Comparison of the interactions of XeF2 and F2 with Si(100)(2 ?? 1)

Article abstract of DOI:10.1021/jp020936p

The interaction of low-energy XeF₂ with Si(100)(2 ?? 1) has been studied and compared to that of F₂. Helium atom diffraction, beam-surface scattering, and thermal desorption measurements are the major techniques used in this study. It is found that XeF₂ dissociatively chemisorbs with high probability solely on the Si dangling bonds up to a coverage of about one monolayer (ML). Molecular fluorine has previously been observed to react similarly, saturating the dangling bonds at 1 ML coverage. The thermal desorption kinetics and products from the fluorinated layer produced by XeF₂ exposure are identical to those produced by F₂ exposure. The interactions of XeF₂ and F₂ are also strikingly similar with respect to the long-range order of the fluorinated Si up to about 1 ML coverage. The order is monitored by He diffraction. In both systems, the diffracted He beams exhibit a sharp decrease in intensity because of the disorder produced by the fluorination of random surface-unit cells as the coverage increases from 0 to about 0.3 ML. The intensity then increases until the fluorine overlayer has fully recovered its (2 ?? 1) periodicity at about 1 ML. This recovery corresponds to the decoration of each Si dangling bond with a fluorine atom. A critical observation of this study is that despite the large exothermicity of the dissociative chemisorption of XeF₂ or F₂ the order of the surface is not destroyed in either system. After saturation of the dangling bonds, F₂ ceases to react with the surface whereas XeF₂ continues to deposit fluorine by reacting with the Si-Si ?? dimer bonds and the Si-Si lattice bonds. The order is destroyed as a result of the continued fluorine deposition, and ultimately, etching occurs by the formation of volatile SiF₄.

Full text of DOI:10.1021/jp020936p

J. Phys. Chem. B 2002, 106, 8399-8406
8399
†
Comparison of the Interactions of XeF2 and F2 with Si(100)(2 × 1)
J. R. Holt, R. C. Hefty, M. R. Tate, and S. T. Ceyer*
Department of Chemistry, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139
ReceiVed: April 9, 2002; In Final Form: June 10, 2002
The interaction of low-energy XeF
atom diffraction, beam-surface scattering, and thermal desorption measurements are the major techniques
used in this study. It is found that XeF dissociatively chemisorbs with high probability solely on the Si
2
with Si(100)(2 × 1) has been studied and compared to that of F
2
. Helium
2
dangling bonds up to a coverage of about one monolayer (ML). Molecular fluorine has previously been
observed to react similarly, saturating the dangling bonds at 1 ML coverage. The thermal desorption kinetics
and products from the fluorinated layer produced by XeF
2 2
exposure are identical to those produced by F
exposure. The interactions of XeF and F are also strikingly similar with respect to the long-range order of
2
2
the fluorinated Si up to about 1 ML coverage. The order is monitored by He diffraction. In both systems, the
diffracted He beams exhibit a sharp decrease in intensity because of the disorder produced by the fluorination
of random surface-unit cells as the coverage increases from 0 to about 0.3 ML. The intensity then increases
until the fluorine overlayer has fully recovered its (2 × 1) periodicity at about 1 ML. This recovery corresponds
to the decoration of each Si dangling bond with a fluorine atom. A critical observation of this study is that
despite the large exothermicity of the dissociative chemisorption of XeF
destroyed in either system. After saturation of the dangling bonds, F ceases to react with the surface whereas
XeF continues to deposit fluorine by reacting with the Si-Si σ dimer bonds and the Si-Si lattice bonds.
The order is destroyed as a result of the continued fluorine deposition, and ultimately, etching occurs by the
formation of volatile SiF
2 2
or F the order of the surface is not
2
2
4
.
I. Introduction
studies of the mechanism and dynamics of the dissociative
21-24
chemisorption of F2 on Si(100)(2 × 1).
The results of these
The integration of conventional electronics with 3D mechan-
ical components within a Si microchip is known as microelec-
tromechanical systems technology or MEMS. The fabrication
of these 3D mechanical structures involves many steps that may
include X-ray lithography, chemical vapor deposition, and
anisotropic chemical etching, but it also requires an isotropic
studies provide a means of calibrating the fluorine coverage.
The calibration in turn enables the absolute probability of the
dissociative chemisorption of XeF2 and the coverage at which
etching begins to be determined.
This investigation also aims to test the notion that the release
of the exothermicity to Si upon XeF2 dissociative chemisorption
disorders the surface periodicity by introducing defects. Disorder
is said to be responsible for the onset and subsequent high rates
1
-5
etching step. One isotropic etchant currently in use is XeF2.
Xenon difluoride is a gas-phase, plasmaless etchant that reacts
rapidly and isotropically with Si at room temperature, neither
leaving an adsorbed byproduct on nor radiatively damaging the
Si. Its etching rates are as high as 40 µ/min. Among other
advantages offered by XeF2 is compatibility with complementary
25-28
of XeF2 etching.
work has shown that F2 incident at low translational energies
<3.8 kcal/mol) on Si(100) reacts only with the Si dangling
This notion is curious because our previous
(
6
bonds. No disorder of the surface is evident, even though the
dissociative chemisorption of F2 on Si can be exothermic by as
metal oxide semiconductor technology processing.
Of interest in the present work is the high reactivity of XeF2
with Si as compared to the reactivity of F2. Given the chemical
similarities of XeF2 and F2, it is curious that the reaction rate
of XeF2 with Si to form volatile etch products such as SiF4 is
21,22
much as 251 kcal/mol.
Indeed, the absence of disorder upon
the interaction of F2 with Si and the absence of F2 etching are
consistent with the notion that disorder is a prerequisite for
etching. However, consider the relative exothermicities of the
interactions of F2 and XeF2 with Si. Although the dissociative
chemisorption of XeF2 is at least 28 kcal/mol less exothermic
than the dissociative chemisorption of F2, XeF2 spontaneously
etches Si whereas F2 does not. It would be of interest to probe
whether the dissociative chemisorption of XeF2 induces disorder,
in contrast to the behavior of the more exothermic interaction
of F2 on Si. The primary roadblock to an assessment of the
structure of a fluorine overlayer is that the readily available
technique, electron diffraction, is not suitable. The large electron-
stimulated desorption cross section of fluorine leads to destruc-
tion of the overlayer before the structure can be determined. In
this study, He atom diffraction is employed to monitor the
structure of the fluorine overlayer as a function of exposure to
3
4
7-11
1
0 -10 times higher than that of F2.
In short, XeF2 etches
Si whereas F2 does not. Since the discovery that XeF2 spontane-
7
ously etches Si under ambient conditions, many investigators
have sought to understand why XeF2 is so reactive with Si.
1
2-20
The vast majority of these investigations have been carried out
with sufficient XeF2 exposure to produce high fluorine coverages
that yield the volatile etch product. In contrast, this investigation
compares the interactions of XeF2 and F2 as the fluorine
coverage builds from zero. It aims to delineate the fluorine
coverage at which the interactions of XeF2 and F2 diverge. This
goal is now possible because of recent extensive and quantitative
†
Part of the special issue “John C. Tully Festschrift”.
Corresponding author. E-mail: stceyer@mit.edu.
*
1
0.1021/jp020936p CCC: $22.00 © 2002 American Chemical Society
Published on Web 07/10/2002

8
400 J. Phys. Chem. B, Vol. 106, No. 33, 2002
Holt et al.
Ar
Ar
XeF2 and F2, thereby testing the notion that disorder resulting
from defects is necessary for the onset of etching.
) 2I V/V . This procedure yields values of 0.097((0.005) and
0.16((0.01) ML F atom/s for the flux of the incident XeF2 and
F2 beams, respectively, where 1 ML is the surface density of
1
4
2
II. Experimental Section
Si atoms on a Si(100) surface, 6.78 × 10 atoms/cm .
The secondary source is aimed at the crystal so that its beam
is at a 20° angle from the primary beam. The secondary beam
is a supersonic expansion of 300 Torr of a mixture of 75% He
The apparatus has been described in detail elsewhere.29
Briefly, the apparatus consists of two differentially pumped
molecular beam sources coupled to an ultrahigh vacuum
(
99.9999%, Spectra Gases)/25% Ar (99.9995%, Spectra Gases)
-
11
chamber (base pressure of 5 × 10
Torr) housing the Si
through a 0.002-in. diameter orifice held at 320 ( 20 K. This
expansion results in nearly monoenergetic He atoms with an
average translational energy of 21.4 meV (fwhm ) 13.3 meV)
and a 0.93-Å de Broglie wavelength. This beam is used for He
diffraction measurements.
B. Si(100) Crystal. The lightly p-doped Si(100) crystal is
mounted between two Ta clamps that are attached to the
manipulator. The Ta clamps are flush with the crystal face,
thereby eliminating the possibility of spurious shadowing effects
of the crystal surface by the beams incident at glancing angles.
The crystal can be cooled to 125 K and heated resistively to
crystal, a cylindrical mirror energy analyzer for Auger spec-
troscopy, an ion-sputtering gun, a residual gas mass analyzer,
and a triply differentially pumped, line-of-sight, rotatable
quadrupole mass spectrometer. The spectrometer is equipped
with a pseudorandom chopper, allowing for time-of-flight
measurements.
A. Molecular Beams. The primary beam is a semieffusive
expansion of neat XeF2 (99% pure by F ion titration, Lancaster
Synthesis) or F2 (97% pure, Air Products, without HF trap). To
avoid decomposition of the XeF2 upon exposure to atmospheric
gases, the XeF2 is stored under nitrogen and transferred to a
stainless steel vessel in a pressurized nitrogen glovebox. The
stainless steel vessel is isolated and then attached to the gas-
handling manifold leading to the beam source without exposing
the XeF2 to the atmosphere. Prior to each day’s experiments,
the XeF2 beam is allowed to flow for 15 min. The manifold is
then evacuated, and the process is repeated. This procedure
serves to degas the XeF2 sample and to passivate the walls of
∼
1100 K, as measured via a W-5%Re/W-26%Re thermocouple
clamped to the back of the crystal. The temperature is held
constant at 250 K during XeF2 or F2 exposure. The temperature
is stabilized using a proportional integral differential (PID)
feedback loop to control the resistive-heating power supply.
Similar results to those shown below are obtained after XeF2
or F2 exposure of Si(100) at 150 and 300 K.
The Si crystal is cut along the (100) plane and cleaned by a
30
the stainless steel gas-handling manifold. To achieve a constant
stagnation pressure, the temperature of the stainless steel vessel
containing the XeF2 solid is held constant at 30((1) °C by
submerging it in a water bath. The remainder of the gas-handling
manifold including the nozzle is also warmed to eliminate “cold
spots” that would condense the XeF2 vapor and cause the
stagnation pressure to fluctuate. Consequently, the stagnation
pressure is constant at 6.40 ( 0.04 Torr as measured by a
Baratron capacitance manometer. The neat F2 is taken directly
from the gas cylinder and regulated at a stagnation pressure of
3
1
wet-etching procedure prior to installation into the vacuum
chamber. The crystal is mounted such that the scattering plane,
defined by the beams, crystal normal, and detector, is along
the (10) direction of the crystal surface. Helium diffraction
confirms the (2 × 1) periodicity of the reconstructed Si(100)
+
surface. The crystal is cleaned by sputtering with 1.5-keV Ar
followed by a 30-min. anneal at 1100 K. This process is repeated
until carbon and oxygen contamination are below the 1%
sensitivity limit of Auger electron spectroscopy. No metal
contamination, such as W, Ta, Cu, or Ni, is observed. Between
experiments, a brief anneal at 1100 K is used to clean the surface
and restore the (2 × 1) surface periodicity. A heating rate of 2
K/s and a cooling rate of 0.7 K/s is used for all anneals. The
crystal is replaced when an etch spot becomes visible, typically
after several months of experiments. No difference in the
reactivity or the diffraction spectra of the Si(100) crystal is
observed over the lifetime of the crystal.
6.4 Torr to match the XeF2 stagnation pressure. The beams used
in most of the experiments presented here are expanded from a
-
9
2
4
.7((0.8) × 10 m orifice of a Ni nozzle and are subsequently
passed through a single differential pumping region, with
collimating slits located at the entrance and exit of the
differential stage. The average translational energies of the XeF2
and F2 beams formed in this fashion are 1.95 ( 0.07 and 1.63
(
0.05 kcal/mol, respectively, as measured by cross-correlation
C. Detection Scheme. Thermal desorption and scattering
measurements employ a triply differentially pumped, rotatable
quadrupole mass spectrometer with electron bombardment
ionization as the detector. The detector rotates in the plane of
the beams around their point of intersection at the crystal sur-
face. The angular range of the detector is 35° to 180.5° with
respect to the primary beam. Its angular resolution in the
scattering plane is 3.52°. It has been shown that the use of a
properly differentially pumped detector is critical, in particular
for the detection of radical species such as SiF2. Spurious
features due to secondary interactions of the radical species with
the chamber walls can obfuscate the signal of the radical species
time-of-flight spectroscopy. The corresponding full energy
widths at half-maximum are 2.30 and 1.99 kcal/mol, respec-
tively. Beams of XeF2 and F2 with average translational energies
of 1.29 and 1.32 kcal/mol were also used in some experiments.
The results are independent of these small changes in incident
energy.
Determination of the absolute beam flux is detailed else-
where.22 Briefly, a beam of Ar at a stagnation pressure of 6.40
(
0.04 Torr is directed into the chamber containing the crystal.
The pressure rise in the chamber, P, is measured with a nude
Bayard-Alpert ionization gauge and calibrated to the absolute
pressure by accounting for the ionization efficiency, C, of Ar.
The absolute flux of Ar impinging on an area, A, of the crystal
22
scattered directly from the surface.
Ar
is given by I ) PCS/kTA, where S is the pumping speed of
III. Results
Ar and T is the temperature of the chamber. The flux of a beam
of pure F2 or XeF2 at a stagnation pressure of 6.40 ( 0.04 Torr
A. Dissociative Chemisorption Probability of XeF2 on
Si(100) at 250 K. Recent studies of the interaction of F2 with
Si(100)(2 × 1) have shown that F2 reacts via atom abstraction
with the Si(100) surface dangling bonds. Once each dangling
bond is bound to a F atom, the reaction ceases. The resulting
Ar
Ar
Ar
is given by I ) I V/V , where V and V are the experimentally
determined average speeds of F2 or XeF2 and Ar, respectively.
Because there are two F atoms per molecule in both F2 and
2
XeF2, the incident flux in terms of F atoms is I [F atoms/m s]

XeF2 and F2 with Si(100)
J. Phys. Chem. B, Vol. 106, No. 33, 2002 8401
Figure 2. Total thermal desorption yield (right axis) and fluorine
coverage in ML F atoms (left axis) as a function of exposure to F
and XeF (O) in ML F atoms. See text for explanation of axes. Solid
lines are exponential fits to the data.
2
(4)
2
chemisorption probability of XeF2 does not approach zero as
the coverage increases to 1 ML.
The total thermal desorption yield at a given exposure or
coverage is the sum of the SiF2 and SiF4 signals in Figure 1
2
4
integrated over temperature. Briefly, the integrated yields of
each product are scaled for the relative detection sensitivities
of SiF2 and SiF4, the factor of 2 more fluorine atoms that SiF4
has relative to SiF2, the different velocity and angular distribu-
tions of the desorbing SiF2 and SiF4 species, and their relative
ionization cross sections and quadrupole transmissions. The SiF4
Figure 1. Thermal desorption spectra measured at m/e ) 66 (a, b)
2 s
and m/e ) 85 (c) after XeF exposure at T ) 250 K to yield the fluorine
coverages in ML F atoms that are shown for each trace. Temperature
ramp rate is 5 K/s.
saturation coverage is about 1 ML (0.94 ( 0.11).22 The partial
pressures of products desorbing from this layer that are
monitored in a thermal desorption experiment and that are
integrated over temperature are a measure of 1 ML of adsorbed
fluorine. This value is used in the present study to calibrate the
fluorine coverage resulting from a known exposure of Si(100)
to XeF2. Knowledge of this fluorine coverage coupled with
knowledge of the absolute flux of the incident XeF2 beam
enables the dissociative chemisorption probability of XeF2 to
be determined as a function of coverage, as follows.
thermal desorption yield as a result of exposure to XeF is small,
never exceeding 9% of the SiF yield.
2
2
The total thermal desorption yield, in arbitrary units, is plotted
on the right-hand abscissa in Figure 2 as a function of exposure
to XeF and F . Both yields increase steadily as the exposure
increases from 0 to 1 ML of F atoms, but beyond this exposure,
the yield resulting from exposure to F2 becomes constant
2
2
whereas the yield resulting from exposure to XeF continues to
increase. Because the fluorine saturation coverage as a result
2
22
The crystal is held at 250 K during exposure to a beam of F2
or XeF2. The beam is incident 20° to the normal of the crystal
surface. The crystal is subsequently rotated so that its normal
is aligned with the axis of the differentially pumped mass
spectrometer. The crystal temperature is then increased at a rate
of exposure to F2 is known to be 0.94 ( 0.11 ML, the abscissa
on the left-hand side has been calibrated such that the average
of the total yields of the nine highest F2 exposures (6.5-ML F
atom exposure) is 0.94 ( 0.11 ML. With this calibration, it is
clear that the fluorine coverage resulting from exposure to XeF2
increases beyond the approximately 1 ML saturation coverage
achieved by exposure to F2. This observation implies that the
dissociative chemisorption probability of XeF2 on Si(100)
covered by 1 ML of fluorine is not equal to zero.
+
of 5 K/s, and the masses at m/e ) 66 (SiF2 ) and m/e ) 85
+
+
+
(
SiF3 ) are monitored. The SiF2 and SiF3 signals originate
from the ionization or dissociative ionization of the SiF2 and
SiF4 parent molecules, respectively, in agreement with previous
3
2
results. Figure 1 shows typical thermal desorption spectra
measured after exposure to XeF2, resulting in a variety of
fluorine coverages measured in monolayers, where 1 ML is
equivalent to one F atom per Si surface atom. The fluorine
coverage determination is described below. The major product,
SiF2, is observed as a single feature around 800 K. The SiF2
desorption rate exhibits second-order kinetics as the coverage
increases to 0.30 ML. Above 0.30-ML F atom coverage, the
SiF2 desorption rate is zeroth order. Desorption of the minor
product SiF4 is observed as two broad features around 500 and
There is a second important difference between the interaction
of XeF2 and F2 with Si(100) at 250 K. For sufficiently long
exposures to XeF2, the fluorine that adsorbs does not have an
infinitely long residence time at 250 K. Instead, the fluorine
desorbs as the volatile etch product, SiF4. This desorption is
seen in Figure 3a, which is a plot of the signal at m/e ) 85
+
(SiF3 ), corresponding to SiF4, as a function of exposure to
XeF2. The XeF2 beam is incident 20° from the normal whereas
the mass spectrometer detector is positioned 15° from the
normal. The ordinate in Figure 3a is converted from exposure
to coverage by means of a calibration similar to that shown in
Figure 2. The resulting plot of the etch product as a function of
fluorine coverage is shown in Figure 3b. For coverages lower
than 0.7 ML, negligible etch product is formed, whereas between
0.8 and 1.0 ML, the amount of desorbing SiF4 grows rapidly.
Beyond 1.0 ML, the observed etch product remains essentially
constant. Observation of this etch product is in striking contrast
700 K in Figure 1c. These desorption products and kinetics are
essentially identical to those that have been previously reported
for the interaction of F2 with Si(100).24 The thermal desorption
traces for the F2 system are not shown here. However, unlike
the behavior observed in the F2 system, desorption of SiF2 and
SiF4 are observed for coverages greater than 1 ML, as shown
in Figures 1a and c, because, as explained below, the dissociative

8402 J. Phys. Chem. B, Vol. 106, No. 33, 2002
Holt et al.
Figure 4. Dissociative chemisorption probability (0.5P
calculated from the derivative of the exponential fits in Figure 2. The
thick and thin lines show the derivative of the exponential fits with
1
2
+ P )
respect to F
solid triangles show 0.5P
0.8 ML XeF exposure produces the SiF
4
2
and XeF
2
exposure, respectively, plotted vs coverage. The
+ P adapted from ref 22. Note that above
etch product, so the
1
2
∼
2
probability cannot be determined by this method. Error bars are 95%
confidence limits and are the propagated errors resulting from the
uncertainties of the fit in Figure 2 of the beam flux and of the saturation
coverage given in ref 22.
+
Figure 3. SiF
3
(m/e ) 85) signal (average of data from five
experiments) as a function of (a) exposure and (b) coverage during
exposure to XeF . The XeF beam is incident at θ ) 20°, and the
detector is at θ ) 15° with respect to the normal. The line marks the
onset of etching.
2
2
i
d
detailed explanation of the definition of the dissociative chemi-
sorption probability is given in the discussion of eq IV.11 in
ref 22. The definition of the dissociative chemisorption prob-
ability is not of consequence in the present work. It is only of
consequence when comparing its value to literature values of
to the interaction of F2 with Si(100), a system in which no
2
2
etching is observed when F2 is incident on Si(100) at 250 K.
The dissociative chemisorption probability of XeF2 and F2
as a function of fluorine coverage is calculated from the
derivatives of the plots in Figure 2. An exponential function
given by
33,34
this quantity.
Our previous determination of the quantity 0.5P1 + P2 by a
22
scattering method for the interaction of F2 with Si(100) has
been included for comparison in Figure 4. There is good
agreement between the F2 dissociative chemisorption probability
that is measured as described here and the measurements made
by a scattering method. The F2 dissociative chemisorption
probability decreases rapidly with increasing coverage, ap-
proaching zero at 0.94 ( 0.11 ML, the saturation coverage. In
contrast, the XeF2 probability is weakly dependent on coverage,
decreasing to only about 0.6 at 0.8 ML coverage, the coverage
at which etching begins to occur.
-0.885ꢀ
coverage [ML] ) 0.950 - 0.970e
for F2 (1)
-0.362ꢀ
coverage [ML] ) 2.26 - 2.28e
for XeF2 (2)
is fit to the measurements of coverage as a function of XeF2 or
F2 exposure in ML F atoms, ꢀ, in Figure 2. The derivative of
these functions with respect to exposure yields the dissociative
chemisorption probability as a function of exposure or coverage.
The dissociative chemisorption probability as a function of
coverage is plotted in Figure 4. Dissociative chemisorption
probabilities for XeF2 are calculated only for coverages below
about 0.8 ML because probabilities determined by this thermal
desorption method are valid only when the residence time of
the adsorbate is effectively infinite. As discussed above with
reference to Figure 3, during exposure to XeF2, the fluorine
begins to desorb as SiF4 as the fluorine coverage builds to 0.8
ML. More precisely, the probability plotted in Figure 4
represents the sum 0.5P1 + P2, where P1 is the probability for
single-atom abstraction and P2 is the probability for two-atom
adsorption. Single-atom abstraction refers to the abstraction of
a single F atom from an incoming F2 or XeF2 molecule by a Si
dangling bond, yielding the complementary fragments F or XeF,
respectively, scattered into the gas phase. Two-atom adsorption
refers to the abstraction of one F atom from the incident F2 or
XeF2 followed by the adsorption of the complementary F atom
in the case of F2 or the abstraction of the second F atom from
the complementary XeF fragment in the case of XeF2. A more
B. Surface Structure Determination by He Atom Diffrac-
tion. Previous reports have detailed the He atom diffraction
35
technique with regard to the reconstructed Si(100) surface and,
specifically, its application to the fluorinated surface.24 Si(100)
reconstructs by forming rows of surface Si dimers, resulting in
one partially filled molecular orbital or dangling bond projecting
into the vacuum for each surface Si atom. The distance between
equivalent Si atoms in adjacent dimers in the same row is the
same as the lattice spacing, whereas the distance between
equivalent Si atoms in dimers from adjacent rows is twice the
lattice spacing, resulting in a (2 × 1) surface unit cell that is
observable by He diffraction.
The characteristic He diffraction spectrum arising from this
reconstructed surface is shown in Figure 5a as a function of
angle from the surface normal. All He diffraction spectra are
measured from a surface at 250 K with the He beam incident
40° from the surface normal. The scattered He signal is
presented in Figure 5a with no background subtraction. The
resulting spectrum consists of specular, half-order, and first-
order diffraction features, as indicated in Figure 5a. The broad

XeF2 and F2 with Si(100)
J. Phys. Chem. B, Vol. 106, No. 33, 2002 8403
Figure 5. Helium signal scattered from Si(100) at 250 K and at θ
i
)
4
0° as a function of the detector angle θ
d
: (a) clean surface; (b) 0.9
ML coverage resulting from F
2
exposure; and (c) 0.9 ML coverage
2
resulting from XeF exposure. Each point represents a measurement
made with a dwell time of 2 s and at angular intervals of 0.25°. Error
bars on the He atom count rate are smaller than the width of each point.
width of the diffraction features is a consequence of the large
acceptance angle of the detector, which was chosen to optimize
the intensity of the reactively scattered signal rather than the
angular resolution of the elastically scattered signal. The widths
of these features can be reproduced well in a simulation that
convolutes the finite size of the incident beam and the detector
chamber entrance slits with the distribution of velocities of the
Figure 6. Intensity of the specular He diffraction feature as a function
of (a) exposure and (b) coverage during exposure to F (b) and XeF
2 2
(O). The He beam is incident at θ ) 40°, and the F or XeF beam is
i
2
2
incident at θ
i
) 20°. The detector is positioned at θ
d
) 40°. Signals
3
6
have been normalized at the maximum intensity of each trace. Error
bars on the He atom count rate are smaller than the width of each point.
The line marks the coverage at which the scattered He signal is a
maximum.
incident beam. The intensity of the specular feature is quite
sensitive to disorder produced by any kind of defect and is used
in this study as a measure of the overall surface order. The half-
order diffraction feature arises from the periodic doubling of
the lattice spacing in the direction perpendicular to the dimer
rows. It is therefore a signature for the presence of Si dimers
and is used in this study as an indicator of the cleavage of Si-
Si σ dimer bonds upon the adsorption of fluorine. The first-
order feature arises from diffraction parallel to the dimer rows.
The Si dangling bonds, which are effectively radical sites
and hence very reactive species, are the logical sites for F atom
reaction. It has been reported that F2 reacts solely at the dangling
bonds on the Si(100) surface and that no Si-Si lattice bonds
or σ dimer bonds are broken.24 Shown in Figure 5b is a He
diffraction spectrum of the Si surface after sufficient exposure
to F2 to saturate the dangling bonds. Although the intensities
of the features are changed upon fluorination from that of a
clean surface, the (2 × 1) periodicity persists. The reaction
produces a fully fluorinated and well-ordered Si(100)(2 × 1)
surface where no Si-Si bonds have been broken, the Si dimers
remain intact, and each surface Si atom is decorated with a single
fluorine atom. F2 does not etch the Si surface.2
decorated with a single fluorine atom. Despite its superior
etching ability and large reaction exothermicity, XeF2 has not
induced any significant disorder of the surface, even though
XeF2 has reacted sufficiently to cover the surface with 0.9-ML
F atoms.
Whereas this experiment provides a snapshot of the fluorine
overlayer at a single coverage, it is informative to probe the
surface periodicity as the fluorine overlayer evolves. To do so,
the intensities of the He diffraction features are monitored as a
function of F2 or XeF2 exposure. The He probe beam is incident
at 40° with respect to the surface normal, while the F2 or XeF2
reactant beam is simultaneously incident at 20° with respect to
the normal. The differentially pumped mass spectrometer is
positioned 40° from the normal angle in the forward scattering
direction. The intensity of the specular feature is then monitored
by the mass spectrometer as a function of exposure. Figure 6a
shows the evolution of the specular feature with respect to
exposure of Si(100) to both F2 and XeF2. The maximum
intensities of both traces have been normalized. Figure 6b shows
the same evolution of the intensity of the specular feature plotted
with respect to coverage. The coverage at a given exposure is
calculated from expressions such as those given in eqs 1 and 2.
The intensity of the specular feature decays rapidly during the
initial exposure of the surface to both F2 and XeF2. This initial
loss of intensity can be understood in terms
2
A He diffraction spectrum measured after exposure of Si(100)
to 1.5-ML F atom of XeF2, which yields a coverage of 0.9 ML,
is shown in Figure 5c. The diffraction spectrum measured after
this exposure to XeF2 is nearly identical to the spectrum
measured after F2 exposure in Figure 5b! The dimer rows are
intact, as indicated by the persistence of the (2 × 1) periodicity,
and at 0.9 ML coverage, almost every surface Si atom is

8404 J. Phys. Chem. B, Vol. 106, No. 33, 2002
Holt et al.
of fluorination of the surface at random sites. The presence of
a F atom bonded to a dangling bond changes the interaction
potential of the He atom with that surface unit cell, thereby
making it different from the surrounding unit cells that do not
yet include a F atom. The surface order is disrupted, thus
2
4,37
resulting in a loss of coherency of the diffracted He beam.
The specular intensity reaches a minimum at about 0.35 ML
and then begins to recover. This recovery signals the transition
to a new order as the periodicity of the fluorinated surface unit
cells begins to predominate over that of the unfluorinated unit
cells. Eventually, the original (2 × 1) periodicity is recovered,
indicating that most of the dangling bonds are fluorinated and
that most unit cells are fluorinated identically to their neighbors.
The intensity of the specular feature fully recovers at 0.9 ( 0.1
ML coverage, corresponding to F2 and XeF2 exposures of 3-5
and 1.3-ML F atoms, respectively. The uncertainty in the
coverage arises from the propagated uncertainty in the fit of
the thermal desorption signal and the uncertainty of the
Figure 7. Intensity of the +1 order (b) and -1/
order (O) He
2
diffraction features as a function of coverage during exposure to XeF
The He beam is incident at θ ) 40°, the F or XeF beam is incident
at θ ) 20°, and the detector is positioned at θ ) 21.5° (+1 order) or
1.25° (- / order). Signals have been normalized at the maximum
2
.
i
2
2
i
d
2
2
previously reported determination of the saturation coverage.
1
5
2
The recovery of the specular feature at a coverage that is
slightly below 1 ML, where the surface is more fully ordered,
can be understood in terms of the limited transfer width of the
apparatus. As discussed elsewhere in detail,24 the transfer width
of the apparatus is estimated to be 35 Å, which means that
ordered regions of the surface greater than 35 Å in length will
intensity of each trace. Error bars on the He atom count rate are smaller
than the width of each point.
and Si-Si lattice bonds because of continued reaction with
XeF2. Indeed, the dissociative chemisorption probability of XeF2
remains high at 0.9 ML, as shown in Figure 4. In addition, the
onset of desorption of the etch product, SiF4, is observed at
this coverage, as shown in Figure 3. Clearly, in contrast to the
interaction of F2 with 1 ML of F adsorbed on Si(100), XeF2
reacts at this coverage, depositing additional fluorine while
cleaving Si-Si bonds. The additional fluorine and the desorption
of the etch product lead to disordering of the surface, as evident
from He atom diffraction.
3
8
not contribute to the intensity of the diffracted beams. That
is, once the ordered regions of the surface have surpassed this
size, the intensity of the diffracted beam will remain constant
as the coverage increases. However, for microscopic 2D ordering
below the transfer width of the apparatus, the He diffracted
intensities are extremely sensitive to surface disorder. As
apparent in Figure 6b, the He diffraction technique is sensitive
to the surface disorder produced by a change in coverage of as
little as 0.01 ML.
IV. Discussion
Figure 6 shows that in the submonolayer coverage regime,
there is a striking similarity between F2 and XeF2. In fact, up
to 0.9 ML coverage, it would be extremely difficult to
distinguish between the two reactants solely on the basis of the
evolution of the surface order. However, beyond 0.9 ML, the
reactivities differ greatly. The intensity of the specular feature
remains essentially constant upon further exposure to F2 because
the F2 dissociative chemisorption probability becomes effectively
zero, as shown in Figure 4. In contrast, the intensity of the
specular feature decreases rapidly upon further exposure to XeF2,
indicating that the periodicity is rapidly destroyed.
This study has shown that dissociative chemisorption of XeF2
proceeds in a manner identical to that of F2 as the coverage
increases from 0 to about 1 ML on Si(100)(2 × 1). Figure 6b
shows that the intensity of the specular He diffraction feature
as a function of fluorine coverage produced by the dissociative
chemisorption of XeF2 is indistinguishable in this coverage
regime from that produced by the dissociative chemisorption
of F2. Both XeF2 and F2 dissociatively chemisorb initially by
reacting solely with dangling bonds at random sites, leading to
a loss of coherence of the zeroth-order diffraction beam. As
the coverage increases and more unit cells are fluorinated, the
coherence of the specular beam recovers, and its intensity is
maximized at about 0.9 ML for both XeF and F . Diffraction
The half-order and first-order diffraction features were also
monitored as a function of coverage during exposure to XeF2.
The half-order feature is of interest because it arises from the
periodicity of the dimer rows and hence is visible only when
the dimer bonds are intact. It is used in this study as an indicator
of the cleavage of Si-Si σ dimer bonds upon the adsorption of
fluorine. The He probe beam is incident at 40° with respect to
the surface normal, while the F2 or XeF2 reactant beam is
simultaneously incident at 20° with respect to the normal. The
intensities of the first- and half-order features are monitored by
the differentially pumped mass spectrometer positioned 21.5°
and 51.25°, respectively, from the normal in the forward
scattering direction. The evolution of these features, whose
intensities are normalized, as a function of coverage resulting
from XeF2 exposure is shown in Figure 7. Both features show
the characteristic initial decrease caused by random fluorination,
recovery at 0.9 ML, and subsequent decay. The decay of both
features beyond 0.9 ML is evidence for the destruction of the
2
2
spectra measured from a fluorinated surface at 0.9 ML coverage
over a wide angular range are shown in Figures 5b and c for
the F and XeF systems, respectively. The spectra reveal that
2
2
the fluorinated surface is well-ordered with a (2 × 1) periodicity.
The dimer rows remain intact, with each surface Si atom
decorated with a fluorine atom. The only difference between
XeF and F in this coverage regime is that the chemisorption
2
2
probability of XeF decreases slowly as the coverage approaches
2
1 ML whereas that of F decreases rapidly and is effectively
2
zero at 1 ML, as shown in Figure 4. The relative constancy of
a chemisorption probability with coverage is usually interpreted
in terms of a long lifetime of an extrinsic precursor, which is a
3
9
species physisorbed on top of filled sites. An extrinsic
precursor diffuses on top of the filled sites until it encounters
an empty site, such as a dangling bond, and reacts or desorbs
without dissociating. In short, the extrinsic precursor makes it
more likely that a XeF2 molecule will encounter an unoccupied
(2 × 1) periodicity and hence the cleavage of the Si-Si σ dimer

XeF2 and F2 with Si(100)
J. Phys. Chem. B, Vol. 106, No. 33, 2002 8405
dangling bond and hence increases the dissociation probability.11
The substantially greater polarizability of XeF2 compared to that
of F2 suggests that a XeF2 extrinsic precursor would have a
longer lifetime than that of F2. A second explanation for the
relative constancy of the dissociative chemisorption probability
is the longer collision time of XeF2 compared to that of F2.
Because it is 4 times more massive than F2, XeF2 spends twice
as long as F2 in the near-surface region where it may encounter
an unoccupied dangling bond. It is not clear from the present
study if either effect or both effects are operative.
coverage provides evidence that even the Si-Si σ dimer bonds
are undisturbed, preserving the original dimer rows. Up until
0.9 ML, the reaction of XeF2 with Si is remarkably similar to
the reaction of F2, especially considering the fact that their
steady-state etch rates differ by 4 orders of magnitude. However,
after saturation of the dangling bonds, F2 ceases to react with
the surface, but XeF2 continues to deposit fluorine on the surface
by reacting with the Si-Si bonds. The surface order is destroyed
as a result of the continued fluorine deposition, and ultimately,
etching occurs by the formation of volatile SiF4.
Once the dangling bonds are saturated at about 1 ML
coverage, the interactions of XeF2 and F2 differ dramatically.
Whereas the reaction of F2 with Si(100) essentially ceases, XeF2
continues to chemisorb dissociatively, as indicated by the
increase in fluorine coverage beyond 1 ML in Figure 2.
Chemisorption of fluorine beyond 1 ML requires cleavage of
the Si-Si σ dimer bonds, the Si-Si lattice bonds, or both. The
simultaneous decay of both the first- and half-order diffraction
features provides evidence that the cleavage of both types of
Si-Si bonds is occurring. The concomitant cleavage of Si-Si
bonds and formation of Si-F bonds leads to disorder of the
surface periodicity, as indicated by the decay of the specular,
first-, and second-order diffraction features as the coverage
increases beyond 1 ML. The increased fluorine coverage results
in the formation of more highly fluorinated species such as SiF2
and SiF3. These species have been identified in XPS measure-
ments of Si(100) at approximately 1.5 ML coverage.40-42 They
eventually form the SiF4 that desorbs. In short, XeF2 etches the
Si surface, but F2 does not.
Acknowledgment. This work is supported by NSF CHE-
091810. We thank D. L. Lahr for critically reading the
manuscript.
0
References and Notes
(1) Ressejac, I. C.; Landsberger, L. M.; Currie, J. F. J. Vac. Sci.
Technol., A 2000, 18, 746.
(2) Tea, N. H.; Milanovi c´ , V.; Zincke, C. A.; Suehle, J. S.; Gaitan,
M.; Zaghloul, M. E.; Geist, J. J. Microelectromech. Syst. 1997, 6, 363.
(3) Chang, F. I.; Yeh, R.; Lin, G.; Chu, P. B.; Hoffman, E.; Kruglick,
E. J. J.; Pister, K. S. J.; Hecht, M. H. Proc. SPIE 1995 Symp. Microelectronic
Struct. Microelectromech. DeVices 1995, 2641, 117.
(4) Chu, P. B.; Chen, J. T.; Yeh, R.; Lin, G.; Huang, J. C. P.; Warneke,
B. A.; Pister, K. S. J. Transducers 97, Proc. Int. Conf. Solid-State Sens.
Actuators 1997, 97, 665.
(
5) Toda, R.; Minami, K.; Esashi, M. Transducers 97, Proc. Int. Conf.
Solid-State Sens. Actuators 1997, 97, 671. Sens. Actuators, A 1998, 66,
68.
(6) Kovacs, G. T. A.; Maluf, N. I.; Petersen, K. E. Proc. IEEE 1998,
86, 1536.
2
(
(
7) Winters, H. F.; Coburn, J. W. Appl. Phys. Lett. 1979, 34, 70.
8) Mucha, J. A.; Donnelly, V. M.; Flamm, D. L.; Webb, L. M. J.
Clearly, the dangling-bond sites on the Si(100) surface are
by far the dominant sites for the dissociative chemisorption of
Phys. Chem. 1981, 85, 3529.
9) Flamm, D. L.; Donnelly, V. M.; Mucha, J. A. J. Appl. Phys. 1981,
52, 3633.
(
2
7,43,44
XeF2, just as they are for F2.
Previous work has
(
(
10) Vasile, M. J. J. Appl. Phys. 1983, 54, 6697.
11) Ibbotson, D. E.; Flamm, D. L.; Mucha, J. A.; Donnelly, V. M. Appl.
demonstrated that the mechanism for dissociative chemisorption
of F2 is atom abstraction. The cross section for F atom
abstraction from F2 by the dangling bonds is measured to be
Phys. Lett. 1984, 44, 1129.
(
(
23.
12) Winters, H. F.; Coburn, J. W. Surf. Sci. Rep. 1992, 14, 165.
13) Tu, Y. Y.; Chuang, T. J.; Winters, H. F. Phys. ReV. B 1981, 23,
2
3
large compared to the cross sectional area of a surface site.
8
The large cross sections suggest the presence of an attractive
(14) Haring, R. A.; Haring, A.; Saris, F. W.; de Vries, A. E. Appl. Phys.
interaction potential that is consistent with the notion of
Lett. 1982, 41, 174.
molecular steering.45,46 Recent work has also provided direct
(15) Vugts, M. J. M.; Hermans, L. J. F.; Beijerinck, H. C. W. J. Vac.
Sci. Technol., A 1996, 14, 2138. Vugts, M. J. M.; Hermans, L. J. F.;
Beijerinck, H. C. W. J. Vac. Sci. Technol., A 1996, 14, 2820.
(16) Chuang, T. J. J. Chem. Phys. 1981, 74, 1461.
4
7
evidence for atom abstraction in the XeF2/Si(100) system.
Upon saturation of the dangling bonds, the only sites available
for reaction are ones that involve cleavage of Si-Si bonds. An
F2 molecule incident at thermal energies does not react with
these sites. In fact, it has been shown that a barrier to that
reaction of 3.8 kcal/mol exists.24 This barrier can be surmounted
by translational activation of F2. In contrast, it is apparent that
no such barrier exists in the case of XeF2. The XeF2 molecule
that is incident at thermal energies reacts readily with these sites.
(
17) Houle, F. A. J. Chem. Phys. 1983, 79, 4237. Houle, F. A. J. Chem.
Phys. 1984, 80, 4851.
18) Houle, F. A. Appl. Phys. Lett. 1987, 50, 1838. Houle, F. A. Phys.
(
ReV. Lett. 1988, 61, 1871. Houle, F. A. Phys. ReV. B 1989, 39, 10120.
(19) Li, B.; Streller, U.; Krause, H. P.; Twesten, I.; Schwentner, N. J.
Appl. Phys. 1995, 77, 350.
(
20) Aliev, V. S.; Kruchinin, V. N. Surf. Sci. 1999, 442, 206.
(21) Li, Y. L.; Pullman, D. P.; Yang, J. J.; Tsekouras, A. A.; Gosalvez,
D. B.; Laughlin, K. B.; Zhang, Z.; Schulberg, M. T.; Gladstone, D. J.; Ceyer,
S. T. Phys. ReV. Lett. 1995, 74, 2603. Ceyer, S. T. Proc. R. A. Welch Found.
Conf. Chem. Res. XXXVIII: Chemical Dynamics of Transient Species;
Welch Foundation: Houston, TX, 1994; pp 156-172. Pullman D. P.; Ceyer,
S. T. Abstr. Pap. Am. Chem. Soc. 1994, 207, 168.
An important point demonstrated by this study is that the
exothermicity released to the surface upon dissociative chemi-
sorption of XeF2 or F2 below about 1 ML coverage does not
(22) Tate, M. R.; Gosalvez-Blanco, D.; Pullman, D. P.; Tsekouras, A.
destroy the surface order, in contrast to the conclusions given
A.; Li, Y. L.; Yang, J. J.; Laughlin, K. B.; Eckman, S. C.; Bertino, M. F.;
Ceyer, S. T. J. Chem. Phys. 1999, 111, 3679.
in previous work.2
5-28
It is also clear that the presence of
disorder is not a precondition for etching to begin. Clearly, XeF2
begins to react with the Si-Si bonds in the presence of an
ordered overlayer of fluorine. It is the reaction of XeF2 with
these Si-Si bonds that leads to the disorder of the surface
periodicity rather than the disorder induced by the exothermicity
release that leads to the reaction of XeF2 with the Si-Si bonds.
(23) Tate, M. R.; Pullman, D. P.; Gosalvez-Blanco, D.; Tsekouras, A.
A.; Li, Y. L.; Ceyer, S. T. J. Chem. Phys. 2000, 112, 5190.
(24) Pullman, D. P.; Tsekouras, A. A.; Li, Y. L.; Yang, J. J.; Tate, M.
R.; Gosalvez-Blanco, D.; Laughlin, K. B.; Schulberg, M. T.; Ceyer, S. T.
J. Phys. Chem. B 2001, 105, 486.
(25) Lo, C. W.; Varekamp, P. R.; Shuh, D. K.; Durbin, T. D.; Chakarian,
V.; Yarmoff, J. A. Surf. Sci. 1993, 292, 171.
(26) Simpson, W. C.; Yarmoff, J. A. Surf. Sci. 1996, 359, 135.
We have shown that the reaction of low-energy XeF2 with
the Si(100)(2 × 1) reconstructed surface occurs solely at the Si
dangling bonds up to a fluorine coverage of 0.9 ML. No Si-Si
bonds are broken, and the ordered (2 × 1) reconstruction is
preserved. The presence of the half-order feature at 0.9 ML
(27) Weakliem, P. C.; Carter, E. A. J. Chem. Phys. 1993, 98, 737.
(28) Carter, L. E.; Khodabandeh, S.; Weakliem, P. C.; Carter, E. A. J.
Chem. Phys. 1994, 100, 2277.
(29) Ceyer, S. T.; Gladstone, D. J.; McGonigal, M.; Schulberg, M. T.
In Physical Methods of Chemistry, 2nd ed.; Rossiter, B. W., Baetzold, R.
C., Eds., Wiley & Sons: New York, 1993; Vol. IXA, p 383.

8406 J. Phys. Chem. B, Vol. 106, No. 33, 2002
Holt et al.
(
(
(
(
30) Qiu, S. R.; Yarmoff, J. A. Phys. ReV. B 2001, 63, 115409.
31) Ishizaka, A.; Shiraki, Y. J. Electrochem. Soc. 1986, 133, 666.
32) Winters, H. F.; Houle, F. A. J. Appl. Phys. 1983, 54, 1218.
33) Engstrom, J. R.; Nelson, M. M.; Engel, T. Surf. Sci. 1989, 215,
(39) Kisliuk, P. J. Phys. Chem. Solids 1957, 3, 95.
(40) Chuang, T. J. J. Appl. Phys. 1980, 51, 2614.
(41) McFeely, F. R.; Morar, J. F.; Shinn, N. D.; Landgren, G.; Himpsel,
F. J. Phys. ReV. B 1984, 30, 764.
(42) Shinn, N. D.; Morar, J. F.; McFeely, F. R. J. Vac. Sci. Technol. A
1984, 2, 1593.
(43) Stillinger, F. H.; Weber, T. A. Phys. ReV. Lett. 1989, 62, 2144. J.
Chem. Phys. 1990, 92, 6239.
(44) Schoolcraft, T. A.; Garrison, B. J. J. Vac. Sci. Technol., A 1990, 8,
3496.
(45) Gostein, M.; Sitz, G. O. J. Chem. Phys. 1997, 106, 7378.
(46) Darling, G. R.; Kay, M.; Holloway, S. Surf. Sci. 1998, 400, 314.
(47) Ceyer, S. T. To be submitted for publication.
4
37.
34) Behringer, E. R.; Flaum, H. C.; Sullivan, D. J.; Masson, D. P.;
Lanzendorf, E. J.; Kummel, A. C. J. Phys. Chem. 1995, 99, 12863.
(
(
35) Cardillo, M. J.; Becker, G. E. Phys. ReV. B 1980, 21, 1497.
(36) Yang, J. J. Ph.D. Thesis, Massachusetts Institute of Technology,
Cambridge, MA, 1997.
37) Poelsema, B.; Comsa, G. Scattering of Thermal Energy Atoms from
Disordered Surfaces; Springer-Verlag: Berlin, 1989.
38) Comsa, G. Surf. Sci. 1979, 81, 57.
(
(