1,l H2dissociation channel are used, much greater primary H2 production would be observed under static conditions than actually is observed.21 Thus, either there is a large isotope effect that causes the ratio, kl/k3, not to equal the ratio, klo/kll,under DRORs experimental conditions, or a large portion of the primary D2 formed in DROR's shock tube reaction experiment must come from other sources than the 1,l D2 disilane dissociation channel (R1 1). Possible sources of primary formation of H2 (and for D2 in the analogous case of SizD6shocks) include the unimolecular dissociation of SiH2and H3SiSiH, generated by R3 or via Si3H8 dissociation reactions. SiH, dissociation is too slow to be a factor under the conditions of DROR's experiment. Under Bowrey and Purnell's (BP)21 or Martin, Ring, and O'Neal's (MRO)6J3low-temperaturestatic disilane dissociation experimental conditions, the contribution of the 1,l H2 channel to the overall rate of disilane dissociation is even less than under DRORs high temperature shock tube condition^.^ R3's critical energy is higher than Rl's critical energy. The predicted primary H2 and Si4Hloformation under static disilane conditions from the 1,l H2 dissociation reaction is far lower than the observed amounts of Bowrey and Purnell. Si4HI0formation results from H3SiSiH insertion into Si2H6. Under their reactor conditions (575 K,62 Torr), k l / k 3= 62.5 is predicted using case A and method 2's assumptions on the high-pressure Arrhenius parameters. Product ratios observed under disilane static pyrolysis conditions (Le., the relative production of SiH, to H2) can serve as upper bounds on
The Journal of Physical Chemistry, Vol. 96, NO. 19, 1992 7699
Estimated Arrhenius Parameters
........
........ ....
m 0
..... c
.........
9
0
60
55
60
66
70
76
80
86
E* (kcal mole') Figure 2. f.,(E,T) = chemical activation distribution function of internal energies for disilane from the bimolecular direction of the main disilane decomposition channel (R-1): -, I200 K, ---,700 K, 300 K. Plotted on the right axis is k,l(E*) (-.-) and k,'(E*) (---), the microscopic reaction rates for the main channel and the 1,l H2 elimination channel, respectively.
..-,
........ . . . . . .
..... 4,,& ........*.
f
.l.
-lo-2
16'
loo
10'
,,1,,1
10"
16
10'
Pressure (Torr) Figure 1. (a, top) Relative rate of stabilization (SiH2 SiH, 2 Si2H6) versus the chemical activation reaction (SiH2 SiH, s Si2H6*f H2 HSiSiH3) as a function of the bath gas pressure at three different 1200 K; ---,700 K, e.., 300 K 0,stabilization temperatures: -, reaction; 0 , chemically activated channel reaction. A silane bath gas with /3uil.nc(300K) = 0.5 is assumed. Method 2's Arrhenius parameters for the 1,l H2 dissociation reaction are used. The rate constant divided by k-,"(T) is actually plotted (k..1"(300K) = 1.3 X lOI4, k1'(700 K) = 8.1 X 1013, k-,"(1200 K) = 1.7 X IO" cm3 mo1-I S-I). (b, bottom) Same calculation as part a, except that method 1's Arrhenius parameters are used for the 1,l H2 dissociation channel.
+
+
+
the possible value of the rate constant for R3 or almost equivalently as a lower bound on the difference in critical energies between the two disilane dissociation channels. Method 2's Arrhenius parameters do not violate the observation from those relatively low temperature disilane pyrolysis studies6J3**'that H2 is not observed in quantity as a primary product. However, it would not take much of an increase in the rate constant from method 2's values for this observation to be violated. Contributionsof the 1,l H2channel from Chemically Activated Disllpae. The 1,l H2dissociation channel does become important via the chemically activated reaction R4 for conditions corresponding to high temperature silane pyrolysis. Figure la displays k-,(T,P) and k4(T,P) versus pressure for three different temperatures, 300, 700, and 1200 K. A SiH4 bath gas using &rilane(300 K) = 0.5 for the collision efficiency of silane with disilane is assumed. Method 2's parameters for the 1,l H2 dissociation channel and case A's parameters for the three-centered H migration channel are used. At the low temperatures of the association reaction experiments of Jasinski and Chu and oth300 K,stabilization dominates over the chemically acers,20*22 tivated channel reaction. However, Figure l a demonstrates that the chemically activated channel will dominate at temperature/pressure conditions roughly above lo00 K and below 1 atm. These conditions are characteristic of those used in the shock tube pyrolysis of SiH:3q24 and in the chemical vapor deposition of epitaxial and polysilicon from a SiH4 source.
Figure l b is identical to Figure la, except that method 1's Arrhenius parameters and transition state is used for the 1,l H, dissociation channel. The differences between parts a and b of Figure 1 demonstratethe effect that the uncertainty in Arrhenius parameters for the 1,l H2dissociation channel has on the predicted rate constant for the chemical activation channel R4. The main effect is in the predicted low pressure limiting rate constant for R4, k40. k40 from method 1's assumptions is roughly an order of magnitude lower than from method 2's assumptions (to be specific, k4O(7O0 K,method 2) = 3.94 X 10l2cm3mol-' s-I, while k40(700 K,method 1) = 3.27 X 10" cm3 mol-' s-'). Therefore, this uncertainty in k40 is of comparable magnitude to the uncertainty in the high-pressure forward rate constant for Si2H6& H2
+ H3SiSiH
(see Table I). Figure 1 demonstrates that uncertainties in the Arrhenius parameters for the 1,l H2channel do not affect the rate constant for the stabilization reaction R-1. Figure 2 illustrates why the temperature plays an important part in determining whether the chemically activated channel (R4) or the stabilization reaction (R-1) dominates. Plotted on the left axis is f-)(E,7') at three different temperatures, 300 K, 700 K, and 1200 K. f-l(E,T) is the chemical activation distribution function. It gives the relative concentration of energized disilane molecules created by R-1 as a function of E*, the energy above the zero point energy level of disilane. It is normalized in the sense that its integral over E* is equal to 1. At 300 K, f-I(E*,T) is sharply peaked at a value roughly 2 kcal mol-' above the critical energy for R1, 50.15 kcal mol-'. A large fraction of the f-,(E*, 300 K)'s distribution lies below method 2's critical energy for the 1,l H2 disilane dissociation channel. For this portion of energized disilane molecules, there is no possibility of reacting along the chemically activated channel. Also plotted in Figure 2 on the right axis are the microscopic rate constants for R1 and R3, k,'(E*) and k,3(E*). These are used in eqs 1 and 2. It can be seen that kal(E*)is greater than &:(E*) over the entire range of E* relevant to thermally driven reactions, due to the lower critical energy for R1 compared to R3. If R3 were postulated to have a larger pre-exponential factor than R1, than a crossing of the k,'(E*) and ka3(E*)curves would be expected at some value of E*. However, this is predicted not to be the case. The significance of the relative magnitudes of k,' and kn3can be seen in Figure la in the low pressure limiting behavior of the chemically activated reaction rate constant for R4, k40(T). Its value is significantly below k-,"(T), especially for low temperatures such as 300 K where a significant portion of the total distribution of energized lies below R3's critical energy. molecules, f-,(E,T), In contrast is the low pressure limiting behavior of the chemically activated channel in Figure 3. Figure 3 and Figure 4 arc direct analogs of Figure l a and 2 for the case of R-3 being the
7700 The Journal of Physical Chemistry, Vol. 96, No. 19. 1992
?y
‘10-2
100
10-1
101
102
104
103
Ressure (Torr)
Figwe 3. Relative rate of stabilization (H2 + H S i i H 3 Si2H6)versus
the chemical activation reaction (H2+ HISiSiH F? Si2H6* SiH2+ SiH,) as a function of the bath gas preasure at three different temperatures: -, 1200 K - - -,700 K -,300 K 0 , stabilization reaction; 0, chemically activated channel reaction. A silane bath gas with @@,(300 K) 0.5 is assumed. Method 2’s Arrhenius parameters for the 1,l H2 dissociation reaction are used. The rate constant divided by k+“( r ) is actually plotted (k-3”(300 K) = 1.8 X 1012,k-3”(700K) = 6.8 X 10l2, k3”(1200 K) = 2.3 X lOI3 cm3mol-’ e l ) .
--
5:
9 8 60
66
60
66
70
76
80
86
E* (kcal mole-’)
= chemical activation distribution function of internal energies for disilane from the bimolecular direction of the 1,l H2disilane decomposition channel (R-3): -, 1200 K; ---,700 K; ..., 300 K. Plotted on the right axis is k,’(E*) (---) and k,3(E*) (---), the microscopic reaction rata for the main channel and the 1,l H2elimination Figure 4. f&T,r)
channel, respectively. main reaction channel which creates internally energized disilane. The same bath gas conditions were used. Figure 3 plots the relative rate of the stabilization reaction (R-3) and the rate of the chemically activated reaction channel (R14) as a function of total pressure. f-&?,7‘) is plotted in Figure 4 for three different temperatures, In this case, the rate of the chemically activated channel (R14) in the limit of low pressure approaches k_,”(T) for all temperatures. In other words, in the absence of collisional stabilization, almost all of the energized disilane that is formed as a result of the H2 H3SiSiH channel will dissociation along the SiH2 SiH4 channel. This is again a manifestation of k,l(E*) being larger than ka3(E*). Equations 1 and 2 demonstrate that the absolute magnitudes of k,’(E*) and k,3(E*) should be compared to the value of &kLJ[M] in order to determine the ratio of the rates for collisional stabilization versus reaction along the chemically activated reaction channel or reaction along the reactants’ channel. &kLJ[M] is independent of E* under the current level of theory, linear in the total pressure, and only weakly dependent on temperature (mostly through the expected temperature dependence of the weak collision efficiency, & and partly through the Lennard-Jones collision frequency, kLj). f(E*,T) is a strong function of temperature, however. Thus, the value of T will have a strong influence on the degree of collisional stabilization because off(E*,T). As an example, at 1 Torr and 700 K,&kw[M] is equal to 7.6 X lo6 S-I (&(700 K) = 0.28). The maximum inf-,(E*,700 K) occurs at roughly E* = 59 kcal mol-].
+
+
Moffat et al. The value of k,3(59 kcal mol-’) is roughly equal to &kLJ[M]. Thus, in Figure la, the rate of collisional stabilization is equal to the rate of the chemically activated channel at 700 K and 1 Torr. At 1200 K, however, the average energy of energized disilane molecules is roughly 73 kcal mol-’. k,3(73 kcal mol?) is roughly 7 X lo9 s-l. Thus, Figure la shows that the chemically activated channel dominates the collisional stabilization channel at 1 Torr and 1200 K. Under Erwin, Ring, and O’Neal’sZ4experimental conditions for the shock tube SiH4pyrolysis in an argon bath gas (1 100 K and 3200 Torr), using method 2’s Arrhenius parameters, k4 is predicted to be 2.1 X cm3 mol-] s-I in contrast to a rate of stabilization of k-l = 0.84 X lOI3cm3 mol-’ s-l. Even if method 1’s parameters are used, k4 = 0.35 X lOI3 cm3 mol-I s-l and k-] = 0.86 X 1013cm3 mol-’ s-l are predicted. Thus, the chemically activated reaction channel (R4) should be very significant under SiH4shock tube pyrolysis conditions. The product stoichiometry observed by Neuman et al.23in their shock tube SiH4 pyrolysis study was AH2/ASiH4 = 1.85. The existence of the chemically activated channel (R4) helps to explain this ratio, which the original authors had difficulty justifying. H3SiSiH produced by R4 can isomerize to HzSiSiHz,which can then undergo additional H2 molecular eliminations with an estimated activation energy of 53 kcal mol-’ that lead to increased formation rates for H2. Typical conditions for atmospheric silicon epitaxy from a SiH4 source are 1300 K, 760 Torr, and a H2 bath gas. Under these conditions, using method 2’s assumptions, k4 = 3.8 X 1013cm3 mol-’ s-l, while k-] = 0.45 X 10” cm3 mol-’ s-I! Even using method 1’s conservative assumptions k4, 0.79 X lOI3 cm3 mol-’ s-I is greater than k-l, 0.47 X 1013 cm3 mol-] s-l. Thus, the formation of Si2H6and probably other higher silanes are significantly affected by the presence of this chemically activated reaction during the epitaxial silicon process. It is not clear what effect this chemical activation reaction has on the full silane pyrolysis mechanism without a complete kinetic modeling study. The H2 bath gas should suppress the formation of H2 denuded silanes. Thus, some H3SiSiH may undergo a recombination reaction with the Hz bath gas to yield SizHs,while some H3SiSiH may undergo further decomposition reactions eventually yielding two silylene molecules and/or silicon atoms. The isomerization reaction producing HzSi=SiH2 has been estimated to have only a 5 kcal mol-l activation en erg^.^,^ The latter scenario may be important for predicting the onset of homogeneous nucleation of silicon powders under silane pyrolysis conditions. Even at the mean reactor conditions of the static silane pyrolysis conducted by Purnell and Walsh (PW) (675 K, 135 Torr, SiH4 bath the chemically activated reaction, k4, can account for a significant amount of the primary production of Si3H8. Purnell and Walsh observed an initial product stoichiometry of [Si3H8]/[Si,H6] = 0.13. Using method 2’s assumptions, k4/kl = 0.06, while using method 1’s assumptions, k4/k-, = 0.005. Thus, method 2’s k4/kdI predicted ratio is of the correct order of magnitude to account for PW’s [Si3H8] primary production amounts, given the assumption that all the H3SiSiH produced by reaction R4 yields Si3H8via an additional SiH4insertion reaction. In summary, the predicted presence of the chemically activated channel (R4) is not in codlict with any previous experiments used to determine Arrhenius parameters for the dissociation reactions of silanes. R4 may actually help to explain some of the observed primary production of Si3H8in PW’s experimentz6and the initial product stoichiometry, AHZ/ASiH4 = 1.85, observed in Newman, Ring, and O’Neal’sz3shock tube silane pyrolysis study. Arrhenius Parameters for the Pyrolysis of Trisilane and
Higher Silanes The same type of transition state occurs for trisilane dissociation reaction as for disilane dissociation. Negative critical energies in the reverse direction appear here, too, when tight transition states are constructed to agree with experiments on the forward and reverse reaction rates. Vanderwielen et al.I4 reported experimental values of log k7( T = 550 K) s-l = (15.69 0.18) (52.99 f 0.43 kcal mol-’)/2.3RT and log k , , ( T = 550 K)s-I =
*
Estimated Arrhenius Parameters (14.68 f 0.23) - (49.24 kcal mol-')/2.3RT), while Jasinski and Chu report k7(300 K) = (1-2.1) X lOI4 cm3mold1s-' depending on the pressure,20and Inoue and Suzuki measured k2(300 K) = 3.4 x lOI4cm3mol-' s-l.= Recently, MOR6redid their trisilane pyrolysis study under silylene trapping conditions (excess trimethylshe or butadiene). For R7, they obtained a slightly lower rate constant and higher,activation energy than under neat pyrolysis conditions, log k7(T = 550 K) s-' = 15.70 - (53.2 f 1.15 kcal molT')/2.3RT. For R13, they obtained a slightly lower rate constant but a significantly different activation energy, log k13( T = 550 K) s-l = 15.41 - (51.17 f 0.43 kcal mo1-')/2,3RT. Transition states for reactions R7 and R13 were constructed using the procedures previously employed in this paper that are consistent with the above experiments. The main results of these calculations are presented in Table I. No internal rotors or corrections for changes in the overall moments of inertia were used. A transition state for R7 was constructed that reproduces MOR'S recent Arrhenius parameters. The reactant vibrational degrees of freedom were set to 2150 (8), 920 (7), 625 (2), 420 (2), 380 (2), 715,400,350, 135, and 121 (2). The transition state vibrational frequencies were 2150 (7), 1200,910 (2), 600 (2), 920 (3), 700,625,420,144,380 (2), 245,137,120,69, and 121. The critical energy for the unimolecular direction was 51.22 kcal mol-', and the statistical factor was 6. Combining this result with our best estimate for APf,298(SiHJ,164.8 kcal mol-', yielded a value for k7(298 K) of 7.6 X lOI4 cm3 mol-' s-I. For comparison, we Used Wf,2g8(Si&) = 28.5 kcal mol-', Wf,p8(Si&j) = 19.1 kcal mol-', and AS7(298 K) = 3 1.77 eu in the above calculation. The predicted activation energy of the reverse bimolecular direction under trisilane pyrolysis temperatures, 567 K,was -0.61 kcal mol-'. The predicted value of k7(298 K) is higher than experimental results indicate. Two possible remedies present themselves. The predicted value of k-,(298 K) may be lowered if a lower value of Wr,298(SiH2)is assumed. We ran a calculation using Wf,298(SiH2)= 64.19 kcal mol-!, within current uncertainty limits of its va1ue.l Using this value of Wf,298(SiH2) has the attribute that the predicted activation energy of the reverse bimolecular direction under trisilane pyrolysis temperatures is equal to zero. The resulting predicted value of k7(298 K) was equal to 2.7 X 1014cm3 mol-l s-I, within the range of experimental measurements. The second possibility is to modify the activation energy of R7, while maintaining the absolute value of the reaction rate at the temperature of the experiment. If the experimental activation energy is increased by 0.6 kcal mol-', Arrhenius parameters of log k7(T = 550 K) s-' = 15.94 - (53.8 kcal mol-')/2.3RTresult. This increase in the activation energy is half the total uncertainty in its value quoted by MOR. This increase also has the advantage of producing a nearly zero activation energy for R-7 under trisilane pyrolysis temperatures. The resulting estimate for k7(298 K), 4.8 X lOI4cm3mol-' s-I, is still roughly a factor of 2 larger than experimental measurements. However, since the k-7(550 K) value is 2.1 X loL4cm3mol-l s-l and the temperature extrapolation of the high-pressure rate constant is uncertain, reasonable compatibility of the experimental data for both the unimolecular and bimolecular direction with the thermochemistry can be claimed. The resulting critical energy for the transition state model in the forward unimolecular direction is 5 1.72 kcal mol-', while the critical energy in the reverse bimolecular direction is -2.91 kcal mol-I, comparable to the disilane dissociation case. The transition state vibrational degrees of freedom are 2150 (7), 1200,910 (2), 600 (2), 920 (3), 700,625,420, 129, 380 (2), 219, 123, 107,61, and 121. Further details on this calculation are presented in Table I. In summary, we have shown by a third law argument that the experimental Arrhenius parameters for the unimolecular direction of R7 are compatible with experimental measurements for the bimolecular direction R-7, given the current level of uncertainty in AHof,,,,(SiH2) and temperature extrapolation of the rate constant for R7. It has been previously noted in the derivation of Arrhenius parameters for the 1,l H2 disilane dissociation reaction that = 74.06 kcal mol-' from a calculation of the AHof,298(H3SiSiH)
The Journal of Physical Chemistry, Vol. 96, No. 19, 1992 7701 equilibrium constant for reaction R13 at 550 K. That calculation assumed that k43(T) was independent of temperature and equal to Jasinski's measured value for SiH2insertion into SiH4,k-,(300 K) = 1.3 X lOI4cm3 mol-l s-I. The consistency of the thermochemistry and assumptions on the reverse reaction rate, kl3( T), can be checked against experiments on the forward reaction6J4 by formulating a transition state model for R13. We used MORS reevaluated Arrhenius parameters for kI3(T ) as the basis for the transition state formulation. The reactant vibrational frequencies were the same as the ones used for R7. The transition state vibrational frequencies were 2150 (7), 1200,910 (2), 600 (2), 920 (3), 625 (2), 420, 168, 380 (2), 286, 160, 140, 80, 121, and the statistical factor was set to 6. Critical energies of 49.28 and -3.96 kcal mol-I for the forward and reverse directions are predicted. Values of kI3(298K) = 4.4 X lOI4 cm3 mol-I s-' and k-13(550 K) = 1.0 X lOI4cm3mol-' s-I with E[I3(550 K) = 4 . 8 kcal mol-I are also predicted. Just as in the previous case for R7, two possibilities for reducing the predicted negative activation energy in the reverse direction arise. One possibility is to lower the estimated value for LW"f,298(H3SiSiH).The other possibility is to increase the activation energy of the forward unimolecular direction, while maintaining the more reliably known value of the overall experimental rate constant. If the experimental activation energy for R13 is increased to 52.0 kcal mol-' (the quoted experimental uncertainty in E[I3(55O K) is f0.43 kcal mol-I), then Arrhenius parameters of log k13(T = 550 K) s-I = 15.74 - (52.0 kcal mol-I)/2.3RT can be used as the basis for constructing another transition state. The resulting transition state vibrational frequencies are 2150 (7), 1200, 910 (2), 600 (2), 920 (3), 625 (2), 420, 144, 380 (2), 245, 137, 120, 69, and 121 with critical energies of 49.97 kcal mol-' in the unimolecular direction and -3.14 kcal mol-' in the bimolecular direction. Values of kI3(298 K) = 2.4 X lOI4 cm3 mol-' s-l and kI3(550K) = 1.0 X 1014cm3 mol-' s-l with E ~ ~ ~ ( K) 5 5=00.0 kcal mol-) are now predicted. Further details of the calculation are given in Table I. In this manner, a unified view of the Arrhenius parameters for SiH4, Si2H6,Si3H8,and probably higher silanes dissociation and their reverse association reactions may be formed. Preexponential factors for 1,l H2 dissociation channels and 3-centered H migration channels leading to the formation of silylenes can be predicted. Activation energies and pre-exponential factors for the reverse recombination reactions also follow a pattern. Activation energies for the bimolecular direction of the 1,l H2 dissociation channels are roughly 0.5-1 kcal mol-I, while for the 3-centered H migration channels, activation energies for the bimolecular direction are -2 to -3 kcal mol-'. Possible effects due to chemically activated reaction channels for the case of trisilane may be explored and contrasted with the analogous cases for disilane. Plotted in Figure 5 at three different temperatures are the relative rate constants for the stabilization reaction, SiH4 + H3SiSiH2 Si3H8,versus the chemically activated reaction channel (R15)
-
SiH4 + H3SiSiH a Si3H8*
IS
SiH2
+ Si2H,
Analogously, Figure 6 plots the relative rate constants for the stabilization reaction, SiH2 + Si2H62 Si3H8,versus the chemically activated reaction channel (R16)
-
SiH2 + Si2H6F! Si3H8*
16
SiH4 + H3SiSiH
The bath gas conditions used were equivalent to those used in K) Figures 1 and 3, Le., a silane bath gas assuming @c,silanc(300 = 0.5. We have shown above that experiments on trisilane dissociation are consistent with R7 having a higher critical energy than R13 by 1.8 kcal mol-I. Therefore, the relative rates for the chemically activated channel versus the stabilization channel are higher in Figure 6 than in Figure 5. However, the differences between Figures 5 and 6 are not as great as the differences between Figures 1 and 3. This is due to the fact that the disparity in critical energies between the two lowest channels for disilane dissociation is 3.09 kcal mol-' whereas it is smaller for the trisilane dissociation are claper in value case. Thus, the values of k,7(Ec)and ka13(,??+)
7702 The Journal of Physical Chemistry, Vol. 96, No. 19, 1992
Moffat et al.
TABLE Ik Lenoard-JoacsParameters molecule o ( A ) c / k (K) molecule SizH, H2 2.827 59.7 SiH4 4.084 Si3H8 207.6
?
y gio-2
10'
1d
id
io3
IO'
Pressure (Torr) Relative rate of stabilization (SiH, + H3SiSiH 2 Si3Hs)versus the chemical activation reaction SiH, + H3SiSiH iSi3H8*s SiH2+ Si2H6)as a function of the bath gas pressure at three different temperatures: -, 1200 K - - -,700 K; ., 300 K 0 , stabilization reaction; 0,chemically activated channel reaction. A silane bath K) = 0.5 is assumed. The rate constant divided by gas with &,h,(300 &-,"(T) is actually plotted (k3'(300 K) = 2.4 X lo1', &-13"(700K) = 1.1 X k-13'(1200 K) = 2.3 X lo1' cm3 mol-l s-l). Figure 5.
E
c/&(K) 301.3 331.0
pected to play a sisnifcant role in high-temperaturepyrolysis cases, such as the shock tube dissocietion of SiH, and atmosphericsilicon CVD from a SiH, source. Their rate constantsunder silane bath gas conditions are estimated. Transition state models for the main decomposition channels of trisilane based on experimental measurements for the forward unimolccular dissociationdirection are presented. The transition state model for the channel which produces H3SiSiH and SiH4 uses Arrhenius parameters of log kI3(T= 550 K) s-I = 15.74 - (52.0 kcal mol-')/2.3RT. These Arrhenius parameters are also consitent with APf,29s(H3SiSiH) = 74.06 kcal mol-' and estimates for the reverse insertion reaction (R-13). The transition state model for the channel which produces SiH, and Si2H6uses Arrhenius parameters of log k,(T = 550 K) s-l = 15.94 - (53.8 kcal mol-')/2.3RT. It is consistent with Wf,298(SiH2)= 64.80 kcal mol-' also and experimental measurements for the reverse insertion reaction (R-7). The rate constants for the chemically activated reactions involving trisilane (R15 and R16) are estimated and shown to be of lea importance than the corresponding reactions for disilane due to the greater number of internal degrees of freedom in trisilane. Acknowledgment. This work was supported in part by NSF DMR 87 04355 and by the Minnesota Supercomputer Institute.
'9
IYC
70-* io-'
.(A) 4.828 5.563
io"
id
io'
id
10'
Pressure (Torr) Figure 6. Relative rate of stabilization (SiH2+ Si2H63 Si3H8)versus the chemical activation reaction (SiH2+ SizHdi- Si3H8*19 SiH, + H3SiSiH)as a function of the bath gas pressure at three different 1200 K;- - -, 700 K, ., 300 K, 0,stabilization temperatures: -, reaction; a, chemically activated channel reaction. A silane bath gas with flcsilpnc(300 K) = 0.5 is assumed. The rate constant divided by k7'(T) is actually plotted (k7"(300 K) = 4.8 X lo", k7"(700 K) = 2.3 X lo", k-,*(l2OO K) = 4.1 X lOI4 cm3 mol-'9-I).
Appendix: Thermochemistry Molecular parameters used in the calculation on SiH4, SiH2, H2, and SizHdhave been previously provided in refs 27 and 15. Vibrational frequency assignments for H3SiSiH were taken from Ho et al." ( u (cm-I) = 380,401,433,724,887,926,945,2005, 2126,2127,2147). Also, one internal free rotation was used with Zrd = 2.73 X lo4 g cm-2 and nsp = 3. Table I1 contains the Lennard-Jones parameters used in the RRKM calculations. Registry No. Si2H6,1590-87-0; H3SiSiH, 50420-90-1.
References end Notes
for any given value of Ec than are k,I(Ec) and k:(E+). The result is a more uniform distribution of decomposition products from an energized trisilane molecule than from an energized disilane molecule. However, the overall rates of both k , 7 ( P ) and k,I3(E+) are lower than k,'(E+)and ka3(E+),because trisilane has more internal degrees of freedom than disilane does. This greatly reduces the rate of the chemically activated reaction channels compared to the stabilization channels, especially at temperatures lower than 700 K. However, Figures 5 and 6 do demonstrate that trisilane dissociation is not always in its high pressure limit under at least some CVD conditions. In particular, for temperatures greater than 1000 K, fall-off effects and chemically activated reaction channels, i.e., R15 and R16, should be taken into account in any detailed reaction mechanism.
(1) Moffat, H. K.; Jensen, K. F.; Carr, R. W. J . Phys. Chem., preceding paper in this issue. (2) Olbrich, G.; Potzinger, P.; Reiman, E.; Walsh, R. Organometallics 1984,3, 1267-1272. (3) Beccrra, R.; Walsh, R. J. Phys. Chem. 1987, 91, 5765-5770. (4) Gordon, M. S.; Truong, T. N.; Bonderson, E. K. J. Am. Chem. SOC. 1986,108, 1421-1427. (5) Dzarnoski, J.; Rickborn, S.F.; ONeal, H. E.; Ring, M. A. Orgunometallics 1982, I 1217-1 220. (6) Martin, J. G.; O'Neal, H. E.;Ring, M. A. Inr. J . Chem. Kiner. 1990, 22,613-632. (7) Reiman, B.; Matten, A.; Laupert, R.; Potzinger, P. Ber. Bunsen-Ges. Phys. Chem. 1977,81, 500-504. (8) Agrawal, P. M.; Thompson, D. L.; Raff, L. M. J . Chem. Phys. 1990, 92, 1069-1082. (9) Schranz, H. W.; Raff, L. M.; Thompson, D. L. J. Chem. Phys. 1991, 94, 42 19-4229. (10) Walsh, R. Acc. Chem. Res. 1981, 14, 246-252. (1 1) Ho, P.; Coltrin, M. E.;Binkley, J. S.;Melius, C. F. J . Phys. Chem. 1986,90, 3399-3406. (12) Ho, P.; Melius, C. F. J. Phys. Chem. 1990, 94, 5120-5127. (13) Martin, J. G.; Ring, M. A.; O'Neal, H. E . Inr. J. Chem. Kiner. 1987, 19, 715-724. (14) Vandenuielen, A. J.; Ring, M. A.; ONeal, H. E.J. Am. Chem. Soc. 1975, 97,993-998. (15) Moffat, H. K.; Jensen, K.F.; Carr, R. W. J. Phys. Chem. 1991,95,
Summary Arrhenius parameters for the 1,l H2 disilane dissociation reaction and trisilane dissociation reactions are evaluated. The 1,l H2 disilane dissociation reaction is estimated (from method 2) to have Arrhenius parameters of log k3(T = 550 K)s-l = 15.81 - (56.31 kcal mol-I)/2.3RT. These Arrhenius parameters are = 74.06 kcal mol-' and esticonsistent with AWf,298(H3SiSiH) mates for the reverse insertion reaction (R-3) based on analogous silylene insertion reactions. The chemically activated reactions (R4 and R14) involving the 1,l H2 dissociation channel are ex-
(16) Robinson, P. J.; Holbrook, K. A. Unimolecular Reactions; John Wiley: New York, 1972. (17) Ho, P.; Coltrin, M. E.;Binkley, J. S.;Melius, C. F.J. Phys. Chem. 1985,89, 4647-4654. (18) Gordon, M. S.;Gano, D. R.;Binkley, J. S.;Frish, M. J. J. Am. Chem. SOC.1986, 108, 2191-2195. (19) Neudorfl, P. S.; Strausz, 0. P. J . Phys. Chem. 1978, 82, 241-242. (20) Jasinski, J.; Chu, J. 0. J . Chem. Phys. 1988, 88, 1678-1687. (21) Bowrey, M.; Purnell, J. H. Proc. R. SOC.London, A 1971, 321, 341-359. (22) Inoue, G.; Suzuki, M. Chem. Phys. Leu. 1985, 122, 361-364. (23) Ncwman, C. G.; ONeal, H. E.;Ring, M. A.; Lcska, F.; Shiplcy, N. I n t . J . Phys. Chem. 1979, 1 1 , 1167-1182.
-.
I
__
145-154. . .
J. Phys. Chem. 1992,96,7703-7708 (24) Erwin, J. W.; Ring, M. A.; ONeal, H.E. In?. J. Chem. Kine!. 1985, 17, 1067-1083. (25) Coltrin, M. E.; Kee, R. J.; Evans, G. H.J . Electrochem. SOC.1989, 136, 819-829. (26) Purnell, J. H.; Walsh, R. Proc. R. SOC.London 1966,293,478-488.
7703
(27) JANAF Thermochemical Tables 1978 Supplement J . Phys. Chem. Ref. Data 1978, 7, 793. (28) Walsh, R. Thermochemistry and Reactivity of Silylenes. In Silicon Chemistry; Corey, E. R., Corey, J. Y., Gaspar, P. P., Eds.; Ellis Horwood Limited: Chichester, UK, 1988.
Low-Energy Electron-Induced Chemistry of Ethylene on Clean and CI- and D-Covered Ag(111) X.-L. Zhou and J. M. White* Department of Chemistry, University of Texas, Austin, Texas 78712 (Received: January 18, 1992)
The chemistry, induced by 50-eV electrons, of ethylene (C2H4) on clean and C1- and D-covered Ag( 1 1 1) has been studied. In the absence of electron irradiation, C2H4 is weakly *-bonded and desorbs, with no thermal decomposition, below 170 K. Consistent with *-donation and *-polarization, one monolayer of ethylene lowers the work function by 0.38 eV. Evidence is presented that adsorbed C2H4, exposed to low doses of electrons, is selectively decomposed to adsorbed H and vinyl (C2H3). The latter leads exclusively to 1,3-butadiene(C4H6)in subsequent temperature-programmeddesorption (TPD). Besides C4H6,higher doses of electrons lead to acetylene (C2H2),precursors to butene and tiny amounts of ethane, but no C, or C3products. In the presence of C1, the parent ethylene is stabilized and the amount of C4H6produced in TPD is enhanced. For ethylene and coadsorbed D, there is, in the absence of electron irradiation, no reaction. With electron irradiation, isotopically labeled dihydrogen and partially deuterated ethylene and ethane form, but there are no C,, C3,or C4 products. We propose that bond-specific decomposition, e.g., C-H bond cleavage, to form vinyl occurs when the weakly bound ethylene is ionized by low-energy electrons. That further C-H bond cleavage occurs with much lower cross section and that there is no C-C bond cleavage may result from relatively effective quenching of excited ionic states of the strongly chemisorbed primary product, C2H3.
Introduction As a part of our continuing investigation of photon’ and electron2+ driven processes at adsorbatemetal interfaces, we report in this paper on the chemistry, induced by 50-eV electrons, of ethylene adsorbed on clean and C1- and D-covered Ag( 11 1) at 100 K. While intrinsically interesting, nonthermal means of activating adsorbed species also offer opportunities to prepare and study intermediates, thought to be catalytically important. For example, heterogeneous hydrocarbon catalysis is a complex multiphase kinetics problem; many intuitively attractive surface intermediates have been proposed, but few of these hydrocarbon fragments have been cleanly synthesized and characterized structurally and kinetically. One elegant approach to adsorbed CH3 fragments involves the dissociation of CH4 during energetic collisions with Ni( 111),5 but generalizing this to species containing more than one carbon atom appears difficult. A second approach, of greater generality but less attractive because a coadsorbate is unavoidable, is the thermal dissociation of alkyl iodides: including vinyl iodide.’ A third approach involves the photodissociation of alkyl halides.* A fourth approach, and the one used here, relies on controlled fluxes of low-energy electrons to drive bond-specific nonthermal adsorbate chemistry. In this paper we provide firm evidence for the production of an important C2intermediate, vinyl (-CH=CHJ? and H by controlled low doses of 50-eV electrons onto r-bonded C2H4 bound to Ag(ll1) at 100 K. H2and residual parent molecules are desorbed by annealing to 200 K,leaving only the vinyl fragments. Ag is an important and interesting catalytic metal. For example, the reaction of ethylene and oxygen to form ethylene oxide on Ag catalysts is well-known.’OJ1While adsorption of CzH4on and C1- and Ooovered Ag(1 10)lohas been studied, little attention has been paid to Ag( 11l),l4 Our earlier studies of indicate that Ag(ll1) is hydrocarbon fragments (C,H,)3*4J5-17 relatively unique among transition-metal surfaces; C-C and C-H bond formation from surface C a mfragments and H atoms have much lower activation energies than C-C and C-H bond dissociation. As one interesting example, the dissociation of benzene
on Ag(l1 l), induced by electrons (150 eV)? leads to surface phenyl and hydrogen, and during sukquent TPD these recombine to form biphenyl and dihydrogen. The results of this paper demonstrate that bond-specific electron-induced dissociation may be a general phenomena that deserves considerable study. Experimental Section We used a UHV chamberi8 equipped with a double-pass cylindrical mirror electron energy analyzer with a coaxial electron gun for Auger electron spectroscopy (AES), a differentially pumped He discharge lamp producing vacuum ultraviolet light (used here for work function change (A@) measurements), a quadrupole mass spectrometer (QMS)for temperature-programmed desorption (TPD), and an ion gun for surface cleaning. The chamber was ion-pumped and had an auxiliary titanium sublimation pump and a 170 L/s turbomolecular pump. The base pressure was 5 X Torr. The clean,19as verified by AES, Ag( 111) crystal (-0.8 cmz surface area) was cooled to 100 K with liquid nitrogen. The temperature, ramped at 2.5 K/s for TPD, was measured with a chromelalumel thermocouple spot-welded to a Ta loop that was pressed into a hole drilled in the edge of the crystal. A@ was determined from the low kinetic energy threshold of secondary emission of the He(1) UPS spectra. Ethylene (99.575, Linde) was dosed, with the Ag( 111) surface at 100 K,through a multichannel array doser positioned about 7 mm from the sample. To prepare for dosing, the crystal was turned away from the doser and the C2H4pressure was increased, Torr at the vacuum using a leak valve, to give AP = 5 X system ion gauge. To initiate the dose, the substrate was then turned to face the doser. While very reproducible, the exposures in langmuirs are not known. Surface C1 was prepared by electron-induced dissociation of monolayer CH3Cl followed by flashing the surface to 600 K to desorb undecomposed CH$l and C, fragments. All the CI fragments desorb as C,H, hydrocarbons, leaving only C1 on the surface.17 Surface D was prepared by exposing the surface at 100
0022-3654/92/2096-7703%03.00/00 1992 American Chemical Society
