You are on page 1of 10

Surface Science 513 (2002) 475–484

www.elsevier.com/locate/susc

C incorporation and segregation during Si1y Cy =Si(0 0 1)


gas-source molecular beam epitaxy from Si2H6 and CH3SiH3
Y.L. Foo, K.A. Bratland, B. Cho, J.A.N.T. Soares, P. Desjardins, J.E. Greene *

Frederick Seitz Materials Research Laboratory and the Materials Science Department, University of Illinois, 104 S Goodwin Avenue,
Urbana, IL 61801, USA
Received 12 December 2001; accepted for publication 23 April 2002

Abstract
We have used in situ D2 temperature-programmed desorption (TPD) to probe C incorporation and surface segre-
gation kinetics, as well as hydrogen desorption pathways, during Si1y Cy (0 0 1) gas-source molecular beam epitaxy from
Si2 H6 /CH3 SiH3 mixtures at temperatures Ts between 500 and 650 °C. Parallel D2 TPD results from C-adsorbed Si(0 0 1)
wafers exposed to varying CH3 SiH3 doses serve as reference data. Si1y Cy (0 0 1) layer spectra consist of three peaks:
first-order b1 at 515 °C and second-order b2 at 405 °C, due to D2 desorption from Si monodeuteride and dideuteride
phases, as well as a new second-order C-induced c1 peak at 480 °C. C-adsorbed Si(0 0 1) samples with very high
CH3 SiH3 exposures yielded a higher-temperature TPD feature, corresponding to D2 desorption from surface C atoms,
which was never observed in Si1y Cy (0 0 1) layer spectra. The Si1y Cy (0 0 1) c1 peak arises due to desorption from Si
monodeuteride species with C backbonds. c1 occurs at a lower temperature than b1 reflecting the lower D–Si bond
strength, where Si represents surface Si atoms bonded to second-layer C atoms, as a result of charge transfer from
dangling bonds. The total integrated monohydride ðb1 þ c1 Þ intensity, and hence the dangling bond density, remains
constant with y indicating that C does not deactivate surface dangling bonds as it segregates to the second-layer during
Si1y Cy (0 0 1) growth. Si coverages increase with y at constant Ts and with Ts at constant y. The positive Ts -dependence
shows that C segregation is kinetically limited at Ts 6 650 °C. D2 desorption activation energies from b1 , c1 and b2 sites
are 2.52, 2.22 and 1.88 eV.
Ó 2002 Elsevier Science B.V. All rights reserved.

Keywords: Thermal desorption spectroscopy; Epitaxy; Surface segregation; Silicon; Carbon; Alloys

1. Introduction and strain-state engineering of layers used in mi-


croelectronic and optoelectronic devices compati-
C-containing group-IV semiconductor alloys ble with Si integrated circuit technology. Adding
are of intense technological and scientific interest C to Si1x Gex alloy layers pseudomorphically
due to the potential they offer for both bandgap grown on Si(0 0 1) rapidly reduces the compressive
strain and leads to increased critical thicknesses
for nucleation of misfit dislocations. The addition
*
Corresponding author. Tel.: +1-217-3331370; fax: +1-217- of C has also been shown to inhibit B diffusion in
2442278. npn bipolar heterojunction transistors [1], while
E-mail address: greene@mrlxp2.mrl.uiuc.edu (J.E. Greene). pseudomorphic Si1y Cy /Si(0 0 1) structures provide

0039-6028/02/$ - see front matter Ó 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 0 3 9 - 6 0 2 8 ( 0 2 ) 0 1 8 2 1 - 6
476 Y.L. Foo et al. / Surface Science 513 (2002) 475–484

sufficient conduction band offset to obtain well- over the entire temperature range investigated
defined two-dimensional electron gases for use in (550 °C 6 Ts 6 650 °C) [9]. The authors proposed
modulation-doped field effect transistors (MOD- that R decreases at Ts ¼ 550 °C due to surface
FETs) [2]. segregation of C to the first-layer which, in turn,
However, the incorporation of even a few atomic decreases the dangling bond coverage hdb (and
percent of C into the Si lattice is highly challenging hence the subsequent SiH4 adsorption rate) since
due to the low equilibrium solid-solubility of C in the H–C bond strength is greater than that of H–
Si, ’1017 cm3 (Ref. [3]), and the extremely large Si. However, at Ts ¼ 650 °C, where the steady-
lattice constant mismatch between diamond (aC ¼ state H coverage is near zero, R still continues to
3:5668 A ) and Si (aSi ¼ 5:4310 A ). Complete in- decrease with increasing y. This was ascribed to a
corporation of C atoms into substitutional sites in different mechanism: namely a decrease in the SiH4
Si1y Cy (0 0 1) has only been achieved at C concen- reactive sticking probability [9]. There are, how-
trations up to ’2 at.%, and only over a very nar- ever, no published data on C surface segregation,
row window of growth conditions [4]. steady-state C coverages hC as a function of pre-
Gas-source molecular-beam epitaxy (GS-MBE) cursor flux and Ts , or the effect of hC on H de-
and ultra-high vacuum chemical vapor deposition sorption rates and precursor sticking probabilities
(UHV-CVD) have been shown to offer poten- with which to test the above hypotheses.
tial advantages over solid-source MBE. These in- In this article, we describe the results of experi-
clude elimination of high-temperature evaporation ments designed to probe C incorporation and
sources, higher sample throughput, better conformal surface segregation kinetics and their effects on
coverage, and the potential for selective epitaxy on steady-state H coverages hH , and hence on film
patterned substrates [5]. A possible disadvantage growth kinetics, during gas-phase Si1y Cy /Si(0 0 1)
for the growth of metastable Si1y Cy (0 0 1) and deposition from Si2 H6 /CH3 SiH3 mixtures. We use
Si1xy Gex Cy (0 0 1) layers from GS-MBE/UHV- GS-MBE in order to take advantage of in situ
CVD, however, is the complexity associated with surface science probes including isotopically
the surface chemical reactions and the dependence tagged TPD, reflection high energy electron dif-
of the overall film growth kinetics on the steady- fraction (RHEED), low-energy electron diffraction
state reaction-product coverage which is in turn (LEED), and Auger electron spectroscopy (AES).
dependent on deposition temperature Ts and alloy For reference, we also carry out TPD experiments
composition. on C-adsorbed Si(0 0 1) as a function of CH3 SiH3
A variety of C-containing precursors, including exposure.
C3 H8 [6], CH4 [7], CH3 SiH3 [8,9], C(SiH3 )4 and D2 TPD spectra from Si1y Cy (0 0 1) layers con-
C(SiH2 Cl)4 [10], have been used for Si1y Cy (0 0 1) tain, in addition to b1 and b2 peaks due to Si
gas-source epitaxy. Of these, monomethylsilane, monodeuteride and dideuteride surface species, a
CH3 SiH3 , is of interest since it is a relatively sim- lower-temperature C-induced monodeuteride c1
ple molecule, is a gas at room temperature, and peak associated with Si surface atoms, labeled Si ,
contains no C–C bonds. CH3 SiH3 is believed to having C backbonds. Increasing the C fraction y in
dissociatively adsorb on Si(0 0 1) through nucleo- Si1y Cy (0 0 1) layers grown at constant Ts results in
philic substitution reactions resulting in Si–C bond a decrease in the intensity of b1 , while the intensity
cleavage and the formation of surface CH3 and of c1 increases such that the total integrated mono-
SiH3 species which further decompose to monohy- deuteride intensity, and therefore the total dan-
drides and dihydrides [11,12]. gling bond density, remains constant. C segregates
Previous reports have described the growth of strongly to the second-layer with Si coverages
Si1y Cy (0 0 1) layers, with y up to 0.018 and 0.03 hSi increasing from 0 to 0.42 ML as y is varied
from SiH4 /CH3 SiH3 mixtures by UHV-CVD [9] from 0 to 0.026 during the growth of Si1y Cy (0 0 1)
and rapid-thermal chemical vapor deposition (RT- at Ts ¼ 600 °C. hSi also increases with Ts at con-
CVD) [8], respectively. The UHV-CVD film growth stant y: for example, hSi ranges from 0.1 ML at
rate R was found to decrease with increasing y Ts ¼ 500 °C to 0.3 ML at Ts ¼ 650 °C for
Y.L. Foo et al. / Surface Science 513 (2002) 475–484 477

Si0:9952 C0:0048 (0 0 1). D2 TPD spectra from Si(0 0 1) There, they were degassed at 600 °C for 4 h, cooled
surfaces exposed to high CH3 SiH3 doses (2:5  to 200 °C, and then rapidly heated at ’100 °C s1
1017 –1:5  1019 cm2 ) at Ts ¼ 650 °C exhibit, in to 1100 °C for 1 min to remove the oxide. The
addition to b1 , b2 and c1 peaks, a higher-temper- maximum pressure rise during oxide desorption
ature feature arising from the desorption of D was 6 3  109 Torr for 6 10 s.
bonded to first-layer C atoms. Substrates processed using the above procedure
exhibited 2  1 RHEED patterns consisting of
well-defined diffraction spots, rather than streaks,
2. Experimental procedure and sharp Kikuchi lines. The layer surfaces were
thus atomically smooth with relatively large ter-
All Si1y Cy (0 0 1) layers were grown in a multi- races. No C or O was detected by in situ AES.
chamber UHV GS-MBE system, described in de- Substrate temperatures were determined using Pt–
tail in Refs. [13,14]. A combination of ion and Rh thermocouples calibrated by optical pyrometry.
turbomolecular pumps provide a base pressure of Si(0 0 1) buffer layers were grown at 800 °C
’5  1011 Torr. The growth chamber, equipped prior to commencement of Si1y Cy (0 0 1) film
with RHEED and a quadrupole mass spectro- growth. The buffer layers serve two purposes. They
meter (QMS), is connected through a transfer cover any remaining surface contamination while
chamber to an analytical station containing pro- simultaneously providing a more uniform dis-
visions for TPD, utilizing a heavily differentially tribution of terrace lengths [16]. The primary
pumped Extrel QMS, AES, electron energy loss Si1y Cy (0 0 1) film growth temperatures were 500,
spectroscopy and LEED. 550 and 600 °C, all in the surface-reaction-limited
Si1y Cy /Si(0 0 1) layers, 0.10–0.45 lm thick, growth regime. The incident Si2 H6 flux was main-
were grown from Si2 H6 and CH3 SiH3 molecular tained constant at JSi2 H6 ¼ 2:2  1016 cm2 s1 for
beams delivered to the substrate through individ- all depositions while the CH3 SiH3 flux JCH3 SiH3 was
ual directed tubular dosers located 3 cm from varied from 7:0  1012 to 4:2  1015 cm2 s1 . In a
the substrate at an angle of 45°. The dosers are parallel series of experiments, clean Si(0 0 1) sur-
coupled to feedback-controlled constant-pressure faces were exposed to a CH3 SiH3 dose rate of
reservoirs in which pressures are separately moni- 4:2  1015 cm2 s1 for time intervals t ¼ 1–60 min
tored using capacitance manometers whose signals (corresponding to doses n of 2:5  1017 –1:5  1019
are in turn used to control variable leak valves. cm2 ) at 650 °C.
Valve sequencing, pressures, gas flows, and film TPD measurements were carried out in situ on
growth temperatures are all computer controlled. all samples. Following Si1y Cy (0 0 1) growth, or
The Si(0 0 1) substrates were 1  3 cm2 plates CH3 SiH3 dosing of a clean Si(0 0 1) buffer layer,
cleaved from 0.5-mm-thick n-type wafers (resistiv- the samples were quenched to <200 °C and ex-
ity ¼ 10–20 X cm, n ¼ 2–4  1014 cm3 ). Initial posed to atomic deuterium until saturation cov-
cleaning consisted of degreasing by successive erage. For this purpose, D2 was delivered through
rinses in warm trichloroethane, acetone, propanol a doser identical to those described above, but
and deionized water. The substrates were then with a hot W filament near the outlet to crack the
subjected to four wet-chemical oxidation/etch cy- gas. All surface H was exchanged for D as dem-
cles comprised of the following steps: 2 min in a onstrated by TPD. The TPD experiments were
2:1:1 solution of H2 O:HCl:H2 O2 , rinse in fresh performed with the sample 2 mm from the 5-mm-
deionized water, and a 30 s etch in dilute (10%) diameter hole in the skimmer cone of a heavily
HF. They were blown dry with ultra-high-purity differentially pumped Extrel QMS. TPD spectra
N2 , exposed to a UV/ozone treatment which con- were acquired using a linear heating rate of 2
sists of UV irradiation from a low-pressure Hg °C s1 . Deuterium was employed rather than hy-
lamp (15 mW cm2 ) for 30 min in air to remove C- drogen in order to suppress the background signal.
containing species [15], and introduced into the Deposited film thicknesses were measured by
deposition system through the transfer chamber. microstylus profilometry while C concentrations
478 Y.L. Foo et al. / Surface Science 513 (2002) 475–484

in as-deposited layers were determined using a


Cameca IMS-5F secondary ion mass spectrometer
(SIMS) operated with a 10 keV Csþ primary ion
beam to detect 12 C. Quantification, with an exper-
imental uncertainty of 10%, was accomplished
by comparison to C ion-implanted bulk Si(0 0 1)
standards. The Si1y Cy (0 0 1) films contained no
detectable impurities.

3. Experimental results

3.1. Si1y Cy (0 0 1) GS-MBE

LEED patterns showed that the surfaces of all


Si1y Cy (0 0 1) GS-MBE layers exhibit 2  1 re-
constructions with 90°-rotated domains. Fig. 1(a)
is an example, obtained in this case from a
Si0:989 C0:011 (0 0 1) layer grown at Ts ¼ 650 °C.
RHEED patterns from layers with y 6 0:0048
grown at Ts 6 550 °C consist of well-defined 2  1
diffraction spots, with nearly equi-intense funda-
mental and half-order reflections, and sharp
Kikuchi lines indicating that the film is atomically
smooth. A typical result from an Si0:9952 C0:0048 (0 0 1)
alloy is presented in Fig. 1(b). At higher C con-
centrations ðy > 0:0048Þ and/or at higher growth
temperatures (Ts > 550 °C), the RHEED patterns
become streaky. The fundamental rods broaden,
half-order to fundamental rod intensity ratios
decrease, and the amount of diffuse scattering in-
creases; all are indicative of atomic scale rough-
ening. Fig. 1(c) is a RHEED pattern from a
Si0:989 C0:011 (0 0 1) layer grown at Ts ¼ 600 °C.
Typical D2 TPD results from GS-MBE Si1y Cy -
(0 0 1) layers are reproduced in Fig. 2. Spectra Fig. 1. (a) LEED pattern from a GS-MBE Si0:989 C0:011 layer,
’1000-A-thick, grown on Si(0 0 1) from CH3 SiH3 /Si2 H6 at Ts ¼
from layers with y 6 0:000060 are identical to those
650 °C. RHEED patterns from GS-MBE Si1y Cy (0 0 1) layers,
from clean Si(0 0 1). An example is shown in Fig. ’1400-A-thick, with C fractions y grown at Ts : (b) y ¼ 0:0048,
2(a) for a y ¼ 0:000056 layer grown at Ts ¼ 600 °C. Ts ¼ 550 °C and (c) y ¼ 0:011, Ts ¼ 600 °C.
The spectra consist of two peaks, labeled b2 and
b1 , due to desorption from the 1  1 Si dideuteride
and 2  1 Si monodeuteride phases, respectively.
The peaks are centered at 405 and 515 °C. While Fig. 2(b)–(d) are D2 TPD spectra from Si1y Cy -
b2 desorption is second-order, b1 follows first- (0 0 1) layers grown at Ts ¼ 600 °C with y ¼
order kinetics, except at very low deuterium 0:00046, 0.0048 and 0.026. As the C concentration
coverages (hD < 0:1 ML) [17], due to p-bonding- is increased at constant Ts , the TPD features
induced pairing of dangling bonds on single di- broaden (suggesting the appearance of additional
mers [18]. desorption peaks), the high-temperature feature
Y.L. Foo et al. / Surface Science 513 (2002) 475–484 479

Fig. 2. D2 TPD spectra from GS-MBE layers Si1y Cy (0 0 1) layers with C fractions y grown on Si(0 0 1) at Ts : (a) y ¼ 0:000056,
Ts ¼ 600 °C, (b) y ¼ 0:00046, Ts ¼ 600 °C, (c) y ¼ 0:0048, Ts ¼ 600 °C, (d) y ¼ 0:026, Ts ¼ 600 °C, (e) y ¼ 0:0048, Ts ¼ 550 °C and (f)
y ¼ 0:0048, Ts ¼ 650 °C.

 n  
decreases in intensity and the intensity of the low- dhD mh Ea
¼ exp  ð1Þ
temperature feature increases. However, the total dT f kT
integrated area remains constant. Fig. 2(e), (c),
and (f), TPD spectra from Si0:9952 C0:0048 (0 0 1) al- where m is the attempt frequency, hD is the instan-
loys grown at Ts ¼ 550, 600 and 650 °C, respec- taneous D coverage, n is the order of the desorption
tively, show that the effects of increasing Ts at reaction, f is the sample heating rate, Ea is the de-
constant y are similar to those corresponding to sorption activation energy and k is Boltzmann’s
increasing y at constant Ts . constant. At high pumping speeds [20]
All TPD spectra from GS-MBE Si1y Cy (0 0 1)  
hD ðT Þ m
layers were fit using standard Polanyi–Wigner ln ¼  IðT Þ ð2Þ
h0 f
analyses in which the desorption rate dhD =dT is
expressed as [19] for first-order desorption and
480 Y.L. Foo et al. / Surface Science 513 (2002) 475–484

h 3.2. C-adsorbed Si(0 0 1)


hD ðT Þ ¼  0 ð3Þ
m
1þ f
h0 IðT Þ
D2 TPD spectra from C-adsorbed Si(0 0 1) sur-
faces, obtained following long exposures of clean
for second-order desorption. h0 in Eqs. (2) and (3) Si(0 0 1) buffer layers to CH3 SiH3 , are quite differ-
is the initial coverage and IðT Þ is given by ent than those acquired from GS-MBE Si1y Cy -
(0 0 1) layers. Fig. 3 shows typical spectra from
 T samples dosed with JCH3 SiH3 ¼ 4:2  1015 cm2 s1
Ea  ee X
1
ð1Þ n! 
nþ1
IðT Þ ¼  2  ð4Þ at Ts ¼ 650 °C for t ¼ 7:5, 15, 30 and 60 min. For
k  e n¼1 en1  short exposure times (t ¼ 1–7:5 min, correspond-
T0
ing to doses n ¼ 2:5  1017 –1:9  1018 cm2 ), the
in which e ¼ Ea =kT . TPD results exhibit a shape which is similar, with
Spectra from Si1y Cy (0 0 1) films with y 6 the exception of a new extended desorption feature
0:000060, including the results shown in Fig. 2(a), at T > 515 °C, to spectra obtained from GS-MBE
were well fit with Ea and m equal to 1.88 eV and Si1y Cy (0 0 1) alloys with y > 0:00060 and com-
1  1013 s1 for b2 and 2.52 eV and 1  1015 s1 for posed of b1 , b2 and c1 peaks. An example, corre-
b1 as previously reported for D2 TPD from clean sponding to t ¼ 7:5 min, is presented in Fig. 3(a).
bulk Si(0 0 1) surfaces [21]. The agreement between Increasing t to 15 min (Fig. 3(b), n ¼ 3:8  1018
measured and calculated spectra is very good ex- cm2 ) results in an increase in the new extended
cept at high temperatures (low D coverages) where high-temperature desorption feature as b1 , b2 and
the measured data for Si(0 0 1) and Si1y Cy (0 0 1) c1 decrease in intensity. Figs. 3(c) and (d) are
layers are higher than the calculated curves due spectra from C-adsorbed samples dosed for t ¼ 30
to the fact that b1 desorption deviates from first- and 60 min (n ¼ 7:6  1018 and 1:5  1019 cm2 ).
order kinetics at low hydrogen coverages where The b1 , b2 and c1 peak intensities have decreased
D2 desorption becomes limited by bimolecular dramatically while the high-temperature feature
recombination and is, therefore, second-order [22]. has evolved into a peak, centered near 670 °C,
D2 TPD spectra from all Si1y Cy (0 0 1) ðy > whose intensity increases with n. We attribute this
0:000060Þ alloys, irrespective of y or Ts , were peak to D2 desorption from first-layer C atoms; it
equally well fit with three peaks: the initial b2 and occurs at higher temperatures since the D–C bond
b1 peaks at 405 and 515 °C, together with a C- is stronger than the D–Si bond. It is important
induced c1 peak at 480 °C. Examples are shown in to note that this feature was not observed in any
Fig. 2(b)–(f). The b1 and b2 peaks have the same of the GS-MBE Si1y Cy (0 0 1) TPD spectra, irre-
physical origin, and hence the same activation spective of y or Ts .
energies and frequency factors, as the equivalent
peaks in TPD spectra from clean Si(0 0 1). c1 cor-
responds to second-order D2 desorption with 4. Discussion
Ea ¼ 2:22 eV and m ¼ 1  1014 s1 . The integrated
intensity Ic1 of the C-induced c1 peak increases The in situ D2 TPD results from GS-MBE
with increasing y at constant Ts (compare Fig. Si1y Cy (0 0 1) alloy layers presented in Section 3.1
2(b)–(d)) and increasing Ts at constant y (Fig. 2(e), exhibit, in addition to b1 Si monodeuteride and b2
(c), and (f)). This behavior is parallel to that ob- dideuteride peaks at 405 and 515 °C, a new c1 peak
served for the B-induced c1 peak in TPD spectra at 480 °C. The intensity of c1 increases with in-
from ultra-highly B doped Si(0 0 1):B GS-MBE creasing y at constant Ts and with increasing Ts at
layers [23]. The c1 to b1 intensity ratio Ic1 =Ib1 in- constant y. A higher-temperature peak near 670
creases rapidly with y, ranging from ’0 with °C, due to D2 desorption from first-layer C atoms,
y 6 0:000060 to 0.73 for y ¼ 0:026 at Ts ¼ 600 °C. was obtained from Si(0 0 1) buffer layers on which
This suggests strong C surface segregation during we had adsorbed C using very high CH3 SiH3 ex-
GS-MBE Si1y Cy (0 0 1) film growth. posures (n > 2:5  1017 cm2 ) at Ts ¼ 650 °C.
Y.L. Foo et al. / Surface Science 513 (2002) 475–484 481

Fig. 3. D2 TPD spectra from clean Si(0 0 1) surfaces exposed to a CH3 SiH3 flux JCH3 SiH3 ¼ 4:2  1015 cm2 s1 at Ts ¼ 650 °C for times
t: (a) 7.5 min, (b) 15 min, (c) 30 min and (d) 60 min. The open circle data points correspond to D2 TPD reference spectra from clean
Si(0 0 1) surfaces.

However, we did not observe this high-tempera- the activation energy for D2 desorption from C-
ture peak in TPD spectra from any of our Si1y Cy - backbonded Si surface atoms is 0.3 eV lower than
(0 0 1) alloy samples. Thus, we conclude that during that from Si surface atoms with Si backbonds.
GS-MBE Si1y Cy (0 0 1) film growth, adsorbed C This is primarily due to charge transfer, resulting
atoms, like B atoms in GS-MBE Si(0 0 1):B [23], from the higher electronegativity of C, from Si
move to the second-layer primarily due to their dangling bonds to the subsurface Si –C back-
small size. We interpret c1 as being due to deute- bonds. Similar findings
p phave been reported for
rium desorption from surface Si atoms with C B-adsorbed Si(111) 3  3 based on electron tun-
backbonds. RHEED and LEED analyses showed neling spectroscopy data [26] and for ultra-highly
that all Si1y Cy (0 0 1) layer surfaces, irrespective of B-doped Si(0 0 1)2  1 layers based upon TPD
the bulk C concentration or film growth temper- [23].
ature, remain 2  1. Fig. 4 is a plot of the coverages h of the b1 and
Si dimer atoms on Si(0 0 1)2  1 are in tension c1 surface monohydride phases, as well as the b2
and have rotated and foreshortened backbonds dihydride phase, obtained from the corresponding
associated with the reconstruction [24]. Placing C integrated TPD peak intensities as a function of
atoms in Si dimer sites would lead to a significant the bulk C fraction y in Si1y Cy (0 0 1) alloys grown
increase in the tensile strain. The small covalent at 600 °C. hb1 decreases with y as hc1 , a direct
radius of C (0.77 A  compared to 1.11 A  for Si), measure of the second-layer C coverage, increases.
combined with the shorter C–Si bond length (1.89 However, the total integrated monohydride
) [25] with respect to Si–Si (2.35 A
A ) [25], results in coverage, hR ¼ hb1 þ hc1 , and hence the surface
the second-layer being the lowest energy site for C. dangling bond density, remains constant at 1
This is consistent with our TPD data showing that ML. Thus, there is no evidence for C-induced
482 Y.L. Foo et al. / Surface Science 513 (2002) 475–484

Fig. 4. b1 and c1 monodeuteride and b2 dideuteride surface Fig. 5. Steady-state surface coverages hSi of C-backbonded Si
coverages h on GS-MBE Si1y Cy (0 0 1) layers grown at Ts ¼ 600 atoms during GS-MBE growth of Si0:9952 C0:0048 (0 0 1) at tem-
°C as a function of y. peratures Ts . JSi2 H6 and JCH3 SiH3 are the incident Si2 H6 and
CH3 SiH3 fluxes.

deactivation of surface dangling bonds. This is


quite different than for the case of ultra-highly B- where Ic1 and Ib1 are the integrated intensities under
doped Si(0 0 1) GS-MBE in which charge transfer the c1 and b1 monodeuteride peaks respectively.
to trivalent second-layer B atoms results in signifi- Fig. 5 shows that hSi , and hence the coverage
cant dangling bond deactivation [23]. The higher of second-layer C atoms, increases continuously
charge transfer from Si to trivalent second-layer B with Ts over the entire range investigated. For
atoms, compared to isoelectronic second-layer C, Si0:9952 C0:0048 (0 0 1) alloys, hSi varies from 0.10 ML
is manifested in the lower B-backbonded Si mono- at Ts ¼ 500 °C to 0.30 ML at 650 °C. This indi-
deuteride desorption temperature, 470 vs 480 °C. cates strong C segregation to the second-layer,
For clean Si(0 0 1), we obtain a hb2 =hb1 coverage while the positive Ts -dependence shows that the
ratio of 0.23 in agreement with previous results segregation is in the kinetically limited regime [27].
[24]. This is far less than the value of unity, which At higher hSi values, it is reasonable to expect
would be expected for complete dideuteride cov- that segregated C in the second-layer will exhibit
erage, due to steric hindrance caused by electronic ordering. While our LEED patterns remains 2  1,
repulsion between adsorbed H nearest neighbors the amount of diffuse scattering increases with in-
[24]. Fig. 4 shows that for Si1y Cy (0 0 1) layers creasing y and Ts . This is consistent with previous
grown at Ts ¼ 600 °C, hb2 increases slowly, as hb1 LEED results for ultra-highly B-doped Si(0 0 1)
decreases, with y. Thus, the ratio hb2 =hb1 continu- layers [21,28] where scanning tunneling micro-
ously increases with y, ranging from 0.26 with scopy (STM) studies by Wang et al. [29] showed
y ¼ 0:0010 to 0.56 with y ¼ 0:026. We attribute that B on Si(0 0 1) moves to second-layer positions
this to the shorter Si –C backbond length result- and induces local ordering with four different
ing in wider separations between dimer rows and polymorphs, each of which include 2  1 dimers in
hence reduced steric hindrance between neighbor- the surface unit cell. Since the ordered domains are
ing monodeuteride species. much smaller than the LEED electron beam co-
The coverage of C-backbonded Si sites on GS- herence length, diffraction patterns remain 2  1.
MBE Si1y Cy (0 0 1) layers can be obtained from TPD spectra from CH3 SiH3 -dosed Si(0 0 1) wa-
our TPD data using the relation fers reveal D2 desorption characteristics which ex-
hibit both similarities to and large differences from
Ic1 those of GS-MBE Si1y Cy (0 0 1) alloy layers grown
hSi ¼ ; ð5Þ
Ic1 þ Ib1 from Si2 H6 /CH3 SiH3 mixtures. After exposures of
Y.L. Foo et al. / Surface Science 513 (2002) 475–484 483

7.5 and 15 min with JCH3 SiH3 ¼ 4:2  1015 cm2 s1 peaks: b1 and b2 due to D2 desorption from Si
at Ts ¼ 650 °C, the TPD spectra (Figs. 3(a) and monodeuteride and dideuteride surface phases, as
(b)) exhibit a rapid decrease in b1 intensity with a well as a new second-order desorption peak c1 .
corresponding increase in c1 as the coverage of C- The b1 , b2 and c1 desorption peaks are centered at
backbonded Si atoms increases, consistent with 515, 405 and 480 °C. No higher-temperature TPD
GS-MBE Si1y Cy (0 0 1) TPD spectra in Fig. 2. peak, corresponding to D2 desorption from sur-
However, at these large CH3 SiH3 exposures face C atoms, as obtained in C-adsorbed Si(0 0 1)
( P 2:8  103 ML), second-layer C coverages are spectra, was observed. The Si1y Cy (0 0 1) c1 peak,
saturated and a broad higher-temperature feature, similar to the case for ultra-highly doped Si-
not observed in Si1y Cy (0 0 1) alloy spectra, emer- (0 0 1):B layers, arises due to D2 desorption from
ges signaling the presence of first-layer surface C. C-backbonded Si monodeuteride species. c1 occurs
The width of this new TPD feature arises from the at a lower temperature than b1 due to the lower D–
heterogeneous environment of the surface C atoms Si bond strength as a result of charge transfer
which have both Si and C in-plane neighbors and from dangling bonds to second-layer C atoms
are also backbonded to both Si and C atoms. which have a higher electronegativity than Si. D2
Thus, the high-temperature feature corresponds to desorption activation energies from b1 , c1 and b2
a convolution of several TPD peaks associated sites are 2.52, 2.22 and 1.88 eV.
with first-layer C atoms. Thus, we have demonstrated, for the first time,
With even larger CH3 SiH3 doses, the high- that C segregates to the second-layer during GS-
temperature TPD feature increases in intensity and MBE Si1y Cy (0 0 1). The total integrated mono-
narrows as first-layer C coverage increases, while deuteride TPD intensity remains constant for all
the intensities of b2 , c1 and b1 continue to decrease. Si1y Cy (0 0 1) films indicating there is no signifi-
In Fig. 3(d), corresponding to the highest CH3 SiH3 cant C-induced deactivation of surface dangling
exposure (n ¼ 1:5  1019 cm2 or 2:3  104 ML), bonds as was observed during ultra-highly B-
the first-layer C peak is centered near 670 °C, well doped GS-MBE Si(0 0 1). The coverage of hSi of
below reported TPD results for H2 desorption C-backbonded Si , and hence the coverage of
from diamond C(0 0 1)2  1 surfaces for which the second-layer C atoms, increases continuously with
dideuteride and monodeuteride peaks occur at 900 Ts over the entire range investigated (500 °C 6 Ts 6
and 970 °C, respectively [30]. This again is due to 650 °C). For GS-MBE Si0:9952 C0:0048 (0 0 1) alloys,
the heterogeneous environment of both first- hSi varies from 0.10 ML at Ts ¼ 500 °C to 0.30
and second-layer C in the present experiments. ML at 650 °C. The positive Ts -dependence shows
Analogous results were demonstrated for D2 de- that C segregation is in the kinetically limited
sorption from Ge-adsorbed Si(0 0 1)2  1 and regime.
Si-adsorbed Ge(0 0 1)2  1 surfaces where both
the Si b2 and b1 peak positions shift down in
temperature due to the presence of in-plane sur- Acknowledgements
face and second-layer Ge neighbors [31].
The authors acknowledge the financial sup-
port of the Materials Science Division of the
5. Conclusions US Department of Energy (DOE) under Award
DEFG02-ER9645439. The authors also appreci-
We have carried out in situ D2 TPD experi- ated the use of the Center for Microanalysis of
ments on both C-adsorbed Si(0 0 1) and GS-MBE Materials at the University of Illinois, which is
Si1y Cy (0 0 1) layers in order to probe pathways for partially supported by the DOE. PD was partially
hydrogen desorption as well as C incorporation supported by the Natural Sciences and Engineer-
and segregation kinetics during GS-MBE Si1y Cy - ing Research Council of Canada. Y.L. Foo is
(0 0 1) deposition. D2 TPD spectra from GS-MBE funded by the Institute of Materials Research and
Si1y Cy (0 0 1) layers consist of three desorption Engineering (IMRE), and the Agency for Science,
484 Y.L. Foo et al. / Surface Science 513 (2002) 475–484

Technology, and Research (A*STAR) of Singa- [13] D. Lubben, R. Tsu, T.R. Bramblett, J.E. Greene, J. Vac.
pore. Sci. Technol. A 9 (1991) 3003.
[14] T.R. Bramblett, Q. Lu, T. Karasawa, M.-A. Hasan, S.K.
Jo, J.E. Greene, J. Appl. Phys. 76 (1994) 1884.
[15] T.R. Bramblett, Q. Lu, N.E. Lee, N. Taylor, M.-A. Hasan,
References J.E. Greene, J. Appl. Phys. 77 (1995) 1504.
[16] P. Desjardins, J.E. Greene, J. Appl. Phys. 79 (1996) 1423.
[1] H.J. Osten, B. Heinenmann, D. Knoll, G. Lippert, H. [17] U. H€ ofer, L. Li, T.F. Heinz, Phys. Rev. B 45 (1992)
Rucker, J. Vac. Sci. Technol. B 16 (1998) 1750. 9485.
[2] K.W. Faschinger, S. Zerlauth, G. Bauer, L. Palmetshofer, [18] J.J. Boland, J. Vac. Sci. Technol A 10 (1992) 2458.
Appl. Phys. Lett. 67 (1995) 3933. [19] P.A. Redhead, Vacuum 12 (1962) 203.
[3] R.I. Scace, G.A. Slack, J. Chem. Phys. 30 (1989) 1551. [20] F.M. Lord, J.S. Kittelberger, Surf. Sci. 43 (1974) 173.
[4] S. Zerlauth, H. Seyringer, C. Penn, F. Sch€affler, Appl. [21] H. Kim, G. Glass, T. Spila, N. Taylor, S.Y. Park, J.R.
Phys. Lett. 71 (1997) 3826. Abelson, J.E. Greene, Appl. Phys. Lett. 69 (1996) 3869.
[5] H. Hirayama, T. Tasumi, N. Aizaki, J. Cryst. Growth 95 [22] G. Boishin, L. Surnev, Surf. Sci. 345 (1996) 64.
(1989) 476. [23] H. Kim, G. Glass, T. Spila, N. Taylor, S.Y. Park, J.R.
[6] S.K. Ray, D.W. McNeill, D.L. Gay, C.K. Maiti, G.A. Abelson, J.E. Greene, J. Appl. Phys. 82 (1998) 2288.
Armstrong, B.M. Armstrong, H.S. Gamble, Thin Solid [24] J.E. Northrup, Phys. Rev. B 44 (1991) 1419.
Films 294 (1997) 149. [25] W. Windl, O.F. Sankey, J. Menendez, Phys. Rev. B 57
[7] J.B. Posthill, R.A. Rudder, S.V. Hattangady, G.G. Foun- (1988) 2431.
tain, R.J. Markunas, Appl. Phys. Lett. 56 (1990) 734. [26] P.J. Chen, M.L. Colaianni, J.T. Yates Jr., J. Appl. Phys. 72
[8] T.O. Mitchell, J.L. Hoyt, J.F. Gibbons, Appl. Phys. Lett. (1992) 3155.
71 (1997) 1688. [27] S.A. Barnett, J.E. Greene, Surf. Sci. 151 (1985) 67.
[9] A.C. Mocuta, D.W. Greve, J. Appl. Phys. 85 (1999) 1240. [28] Y. Wang, R.J. Hamers, Appl. Phys. Lett. 56 (1995) 2057;
[10] D. Chandrasekhar, J. McMurran, D.J. Smith, J.D. Lo- Y. Wang, R.J. Hamers, J. Vac. Sci. Technol. A 13 (1995)
rentzen, J. Menendez, J. Kouvetakis, Appl. Phys. Lett. 72 1431.
(1998) 2117. [29] Y. Wang, R.J. Hamers, E. Kaxiras, Phys. Rev. Lett. 74
[11] J. Xu, W.J. Choyke, J.T. Yates Jr., J. Phys. Chem. B 101 (1995) 403.
(1997) 6879. [30] C. Su, J.-C. Lin, Surf. Sci. 406 (1998) 149.
[12] M. Shinohara, T. Maehama, M. Niwano, Appl. Surf. Sci. [31] H. Kim, P. Desjardins, J.R. Abelson, J.E. Greene, Phys.
162–163 (2000) 161. Rev. B 58 (1998) 4803.

You might also like