Temperature programmed desorption studies of deuterium passivated silicon nanocrystals

Article abstract of DOI:10.1016/j.susc.2009.02.033

As grown silicon (Si) surfaces are known to reconstruct in order to reduce the number of dangling bonds. Surface reconstructions of hydride-terminated Si(1 0 0) and Si(1 1 1) surfaces have already been extensively studied using temperature programmed deso

Full text of DOI:10.1016/j.susc.2009.02.033

Surface Science 603 (2009) 1121–1125
Contents lists available at ScienceDirect
Surface Science
journal homepage: www.elsevier.com/locate/susc
Temperature programmed desorption studies of deuterium
passivated silicon nanocrystals
*
Navneethakrishnan Salivati, John G. Ekerdt
Department of Chemical Engineering, University of Texas at Austin, Austin, TX 78712, United States
a r t i c l e i n f o
a b s t r a c t
Article history:
As grown silicon (Si) surfaces are known to reconstruct in order to reduce the number of dangling bonds.
Surface reconstructions of hydride-terminated Si(100) and Si(111) surfaces have already been exten-
sively studied using temperature programmed desorption (TPD). The surfaces of nanocrystals, are yet
to be probed using TPD. Si nanocrystals less than 8 nm and ranging from 50 to 200 nm in diameter are
grown on SiO2 surfaces in an ultra high vacuum chamber and the as grown surfaces are exposed to atomic
deuterium. Desorption spectra are interpreted using analogies to Si(100). TPD spectra show that that the
nanocrystals surfaces are covered by a mix of monodeuteride, dideuteride and trideuteride species.
Monodeuteride species can be isolated by selectively annealing away the dideuteride and trideuteride,
monodeuteride and dideuteride species can be isolated by annealing away the trideuteride. The relative
populations of the deuterides depend on particle size, and their manner of filling on nanoparticles differs
from that for extended surfaces. Etching of the nanocrystal surface is observed during TPD, which is a
confirmation of the presence of trideuteride species on the nanocrystal surface.
Received 11 December 2008
Accepted for publication 24 February 2009
Available online 6 March 2009
Keywords:
Silicon nanocrystals
Temperature programmed desorption
Etching
Passivation
Scanning electron microscopy
Ó 2009 Elsevier B.V. All rights reserved.
1. Introduction
hence there will be some defect states due to the remaining dan-
gling bonds. Another way of terminating the dangling bonds is
Bulk Si being an indirect semiconductor is a poor emitter of
light. However, systems containing nanometer sized features of
Si such as porous silicon [1], silicon/silicon dioxide superlattices
[2], silicon nanoparticles [3,4] and silicon nanowires [5] have been
known to emit light in the visible region of the spectrum with the
wavelength being dependent on the size of the Si nanostructure.
The ratio of surface atoms to the total number of atoms can be
quite large in nanoscale systems. A Si atom in the bulk is bonded
to four other Si atoms. At the interface however, there are unsatis-
fied (dangling) bonds. These uncompensated bonds establish local-
ized defect states within the forbidden gap of Si nanocrystals,
providing sites for non-radiative recombination of excitons [6].
Earlier studies with extended Si surfaces show that the unpassivat-
ed surface tends to minimize its energy by reducing the number of
dangling bonds through relaxation and surface reconstructions
[7,8]. Theoretical studies of similar reconstructions on nanocrystal
surfaces show that such reconstructions result in severe distortion
of the surface bonds [9]. Defect states associated with bond distor-
tions are expected to appear in the band gap [10]. Puzder et al. [11]
have predicted that surface step reconstructions unique to the
highly curved surfaces of nanoparticles reduce the optical gap sub-
stantially. Relaxation will not eliminate all dangling bonds and
by capping them with hydrogen.
Hydrogen-terminated extended Si(100) and Si(111) surfaces
have been studied using TPD [7,12,13]. A monohydride-terminated
Si(100) surface (one H atom per Si atom) can be obtained even
with saturation doses of atomic hydrogen at elevated temperatures
(ꢀ650 K). If H-dosing is done around 400 K, the surface first satu-
rates with monohydride species after which dihydride species
(two H atoms per Si atom) begin to form [14,15]. If dosing is done
at low temperatures (ꢀ210 K), a trihydride species is formed. In
TPD spectra, hydrogen desorption from the hydride species occurs
at different temperatures [13]: the monohydride (b1) desorbs
around 780 K, the dihydride (b2) desorbs around 680 K and the tri-
hydride (b3) appears as a broad feature <600 K. Trihydride species
can abstract an H atom from a neighboring hydride and form SiH4,
which desorbs around 600 K. This results in etching of the Si
surface.
The various hydride species on porous silicon surfaces and their
desorption kinetics have been investigated using FTIR by Gupta
et al. [16] and using TPD by Martin et al. [17]. They conclude that
the as-prepared surfaces are covered by a mixture of mono-, di-
and trihydride species. FTIR studies conducted on aerosolized Si
nanocrystals have shown that nanocrystal surfaces predominately
contain trihydride and dihydride species [18].
We use TPD to probe the nature of the hydride termination on
nanocrystal surfaces. Si nanocrystals are grown on SiO2 surfaces at
* Corresponding author. Tel.: +1 512 471 4689; fax: +1 512 471 7060.
E-mail address: ekerdt@che.utexas.edu (J.G. Ekerdt).
0039-6028/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2009.02.033

1122
N. Salivati, J.G. Ekerdt / Surface Science 603 (2009) 1121–1125
875 K. Large particles (50–200 nm) were grown using chemical va-
por deposition (CVD) to see how they mimic the desorption fea-
tures from extended Si surfaces. Nanoparticles (<8 nm) were
grown using a mixed process of hot wire chemical vapor deposi-
tion (HWCVD) and CVD. Similar to extended surfaces, we find
nanocrystal surfaces are terminated by a combination of deuteride
species; however, in contrast to extended surfaces dideuteride and
trideuteride species start forming before the monodeuteride state
saturates.
2. Experimental
The experimental apparatus consists of a growth chamber, a
heating furnace and an analytical chamber, all connected to each
other via an intermediate transfer chamber. The base pressure of
the growth and analytical chambers is 5 ꢁ 10ꢂ9 Torr. The transfer
chamber and furnace have a base pressure of 3 ꢁ 10ꢂ8 Torr. More
details of the system are available elsewhere [19,20]. Si particles
were grown in the growth chamber with disilane (Voltaix; 4% in
He). A hot tungsten filament situated 3 cm away from the substrate
is used in HWCVD; a constant filament current of 4 A was used that
led to an estimated filament temperature of 1500 °C [21]. Disilane
partial pressures during CVD and HWCVD were 10ꢂ4 Torr and
8 ꢁ 10ꢂ8 Torr, respectively. The samples were transferred in situ
to the analytical chamber where a differentially pumped Thermo
Electron Smart IQ+ mass spectrometer covered by a nose cone with
a 0.5 cm opening was used for TPD. Mass alignment of the mass
spectrometer was done using He+ (mass/charge 4 [m/e 4]), Ar+
(m/e 40), and Xe+‘ (m/e 129). All TPD spectra were collected using
an ionizing energy of 57 eV and a constant dwell time. Particle size
was determined using scanning electron microscopy (SEM) [Zeiss
Supra 40VP] under an accelerating voltage of 15 kV after coating
the sample with 1 nm of Ir. Secondary electrons are detected and
the smallest particles that can be resolved are ꢀ4 nm in diameter.
Si(100) wafers with 10 nm of thermal oxide were supplied by
Freescale, Inc. These wafers were cut into squares of
1.6 cm ꢁ 1.6 cm. Each piece was sequentially rinsed with acetone,
ethanol and deionized (DI) water and then dipped in piranha solu-
tion (60% DI water, 10% H2O2 and 20% H2SO4) for 15 min. Subse-
quently, the samples were dipped in 2% HF for 4 s removing
ꢀ1 nm of the oxide, thoroughly rinsed in DI water and then dried
using high purity helium. The cleaned samples were mounted on
a 1 in. diameter molybdenum ring and inserted into an ultrahigh
vacuum system using a load lock.
The samples were then annealed in the furnace at 975 K for
15 min and allowed to cool to ꢀ375 K and baseline spectra were
taken for various m/e signals in the analytical chamber. Baseline
spectra were taken by heating the sample at 3 K/sþfrom þ375 K
to 1000 K while monitoring signals Dþ2 , SiD+, SiD2 , SiD3 and
SiDþ4 corresponding to m/e ratios of 4, 30, 32, 34 and 36, respec-
tively. The sample was then transferred to the growth chamber
and heated to the growth temperature of 875 K and allowed to
equilibrate thermally for 15 min. A 17 min CVD time was used
to grow 50–200 nm sized Si islands (Fig. 1). Small particles
<8 nm were grown using a two step process: HWCVD for 5 min
followed by CVD for 4.5 min (Fig. 2). At a growth temperature
of 875 K, Si nanoparticles are expected to be crystalline [22,23].
The inset of Fig. 2 presents a transmission electron microscopy
image (JOEL 2010F operated at 200 keV) of a 5 nm nanocrystal
that was grown by the two step process of HWCVD seeding and
CVD on a 50 nm silicon dioxide membrane obtained from Struc-
ture Probe, Inc.
Fig. 1. SEM image of silicon nanoparticles after 17 min CVD.
Fig. 2. SEM image of silicon nanoparticles after 5 min HWCVD seeding and 4.5 min
CVD. Inset-TEM image of crystalline particle.
any disilane fragments that may have readsorbed on the nano-
crystal surface while the growth chamber was being evacuated
to base pressure. The sample was transferred back to the growth
chamber after it cooled to ꢀ375 K and the cracking filament,
which was ꢀ3 cm away from the sample was turned on under
a D2 pressure of 2 ꢁ 10ꢂ6 Torr. During dosing, the sample temper-
ature rose by no more than 15–20 K. The sample temperature
during dosing was estimated to be around 375–395 K. After dos-
ing is complete, the sample was transferred to the analytical
chamber and another TPD spectrum was collected. Sample tem-
perature was determined using a type-K reference thermocouple
that was fixed between the heating bulb and the sample. In sep-
arate experiments this reference thermocouple was calibrated
against an instrumented sample that had a type-K thermocouple
attached to the sample surface.
3. Results and discussion
Over Si(100) the monohydride and dihydride are denoted by b1
and b2, which desorb around 800 and 680 K respectively; the trihy-
dride denoted by b3 appears as a broad feature below 600 K [13].
We use similar nomenclature as that used for the hydride states
of Si(100), namely: b1 for monohydride species, b2 for dihydride
species, and b3 for trihydride species. D2 exposures are reported
in Langmuir (1 L = 10ꢂ6 Torr s); the efficiency of D2 cracking and
the absolute D atom flux are unknown.
After growth was complete, the sample was allowed to cool to
ꢀ375 K. It was then transferred to the analytical chamber where a
3 K/s ramp was employed to flash the sample at 975 K to remove

N. Salivati, J.G. Ekerdt / Surface Science 603 (2009) 1121–1125
1123
3.1. Experiments with large particles
Large Si particles were grown using a 17 min CVD procedure
and subjected to various doses of atomic deuterium. In between
each dose, the samples were annealed at 975 K for 15 min. Fig. 3
shows Dþ2 (m/e 4) TPD spectra for various exposures starting from
120 L. Two peaks stand out clearly, one around 800 K and the other
around 680 K and given similarities to TPD from Si(100), we asso-
ciate these with monodeuteride and dideuteride species, respec-
tively. Unlike extended surfaces [14,15], the dideuteride state
starts filling in before the monodeuteride saturates. With increas-
ing exposures, the intensities of both features increase indicating
that both monodeuteride and dideuteride species form simulta-
neously on the nanocrystal surface. The intensity of the dideuteride
is always lower than the monodeuteride. With increasing expo-
sure, the leading edge of the b2 feature shifts to a lower tempera-
ture raising the possibility of a new species, such as trideuteride
(b3), which can form at higher exposures.
Cheng et al. [13] have demonstrated with Si(100) that trihy-
dride species can desorb as silane by extraction of an H-atom from
a neighboring Si-hydride thus etching the Si surface. If trideuteride
species (SiD3) formed on the nanocrystal surface, SiD4 evolution
should be detected during TPD. SiD+ (m/e 30), SiD2þ (m/e 32) and
SiDþ3 (m/e 34) and SiDþ4 signals were monitored during TPD. The
SiDþ4 signal has a very low intensity, so other fragments are used
as an indication of trideuteride formation on the nanocrystal sur-
face. Fig. 4 shows the D2 TPD spectrum along with the correspond-
ing etching products from an experiment in which the particles
were subjected to 1800 L D2. The b3 feature is more apparent than
in Fig. 3. The presence of b3 and etching products confirms the
presence of trideuteride species over the nanoparticles.
Fig. 4. D2 desorption spectrum along with the corresponding etching products
observed from large particles. The lines below each spectrum indicate the baseline
signal intensity, which has been truncated for the sake of clarity.
the nature of the deuteride termination. The results are displayed
in Fig. 5. The TPD spectra are identical and there is no change in
the intensities or locations of the peaks. Additionally there were
no differences in the SEM images before and after the experiments,
indicating that the intermediate annealing does not substantially
affect the size of big particles.
3.2. Experiments with small particles
With extended Si surfaces, trihydride formation is not observed
except at low temperature (T = 210 K) H-dosing [13]. On the con-
trary, there is substantial trideuteride formation on nanocrystal
surfaces when subjected to atomic deuterium doses at 375 K. The
Mo ring can introduce desorption features [24]. In order to ensure
that none of these signals originated from the Mo ring, control
experiments were conducted in which the Mo sample support ring
was subjected to 17 min exposures to disilane at 875 K followed by
atomic D doses. TPD spectra did not show the presence of any etch-
ing products, confirming that nanoparticles were solely responsi-
ble for the etching products.
A mixed process involving HWCVD seeding for 5 min followed
by CVD for 4.5 min was employed to grow small Si particles
<8 nm in diameter. After each TPD, the sample was flashed at
975 K using a 3 K/s ramp instead of the 15 min anneal in order to
minimize agglomeration of the particles. The TPD spectra from re-
peated 1800 L doses are shown in Fig. 6. Unlike with large particles,
there is an increase in b3 with each dose, and b1 and b2 decrease in
intensity. The total area decreases with each TPD such that after
TPD 5 the area is 84% of the area associated with TPD 1.
The decrease in TPD area, primarily of the b1 and b2 features,
likely results from a loss of Si nanocrystal surface area caused by
both etching and agglomeration. Similar to the large Si particles,
The nanoparticles were also repeatedly subjected to a constant
exposure of 1800 L D2 with intermittent annealing at 975 K for
15 min to check if the intermediate annealing had any effect on
Fig. 5. D2 desorption spectra from large Si particles subjected to repeated doses of
Fig. 3. D2 desorption from large Si particles.
1800 L.

1124
N. Salivati, J.G. Ekerdt / Surface Science 603 (2009) 1121–1125
Fig. 6. D2 desorption spectra from small Si particles subjected to repeated doses of
1800 L. The thin line corresponds to a control experiment in which nanocrystals
were not grown on the SiO2/Si(100) substrate.
Fig. 8. D2 desorption spectra from small particles. Inset: D2 desorption from small
particles subjected to 15 L exposure.
From experiments with large particles, we know that the differ-
ent deuteride species form simultaneously unlike with extended Si
surfaces; the same was observed with small particles. TPD spectra
from smaller doses are displayed in Fig. 8. The 15 L dose is shown
in the inset for clarity. The b2 feature is present even at low expo-
sures of 15 L and at higher exposures, both b1 and b2 increase in
intensity illustrating that the monodeuteride does not saturate be-
fore the dideuteride fills in. Since no etching products are observed
at such low exposures, we conclude that the feature at 450 K is
from the Mo ring on which the sample is mounted.
With extended Si surfaces, it is possible to selectively desorb a
particular hydride species, isolating the hydrides that desorb at
higher temperatures. This technique may be useful in isolating a
specific deuteride species on the nanocrystal surface. A 1800 L dose
was employed, which was shown to create a surface covered by a
mix of deuteride species (Fig. 7). The sample was then annealed
at different temperatures for 15 min and allowed to cool to around
375 K following which a TPD spectrum was recorded. The ‘‘no an-
neal” spectrum in Fig. 9 is the initial TPD spectrum from the
1800 L dose. By annealing at 650 K, it is possible to completely iso-
late the monodeuteride. Annealing at 500 K results in a surface ter-
etching products are observed as shown in Fig. 7, and the b3 feature
in Fig. 6 is associated with etching. SEM images of the sample after
the experiments (not shown) indicate some coalescence of parti-
cles; however, the resolution of the SEM does not reveal an average
smaller size as a consequence of etching in those particles that
have not coalesced. While the b1 and b2 area decrease can reason-
ably be connected to agglomeration, the exact cause for an increase
in the b3 feature with repeated TPD cycles remains unresolved. One
possibility could be a progressive roughening of nanocrystal sur-
faces with TPD cycles that acts to increase the surface concentra-
tion of sites where trihydride species form.
The thin line in Fig. 6 is from a control experiment in which the
Mo ring and silicon wafer were subjected to 1800 L D2. Winken-
werder et al. [24] have shown that the desorption feature at
800 K is due to D2 desorption from silicon dioxide. Using control
experiments (not shown), we conclude that the feature around
450–600 K is from the Mo ring. The intensities of the features from
the Mo ring and silicon dioxide are much lower than those from
the particles.
Fig. 7. D2 desorption spectrum along with the corresponding etching products
observed from the small particles subjected to the first cycle of D2 exposure. The
lines below each spectrum indicate the baseline signal intensity, which has been
truncated for sake of clarity.
Fig. 9. D2 desorption spectra from large Si particles subjected to repeated doses of
1800 L and subsequently annealed at 650, 550, 500, 450 and 700 K for 15 min.

N. Salivati, J.G. Ekerdt / Surface Science 603 (2009) 1121–1125
1125
mination devoid of trideuteride species. A slightly lower anneal
References
temperature of 450 K results in the retention of the trideuteride
species and a 700 K anneal desorbs all the deuteride species leaving
behind a small feature that is due to deuterium desorption from the
SiO2 surface [24].
[1] L.T. Canham, Appl. Phys. Lett. 57 (1990) 1046.
[2] Z.H. Lu, D.J. Lockwood, J.M. Baribeau, Nature 378 (1995) 258.
[3] N.M. Park, C.J. Choi, T.Y. Seong, S.J. Park, Phys. Rev. Lett. 86 (2001) 1355.
[4] C.S. Yang, R.A. Bley, S.M. Kauzlarich, H.W.H. Lee, G.R. Delgado, J. Am. Chem. Soc.
121 (1999) 5191.
[5] J.D. Holmes, K.P. Johnston, R.C. Doty, B.A. Korgel, Science 287 (2000) 1471.
[6] F. Koch, V. Petrova-Koch, T. Muschik, A. Nikolov, V. Gavrilenko, Mat. Res. Soc.
Symp. Proc. 283 (1992) 197.
4. Summary
[7] H.N. Waltenburg, J.T. Yates Jr., Chem. Rev. 95 (1995) 1589.
[8] A. Ramstad, G. Brocks, P.J. Kelly, Phys. Rev. B 51 (1995) 14504.
[9] E. Kaxiras, Phys. Rev. B 56 (1997) 13455.
[10] L. Liu, C.S. Jayanthi, S.Y. Wu, J. Appl. Phys. 90 (2001) 4143.
[11] A. Puzder, A.J. Williamson, F.A. Reboredo, G. Galli, Phys. Rev. Lett. 91 (2003)
157405.
[12] C.M. Greenlief, S.M. Gates, P.A. Holbert, J. Vac. Sci. Technol. A 7 (1989) 1845.
[13] C.C. Cheng, Y.T. Yates Jr., Phys. Rev. B 45 (1991) 4041.
[14] G. Hess, M. Russell, B. Gong, P. Parkinson, J.G. Ekerdt, J. Vac. Sci. Technol. A 15
(1997) 1129.
[15] P.S. Parkinson, Ph.D. dissertation, University of Texas at Austin, 2000.
[16] P. Gupta, V.L. Colvin, S.M. George, Phys. Rev. B 37 (1988) 8234.
[17] P. Martın, J.F. Fernandez, C. Sanchez, Mater. Sci. Eng. B 108 (2004) 166.
[18] J. Holm, J.T. Roberts, J. Am. Chem. Soc. 129 (2007) 2496.
[19] W.T. Leach, J.H. Zhu, J.G. Ekerdt, J. Cryst. Growth 240 (2002) 415.
[20] W.T. Leach, J.H. Zhu, J.G. Ekerdt, J. Cryst. Growth 243 (2002) 30.
[21] S.S. Coffee, D. Shahrjerdi, S.K. Banerjee, J.G. Ekerdt, J. Cryst. Growth 308 (2007)
269.
TPD experiments reveal that Si nanocrystals both big and small
behave very differently from extended Si surfaces. Formation of the
dideuteride and trideuteride species seems to be facilitated by the
curvature of the nanocrystal. The presence of SiD4 fragments in the
TPD spectra is confirmation of the presence of trideuteride species.
Interestingly, we find that unlike extended surfaces, trideuteride
species are prevalent on the nanocrystal surface even when the
dosing temperature is around 375 K. Tsai et al. [25] have shown
with FTIR that porous silicon gives only a weak luminescence when
SiH species are present on the surface; the luminescence starts to
increase only when SiH2 species are present. Our results indicate
that it is possible to control the type of hydride termination on
nanocrystal surfaces.
[22] M. Fukuda, K. Nakagawa, S. Miyazaki, M. Hirose, Appl. Phys. Lett. 70 (1997)
2291.
Acknowledgements
[23] T. Baron, F. Martin, P. Mur, C. Wyon, M. Dupuy, J. Cryst. Growth 209 (2000)
1004.
[24] W.A. Winkenwerder, J.G. Ekerdt, Surf. Sci. 602 (2008) 2796.
[25] C. Tsai, K.-H. Li, D.S. Kinosky, R.-Z. Qian, T.-C. Hsu, J.T. Irby, S.K. Banerjee, A.F.
Tasch, J.C. Campbell, B.K. Hance, J.M. White, Appl. Phys. Lett. 60 (1992) 1700.
The authors thank the Welch Foundation (Grant No. F1502) and
D.C. Sirica Corporation for funding.