J. Phys. Chem. 1994,98, 1719-1731
Mechanisms of Halogen Chemisorption upon a Semiconductor Surface: Chemisorption upon the Si( 100) (2x1) Surface
1719
12,
Br2, Cl2, and CsHsCl
Harris C. Flaum, Daniel J. D. Sullivan, and Andrew C. Kummel’ Department of Chemistry, University of California, San Diego, La Jolla, California 92093 Received: October 18, 1993; In Final Form: November 29, 1993’
Chemisorption probabilities ( S ) of monoenergetic 12, Br2, Cl2, and C6HsC1 beams have been measured on the Si(100) (2X 1) surface. The sticking probabilities ( S ) were measured as a function of the incident translational energy (Ei), the surface temperature (Ts), the angle between the incident beam and the surface normal (ei), and the surface adsorbate coverage (e). All three diatomic halogens can adsorb via both precursor-mediated and direct-activated chemisorption, while CaHsCl adsorbs only by precursor-mediated chemisorption. For the diatomic halogens, precursor-mediated chemisorption is most significant for 12, followed by Br2, and Cl2. The barriers to direct-activated chemisorption are seen to increase from -0.055 eV for Cl2, to -0.12 eV for Br2, to -0.14 eV for 12. For the diatomic halogens, the largest initial sticking probabilities are obtained a t the highest incident energies, with SOequal to 95-100% for I2 and Br2 and 80435% for Cl2. The initial sticking probability, SO,is independent of the incident angle of the molecular beam for all incident energies for Cl2 and for C6HsC1. For all incident kinetic energies, the chemisorption probability decreases linearly with coverage (S = SO(1 - e))for the three diatomic halogen gases. This suggests that a single site is required to initiate the chemisorption process.
I. Introduction In this paper, we will elucidate the basic mechanisms of chemisorption for three diatomic halogen species and for monochlorobenzene (MCB) upon the Si(100) (2x1) surface. Diatomic halogens are among the most common gas-phase species used in semiconductor dry etch processing. Understanding and quantifying the chemisorption mechanisms for diatomic halogen chemisorption may be of great use especially in regard to choosing a particular halogen and or set of operating conditions (i.e., surface temperature, gas pressure, etc.) for a specific etch process. Comparing the relative contributions from precursor-mediated and direct-activated chemisorption among the three diatomic halogens studied here is of interest from a fundamental viewpoint as well as from a practical viewpoint. Contrasting the chemisorption mechanisms of the diatomic halogens to a larger halogencontaining molecule such as MCB will also help clarify the dynamical process of halogen chemisorption upon the Si( 100) (2x1) surface. In general, the two basic mechanisms for chemisorption of a gas-phase species upon a surface (semiconductor or otherwise) are precursor-mediatedchemisorptionand direct-activated chemisorption. “Precursor Chemisorption”denotes a process in which an incident molecule first becomes trapped in a molecular physisorption well and then chemisorbs (often dissociatively) by migrating to defects, utilizing the thermal energy of the surface, or by repartitioningmolecular energy. For “direct chemisorption”, molecules instantaneously chemisorb upon hitting the surface. The dangling bond character of semiconductor surfaces may allow direct chemisorption of the diatomic halogen species without overcoming a barrier. “Direct activated chemisorption” refers to a chemisorption process in which an incident molecule requires a minimum incident energy in order to surmount a barrier prior to reaching a stable chemisorption well. The Si(100) (2x1) surface has the top layer of atoms arranged in dimer pairs with all the atoms in equivalent sites within the unit cell,’ in contrast to the Si( 1 11) (7x7) surface with 13 distinct surface atom sites2 However, due to defects, the sites on the Si( 100) (2X 1) surface may have different dangling bond order, and thus it is reasonable e
Abstract published in Advance ACS Abstracts, January 15, 1994.
0022-3654/94/2098-17 19$04.50/0
to expect a variety of barrier heights to chemisorption to be present on this surface. The chemisorption mechanism(s) can be determined by measuring the chemisorptionprobability versus the incident kinetic energy of the gas molecule^.^-^ For precursor-mediated chemisorption, the sticking probability decreases with increasing molecular beam energy because high-energy molecules cannot dissipate enough of their kinetic energy into either the surface lattice or rotational/vibrational excitation to fall into the physisorption well. In addition, for precursor-mediated chemisorption, the zero coverage (“initial”) sticking probability normally decreases with increasing surface temperature since the barrier to desorption from the physisorption well is usually greater than the barrier to chemisorption from the physisorption well.).6 For direct chemisorption, molecules chemisorb instantaneously upon hitting the surface. If this process is barrierless, the initial sticking coefficient will be independent of the translational energy of the molecular beam (0.02-4.0 eV). In general, for direct-activated chemisorption, the initial stickingcoefficient increases with kinetic energy until it approaches a constant value. For most direct chemisorption processes, the initial sticking probability is independent of the surface t e m p e r a t ~ r e . ~ ~ Although chemisorption and physisorption both result in dissociation of the diatomic halogen species, an intermediate metastable molecular chemisorption state may precede dissociation. For example, J. Grimbolt, A. C. Luntz, and D. E. Fowlers and C. T. Rettner and C. B. Mullins9 show that both precursormediated and direct-activated chemisorption occur through an intermediate 0 2 - chemisorption state for the chemisorption of 0 2 on Pt( 1 1 1). Therefore, in direct-activated chemisorption, the barrier being probed may reside between the gas-phase and the molecular chemisorption state or between the gas-phase and the dissociative chemisorption state. The presence of molecularly adsorbed states is suggested for HCl and HBr on Si( 11 1) by M. Miyamura, Y.Sakisaka, M. Nishijima, and M. Onchi.lo The presence of molecularly chemisorbed states for these species on the Si( 11 1) surface would indicate a strong possibility for the existence of molecularly chemisorbed states for Clz, Br2, and 12 ontheSi(100) (2Xl)surface. C.C.Cheng,Q.Gao, W. J.Choyke, and J. T. Yates, Jr.11 show that at least two states are accessed for C12 adsorbed on Si( loo), and Miyamura et al. show that HCI 0 1994 American Chemical Society
1720 The Journal of Physical Chemistry, Vol. 98, No. 6,1994
Flaum et al.
also accesses at least two states on the Si(ll1) surface. The two states observed by Cheng et al. are both dissociative with C1 in either a bridging site, bonding with two silicon surface atoms, or a tilted atop site, bonding with a single silicon atom. Miyamura et al. suggest that one of the two states accessed by HCl is molecular and the other is dissociative. The interaction of chlorine with silicon is important in the etching procedures used in device fabrication.I2-I5 The literature contains several studies on the interaction of molecular chlorine with silicon ~urfaces.12-1~J61* These studies have focused on the products of reaction,17J9-21 the nature of the bonding on the ~urface,*2+~~ the reaction sites,19-29 and reactions at various coverages (9cl).17J9 In these references, all of the surfaces have already undergone reaction and have chlorine present on the surface (&I > 0) when the experiments are performed. Conversely, the experiments described in this paper probe the reactions on unreacted surfaces with a well-characterized structure with zero halogen coverage (9= 0) and with known coverages. More generally, halogen chemisorption and etching (including fluorine, bromine, and iodine) of semiconductor surfaces have been studied by several other groups. S. M. Mohapatra, N. Sahoo, K. C. Mishra, B. N. Dev, W. M. Gibson, and T. P. Das30.31 have performed self-consistent Hartree-Fock cluster calculations for halogens adsorbed upon Si( 1 11) surfaces generating results for bond lengths, UPS and XPS spectra, and vibrational frequencies and amplitudes. V. Barone, F. Lelj, N. Russo, and M. T o ~ c a n o ~ ~ performed MNDO calculations for all the halogens upon the Si(ll1) and Si(100) surfaces with results for bond lengths, chemisorption energies, and charge exchange between halogen Figure 1. (A) A side-view schematic of experimental apparatus. The and surface. Z. H. Walker and E. A. Ogryzl03~73~ and E. A. apparatus consists of two major parts, the molecular beam source and Ogryzlo, D. L. Flamm, D. E. Ibbotson, and J. A. Muchajs have the ultrahighvacuum experimentalchamber. The molecular beam, quartz performed several kinetics studies upon halogen (both atomic flag, sample, and the quadrupole mass spectrometer are collinear. The quartz flag and the sample may be moved from the path of the molecular and diatomic) etching of both silicon and gallium arsenide; they beam to allow the quadrupole mass spectrometer to sample the direct reported a half-order reaction rate for silicon and inferred from (B) An internal view of the molecular beam source. The molecular beam. this a two-step etching process consisting of adsorption followed beam originates from the pulsed valve and travels through the skimmer, by dissociation and desorption of etch products. R. B. Jackman, a slit on the chopper wheel, the gate valve, and finally through the R. J. Price, and J. S. F ~ o r studied d ~ ~ the thermal desorption of collimator to enter the experimental chamber. bromine-saturated Si( 100) surfaces and identified two chemisorbed surface states depending upon the surface coverage. G. products and the second being the direct process of induced C. Tyrrell, I. W. Boyd, and R. B. Jackman37 studied the affect desorption caused by impact of the hyperthermal particles from of ion bombardment upon thesesurface sites: namely, they report the beam. a chemical interconversion between these states as a result of ion We have measured the chemisorption probability (both initial bombardment. and as a function of surface coverage) of molecular chlorine, iodine, bromine, and monochlorobenzene on the 300 K Si( 100) The groups of F. X.Campos, G. C. Weaver, C. J. Waltman, (2x1) surface at normal incidence as a function of molecular and S. R. Leone38 and A. Szabo, P. D. Farrall, and T. EngeP9 beam energy. In addition, we have determined the initial recently reported the effects of increased incident energy on the chemisorption probability as a function of the surface temperature etching of Si( 100) with atomic and molecular chlorine. In both and the angle of incidence. of these studies, the etch rates are determined by monitoring the etch products with mass spectroscopy. The molecular beams in 11. Experimental Section both of these studies produceatomic as well as molecular chlorine and thus may influence the etch rates and etch products. Campos The experiments are performed in the vacuum chamber shown et al. use a molecular beam produced by laser desorption of schematically in Figure 1. The machine consists of two major cryogenic C12 films. This source provides a beam with a wide parts, a molecular beam source and an ultrahigh Vacuum (UHV) distribution of kinetic energies. They report a factor of 10increase experimental chamber. The apparatus is described in detail in in the etch rate of Si(100) surfaces (at 317 K) when the mean a previous paper.40 incident energy is increased from that of a thermal beam (-0.025 A. Molecular Beam Source. The molecular beam source eV) to 0.4 eV. Campos et al. observe that the etch rate increases (Thermionics) is divided into three independently pumped with increasing incident energy when a substantial fraction of chambers (see Figure 1B): chamber 1 houses the pulsed valve the incident beam flux has kinetic energies above 3 eV, with etch and skimmer; chamber 2 contain the chopper wheel, chopper rates 30 times that observed for thermal beams. Campos et al. motor, and gate value; and chamber 3 holds the collimator. suggest that the increased etch rate is due to the kinetic energy Chamber 1 is pumped by a water-baffled 16-in. diffusion pump of the incident molecular chlorine being used to break the silicon(Varian VHS-10, pumping speed 2120 L/s). Chambers 2 and silicon bonds on the surface. 3 are each pumped by a water-baffled 6-in. diffusion pump (Edwards 100F, pumping speed 322 L/s). Typical pressures for In the study reported by Szabo et al., hyperthermal beams of Cl/C12 are shown to have much higher etch rates on Si(100) chambers 1, 2, and 3 are lo-’, 10-8, and legTorr, respectively. surfaces at low temperatures (180K) I than thermal beams. Szabo The beam is composed of Clz, Br2,12, or C6HsCl seeded in He, Ne, Ar, Kr, or Xe expanded through a pulsed valve nozzle (General et al. suggest that two pathways for desorption of the etch products Valve) with a 2.0-mm orifice. Cl2 is obtained from Matheson are active, one pathway being the thermal desorption of volatile
Halogen Chemisorption upon a Semiconductor (99.9995% purity), while Br2 (99.99%), 12 (99.99%), and C6H5C1(99.9%) are obtained from Aldrich. Chlorine is used from a standard gas cylinder and mixed with the appropriate inert gas in an intermediate cylinder reservoir. Bromine, iodine, and MCB are placed in a 1-L Teflon-lined stainless steel cylinder. The air is removed by repeated freeze-thaw pump-out cycles with a dry ice-methanol bath. The appropriate inert gas is then added to this reservoir a t the desired regulated pressure. The translational energy of the diatomic halogen or MCB molecules in the beam is varied between 0.038 and 2.0 eV (the energy range depending upon the particular species used) by expanding mixtures of the diatomic halogen or MCB and He, Ne, Ar, Kr, or Xe. The monoenergetic beam passes through a skimmer (Beam Dynamics), an open slit of the chopper wheel, and a collimator. The chopped beam pulses can be measured by a quadrupole mass spectrometer and are averaged/displayed on a digital oscilloscope (LeCroy, 9400A). Thevelocity of the molecules is determined by measuring the flight time from the chopper wheel to the quadrupole. From the rotational spectra and fast time-of-flight measurements (time response < 0.1 ps) using resonantly enhanced multiphoton ionization detection (REMPI), our N2 and CO molecular beams are known to be nearly monoenergetic, Avlv < 1150.41 For the diatomic halogens and the MCB, the uncertainty in flight times ( f 1 0 ps) results from the slow response of the quadrupole mass spectrometer and causes larger uncertainties of the reported translational energies with increasing molecular velocity. Auger spectroscopy is used to determine the size of the molecular beam at the sample. The molecular beam is determined to be approximately 2-3 mm in diameter when it strikes the sample. The sample itself is 7 X 10 mm. The beam size can easily be changed through the use of different sized collimators. The choice of collimator for these experiments provides a maximum in signal while avoiding any edge effects or requirements for highly precise alignment of the sample. In our experiments we employ pulsed molecular beams to measure sticking probabilities. The normal practice in the literature is the use of continuous or continuous chopped beams.3.5-' It is necessary to obtain a steady background of the diatomic halogenor MCB in the experimentalchamber to perform sticking measurements using the technique of King and Wells. Unless the pulsed valve operates at greater than 10 Hz, the gas capacitance of the chamber may be observed on the quadrupole as a decay in the partial pressure of the diatomic halogen or the MCB after every pulse. Data are acquired with the pulsed valve running between 15 and 60 Hz; this provides a quasi-continuous beam. All experiments are done by operating the pulsed valve with the quartz beam block in the beam path for 15 minutes, or until the quadrupole signal is steady (f5% over 1 min), to ensure a nearly constant pumping speed of the chamber walls. To ensure that the adsorption of background and scattered diatomic halogen or MCB is negligible for each surface, a beam of pure diatomic halogen is allowed to enter the chamber and strike the quartz beam block, located directly in front of the sample, for 30 min. The sample is then checked by Auger spectroscopy to ensure that no Cl2, Br2, 12, or C6HsCl is present on the surface. The rise in diatomic halogen partial pressure in the experimental chamber due to the molecular beam is approximately 1 X 10-12Torr (see Figure 2B), while MCB produces a rise of 1 X 10-l1 Torr. The very low partial pressure of the diatomic halogen or MCB used allows 1 h before 0.01 monolayer can adsorb, assuming a sticking probability of 100%. Methods of measuring effusive sources which would interfere with the sticking measurements are described in a previous paper.@ To check for corrosion in the pulsedvalve from the highly reactive halogens (which would affect the supersonic expansion), the velocity of the molecular beam is measured before and after every experiment. In addition, the pulsed valve is checked for leakage before and after each experiment. B. UHV Chamber. The UHV experimental chamber is equipped with a cylindrical mirror analyzer Auger (Perkin Elmer,
-
-
-
The Journal of Physical Chemistry, Vol. 98, No. 6,1994 1721
'1 1
QUARE? FLAG
A)
MOLECULAR BEAM
A)
B)
C)
p(c,2)
kAP .......
Y
.. .. .'.
!$+ c
I
.::
B 0
5
IO
10
I3
25
30
31
u)
Time (sec.)
Figure 2. Experimental procedure of King and Wells: (A) Schematic of the experimental procedure with the three steps labeled A, B, and C. In step A the partial pressure of (212, P(C12), is measured when the beam strikes an object upon which S = 0. In step B the P(C12) drops by AP when the beam is exposed to the clean sample surface. In step C the beam is turned off and the background P(C12) is measured. The pressure rise due to the molecular beam, P,is defined as the difference in P(C12) during steps A and C. The initial sticking probability,SO,is then defined as AP/P. (B) Raw data from an initial sticking probability experiment. The beam block is initially in the path of the molecular beam. The beam block is removed from the path of the beam at 7 s, allowing the beam tostrike thecleansurface. ThebeamMockisreinserredat 12s,withdrawn at 19 s, and reinserted again at 23 s. The beam is then turned off at 29 s. The positions labeled A, B, and C on the raw data correspond to the actions shown in the schematic of the experimental procedure (part A). The units of pressure are approximately 1X10-12 Torr.
10-1 5 9 , low-energy electron diffraction (LEED) (Princeton Research Instruments, RVL 8-120), a sputter gun, and a quadrupole mass spectrometer (QMS) (UTI, 100-C). The quadrupole mass spectrometer is directly opposite of the molecular beam source (see Figure 1). This geometry allows the direct beam to be measured by the quadrupole. The UHV experimental chamber is pumped by a liquid nitrogen trapped, water-baffled &in. diffusion pump (Edwards 160, effective pumping speed 500 L/s) and a liquid nitrogen trapped titanium sublimation pump (effective pumping 4000 L/s). The pumping speed for the diatomic halogens is much larger due to the adsorption of the diatomic halogens onto the stainless steel walls of the vacuum chamber. This high but stable pumping speed allowsus to measure very small rises in the diatomic halogen pressure, 1 X 10-12 Torr, with signal to noise levels of L8:l. The chamber consistently reaches pressures of 1 X 10-10 Torr after bakeout. The Si( 100) samples are mounted on a manipulator equipped with electron bombardment and photon heating and liquid nitrogen or water cooling. The manipulator allows x , y , and z motions, a flip motion out of the sample plain, and a rotation about the vertical axis. The manipulator is described in detail el~ewhere.~z C. Sample Preparation. The silicon (100) wafers are p-type, boron doped, with resistivities of 3.0-6.6 and 0.0084.020 0 cm,
-
-
Flaum et al.
1722 The Journal of Physical Chemistry, Vol. 98, No. 6, 1994
respectively. The wafers are cut into 10 X 7 mm sections and chemically cleaned before placement into vacuum. The following cleaning procedure is employed: etching in hot HzS04/H202 1:2 for 15 min followed by sonication in HF/H20 1:25 for 15 s and sonication in HPLC water for 15 s; this leaves the surface with a thin protective oxide layer.43 The sample is mounted on a tantalum plate and placed into a load-lock chamber to await transfer into the UHV experimental chamber. The sample is initially sputtered with 2 or 3 keV Ar+ with the surface at 500600 OC (773-873 K). The surface is then flashed to 850 OC (1 123 K) for 1 min to anneal the surface. After several sputter and anneal cycles, very sharp (2X 1) LEED patterns are observed, and the carbon and oxygen Auger signals are below detection limits (-0.1%). The sample must be sputtered, annealed, and characterized by LEED before each set of measurements, as well as periodically checked for carbon and oxygen contamination by Auger. D. Data Collection. The initial sticking probabilities (SO) are determined by the method of King and well^.^^,^^ The King and Wells measurements require the molecular beam, quartz beam block, and the sample to be collinear (see Figure 1A). During an experiment, the partial pressure of molecular chlorine, mass 70, is monitored by the quadrupole mass spectrometer and recorded with the data acquisition program, LabView 2, on a Macintosh IIci. As shown in Figure 2A, the diatomic halogen or MCB partial pressure is monitored in the experimental chamber: (A) when the monoenergetic beam is incident upon a quartz beam block (zero sticking probability), (B) when the monoenergetic beam is incident upon the clean sample, and (C) when the beam is off. The initial sticking probability is then calculated: SO= A P / P , where AP is the pressure drop between times "A" and "B" while P is the pressure drop between times "A" and "C". The data obtained in a typical initial sticking probability experiment are shown in Figure 2B, with A, B, and C corresponding to the descriptions given above. This method for determining SOis discussed by C. T. Rettner, L. A. DeLouise, and D. J. Auerbach3 as a reasonable measurement for an experiment in which the vacuum time constant (7)is negligible, the pumping speed is stable, and the observed partial pressure is entirely due to the direct beam. These conditions, although very close to realization, are not fully achieved in our experiments. Several tests are performed to eliminate systematic errors in the sticking measurements and are discussed in our previous paper.40 We also monitor the concentrations of clusters in the beam (for Clz and C6HsCl) and operate under conditions where the detected cluster concentrations are less than 0.05% of the Clz signal and less than 5% of the C&Cl signal. Although the C&Cl cluster concentrations were much higher than the C12 cluster concentrations, SOis found to be independent of the cluster concentration for C6H5Cl. It was not possible to monitor clusters for 12 or Brz since our quadrupole cannot detect species greater than 300 amu. However, in order to minimize cluster production, the nozzle backing pressure is kept as low as possible and the pulsed valve is operated at opening times for the Br2 and I2 beams known to minimize cluster production for Clz beams. Sources of error in the measurements that we are not able to completely eliminate include slight variations in the defect density due to nonreproducible surface preparation, noise in the quadrupole signal, and variability in the pumping speed of the chamber walls. Different samples cut from the same wafer always exhibit the same trends in SOversus Ei but with small (*5%) global shifts in SO.The data reported here are averages of at least three experiments with the standard deviations shown as vertical error bars, one standard deviation shown on each side of the data point. The C12 flux, F, is calculated from saturation measurements as 0.076 f 0.022 ML/s (ML = monolayer). The method of calculating the flux is discussed previously.46 Bromine and iodine molecular beam fluxes are the same within a factor of 2. Saturation experiments are run with fluxes within a factor of 2 of this value and initial sticking measurements with slightly lower
fluxes. MCB molecular beam fluxes are 10-20 times smaller for initial sticking probability measurements: this is possible due to the much reduced pumping speed for MCB which results in much better signal to noise. Saturation measurements for MCB are performed at fluxes 10-20 times higher; otherwise, the time to complete the saturation measurement would be inordinately long. Checks are also made to ensure that SOis not dependent upon the incident flux. The initial sticking probability is not affected, within experimental uncertainties, by changes in the molecular flux of up to a factor of 4. The incident flux may be increased up to -0.30 ML/s before it is impossible to determine the initial sticking probability due to the rapid saturation of the surface. Otherwise, as discussed in a previous paper,40we must extrapolate the data to zero coverage. Sticking measurements with incident fluxes below 0.02 ML/s (except for MCB) are very difficult due to low signal to noise. The data for elevated surface temperatures are obtained by heating the sample with radiative emission and monitoring the sample temperature with an infrared thermometer (Mikron M90H). The temperatures reported are determined by measuring the current passing through the heater filaments. The heater filament current is calibrated by use of thermocouples attached directly to the face of a sample. The values obtained by the infrared thermometer are slightly higher than the thermocouple values (+8%). When the temperature read by the infrared thermometer, monitoring 0.78-1.06 pm with an assumed emissivity of 0.7, is 600 OC (873 K), the thermocouple reads 807 K. The thermocouple values are considered more accurate because the emissivity for silicon changes with temperature and the viewport window may affect the infrared transmission. We have also performed experiments with molecular chlorine and GaAs(100) crystals. We have detected the onset of desorption of GaCl, which is reported in the literature to occur a t 650 K,47,48by thermocouple measurements at 653 K. This close agreement with the literature values re-enforces our belief in the accuracy of the thermocouple measurements. For C12, the highest surface temperature used in initial sticking measurements for all incident energies is 807 K (with the exception of 0.038-eV incident energy where an additional data point is taken at 900 K for the purpose of analyzing sticking versus surface temperature as discussed in the next section). For Brz and 12, the highest surface temperature used in initial sticking measurements is 900 K. The Clz high temperature sticking measurements are at a lower temperature than those of Br2 or 12, because the infrared thermometer was initially used for temperature determination with Clz prior to thermocouple calibration. Subsequent high surface temperature sticking measurements with the other diatomic halogens were conducted at 900 K (as determined by thermocouple). 111. Results
A. Initial Sticking versus Translational Energy. The dependence of the initial sticking probability on the incident energy can reveal information about the active mechanisms for chemisorption. Precursor-mediated chemisorption is expected to exhibit a decrease in SOwith increasing incident energy. This is due to the inability of the incident molecule to dissipate sufficient translational energy into either the lattice or internal energy of the molecule to allow it to trap into the physisorption well. In contrast, a direct nonactivated mechanism will exhibit no dependence of Sowith changing incident energy, and for a directactivated mechanism, SOwill increase rapidly when the incident energy of the molecule exceeds the barrier to reaction. The initial sticking probability (SO) measurements for the Si(100) surfaces are performed using several different samples cut from the same wafers. Plots of the initial sticking probabilities (SO) for C12, Brz, 12, and C6HsCl versus the incident energy (Ei), for T, = 300 and 807 K (for C12) or 900 K (for Br2 and 12) at normal incidence, are depicted in Figure 3. The data shown in Figure 3 are all taken from one sample.
The Journal of Physical Chemistry, Vol. 98, No. 6, 1994 1723
Halogen Chemisorption upon a Semiconductor
-ep
,
,
.
100
, . .a .... ,.... ,.
.
,.
100 r . . .
CIS on Si(l00) (2x1)
.,
,
,. . . . , . . . . . . . _ ,. . . , . .
. . . . . ,J I
200
300
500
400
600
700
900 1000
800
T,, Surface Temperature (K) 100
Brz on Si(100) (2x1)
60 E M
'g
40
2
L
0
0.1
0
0.2
By E,
3WK 900K
0
r
m-
......................................
I00
0.3
0.4
0.5
1,.
YP
0.6
0 200
E,. Incident Energy (eV)
, ,
-
on S i ( 1 0 0 ) (2x1)
,
0.087 CV
, ,, ,
300
, ,-,, ,
500
400
, ,,
,
, ,
600
,,
,, ,, ,, , ,, ,, ,
700
800
,
,
,]
900 1 0 0 0
"3
I P
0
I= on Si(100) (2x1) 0
0
300K WOK
2
1,.
200
,
,
,
,, , ,
300
,
400
,
-. SO0
, ,
, ., , ,
600
700
, ,
. ,, 800
,
,
,
:
j
900 1000
Ts, Surface Temperature (K)
i pa'\+ i '*\
.
100
Monofhlorobenzenc on S i ( 1 0 0 ) (2x1) T. 300 K
-&--I--,-
\
-.
,._..,.
. . . . . . . . . . . . . . . . . .
,
1 T', Surface Temperature (K)
Figure 3. Plot of the initial sticking probability (So) of normal incidence of (a) Clz, (b) Brz, (c) 12, and (d) C&C1 upon the Si(100) (2 X 1) surface as a function of incident energy, Ei,at 300 K and at some elevated surface temperature (807 K for Clz and 900 K for Br2 and 12). The
increasein chemisorption probabilitywith incident energy (most notable at high temperature) for C12, Brz, and 12 indicates the presence of directactivated chemisorption. The decrease in chemisorption probability at elevated surface temperature for Cl2, Brz, and 12 at the lowest incident energies indicates precursor-mediated chemisorption. In part d, the decrease in chemisorption probability with increasingincident energy is evidence of precursor-mediated chemisorption for C&Cl; no direct chemisorption is present. The data points are averages of at least three experiments with the standard deviations shown as vertical error bars. The horizontal error bars represent the uncertainty in the translational energies. For the purposes of comparison, in Figure 3a, we reproduce Clz data from a previous paper.46 The SOfor Clz is shown as a function of incident energy a t 300 and 900 K. At 300 K, the initial sticking probability (at normal incidence) of C12 is a strong function of the incident molecular beam energy. There is a decrease in SOfrom 58% at 0.038 eV to 42% at 0.045 eV and a sharp increase in SOfrom 42% a t 0.045 eV to 72% a t 0.16 eV. Above 0.16 eV, the initial sticking probability increases slowly with increasing incident energy. The significant drop in SOwith increasing incident energy from 0.038 to 0.045 eV strongly indicates the presence of a precursor-mediated chemisorption
Figure 4. Initial sticking probability,SO,for normal incidence of (a) Clz, (b) Brz, (c) 12, and (d) C&Cl upon the Si(100) (2x1) surface as a
function of surfacetemperature, Ts.With the exception Of C6H&1 (where a higher incident energy is also shown), data shown are for the lowest incident energy for each diatomic halogen. The data points are averages ofat least three experimentswith the standard deviations shown as vertical error bars. Fits to a standard precursor model are shown. Reprinted from J. Phys. Chem. 1993, 97, 12051-12060. channel. The sharp increase in So with increasing incident energy from 0.045 to 0.10 eV reveals the presence of a direct activated chemisorption channel. At 807 K, SOis significantly less at the lowest incident energies where precursor-mediated chemisorption is strongest and is approximately the same as SO(300 K) at the highest incident energies where direct activated chemisorption is the dominant mechanism. The vertical error bars represent the standard deviations of a t least three independent measurements of the sticking probability for each data point. The horizontal error bars represent the uncertainties in the measurements of the translational energies of the molecular beam. We note that the uncertainty in the translational energy increases a t higher energies due to the modest distance between the chopper and the quadrupole mass spectrometer. All measurements are made close to zero coverage (e/€), = 0-595). In Figure 3b, SO(normal incidence) at 300 and 900 K for Br2 is shown. Direct-activated chemisorption is evidenced by an
Flaum et al.
1124 The Journal of Physical Chemistry, Vol. 98,No. 6, 1994
a> 5.5
.....
I
.,
. . . . . . . . . . . . . . .
I
0.8
0.6
0.4
0.1
....... .'
3
c>
'
.
.
.
.
.
.
.
I
50
0
.
. . . .
.
0 150
100
0
0.2
Time (sec) . . . . . . . . . . . . . . . . .
0.6
0.4
(1
- @/e>*
0.8
1
I
0.8
0 6
0 .
0.2
0
IO
30
10
0
50
10
60
70
SO
Time (sec) I
,
0
.
,
20
.
,
40
.
,
60
.
,
EO
.
.
,
.
I ,
. . . . . . . . . . . . I, on Si(100)( 2 x 1 ) E, 0.107 C V
-
IM
Time (sec)
(1
-
.........
@/@>d
I
0 8
0.6
0 4
0.1
0 30
do
60
80
100
0
0.:
0.4
06
0.8
I
Time (sec)
Figure 5. Partial pressure versus time and the normalized sticking probability at normal incidence (S/So) versus (1 is shown for (a,b) Clz, (c,d) Br2, (e$) 12, and (g,h) CaHsCl at low incident energies. For the diatomic halogens, (S/So) is shown versus (1 - e/ec) and (1 - e/ec)*,with a linear decrease in (S/So) with coverage yielding the best fit. For MCB, (S/So) is shown with d = 1 , 3/2, and 2; d = 3/2 yields the best fit.
TABLE 1: Initial Sticking Probabilities of C&CI upon Si(100) (2x1) for One Incident Energy at Different Angles of Incidence C6HsCl incident energy (eV) angle of incident (deg) 0.95 00
33.2 (11.5)
31° tilt
32.3 (11.5)
increase in SOwith increasing incident energy at both 300 and 900 K. At 900 K, SOshows the greatest change, increasing from -38% at the lowest energy (0.087 eV) to -95% a t the highest energies (-0.55 eV). Precursor-mediated chemisorption is evidenced by the decrease in SOa t elevated temperature, especially at the lowest incident energies (less than 0.3 eV). At the lowest incident energies, SOat 300 K does not increase with decreasing incident energy as in the case of CI2. This is most likely due to
the fact that the lowest Brz incident energy is only 0.087 eV while the lowest Clz incident energy is 0.038 eV. In Figure 3c, SOat 300 and 900 K for 12 is shown. Directactivated chemisorption is evidenced by an increase in SOwith increasing incident energy a t 900 K. SOat 300 K exhibits little change with incident translational energy. This is result of strong precursor-mediated chemisorption at the lowest incident energies where the direct-activated component is diminishing with decreasing incident energy. Thus the two mechanisms are complementary, with the result being an almost constant value for SOover the range of incident energies studied at 300 K. In Figure 3d, SOis shown for MCB at T, = 300 K. Data for 900 K are not shown since SOis equal to zero at this temperature for all incident energies. The decrease in SO (300 K) with increasing incident energy is indicative of precursor-mediated chemisorption. In addition, the zero chemisorption probability at T, = 900 K is also characteristic of precursor-mediated
The Journal of Physical Chemistry, Vol. 98, No. 6,1994 1725
Halogen Chemisorption upon a Semiconductor
a>
4
r
,
. , . . . . ,
.
.
. . . , .
,
I
0.8
5 3.5 Gs
0.6
3
b
0.4
t
3
2.5
0.2
Q
0 50
150
100
200
Tiins (sec)
A.8 0..
-13
a
,--
Y
4.4
c)
. . .
L
Y =
I
d
2
E e
3.6 3.2
20
60
40
80
Time (sec)
0.85
L .
,
.,
1
0..
0.6
0.4
0.2
0
20
40
60
80
100
120
140
0
Time (sec)
0.8
0.6
0.4
0.2
0
0
0.:
0 4
0.6
08
1
Figure 6. Partial pressure versus time and the normalized sticking probability at normal incidence, @ / S O )versus , (1 - €)/e# for (a,b) (212,.(c,d) Brz, (e,f) 12, and (g,h) CsHsCl at high incident energies. For both the diatomic halogens and for CsHsC1, (S/So)is shown versus (1 -e/€),) and (1 - e/e,)z, with a linear decrease in @/SO)with coverage yielding the best fit for both.
chemisorption. It is interesting to note that even for the highest incident energies SO(900 K) = 0 f 2%; this shows that directactivated chemisorption is nonexistent for this molecule/surface system over the range of energies studied. Auger measurements following thermal desorption of the MCB at Ts 700 f 50 K show a silicon(92)/carbon(272) ratio typically greater than 50, supporting the notion that the MCB resides on the surface in either a weakly bound precursor well or a nondissociative chemisorption state, with no direct dissociative chemisorption occurring at the highest incident energies (which would break the C-Cl bond, resulting in a strong Si-C bond at the surface and thus a residual carbon peak in the Auger spectra). B. versus Surface Temperature. At the lowest incident energies, Sois measured as a function of T8in order to elucidate and compare the precursor kinetics for the diatomic halogens and MCB on the Si( 100) surface. The lowest incident energies are used since the precursor is largest here, and SOchanges the most with temperature. For these intermediate surface temperatures, the sample temperature cannot be measured with an
-
infrared thermometer, and therefore a calibration procedure is employed. This procedure is described in the Experimental Section of this paper. Following the calibration, the filament current is employed as a measure of the sample temperature. The accuracy of this technique is confirmed by using the infrared thermometer at surface temperatures above 600 OC and measuring the desorption temperatureof GaCl from GaAs( 100)at 650 K.47,48 In Figure 4a, reproduced from a previous paper,& the measurements for So increases from -44% to 54%. At 250 K, So increases to -72%. In the following section, we discuss the fitting of these data along with the other diatomic halogen data to a standard precursor model. In Figure 4b, analogous data are shown for Br2, except that the incident energy is equal to 0.087 eV and there is no data point for T*equal to 250 K. Here, SOincreases from -40% at 900 K to -55% at 300 K. Since these data are quite similar to the C12 data (except that it is taken at 0.087 eV), the precursor is a larger component of chemisorption at the lowest incident energies for Br2 than for C4.
-
1726 The Journal of Physical Chemistry, Vol. 98, No. 6, 1994
In Figure 4c, SOversus surface temperature is shown for I2 at 0.106-eV incident energy. SOincreases from -40% a t 900 K to -80% at 300 K. This increase shows that precursor-mediated chemisorption is more significant for I2 than for Br2 or C12, especially considering that the incident energy (0.106 eV) for the 12 measurements in Figure 4c is larger than the incident energies for Cl2 or Br2 in Figure 4a and 4b. The data in Figure 4a-c for the diatomic halogens clearly indicate the presence of precursor-mediated chemisorption, with a residual component of chemisorption at the highest surface temperatures mostly due to direct-activated chemisorption. Finally, in Figure 4d, SOversus T,is shown for MCB. Here, SOdecreases from -70% at 300 K to 0%a t 900 K. This is clearly the largest change in SOversus surface temperature that we have observed for any molecule upon the Si(100) (2x1) surface. This is mainly due to the absence of any direct component to chemisorption, and thus at the highest surface temperature (900 K, where the precursor does not contribute to chemisorption) there is no residual direct component to chemisorption as in the case of the diatomic halogens. C. S, versus Incident Angle. Initial sticking probabilities as a function of the angle between the molecular beam and the surface normal for MCB are shown in Table 1. We define the angle of incidence as the angle between the incoming molecular beam and the surface normal, so Oo is normal incidence between the sample surface and the incident molecular beam. In the geometry used in these experiments, the sample may be rotated about two independent axes. The sample is rotated "out of plane", about the x axis (vertical axis), for the 25' measurements and is "tilted", about the y axis (horizontal axis), for the 37' measurements. The angles are the maximum allowed by the sample and beam sizes. As discussed in a previous paper,4O~~~ the useof both out of plain and tilt motions for the angle measurements ensures that no spurious results are obtained due to the geometry of the experiments. The SOis found to be independent of the angle of incidence for MCB at 1-0-eV incident energy with a 25O tilt angle. For Cl2, SOis observed to be independent of the incident angle at all incident energies for both tilt and out of plane measurements (as discussed in a previous paper).& In a subsequent paper, we will report initial sticking measurements for 0 2 and F2 upon the Si( 100) (2X 1) surface where we observe normal energy scaling at higher incident energies. D. LEED and Auger Measurements. Auger measurements are performed on the clean and on the halogen- or MCB-saturated Si(100) (2x1) surfaces. Oxygen and carbon contamination is always less than 1% on the freshly sputtered and annealed Si(100) (2x1) surface. Prior to Auger measurements, the entire surface is saturated by translating the manipulator through the limits of the sample dimensions. A molecular beam integrated flux of 2-5 times the normal integrated flux required for a saturation measurement is used to guarantee a completely saturated surface. On the Cl2 saturated surface, the Si(92 eV)/ Cl( 18 1 eV) ratio (uncorrected for elemental sensitivity) is 2.0. On the Brrsaturated surface, the Si(92 eV)/Br(55 eV) ratio is -5.8. For 12, the Si(92 eV)/I(5ll eV) ratio is -6.5. On the MCB-saturated surface, the Si(92 eV)/C(271 eV) ratio is 13, and the Si(92 eV)/C1(181 eV) ratio is 5.0. LEED patterns of (2X 1) are observed for all the diatomic halogen-saturated surfaces (without annealing). A (2x1) LEED pattern is also observed when the Si(100) surface is saturated with MCB at 0.15-eV incident energy. For the diatomic halogens, this may indicate either that (2x1) ordered overlayers are formed with no silicon dimer breaking occurring or that disordered regions of (1x1) order are formed with dimer breaking, with a (2X 1) pattern always observed. Cheng et al." report ESDIAD (electron stimulated desorption ion angular distributions) experiments which support the dimer breaking picture. For the diatomic halogens, LEED patterns are observed showing a weak (1 X 1) order following a saturation and short anneal (keeping the temperature below that for desorption). This may be due to partial etching either from
-
Flaum et al. limited desorption (which may occur when close to the desorption temperature) or from an ordering of the hypothetical (1 X 1) broken dimer regions. Unfortunately, a t present, we cannot distinguish among these two possibilities. E. Sticking versus Coverage. In the previous two sections, we have discussed the effects of incident translational energy and surface temperature upon the initial sticking probability, SO, Le., the chemisorption probability for a clean well-ordered surface. We have also performed room temperature saturation sticking measurements for all the diatomic halogens and for MCB. During a saturation measurement, the quartz flag is removed and the clean sample is exposed to the molecular beam until the sticking probability, S , has become zero to within the signal to noise of the technique. In Figure 5a,c,e,g, the partial pressure versus time is shown for (212, Br2, 12, and C6H&l, respectively, at low incident energies. Likewise, we have plotted saturation data at higher incident energies in Figure 6a,c,e,g for C12, Br2, 12, and CbHsCl, respectively. By analyzing the data for partial pressure versus time, we are able to generate plots of S/So (sticking normalized by initial sticking) versus (1 - €)/e,)", where 0 is the surface coverage, 8, is the critical surface coverage where S = 0, and d is an exponent describing thesaturation behavior. In what follows, we describe the procedure used to generate these plots. The sticking at any time, S ( t ) , is equal to AP(t)/Pt,,l (where AP(t) is the instantaneous partial pressure drop when the sample is exposed to the molecular beam and P,,, is the partial pressure rise due to the molecular beam). The data may then be replotted as S(t) versus time. The surface coverage at any time, e(t),may be obtained by integrating thedata for S ( t ) ,thesticking probability versus time:
8 = l F S ( t )dt
(1)
We do not need to know the flux F and arbitrarily set it equal to 1 when integrating the data for S ( t ) ,as it will later cancel out in the equation for e/ec. The surface coverage, e(t),may then normalized by dividing by the coverage 8, at which S has become zero:
e, = c F S ( r ) dt
(2)
Again, we set Fequal to 1 when calculating 8,. If e(t)is divided by e,, we obtain
Thus the actual value for F is not important, as it cancels out in the normalization process. S(t) may now be plotted versus e/€&. If 0/e, is now subtracted from 1 and raised to some power d, the plot becomes S { t ) versus (1 - e/e,)d. Finally, if S(t) is now divided by SO,the plot becomes S(t)/So (normalized sticking) versus (1 - e/e,)d.In Figure Sb,d,f, we have plotted S/Soversus (1 - @/e,) and (1 - €l/e,)2 for C12, Brz, and 12,respectively, at lower incident energies. In Figure Sh, S/So versus (1 - e/e,), (1 - 8 / 8 , ) 3 / 2 , and (1 - e/e,)2are plotted for C6H5Cla t a lower incident energy. In Figure 6b,d,f,h, we have plotted S/So versus (1 - e/e,) and (1 - e/eJ2for C12, Br2, 12, and C6H5Cl, respectively, a t higher incident energies. We find that for the diatomic halogens, at both low and high incident energies, the sticking probability decreases linearly with coverage; i.e., S / & = (1 - e/e,). For C&Cl, S/So decreases linearly with coverage at higher incident energy but decreases as (1 - 8 / 8 , ) 3 / 2 at the lower incident energy. In Table 2, we show the regression coefficients for the fits to (1 with d equal to 1/2, 1, and 2 (with the exception of MCB at low incident energy where we show d = 1,3/2, and 2). This table shows that, to within the accuracy of these saturation measurements, the
The Journal of Physical Chemistry, Vol. 98, No. 6, 1994 1727
Halogen Chemisorption upon a Semiconductor
TABLE 2 Regression Coefficients for Linear Fits to S/& versus (l-e/e,)d Ei (ev) d = 112 d=l d-2 c12
Br2 I2
0.038 0.56 0.086 1.5 0.107 2.0
C6H&l CsHsCl
0.93645 0.9374 0.9478 0.971 1 0.9326 0.9309
0.9146 0.9 127 0.9425 0.9677 0.9129 0.8816
0.05
2.0
0.8926 0.8875 0.8744 0.8908 0.8815 0.9106
d=l
d = 312
d=2
0.9904
0.9926
0.973 1
d = 112
d=l
d=2
0.9589
09765
0.9086
TABLE 3 Incident Energies for Specific Values of the Ratio &(807 K)/&(maximum) for Cl2 and &(WO K)/ &(maximum) for Br2 and 12' SoISdmax) Ch (eV) Br2 (ev) 12 (ev) 40% 50% 60% 70% a
0.030 f 0.01 0.055 f 0.01 0.083 f 0.01 0.106 f 0.01
0.083 f 0.02 0.120 0.02 0.165 0.02 0.210 f 0.02
0.105 f 0.02 0.144 0.02 0.185 f 0.02 0.227 f 0.02
&(maximum) is the maximum value of SO.
sticking always decreases linearly with coverage (with the exception of MCB at low incident energy). Although we do not show the data here, it is found that a high incident energy, S/So (1 - e/€),)for Cl2 upon the Si( 111) (7x7) surface, while at low incident energy, we find d = 1 or 1/2 give the best fits. At low incident energy, we find regression coefficients of 0.9486,0.9352, and 0.8565 for d = 1/2, 1, and 2, respectively. At high incident energy, we find regression coefficients of 0.9304,0.97056, and 0.9328 for d = 1/2, 1, and 2, respectively. Thus, a t high incident energy, the sticking clearly decreases linearly with coverage, while a t low incident energy, the sticking decreases either as (1 - e/e,) or as (1 - 8/0,)1/2.
-
IV. Discussion
A. Estimate of Activation Barriers for Direct Chemisorption. By inspecting Figure 4a-q one can see that SOapproaches a constant value with increasing surface temperature for the diatomic halogens on the Si( 100) surface. This trend allows us to make the assumption that the precursor contribution to sticking at the highest surface temperature is sufficiently suppressed to be considered negligible ( Br2 > Clz), we notice there is no correlation with diatomic halogen bond strength or reaction exothermicity (using X2 + Si 2SiX). Therefore, we need to consider how bond lengths might affect the barriers to chemisorption. A plausible explanation for the trend of increasing barrier height with diatomic halogen mass may be found if we consider the affects of the physisorption well depth E,, the physisorption well distance dp, the chemisorption well depth E,, and chemisorption well distance d, upon the activation barrier Enn. In Figure 7a, we show a potential energy diagram for a directactivated chemisorption process in which the activation barrier to chemisorption is located at the crossing between the chemisorption and physisorption potential energy curves. If either E,, d,, or E, increases, then the activation barrier E,,, will decrease, while if d, increases, E,,, will increase. As discussed later, there is likely an intermediate molecular chemisorption state which exists prior to dissociative chemisorption. For the sake of simplicity, we only show two wells, with the understanding that the chemisorption well may represent either a dissociative state or a molecular chemisorption state. In either case, the arguments used to explain the trend in barrier heights with halogen mass do not change. If E,, the chemisorption well depth, is correlated with the Si-X bond energy, which is seen to decrease with increasing diatomic halogen mass (see Table 4) then we may assume that E, also decreases with increasing diatomic halogen mass. Since a decrease in E, will increase Ea,,, this may partially explain the observed increase in barrier height, Ean,with diatomic halogen mass. It is seen that d,, the chemisorption well distance, increases with increasing diatomic halogen mass if d, is correlated with Si-X bond length in Table 4. Since the barrier height decreases with increasing d,, we may conclude that d, cannot be an important factor in the observed trend of barrier height with diatomic halogen mass. If E,, the physisorption well depth, is correlated with the molecular polarizability, which increases with increasing diatomic halogen mas~,~Oone would expect E,,, to decrease with increasing diatomic halogen mass, which is opposite of the observed trend. Thus, we may conclude that E , is not likely an important factor in the consideration of barrier height versus halogen mass trend. Since dp, the physisorption well distance, increases with increasing diatomic halogen mass (if d, is correlated with the van
-
Flaum et al.
1728 The Journal of Physical Chemistry, Vol. 98, No. 6, 1994
r 2 d p - d
r, distance from surface
2,
F 8 w
defects. The total energy scaling found for Clz is also consistent with a highly corrugated gas-surface potential. B. Precursor-Mediated Chemisorption. The data in Figure 4 may be fit with appropriately modified forms of a standard precursor model, as discussed in the Appendix of our previous paper.40 For the Clz and Br2 data in Figure 3a,b, we use a modified form of the precursor mode140 which yields estimates for the following three parameters: the direct contribution to sticking (SD), the physisorption well trapping probability times the ratio of the pre-exponentials for chemisorption and desorption (Ptrspva/ vd), and the difference in barrier heights for desorption and chemisorption (Ed - Ea). These parameters are incorporated in the following equation for the initial sticking probability, SO,as a function of surface temperature, Ts:
From our previous paper,46 we find SDto be -28% 4% and (PtrapValVd) and (Ed - Ea) equal to 0.093 f 0.032 and 0.032 f 0.006 eV, respectively at 0.038-eV incident energy for Clz (Figure 4a). The uncertainties assigned to (PtrapVa/Vd) and (Ed - Ea) are due to the uncertainty in SD,*4%, and to the uncertainty associated with the least-squares fit. As discussed in our previous we find Vd/Va 5 assuming a value for Ptrap of -50% at 0.038-eV incident energy, which is not uncommon for precursor wells.51 For T, = 807 K and Ei = 0.038 eV, the precursor contributes -15% to Clz chemisorption on the Si(100) (2x1) surface. For Br2, assuming that SDis approximately equal to So(900 K) which is 40% at 0.087 eV incident energy, we perform a twoparameter fit to the data in Figure 4b and find values for (Ptrapva/ Vb) and (Ed - Ea) Of 0.038 f 0.038 and 0.032 0.008 e v , respectively. We note the similarity in the values for (Ed - E,) of 0.032 eV for Clz and Br2. The smaller value of (PtrapValVd) for Br2 as compared to that for Clz may be largely due to the smaller trapping probability at 0.087-eV incident energy, where the data in Figure 4b for Br2 are taken (as compared to 0.038-eV incident energy for the c12 data). It is also possible that Va/Vd is smaller for Br2 than for C12. The following modified form of the standard precursor model is used to fit the data for 1 2 and MCB in Figure 4c,d:
-
r, distance from surface Figure7. Diagramsfor thechemisorptionand the physisorption potential energy curves for (a) direct-activated chemisorption and (b) precursormediated chemisorption. Shown in the figure are the chemisorption well distance d,, the physisorption well distance dp, the physisorption well depth E,, the chemisorption well depth, E,, the activation barrier in part a (Eaa) and (Ed - E,) in part b. (Ed - E,) is the difference in energy for desorption from the physisorption well and adsorption into the chemisorptionwell. For the caseof directchemisorption,the physisorption well crosses the chemisorption curve at a higher point and thus in a steeper region than for part b in the case of precursor-mediated chemisorption. As discussed in the text, the different curve-crossing regions will determine which of the four factors (d,, d,, E,, EP)will most strongly determine the trends of barrier height Ean and (Ed - E,) with diatomic halogen mass.
der Waals radii in Table 4) and Eactincreases with dp,this may explain the observed trend in Eactwith diatomic halogen mass. This effect is shown in Figure 7a, where two physisorption potential energy curves are shown and the curve with a greater dpis seen to result in a larger E,,,. The crossing point between the physisorption and chemisorption potential energy curves will strongly influence which of the four parameters (Ec, d,, Ep,dp) determines the trend of barrier height versus diatomic halogen mass. For example, the influence of chemisorption well distance, d,, may be less significant than physisorption well distance, d,, if the potential curve crossing occurs where the chemisorption potential curve has a steeper gradient than the physisorption potential curve. Thus, we hypothesize that the equilibrium distance of the physisorption well and the chemisorption well depth are the determining factors in the trend of chemisorption barrier height versus diatomic halogen mass. The width of the barriers is seen to be -0.15 eV for all three diatomic halogens, where the width is defined to be the difference in energies around Ean for which SOvaries by *25% of one-half the maximum SO. Such broad distributions in barrier heights suggest the presence of a highly corrugated gas surface potential for the diatomic haloge@( 100) system. This corrugation may be due to a dependence of chemisorption barrier upon molecular orientation, the gas-surface impact coordinate, and/or surface
*
As discussed in the Appendix in our previous paper,@ in this equation no approximation is used. We are able to use this form since the data in Figure 4c,d may be fit with less ambiguity due to the inflection points in both data sets. As discussed in the Appendix of our previous paper,@ for systems where (Ed - E,) >> keT at the lower temperatures, a “freezing out” occurs, and the precursor-mediated chemisorption saturates at the lowest temperatures. We find this to be the case for 12 and MCB. For 12, if we allow SDto vary from 0% to 35% and perform a threeparameter fit for each value of SDat 0.106-eV incident energy, we find that Ptrap ranges from -90% to 50%, vd/va ranges from -6 to -80, and (Ed - Ea) varies from 0.14 to 0.25 eV. The values obtained for PtmP are quite reasonable. Although the range of values obtained for VdIVa is quite large, these values are within the range of values found for other physisorption ~ystems.5~ The I2 values for (Ed - E,) are considerably larger than the valuesobtained for Cl2 and Brz (0.032 eV). This may be explained by considering the potential energy diagram for precursormediated chemisorption shown in Figure 7b. As for the case of direct-activated chemisorption discussed previously, the parameters E, (chemisorption well depth), d, (chemisorption well distance), Ep (physisorption well depth or desorption energy), and ds (physisorption well distance) will determine where the potential energy curve crossing occurs and thus the value of (Ed
Halogen Chemisorption upon a Semiconductor
d,, or E , increases, then (Ed - E,) will increase, while if d, increases, (Ed - E,) will decrease. The potential energy curve crossing region will determine which of the four parameters (E,, d,, E,, d,) will most strongly determine the value of (Ed - E,). The equilibrium physisorption well in Figure 7b crosses the chemisorption curve at a lower energy and thus at a much less steep part of the chemisorption curve than in Figure 7a. The equilibrium physisorption well is deeper in Figure 7b than in Figure 7a, because the molecule has thermally accommodated to the surface and has thus reached a minimum in energy corresponding to the deepest physisorption well. This lower crossing point may allow the chemisorption bond length, d,, to be the determining factor in (Ed - Ea). The physisorption well distance, d,, may not be an important factor in this case since the gradient in the chemisorption potential energy curve with distance is smaller at the bottom of the chemisorption potential energy curve. If E,, the chemisorption well depth, is correlated with the Si-X bond energy, which is seen to decrease with increasing diatomic halogen mass (see Table 4), then we may assume that E, also decreases with increasing diatomic halogen mass. We would then expect (Ed - E,) to decrease with increasing halogen mass, an effect not observed. Thus, E, is not likely to be an important factor in determining the correlation between (Ed - E,) and diatomic halogen mass. Since E , (physisorption well depth) increases with increasing diatomic halogen mass50 (if E , is correlated with the molecular polarizability), one expects (Ed - E,) to increase with increasing diatomic halogen mass. However, no increase in (Ed -Ea) is seen from Cl2 to Brz, and thus this effect is not dominant. Although we may expect d, (physisorption well distance) to increase with increasing diatomic halogen mass (if d, is correlated with the van der Waals radii in Table 4), an expected decrease in (Ed-&) with increasing diatomic halogen mass is not observed. Thus, d , is not likely to be an important factor influencing the trend of (Ed - E,) versus diatomic halogen mass. Since d, (or Si-X bond length) increases with increasing diatomic halogen mass (see Table 3), we expect (Ed - E,) to increase with increasing halogen mass. Indeed, this trend is observed and is consistent with the relative increases in bond length for the Si-X system. In Table 3 a larger increase in bond length occurs between Br2 and I2 than between Cl2 and Brz. The silicon-iodine bond length is 0.28 A larger than the siliconbromine bond length, while the silicon-bromine bond length is 0.14 A larger than the silicon-chlorine bond length. This nonlinear increase is consistent with the sudden increase in (Ed - E,) for 12. In applying eq 5 above to the data in Figure 4d for MCB, we set SD equal to zero as there is no direct component to chemisorption. Performing a three-parameter fit, we find that a t 0.146-eV incident energy Ptrap= 72% f 9%, Vd/V, = 620 f 600,and (Ed-&) = 0.26eVf0.12eV. Thelargevalueobtained for ud/va is expected (we note the large uncertainty and thus we caution against placing too much emphasis on this discussion), considering that gas-phase MCB has a large number of degrees of freedom (compared to a diatomic halogen for example).50The value of (Ed - E,) for C&Cl is larger than that of c12 or Br2 but comparable to thevalue for 12. Using theargumentsdescribed in the previous paragraph, either the physisorption wells for CsH5C1 and 1 2 are closer to the silicon surface than those for Br2 or C12 or the silicon-C6H+21 (or 12) chemisorption potential well is farther from the surface. No direct dissociative chemisorption is observed for MCB; this is not surprising considering that the carbon-chlorine bond is 3.3 eV which is significantly larger than the diatomic halogen bond energies (C12 being the largest a t 2.4 eV). This is confirmed by saturating the surface at high incident energy (where any direct dissociative chemisorption would be expected), heating the surface to 600 “C, and then measuring the carbon to silicon Auger ratios. C(272 eV)/Si(92 eV) is always - E,) for precursor-mediated chemisorption. If either E,,
The Journal of Physical Chemistry, Vol. 98, No. 6, 1994 1729
TABLE 5: Sticking due to Precursor-Mediated Chemisorption versus Incident Energy Where S(300 K) - S(900 K) Except for Clz Where%=, S(300 K) - Sot807 K)
-
--
=
sticking at Ei 0.1 C12 Br2 I2 C6HsCI
0.2
0.3
0%(f6%) 15%(17%) 12%(17%) 6% (17%) 47% (18%) 35% (18%) 7% (110%) 80%(15%) 74% (45%) 69 (15%) 11% (16%) 5% (16%)
-
0.4
0.5
0%(16%) 0%(+7%) 0%(110%) 69 (15%)
0% (16%) 0%(&7%) 0%(110%) 58% (15%)
less than 1/50, which proves negligible dissociative chemisorption is present for this system. Thestate that the MCB adsorbs into from the physisorbed precursor must therefore be a molecular chemisorption state. A similar precursor-mediated molecular chemisorption state has been found for the case of CO chemisorbing upon Pt( 11 1) by J. Harris and A. c. L u n t ~ . ~ ~ The extent to which precursor-mediated chemisorption contributes to the sticking in the diatomic halogens and in MCB can be seen by inspecting Table 5. Table 5 shows the precursor contribution to chemisorption for the three diatomic halogens and MCB as a function of incident energy. Although it is difficult to estimate the exact precursor contribution due to the uncertainty in the direct contribution, we may use an operational definition for the sake of comparison. Thus we define Sprccursor equal to So(300 K)minusSo(900 K),whereSo(900 K) isanapproximation to the direct contribution to chemisorption. In moving down the table from chlorine to MCB, the precursor contribution to chemisorption increases at every incident energy. The increase in precursor-mediated chemisorption from chlorine to iodine is not surprising, as the molecular mass has increased in moving from chlorine to iodine, and we expect a more inelastic collision as the incident molecule’s mass approaches that of the effective silicon surface mass. A larger inelasticity would lead to a higher trapping probability. Precursor-mediated chemisorption is the strongest for MCB even though MCB is not the most massive molecule. However, MCB has the largest molecular polarizability50 which should allow for a deeper physisorption well and thus a higher probability of precursor-mediated chemisorption for MCB. C. Energy Scaling. The initial sticking probabilities typically depend on the angle of incidence in one of two ways. The initial sticking probability can obey total energy scaling (SOis not a function of the incident angle) or normal energy scaling (SO depends on the normal component of the incident energy and thus So scales with Ei cos2 e). Total energy scaling is associated with either high corrugation in the gas-surface potential interaction potential or a “late barrier” for chemisorption when the surface corrugation is ~ma11;53.~~ a “late barrier” would imply that the barrier to chemisorption is for the breaking of the diatomic halogen bond as opposed to the formation of any Si-halogen bonds. Normal energy scaling is associated with an early barrier to chemisorption for a gas-surface potential of modest c o r r u g a t i ~ n . ~ ~ - ~ ~ The sticking probability of Cl2 (as discussed in our previous paper&) andMCBontotheSi(100) (2x1) surfacedoesnotchange by more than 10% in going from normal to glancing incidence. This was found to be true for all incident energies for C12. At low incident energies, where precursor-mediated chemisorption is dominant, the gas-surface interaction cannot be sensitive to a late barrier, suggesting that the gas-surface potential is corrugated for the C12/Si(100) (2X 1) system. Summarizing our previous results, total energy scaling for the Clz chemisorption probability on the Si(100) (2x1) surface indicates that the gas-surface potential is corrugated, and there may be a late barrier to chemisorption. Since MCB does not dissociatively chemisorb upon the Si(100) (2x1) surface, it is not meaningful to discuss an “early” or “late” barrier. However, a highly corrugated gas surface potential which effectively scrambles the parallel and perpendicular components of momentum can be inferred from the total energy scaling for MCB. In the case of CO upon Pt-
1730 The Journal of Physical Chemistry, Vol. 98, No. 6, 1994
Flaum et al.
(1 1l),52 where a precursor-mediated molecular chemisorption we see no evidence among the diatomic halogens for a long-lived state is found, total energy scaling is also observed, and Harris extrinsic precursor state; such a state would be most likely for and Luntz conclude in a similar fashion that scrambling among I2 with the surface slightly below room temperature. Dissociative the parallel and perpendicular components of translational energy chemisorption from the physisorbed precursor is also consistent is due to a highly corrugated interaction potential. with this result, given an intrinsic precursor which requires a D. Coverage-DependentKinetics. As discussed by W e i ~ ~ b e r g , ~ ~single site as well as a rate of chemisorption from the precursor for the case of direct chemisorption, S/So will decrease as (1 well much greater than the rate for desorption. Since it is known 0/0,)d where the exponent d is simply equal to the number of that dissociative chemisorption is the final step in diatomic halogen sites required for chemisorption (whether it be molecular or chemisorption upon the Si(100) (2x1) surface, we must include dissociative chemisorption). A site may be, for example, either the final dissociative step in any complete kinetic model. However, a single surface atom or a surface dimer pair.59 In the case of the coverage results at high incident energy do suggest the presence a dimer pair, it is required that the dimer pairs are correlated of a molecularly chemisorbed state which the diatomic halogen with a single configuration. The relationship between sticking molecule must pass thru prior to dissociative chemisorption. probability and adsorbate coverage is modified given the presence The recent STM work of C. Yan, J. A. Jensen, and A. C. of either an intrinsic or extrinsic precursor. Kummel61 shows island formation at low incident energies for We briefly summarize the results for coverage-dependent Clz upon Si( 11 1) (7x7) suggesting the presence of a mobile sticking with precursor-mediated chemisorption as discussed by precursor. At high incident beam energies, they find only isolated Weinberg. We treat the case of molecular chemisorption first chemisorption sites, suggesting that if a molecular chemisorption (with no dissociative chemisorption). For an intrinsic precursor, state exists, it is short lived and thus not mobile. Since weobserve the sticking decreases linearly with coverage (d = l), assuming the same linear decrease in sticking with coverage for the Si( 1 11) that the rate constants for chemisorption and desorption from (7x7) surface, it is reasonable to assume that the same kinetic the precursor well do not depend on coverage. If both intrinsic mechanisms apply to both surfaces. Thus if we extrapolate the and extrinsic precursors exist (and are energetically equivalent STM results of Yan et al. to the Si(100) (2x1) surface, we may as discussed by Weinberg), the solution for sticking versus coverage infer that the intrinsic precursor is mobile on this surface, while is more complicated but may be simplified under certain limits. If the rate at which molecules are chemisorbing from the precursor the hypothesized molecular chemisorption state is short lived and well is much greater than the rate at which they are desorbing, not mobile. then S is independent of coverage. On the other hand, if the For MCB, we find that S/So decreases linearly with coverage reverse is true, then S will still decrease linearly with coverage. a t high incident energy. Since dissociative chemisorption does In the case of dissociative chemisorption with an intrinsic not occur here, chemisorption via an intrinsic precursor (with no precursor which requires a single site, the interpretation of restriction on the number of sites) followed by molecular measurements of sticking versus coverage is complicated. In the chemisorption is most likely. An extrinsic precursor is also possible limit that the rate for chemisorption is much greater than the if the rate for desorption is much greater than the rate for rate for desorption of the precursor, then S decreases linearly adsorption, but again Ptrap SOso an extrinsic precursor is very with coverage. In the opposite limit, then S/So (1 unlikely. At low incident energy, we find that S/So (1 If the intrinsic precursor requires a nearest neighbor pair of sites, 8/8,)3/2. This might be explained by the presence of secondor then S/So (1 - 0/8,)2. If both intrinsic and extrinsic precursors third-layer adsorption, with MCB molecules physisorbing upon exist, and thedesorption rate is much larger than thechemisorption islands of previously adsorbed MCB with the sticking probability rate for the precursor, then S/SO (1 - 0/0J2. on the second or third layer being coverage dependent. If we To within the accuracy of our measurements, it is found that integrate the saturation coverage (0,) at low and high incident S/So (1 - 0/8,) for diatomic halogen chemisorption upon the energy, we find that the ratio of 8, at low incident energy to 8, Si(100) (2x1) surface at all incident energies. This result is at high incident energy is -2.6 f 0.5 for MCB but 1.0 f 0.3 consistent with several cases discussed above. At high incident for Clz. This is consistent with the hypothesis of limited multilayer energies, where a physisorbed precursor is not a factor, the linear adsorption. decrease in S with coverage may be explained by the presence of a short-lived molecular chemisorption state which exists prior E. Ramifications for Practical Etching. Our results show that to dissociation. The molecular chemisorption state would be in a thermal etch process which requires minimum sidewall etching equivalent to an intrinsic physisorbed precursor requiring a single accelerated beams of Brz or 1 2 are preferable to Clz. This is site (thus the (1 - @/e,) scaling) followed by dissociative because at surface temperatures of 900 K and at incident energies chemisorption (with a much larger rate of dissociative chemiabove 0.5 eV SO 95% for Brz and IZwhile SO 75% for Clz. sorption versus desorption from the molecular chemisorption well). The higher initial sticking probability results in a fivefold reduction Dissociation at two correlated sites of a single type would also in scattered flux and thus a possible fivefold reduction in sidewall yield a (1 - e/&) dependence. However, the STM work of J. etching.62 Additionally, accelerated beams of 1 2 may be superior Boland60 shows at least five different types of correlated sites for to Br2 beams, since those molecules that due scatter are less likely C12/Si( 100) (2x1). At higher coverages, thesticking probability to chemisorb in the case of 12 due to the slightly greater activation would then deviate from a simple (1 -@/e,)dependence. Within barrier for 12. As a practical matter though, it is easier togenerate the experimental uncertainties of our technique, no deviation is molecular beams of Br2 due to the higher vapor pressure of Brz observed, and thus direct dissociation with no intermediate as compared to 12. molecular chemisorption state, into a correlated pair of sites, is In a digital etch process, where monolayer halogen adsorption not consistent with our results. at room temperature or lower is followed by electron-, photon-, At lower incident energies where precursor-mediated chemior ion-stimulated desorption, a halogen with an extrinsic precursor sorption is significant, there are several cases as discussed above would be preferred.63 In the absence of an extrinsic precursor that are consistent with the linear decrease in sticking with the sticking probability decreases linearly with coverage, and coverage. A molecular chemisorption state with an intrinsic thus the time to cover the surface with adsorbates will be lengthy. precursor would certainly be consistent with this result. An Although we see no evidence of an extrinsic precursor at room extrinsic precursor would also be possible, if the rate for desorption temperature for the diatomic halogens, a t temperatures slightly were much greater than the rate for chemisorption from the below room temperature 12 is the most likely to possess an extrinsic precursor well. However, since we observe that Ptrap is approxprecursor, since intrinsic precursor-mediated chemisorption is imately equal to SOat 300 K, the opposite is true; that is, the rate for desorption is much less than the rate for chemisorption. Thus strongest for 12.
-
-
-
-
-
-
-
-
-
Halogen Chemisorption upon a Semiconductor
V. Conclusion We have measured the sticking probabilities of C12, Brz, 12, and CsHSCl upon the Si(100) (2x1) surface versus incident translational energy, surface coverage, surface temperature, and angle of incidence. For the diatomic halogens, both precursormediated and direct-activated chemisorption are present. The average barrier to direct chemisorption is found to increase in going from c12(0.055 eV) to Brz(0.120 eV) to 12(0.144 eV). This may be caused by an increase in the physisorption well distance or chemisorption well depth with diatomic halogen mass, which has the effect of increasing the barrier to direct chemisorption by shifting the crossing between the physisorption and chemisorption potential energy curves. The precursor mechanism is very similar for C 4 and Br2, with the difference in the activation energy for desorption and adsorption from the precursor well, (Ed - Ea),equal to 0.032 eV. In contrast, we find for 1 2 a (Ed - E , ) value of -0.2 f 0.05 eV, which might be explained by the significantly larger silicon-iodine chemisorption well distance. Direct chemisorption is absent for MCB, with precursor-mediated chemisorption persisting to -2.0 eV. For MCB we find (Ed E,) equal to 0.26 f 0.12 eV which is larger than the values of (Ed - E,) for c12 and Brz. Total energy scaling for chemisorption is observed at all incident translational energies for chlorine and at 1.O eV incident energy for MCB, which strongly suggests highly corrugated gas-surface potentials. The sticking probability decreases linearly with coverage for the three diatomic halogens at all incident energies, implying that a single site is required for chemisorption. At low incident energies, where precursormediated chemisorption is significant, the physisorbed intrinsic precursor requires a single site. Since the final step in diatomic halogen chemisorption is dissociation (at the temperatures where sticking measurements are performed in these experiments), we hypothesize the existence of a short-lived molecular chemisorbed state requiring one site for adsorption, which the molecule passes through prior to dissociation.
Acknowledgment. This work was supported by the Air Force Office of Scientific Research (AFOSR) under Grant No. 890390. We thank Tom Sage for his illustrations. We thank John Jensen for his help in assembling the apparatus. We also thank Michael Ricks, Carl Faria, Mark Trujillo, and John Trujillo of Thermionics for their help in designing and building the molecular beam source and Alan Luntz for useful discussions. References and Notes (1) Chadi, J. Phys. Rev. Lett. 1979,43,43. (2) Takayanagi, K.; Tanishiro, Y.; Takahasi, S.; Takahashi, M. Surf. Sci. 1985. 164. 367. (3) Rettner, C. T.; DeLouise, L. A.; Auerbach, D. J. J . Chem. Phys. 1986,85,113 1. (4) Ceyer, S . T. Langmuir 1990,6,82. (5) DEveln, M. P.; Hamza, A. V.; Gdowski, G. E.; Madix, R. J. Surf. Sci. i986,167,451. (6) Luntz, A. C.; Williams, M. D.; Bethune, D. S. J. Chem. Phys. 1988, 89,4381. (7) Hamza, A. V.; Madix, R. J. Surf. Sci. 1987, 179,25. (8) Grimblot,J.;Luntz,A. C.;Fowler,D.E.J.Elect.Spec. Relat. Phet". 1990,52, 161. (9) Rettner, C. T.; Mullins, C. B. J . Chem. Phys. 1991,94,1626. (10) Miyamura, M.; Sakisaka, Y.; Nishijima, M.; Onchi, M. Surf. Sci. 1978,72,243. (11) Cheng, C. C.; Gao, Q.; Choyke, W. J.; Yates, J. T., Jr. Phys. Rev. E, in press. (12) Mogab, C. J.; Levinstein, H. J. J. Vac. Sci. Technol. 1980,17,721.
The Journal of Physical Chemistry, Vol. 98, No. 6, 1994 1731 (13) Geis, M. W.; Efrermow, N. N.; Lincoln, G. A. J . Vac.Sci. Technol. B 1986,4, 315. (14) Aoto, N.; Ikawa, E.; Kuogi, Y. J. Vac. Sci. Technol. B 1986,4,806. (15) Geis, M. W.; Efrermow, N. N.; Pang, S. W.; Anderson, A. C. J . Vac. Sci. Technol. B 1987,5, 363. (16) McNevin, S. C.; Becker, G. E. J . Vac. Techno. B 1985,3,485. (17) Florio, J. V.; Robertson, W. D. Surf. Sci. 1969,18,398. (18) Madix, R. J.; Schwarz, J. A. Surf. Sci. 1971,24, 264. (19) Gupta,P.;Coon,P.A.;Koehler,B.G.;George,S.M.Surf.Sci.1991,
249,93. (20) Jackman, R. B.; Ebert, H.; Foord, J. S.Surf. Sci. 1986, 176, 183. (21) Aoto, N.; Ikawa, E.; Kuogi, Y. Surf. Sci. 1988,199,408. (22) Schluter, M.; Rowe, J. E.; Margaritondo, G.; Ho, K. M.; Cohen, M. L. Phys. Rev. Lett. 1976,37, 1632. (23) Rowe, J. E.; Margaritondo, G.; Christman, S. B. Phys. Rev. B 1977, 16,1581. (24) Larsen, P. K. et al. Phys. Rev. B 1978,17,2612. (25) Bachelet, G. B.; Schluter, M. Phys. Rev. E 1983,28, 2302. (26) Thornton, G.; Wincott, P. L.; McGrath, R.; McGovern, I. T.; Quinn, F. M.; Norman, D.; Vvedensky, D. D. Surf. Sci. 1989,211,959. (27) Citrin, P. H.; Rowe, J. E.; Eisenberger, P. Phys. Rev. B 1983,28, 2299. (28) Seel, M.; Bagus, P. S. Phys. Rev. B 1983,28,2023. (29) Pandey, K. C., Sakurai, T., Hagstrum, H. D. Phys. Reu. B 1977 16, 3648. (30) Mohapatra, S. M.; Dev, B. N.; Mishra, K. C.; Sahoo, N.; Gibson, W. M.; Das, T. P. Phys. Rev. B 1988,38, 12556. (31) Mohapatra, S. M., Dev, B. N., Mishra, K. C., Sahoo, N., Gibson, W. M.. Das. T. P. J. Vac. Sci. Technol. A 1986. 4. 2441. (32) Barone, V., Lelj, F., Russo, N. Toscano,' Solid State Commun. 1986,59,433. (33) Walker, Z . , Ogryzlo, E. A. J. Chem. SOC.,Faraday Trans. 1991,87,
hi.
.".
AT
(34) Walker, Z., Ogryzlo, E. A. J. Appl. Phys. 1991,69,2635. (35) Ogryzlo, E. A., Flamm, D. L., Ibbotson, D. E., Mucha, J. A. J . Appl. Phys. 1988,64,6510. (36) Jackman, R. B., Price, R. J., Foord, J. S. Appl. Surf. Sci. 1989,36, 296. (37) Tyrrell, G. C., Boyd, I. W., Jackman, R. B. Appl. Surf. Sci. 1989, 43,439. (38) Campos, F. X., Weaver, G. C., Waltman, C. J., Leone, S. R. Mat. Res. SOC.Symp. Proc. 1992,236, 177. (39) Szabo, A., Farrall, P. D., Engel, T. Unpublished results. (40) Flaum, H. C., Sullivan, D. J. D., Kummel, A. C. J. Chem. Phys., in
press. (41) 1484. (42) (43) (44) (45) 245. (46)
Hanisco, T. F., Yan, C., Kummel, A. C. J . Chem. Phys. 1992,97, Sullivan, D. J. D. Kummel, A. C. Rev. Sci. Instrum. 1992,63,4285. Kern, W. Semiconductor Internat. 1984,7,94. King, D. A., Wells, M. G. Surf. Sci. 1972,29, 454. King, D. A., Wells, M. G. Proc. R . SOC.London Ser. A 1974,339,
~~
Sullivan, D. J. D., Flaum, H. C., Kummel, A. C. J . Phys. Chem. 1993,97, 12051-12060. (47) Su, C., Xi, M., Dai, Z., Vernon, M. F., Bent, B. E. Surf. Sci. 1993, 282,357. (48) Ludviksson, A., Xu, M., Martin, R. M. Surf. Sci. 1992,277,282. (49) Huheey, J. E. Inorganic Chemistry; Harper and Row: New York, 1983. (50) CRC Handbook of Chemistry and Physics; 1991. (51) Weinberg, W. H. Advances in Gas-Phase Photochemistry and
Kinetics, Dynamics of Gas Surface Interactions; Rettner, C . T., Ashford, M. N. R., Eds.; 1991; p 171. (52) Harris, J., Luntz, A. C. J. Chem. Phys. 1989,91,6421. (53) Kara, A., Depristo, A. E. Surf. Sci. 1988,193,455. (54) Holloway, S., Gadzuk, J. W. 1985,82,5203. J. W.. Hbllowav. S. Chem. Phvs. Lett. 1985. 114. 314. (55) Gadzuk. (56) DeLouise, L. A. Chem. phys. Lett. 199i,180, 149. (57) Darling, G. R., Holloway, S. Surf. Sci. Lett. 1991, 244,L81. (58) Weinberg, W. H. In Kinetics of Interface Reactions Springer Series in Surface Science, Vol. 8; Grunze, M., Kreuzer, H. J., Eds.; Springer: . - Berlin, 1987; p 94. (59) Metiu, Horia Personal Communication. (60) Boland, J. J. Vac. Sci. Technol. Proceedings, to be published. (61) Yan, C., Jensen, J. A., Kummel, A. C. Phys. Rev. Lett., submitted. (62) Teraoka, Y., Uesugi, F., Nishiyama, I. Mater. Res. Symp. Proc. \ - - ,
I~
1992,236, 183. (63) Meguro, T., Hamagaki, M., Modaressi, S., Hara, T., Aoyagi, Y., Ishii, M., Yamamoto, Y. Appl. Phys. Lett. 1990,56, 1552.