Phosphine and tertiarybutylphosphine adsorption on the indium-rich InP (0 0 1)-(2 × 4) surface

Article abstract of DOI:10.1016/j.susc.2006.08.014

Phosphine and tertiarybutylphosphine adsorption on the indium-rich InP (0 0 1)-(2 × 4) surface at 25 °C have been studied by internal reflection infrared spectroscopy, X-ray photoelectron spectroscopy, and low energy electron diffraction. Both molecules form a dative bond to the empty dangling bonds on the In-P heterodimers and the second-layer In-In dimers and vibrate symmetrically at 2319 (2315) and 2285 (2281) cm^-1 and asymmetrically at 2339 (2339) and 2327 (2323) cm^-1. A fraction of these species dissociate into adsorbed PH₂ with the hydrogen and tertiarybutyl ligands transferring to nearby phosphorus sites. The calculated energy barriers for desorption (<11 kcal/mol) of these molecules is less than that for dissociation (>17 kcal/mol) and explains their low sticking probabilities at elevated temperatures under InP growth conditions.

Full text of DOI:10.1016/j.susc.2006.08.014

Surface Science 600 (2006) 4888–4895
www.elsevier.com/locate/susc
Phosphine and tertiarybutylphosphine adsorption
on the indium-rich InP (001)-(2 · 4) surface
a
b
a
a
b
a,
*
R.L. Woo , U. Das , S.F. Cheng , G. Chen , K. Raghavachari , R.F. Hicks
a
Department of Chemical and Biomolecular Engineering, University of California, Los Angeles, CA 90095, United States
b
Department of Chemistry, Indiana University, Bloomington, IN 47405, United States
Received 10 February 2006; accepted for publication 11 August 2006
Available online 1 September 2006
Abstract
Phosphine and tertiarybutylphosphine adsorption on the indium-rich InP (001)-(2 · 4) surface at 25 ꢀC have been studied by internal
reflection infrared spectroscopy, X-ray photoelectron spectroscopy, and low energy electron diffraction. Both molecules form a dative
bond to the empty dangling bonds on the In–P heterodimers and the second-layer In–In dimers and vibrate symmetrically at 2319
ꢀ1
ꢀ1
(
2315) and 2285 (2281) cm and asymmetrically at 2339 (2339) and 2327 (2323) cm . A fraction of these species dissociate into
adsorbed PH with the hydrogen and tertiarybutyl ligands transferring to nearby phosphorus sites. The calculated energy barriers for
2
desorption (<11 kcal/mol) of these molecules is less than that for dissociation (>17 kcal/mol) and explains their low sticking probabilities
at elevated temperatures under InP growth conditions.
ꢁ
2006 Elsevier B.V. All rights reserved.
Keywords: Indium phosphide (InP); Phosphine; Tertiarybutylphosphine; Infrared spectroscopy; Cluster calculations; Density functional theory
1
. Introduction
Phosphine (PH ) and tertiarybutylphosphine (TBP) sup-
3
ply the group V sources to the semiconductor crystal dur-
ing MOCVD growth [8–12]. Using in situ monitoring of
the phosphorus coverage by reflectance difference spectros-
copy (RDS), Sun et al. [13,14] measured the reaction kinet-
Metalorganic chemical vapor deposition (MOCVD) is
one of the essential technologies for manufacturing opto-
electronic devices such as light emitting diodes, solid-state
lasers, high efficiency multijunction solar cells, and hetero-
junction bipolar transistors [1–5]. The performance of these
devices is affected in a major way by the deposition process.
For example, the emission intensity of long wavelength
InGaAsP/InP lasers depends highly on the interface
abruptness during epitaxial growth [6]. The most important
factors influencing this abruptness is the mass transport in
the boundary layer coupled with the chemical reactions
taking place on the film surface [6,7]. A better understand-
ing of the heterogeneous reaction chemistry could lead to
further improvements in the MOCVD process.
ics of PH and TBP with the InP (001) surface. They
3
observed that the sticking probability of TBP decreases
from 0.007 to 0.001 with increasing temperature from 420
to 520 ꢀC; whereas the sticking probability of PH3 is
approximately equal to 0.001 over the same range. Based
on the kinetic measurements, Sun proposed the following
mechanism for the decomposition of the group V sources
on indium phosphide (using TBP as an example):
ꢁ
ꢁ
ꢁ
ꢁ
In þ TBPðgÞ $ InTBP
ð1Þ
ð2Þ
ð3Þ
ꢁ
ꢁ
ꢁ
þ
InTBP !
L þ InPL
P
P
2
ꢁ
InPL
2
!
P
þ products
*
where L is an alkyl group or a hydrogen atom. Initially, the
TBP molecule is believed to form a dative bond with the
empty dangling orbital on an exposed indium atom. This
Corresponding author. Tel.: +1 310 206 6865; fax: +1 310 206 4107.
E-mail addresses: kraghava@indiana.edu (K. Raghavachari), rhicks@
ucla.edu (R.F. Hicks).
0
039-6028/$ - see front matter ꢁ 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2006.08.014

R.L. Woo et al. / Surface Science 600 (2006) 4888–4895
4889
reaction is reversible. Then, TBP undergoes dissociative
adsorption and decomposes to form phosphorus dimers.
Sun and coworkers did not explain why the sticking prob-
ability of the group V precursors is so low on InP (001). In
addition, the datively bonded state has not been observed
experimentally.
chamber were turned off in order to avoid exposing the sur-
face to dissociated radicals.
The calculations were performed using density func-
tional theory (DFT) with the Becke three-parameter ex-
change functional and the Lee–Yang–Parr correlation
functional (B3LYP). We chose the (18s/14p/9d)/[6s/5p/
3d] contracted basis set for indium, the Dunning–Huzinaga
In this report, we present a study of phosphine and ter-
tiarybutylphosphine adsorption on the indium-rich InP
*
*
(11s/7p/1d)/[6s/4p/1d] contracted basis set (D95 ) for
phosphorus, the D95** polarized double-f basis set for
the surface hydrogen atoms and the D95 double-f basis
set for the terminating hydrogen atoms. In previous work
on hydrogen adsorption on P-rich indium phosphide sur-
face, Fu et al. have successfully applied a single frequency
shift for each mode to map the calculated vibrational fre-
quencies onto those observed by experiment [15]. We have
used the same approach here: the predicted P–H stretching
(
001)-(2 · 4) surface. Using vibrational spectroscopy, we
have identified the reaction intermediates, and have shown
that the datively bonded molecule plays a significant role in
the adsorption process.
2
. Experimental and theoretical methods
ꢀ
1
An indium phosphide film, 35 nm thick, was grown on
frequencies have been uniformly shifted down by 110 cm
an InP (001) substrate by MOCVD. The following
to correct for systematic errors in the calculations due to
deficiencies in the cluster models, neglect of anharmonicity,
and so forth.
conditions were used during growth: 530 ꢀC, 60 Torr of
ꢀ
4
hydrogen, 6.7 · 10 Torr of trimethylindium (TMIn),
ꢀ
3
7
.9 · 10 Torr of TBP (V/III ratio of 117), and a total
flow rate of 33 L/min (at 25 ꢀC and 760 Torr). Once the
growth was completed, the TBP and the hydrogen flows
were maintained until the sample was cooled down to
3
. Results
3.1. Phosphine adsorption
3
00 and 24 ꢀC, respectively. Then, the sample was trans-
ferred to an ultrahigh vacuum chamber without air expo-
Shown in Table 1 are the atomic percentages of indium
sure. In vacuum, the sample was annealed at 500 ꢀC for
and phosphorous measured by XPS on the clean and cov-
ered InP (001) surfaces. Since the InP (001)-(2 · 1) and
InP (001)-d(2 · 4) surfaces have known phosphorus cover-
ages of 1.00 and 0.125 [16], we used the atomic % of In and
P on the clean surfaces to estimate the coverages on those
1
5 min to obtain the (2 · 4) reconstruction.
After cooling the samples to 25 ꢀC, the surface structure
and composition were characterized by low energy electron
diffraction (LEED) and X-ray photoelectron spectroscopy
(
XPS). Core level photoemission spectra of the P 2p and
dosed with PH and TBP. Phosphine adsorption onto the
3
In 3d lines were collected with a PHI 3057 spectrometer
using magnesium Ka X-rays (hm = 1286.6 eV). All XPS
spectra were taken in small area mode with a 7ꢀ acceptance
angle and 23.5 eV pass energy. The detection angle with re-
spect to the surface normal was 25ꢀ. The P and In atom%
were determined from the integrated intensity of the P 2p
and In 3d photoemission peaks, dividing by their sensitivity
factors, 0.49 and 4.36, respectively. The P/In ratio was ob-
tained by dividing the P atom% by the In atom%.
InP (001)-(2 · 4) surface at 25 ꢀC caused the P/In ratio
measured by XPS to increase from 0.85 to 1.20. This sug-
gests that the surface contains over 2.0 monolayers of ad-
sorbed phosphorus. At the same time, the (2 · 4) LEED
pattern gradually converted to a (1 · 1), indicating that
the phosphine molecules reacted with the exposed In–P
and In–In dimers and disrupted the local ordering.
Shown in Fig. 1 are a series of infrared reflectance spec-
tra for different phosphine dosages on the (2 · 4) at 25 ꢀC
The infrared spectra were recorded by multiple internal
ꢀ6
(
1 L = 1 · 10 Torr s). In these spectra, there is a set of
reflections through a trapezoidal InP crystal, 40.0 ·
ꢀ1
sharp peaks between 2350 and 2200 cm
broad bands between 1750 and 1000 cm . The higher fre-
quency peaks are due to P–H stretching modes, while the
and several
3
1
[
0.0 · 0.64 mm . The long crystal axis was parallel to the
ꢀ1
110] direction. This crystal provided for 31 reflections
off the front face, which significantly enhanced the signal-
to-noise ratio. The group V molecules were introduced into
ꢀ
6
Table 1
the UHV chamber at 5.0 · 10 Torr through a leak valve.
Dosing was continued for up to 45 min to ensure that the
InP (001)-(2 · 4) surface was completely saturated with
The percentages of In and P atoms observed by XPS on the clean, PH
3
and TBP covered InP surfaces
a
Surface
In % P %
P/In
p
h (ML)
TBP or PH . A series of infrared spectra was collected be-
3
InP (001)-(2 · 1)
InP (001)-(2 · 4)
49.9
54.0
45.3
48.7
50.0
55.1
50.2
46.0
54.7
51.3
50.0
44.9
1.00
0.85
1.20
1.05
1.00
0.81
1.00
0.13
2.20
1.30
1.00
0.13
fore and during TBP and PH adsorption. These spectra
3
ꢀ
1
were recorded by taking 1024 scans at 8 cm resolution.
The spectra presented in this paper show the change in
reflectance (per reflection) that results from taking the ratio
of the sample spectrum after dosing to that of the clean sur-
face. During these experiments, all of the filaments in the
3
PH : InP (001)-(2 · 4) @ 25 ꢀC
TBP: InP (001)-(2 · 4) @ 25 ꢀC
PH
PH
a
3
: InP (001)-(2 · 4) @ 180 ꢀC
: InP (001)-(2 · 4) @ 270 ꢀC
3
Based on linear extrapolation.

4890
R.L. Woo et al. / Surface Science 600 (2006) 4888–4895
P-H
2285
Dose
2319
In-Ht,b
o
2
8600 L
25 C
1
8600 L
2600 L
900 L
500 L
o
1
80 C
1
Difference
6
1
4000
3000
2000
1000
2500
2375
2250
2125
2000
-
1
-1
Wavenumber (cm )
Wavenumber (cm )
Fig. 1. Infrared reflectance spectra of adsorbed phosphine on InP (001)-
2 · 4) at 24 ꢀC and different phosphine dosages.
Fig. 3. Infrared reflectance spectra of adsorbed phosphine on the InP
(001)-(2 · 4) surface at different temperatures.
(
Presented in Fig. 3 are infrared reflectance spectra taken
after dosing 20,000 L of phosphine at 25 and 180 ꢀC, and
their difference in peak intensity. Note that larger dosages
do not change the spectra. Inspection of the figure reveals
that the intensities of the P–H bands are reduced at ele-
vated temperature. Among these, the peaks at 2319 and
2
285
2
319
301
2
262
2
ꢀ1
2285 cm have decreased by the largest amount, indicat-
2
242
226
219
ing that the adsorbed species corresponding to these peaks
are the most sensitive to changes in temperature. No
absorption bands were observed at 270 ꢀC as expected.
2
339
327
2
2
2
3.2. Tertiarybutylphosphine adsorption
2
400 2350 2300 2250 2200 2150
Upon dosing indium phosphide with 25,000 L tertia-
-1
Wavenumber (cm )
rybutylphosphine at 24 ꢀC, the (2 · 4) reconstruction was
replaced by the (1 · 1), and the P/In ratio obtained from
the XPS spectra increased from 0.85 to 1.05. The latter
P/In ratio yields an estimated phosphorus coverage of
Fig. 2. Infrared reflectance spectra of adsorbed phosphine on InP (001)-
2 · 4) showing the deconvolution of the bands.
(
1
.3 ML. Shown in Fig. 4 are a series of infrared spectra
bands at lower frequency result from indium hydrides [15–
9]. The intensity of the In–H vibrational modes is small
taken at different dosages of TBP at room temperature.
In these spectra, there are five main sets of bands appearing
1
compared to the P–H peaks. Shown in Fig. 2 is the decon-
voluted infrared spectrum of the phosphorus–hydrogen
stretching modes at saturation coverage. Nine overlapping
peaks are observed at 2339, 2327, 2319, 2301, 2285,
C-H
-
CH3
ꢀ
1
Dose
P-H
C-C
2
262, 2242, 2226, and 2219 cm . The full-width-at-half
2
8200 L
maximum (FWHM) of these bands varies from 9.0 to
ꢀ1
1
8.0 cm
.
1
8600 L
Phosphine adsorption experiments were also performed
at 180 and 270 ꢀC. In the former case, the (2 · 4) LEED
pattern gradually converted to a weak (1 · 2) whereas in
the latter case, the (2 · 4) pattern remained but the spot
intensity became weaker. The P/In ratio calculated from
the XPS spectra was 1.00 and 0.81 at 180 and 270 ꢀC,
respectively. As shown in Table 1, these P/In ratios indicate
that the phosphorus coverage is 1.00 ML at 180 ꢀC and
12600 L
6
1
900 L
500 L
3
500 3000 2500 2000 1500 1000
–
1
Wavenumber (cm )
0
.13 ML at 270 ꢀC. Evidently, phosphine does not stick
to the surface at 270 ꢀC and a PH partial pressure of
3
Fig. 4. Infrared reflectance spectra of adsorbed tertiarybutylphosphine on
InP (001)-(2 · 4) at 24 ꢀC and different TBP dosages.
ꢀ
6
5
.0 · 10 Torr.

R.L. Woo et al. / Surface Science 600 (2006) 4888–4895
4891
2
281
2
261
242
226
212
2
2
301
2
2
2
315
2
323
2339
2
400 2350 2300 2250 2200 2150
-
1
Wavenumber (cm )
Fig. 5. Infrared reflectance spectra of adsorbed tertiarybutylphosphine on
InP (001)-(2 · 4) showing the deconvolution of the bands.
ꢀ1
in the following frequency ranges: 3000 to 2800 cm , 2350
ꢀ
1
ꢀ1
ꢀ1
to 2200 cm , 1480 to 1435 cm , 1385 to 1350 cm , and
ꢀ1
1
230 to 1000 cm . The peaks in the frequency range be-
tween 2350 and 2200 cm
modes [15–19]. The bands between 3000 and 2800 cm re-
ꢀ
1
are due to P–H stretching
ꢀ
1
sult from CH stretching modes, while those between 1480
3
ꢀ
1
ꢀ1
and 1435 cm , and 1385 and 1350 cm are due to CH3
bending modes [20–22]. Finally, the peaks between 1230
ꢀ
1
and 1000 cm are assigned to C–C skeletal vibrations.
Shown in Fig. 5 is the deconvoluted infrared spectrum
of the P–H stretching modes obtained at saturation cover-
age. Individual peaks are observed at 2339, 2323, 2315,
Fig. 6. (a) Filled-states STM image of the InP (001)-(2 · 4) reconstruction
˚
2
˚
2
ꢀ
1
(440 · 440 A ; (inset) 40 · 40 A ) and (b) ball-and-stick model [16,17].
2
301, 2281, 2261, 2242, 2226, and 2212 cm . Their
ꢀ
1
FWHM varies from 7.0 to 16.0 cm . The number of peaks
seen for TBP adsorption is the same as that seen for PH3
adsorption (refer to Fig. 2). Moreover, the position and
relative intensities of the peaks recorded for TBP are only
slightly different from those seen for PH3.
tion. In this reconstruction, the empty dangling bonds
associated with indium on the In–P heterodimer and the
second-layer In–In dimers are available for chemical bond-
ing [16,24].
In order to assign the vibrational modes to specific bond
configurations, molecular cluster calculations using the
GAUSSIAN 03 quantum chemistry suite have been per-
formed. To mimic the (2 · 4) reconstruction, cluster 1
was assembled as shown in Fig. 7. It contains one In–P
mixed dimer connected to two indium dimers in the next
layer. The two In–In dimers are tethered to eight ‘‘bulk’’
P atoms and three ‘‘bulk’’ In atoms. The cluster termina-
tions have been carefully chosen to achieve a proper bal-
ance between covalent and dative bonding appropriate
for compound semiconductors [18,26]. Note that cluster 1
represents roughly one half of the (2 · 4) unit cell and is
of sufficient size to investigate the adsorption mechanisms
4
. Discussion
4
.1. Peak assignments
To interpret the vibrational spectra, we must understand
the atomic structure of the (2 · 4) reconstruction. A filled-
states scanning tunneling micrograph of the In-rich (2 · 4)
surface is shown in Fig. 6 [16,17]. In this reconstruction,
uniform gray rows extend along the [ꢀ110] direction,
and the step edges lie straight along the rows. In the inset
picture, a close up of the (2 · 4) unit cell is highlighted with
a white rectangle. Within each cell, one can see three bright
spots forming an equilateral triangle. These spots result
from the filled phosphorus dangling bond and the two
In–In back-bonds of each In–P dimer [17]. A ball-and-stick
model of the (2 · 4) is presented in Fig. 6b. A single In–P
heterodimer straddles four In dimers located in the next
lower layer, with a phosphorus coverage of 0.125 monolay-
ers (ML). In addition, each set of four indium dimers is
separated by a trench that is parallel to the [ꢀ110] direc-
of PH and TBP.
3
To simulate phosphine adsorption on the indium-rich
(2 · 4) reconstruction, phosphine was added to cluster 1.
Ab initio calculations were performed to determine the
optimum configurations and the optimized molecular clus-
ters, 2, 3, 4, and 5, are presented in Fig. 7. Cluster 2 reveals
the initial stage of phosphine adsorption onto the empty In
dangling bond in the heterodimer. In this case, a lone pair

4892
R.L. Woo et al. / Surface Science 600 (2006) 4888–4895
Fig. 7. Ball-and-stick models of the optimized clusters illustrating the different ways phosphine molecule coordinate to adsorption sites and surface species
during dissociative adsorption of phosphine. The In, P, and H atoms are the large red, large green, and small white spheres, respectively.
of electrons from the phosphine molecule is donated to
form a weak dative bond between phosphine and the in-
dium atom. In cluster 3, the phosphine molecule has disso-
ciatively adsorbed onto the InP heterodimer, forming
phosphorus dihydrogen and monohydrogen species. Note
layer. This is illustrated by cluster 4. Two phosphine mol-
ecules are shown datively bonding to the indium atoms
(second dashed line is omitted for ease of viewing). The da-
tively bonded phosphine molecules in cluster 4 may disso-
ciate into phosphorus dihydrogen species and transfer the
hydrogen to the bridging position between the In dimer
atoms. This configuration is illustrated in cluster 5 for
one of the PH molecules. The PH species is covalently
that the phosphorus atom associated with the PH species
2
contains a lone pair of electrons.
The phosphine molecules also may form dative bonds to
the empty dangling bonds on the In–In dimers in the next
3
2
bonded to the indium atom and the P atom still retains

R.L. Woo et al. / Surface Science 600 (2006) 4888–4895
4893
Table 2
Although the theory predicts the symmetric and asym-
metric stretching modes of PH and PH species on the sur-
Comparison of the P–H vibrational frequencies and relative intensities
determined by experiment and theory
3
2
face very well within a difference of a few wavenumbers, it
predicts the isolated P–H stretching mode on the adatom
Assignments
Theory
Experiment
ꢀ
1
PH
3
TBP
dimer with a difference of 18 cm . This is clearly an arti-
fact of our present cluster model, since it does not include
the effect of the In–In dimers on the other side of the In–P
adatom dimer. More detailed calculations are needed with
larger cluster models to assign this mode more definitively.
a
m
I
Model
m
I
m
I
s
PH (In–P heterodimer) 2318 0.7
2
2
3
3
3
2319 0.7 2315 0.6
2339 0.2 2339 0.1
2301 0.8 2301 0.7
2242 0.8 2242 0.9
2262 0.9 2261 0.9
3
as
PH3
2338 0.3
2319 0.4
2245 0.7
2255 0.8
P–H
s
PH
ꢀ1
2
The 2219 cm frequency mode has not been assigned thus
as
PH2
ꢀ1
far. While it is close enough to the 2226 cm mode to be
s
3
PH (In–In dimer)
2296
1
4
4
5
5
2285
1
2281
1
assigned to the symmetric stretch of a PH group, its low
2
as
3
PH
2329 0.2
2233 0.8
2242 0.7
2327 0.3 2323 0.2
2226 0.3 2226 0.5
2242 0.8 2242 0.9
frequency may also be due to a longer P–H bond associ-
ated with the presence of a lone pair on the phosphorus
atom. It should be noted that a similar stretching mode
has been observed previously in the case of hydrogen
adsorption on the P-rich InP surface [15].
s
PH
2
as
PH2
ꢀ1
m = frequency (cm ).
I = relative intensity.
as = asymmetric, s = symmetric.
a
ꢀ1
Refer to Fig. 7.
The assignment of the peaks at 2319 and 2285 cm to
the symmetric stretching modes of datively bonded PH3
and TBP is further supported by the annealing experi-
ments. Since this bond is weak, the molecule should desorb
at a relatively low temperature. As seen in the difference
spectrum in Fig. 3, the intensities of the peaks at 2285
and 2319 cm decrease significantly more than the other
P–H peaks as the temperature is raised from 25 to 180 ꢀC.
its lone pair of electrons. The H atom is predicted to form a
3
-center-2-electron bond with the In dimer atoms.
The P–H vibrational modes calculated for clusters 2
ꢀ
1
through 5 are listed in Table 2 along with the P–H vibra-
tional modes observed in the experiments. Also shown in
the Table is the relative intensity of each infrared band,
which is defined as follows:
4.2. Adsorption sites
Max
mðPꢀHÞ
I ¼ jImðPꢀHÞj=jI
j
The adsorption mechanism proposed by Sun et al. [13],
reactions (1)–(3) in the Introduction, is consistent with the
results obtained in the present study. Shown in Fig. 8 is a
potential energy diagram for phosphine adsorption, ob-
tained from molecular cluster calculations [26]. In reaction
(1), the group V precursor makes a dative bond to the
empty dangling orbital of the indium atom with a binding
energy of ꢀ11 kcal/mol. In reaction (2), the molecule disso-
Max
where Im(P–H) and I
are the heights of the given P–H
mðP–HÞ
band and the most intense P–H band, respectively. A com-
parison of the theory to experiment reveals that good
agreement has been achieved in most cases. Datively
bonded PH adsorbed onto an In–P heterodimer (cluster
3
2
) is predicted to yield symmetric and asymmetric stretch-
ꢀ
1
ing modes at 2318 and 2338 cm , respectively. These
modes correspond to the experimentally observed peaks
ciates into adsorbed PH by transferring hydrogen or a t-
2
ꢀ
1
at 2319 and 2339 cm . Dissociative adsorption of PH3
on the In-P heterodimer (cluster 3) generates P–H and
PH species. The cluster calculations predict a stretching
vibration at 2319 cm for the isolated P–H bond, and at
butyl ligand to a nearby phosphorus adsorption site.
According to the theory, this reaction requires activation
energy of 17 kcal/mol. In order to conserve electrons, the
2
ꢀ
1
PH group will have a lone pair of electrons attached to
2
ꢀ
1
2
245 and 2255 cm for the symmetric and asymmetric
the P atom. This gives rise to the stretching modes at
ꢀ
1
stretching modes of the PH group. The last two modes
2242 and 2262 cm . The cluster calculations further pre-
2
most likely correspond to the experimental infrared peaks
dict that dissociation of the PH species is not possible at
room temperature. The reader is referred to Ref. [26] for
a detailed discussion of the reaction mechanism.
2
ꢀ
1
at 2242, and 2262 cm , respectively. The isolated P–H
ꢀ1
mode calculated at 2319 cm is tentatively assigned to
ꢀ1
the experimental peak at 2301 cm (vide infra).
During the dissociative adsorption of PH on the In–In
3
Cluster 4 exhibits symmetric and asymmetric stretching
dimer, the hydrogen ligand is predicted to insert into the in-
dium bridge bond. However, here the calculated activation
energy is 30 kcal/mol [26]. This energy barrier is too high
for the reaction to occur at room temperature. However,
modes for PH datively bonded to the In–In dimer at 2296
3
1
ꢀ
and 2329 cm , respectively. The corresponding peaks ob-
ꢀ
1
served by experiment occur at 2285 and 2327 cm . Disso-
ciative adsorption of phosphine at this site results in a PH2
group and bridging indium hydride as illustrated by cluster
vibrational modes ascribable to PH bound to the In–In di-
2
mer are observed experimentally, i.e., at 2219 (2212) and
ꢀ1
5
. According to the theory, the symmetric and asymmetric
2226 cm . Surface diffusion of the PH species from the
2
stretching modes of the PH groups are at 2233 and
In-P heterodimer to the In–In dimer could explain this
result, as this process would have a much lower energy
barrier. Similarly, hydrogen atoms may diffuse from
2
ꢀ
1
2
2
242 cm , whereas the experiment yields frequencies of
226 and 2242 cm , respectively.
ꢀ1

4894
R.L. Woo et al. / Surface Science 600 (2006) 4888–4895
2
Fig. 8. Potential energy diagram for phosphine adsorption and dissociation into PH .
phosphorus adsorption sites to In–In bridging sites and
produce indium hydrides band with vibrational frequencies
between 1750 and 1000 cm . These bands can be observed
in the infrared spectra presented in Fig. 1.
so that at elevated temperatures the sticking probability
of the group V molecules is low, i.e., between 0.001 and
0.007 [13,14]. The calculated binding energy of TBP on
the indium phosphide surface is about 3–4 kcal/mol higher
ꢀ
1
Listed in Table 1 are the estimated percentages of P and
In atoms observed by XPS following phosphine and tertia-
rybutylphosphine adsorption on InP (001). A comparison
of the P/In ratio indicates that the saturation coverage of
than that of PH , consistent with the slightly higher stick-
ing probability of TBP.
3
5. Conclusions
PH is higher than that of TBP. The calculated van der
3
Waals diameters from the critical pressures and tempera-
We have characterized the adsorption of phosphine and
tertiarybutylphosphine on the indium-rich InP (001)-
(2 · 4) surface by vibrational spectroscopy, X-ray photo-
emission, and low-energy electron diffraction. The group
V molecules first form dative bonds to exposed indium
atoms on the surface, and then undergo dissociation into
˚
tures for PH and TBP are 4.41 and 6.01 A, respectively
3
[
4
23,25]. With
a fixed InP unit cell dimension of
.0 · 4.0 A˚ 2 for the (1 · 1), more phosphine molecules
should be able to adsorb than TBP molecules, as observed.
Nevertheless, we do not believe that the phosphorus cover-
age is truly 2.2 for phosphine adsorbed on the InP (001)-
PH , with transfer of the H or tertiarybutyl ligands to adja-
2
(
2 · 4) surface. Since the number of photoelectrons
cent phosphorus sites. The weakly held molecular state
provides an explanation for the low sticking probability
of PH and TBP under MOCVD growth conditions.
3
detected by XPS decreases exponentially with depth, the
datively bonded phosphine molecules protruding out of
the surface would tend to over-emphasize the abundance
of these species relative to P and In in the lower layers.
Acknowledgements
˚
For example, with an escape depth of 12 A, photoelectrons
from the 4 top most P layers and the 4 top most In layers of
the (2 · 1) surface will be detected by XPS. By contrast, for
the phosphine covered (2 · 4) surface, photoelectrons from
the 4 top most P layers and 3 top most In layers will be
detected.
The infrared study presented here provides solid evi-
dence for the formation of datively bonded phosphine
and tertiarybutylphosphine on InP (001). This reaction is
reversible with molecular desorption increasing rapidly
with surface temperature. At the same time, a portion of
Funding for this research was provided by Spectrolab,
Inc. (a division of Boeing), Epichem, Inc., SAES Getters,
Inc., and the UC Discovery program. K. Raghavachari
and U. Das acknowledge support from American Chemical
Society, Petroleum Research Fund grant (PRF 43465-
AC10).
References
[
1] H.-G. Bach, W. Schlaak, G.G. Mekonnen, A. Seeger, R. Steingruber,
C. Schramm, G. Jacumeit, R. Ziegler, A. Umbach, G. Unterborsch,
W. Passenberg, W. Ebert, T. Eckardt, in: Trends in Optics and
Photonics, Optical Fiber Communication Conference, vol. 37, Bal-
timore, MD, USA, 2000.
the adsorbed PH and TBP undergo dissociative adsorp-
3
tion on the In–P heterodimer. Molecular desorption com-
petes effectively with sequential dissociation adsorption,

R.L. Woo et al. / Surface Science 600 (2006) 4888–4895
4895
[
[
[
[
[
[
2] F. Alexandre, J.L. Benchimol, J. Dangla, C. Dubon-Chevallier, V.
Amarger, Electron. Lett. 26 (1990) 1753.
3] M.T. Camargo Silva, J.E. Zucker, L.R. Carrion, C.H. Joyner, A.G.
Dentai, IEEE J. Sel. Top. Quant. 6 (2000) 26.
4] Y. Akage, K. Kawano, S. Oku, R. Iga, H. Okamoto, Y. Miyamoto,
H. Takeuchi, Electron Lett. 37 (2001) 299.
5] G.B. Stringfellow, Organometallic Vapor-Phase Epitaxy: Theory and
Practice, Academic press, San Diego, 1989.
[15] Q. Fu, E. Negro, G. Chen, D.C. Law, C.H. Li, R.F. Hicks, K.
Raghavachari, Phys. Rev. B 65 (2002) 075318.
[16] L. Li, Q. Fu, C.H. Li, B.-K. Han, R.F. Hicks, Phys. Rev. B 61 (15)
(2000) 10223.
[17] W.G. Schmidt, Phys. Rev. B 57 (1998) 14596.
[18] K. Raghavachari, Q. Fu, G. Chen, L. Li, C.H. Li, D.C. Law, R.F.
Hicks, J. Am. Chem. Soc. 124 (2002) 15119.
[19] G. Chen, S.F. Cheng, D.J. Tobin, L. Li, K. Raghavachari, R.F.
Hicks, Physical Review B, Rapid Comm. 68 (2003) 121303 (R).
[20] G. Socrates, Infrared Characteristic Group Frequencies, second ed.,
Wiley, Chichester, England, 1994.
6] <http://www.compoundsemiconductor. net/articles/magazine/8/1/6/
1
>.
7] S.D. Adamson, B.-K. Han, R.F. Hicks, Appl. Phys. Lett. 69 (1996)
236.
3
[21] M. Sugiyama, K. Kusunoki, Y. Shimogaki, S. Sudo, Y. Nakano, H.
Nagamoto, K. Sugawara, K. Tada, H. Komiyama, Appl. Surf. Sci.
117/118 (1997) 746.
[
[
8] G. Miller, Solid State Technol. 32 (1989) 59.
9] K.L. Hess, R.J. Riccio, J. Crystal Growth 77 (1986) 95.
[
[
10] G. Chen, D. Cheng, R.F. Hicks, A.M. Noori, S.L. Hayashi, M.S.
Goorsky, R. Kanjolia, R. Odedra, J. Crystal Growth 270 (2004) 322.
11] M.J. Begarney, C.H. Li, D.C. Law, S.B. Visbeck, Y. Sun, R.F. Hicks,
Appl. Phys. Lett. 78 (2001) 55.
12] G. Kaneda, N. Sanada, Y. Fukuda, Appl. Surf. Sci. 142 (1999) 1.
13] Y. Sun, D.C. Law, S.B. Visbeck, R.F. Hicks, Surf. Sci. 513 (2002)
[22] G.H. Fan, R.D. Hoare, M.E. Pemble, I.M. Povey, A.G. Taylor,
J. Crystal Growth 124 (1992) 49.
[23] <http://webbook.nist.gov/chemistry/>.
[24] C.H. Li, L. Li, D.C. Law, S.B. Visbeck, R.F. Hicks, Phys. Rev. B 65
(2002) 205322.
[25] D.A. McQuarrie, J.D. Simon, Physical Chemistry: A Molecular
Approach, University Science Books, Sausalito, 1997.
[26] U. Das, R.L. Woo, R.F. Hicks, K. Raghavachari, in preparation.
[
[
256.
[
14] Y. Sun, D.C. Law, R.F. Hicks, Surf. Sci. 540 (2003) 12.