Download Characterization of boron charge traps at the interface of Si/SiO 2

APPLIED PHYSICS LETTERS 95, 062102 共2009兲
Characterization of boron charge traps at the interface of Si/ SiO2 using
second harmonic generation
H. Park,1,a兲 J. Qi,1 Y. Xu,1 K. Varga,1 S. M. Weiss,1,2 B. R. Rogers,3 G. Lüpke,4 and
N. Tolk1
1
Department of Physics and Astronomy, Vanderbilt University, Nashville, Tennessee 37235, USA
Department of Electrical Engineering and Computer Science, Vanderbilt University, Nashville,
Tennessee 37235, USA
3
Department of Chemical and Biomolecular Engineering, Vanderbilt University, Nashville,
Tennessee 37235, USA
4
Department of Applied Science, College of Mary and Williams, Williamsburg, Virginia 23187, USA
2
共Received 25 June 2009; accepted 17 July 2009; published online 10 August 2009兲
We report results from optical second harmonic generation studies of boron charge traps near the
interface of Si/ SiO2. Our data suggest that a static electric field at the interface is formed during the
oxide growth process due to the presence of negative boron ions 共B−兲 in the silicon substrate and
positive boron ions 共B+兲 in the oxide. We demonstrated that the B+ state traps could be filled through
the creation of neutral boron states created by internal photoelectron emission. By fitting our data,
we found that the effective interface susceptibility 兩␹共2兲兩 depends on doping concentration. © 2009
American Institute of Physics. 关DOI: 10.1063/1.3202420兴
Electric-field-induced second harmonic generation
共SHG兲 is useful in the study of the Si/ SiO2 system
interface.1–4 When the oxide is less than 10 nm thick, the
typical response of time-dependent SHG 共TD-SHG兲 is a
slow monotonic increase, which saturates after a long time.5,6
The TD-SHG intensity can be described in general by6
I2␻ = 兩␹共2兲 + ␹共3兲E共t兲兩2共I␻兲2 .
共1兲
␹共2兲 and ␹共3兲 are effective nonlinear second- and third-order
electric susceptibilities in the near Si/ SiO2 interface region,
respectively. E共t兲 is a TD electric field across the interface. If
the oxide is relatively thin, electrons from the silicon valence
band may be excited by three-photon absorption and travel
through the oxide’s conduction band to the surface of the
oxide. Electrons reaching the oxide surface may be captured
by ambient oxygen molecules. The electrons trapped on the
surface and the holes remaining in the silicon substrate form
a capacitorlike field, which extends over the interface in the
direction of the surface normal.3–6
In this paper we present an additional electric field formation mechanism. This mechanism results in an electric
field forming at the interface of highly boron-doped silicon
and SiO2 due to boron impurities in the oxide and in the
silicon substrate. The TD-SHG measurements are electric
field sensitive at the interface and show behavior that can be
attributed to the presence of the boron charge traps. Specifically, we have identified an initial sharp decrease in the TDSHG measurement 关Fig. 1共a兲兴 arising from boron charge
traps. In contrast, highly antimony-doped wafers did not
show a decrease in the SHG measurement 关inset of Fig.
1共a兲兴, which indicates that no significant charge traps were
formed by antimony doping.
In this study, the interface between the native oxide and
silicon substrate is investigated under the following optical
experimental conditions. A Mira Ti:sapphire laser is used to
generate a P-polarized 800 nm beam, which is focused onto
the samples at 45°. The laser repetition rate is 75 MHz with
a兲
Electronic mail: [email protected].
0003-6951/2009/95共6兲/062102/3/$25.00
a pulse width of about 150 fs. Reflected SHG beam 共400 nm兲
intensity is measured by a photomultiplier tube with a photon
counter. The samples are aligned so that the plane-ofincidence is always parallel to the 具100典 crystal plane. We
observe the same result when the sample is rotated to 具110典
crystal orientation. All measurements are performed in air at
room temperature and in dark conditions. Following is a list
of the silicon 共100兲 wafers used in this study. For each wafer,
the dopant species, resistivity, and dopant concentration are
shown.
• Wafer A: boron, 5 – 15 ⍀ cm, 2.7⫻ 1014 – 9 ⫻ 1015 cm−3.
FIG. 1. 共a兲 TD-SHG signal from B-doped silicon 共0.01– 0.02 ⍀ cm兲. Inset
is from Sb-doped silicon 共0.005– 0.02 ⍀ cm兲. At the beginning of the each
measurement the beam was blocked, and later it was blocked again for a
short period. 共b兲 The interpretation of the SHG data corresponding to the
interface electric fields. White arrow: electric field due to boron charge traps
and dark arrow: electric field due to oxygen traps on the surface.
95, 062102-1
© 2009 American Institute of Physics
Downloaded 10 Aug 2009 to 129.59.73.55. Redistribution subject to AIP license or copyright; see http://apl.aip.org/apl/copyright.jsp
062102-2
• Wafer
⫻ 1018
• Wafer
⫻ 1020
• Wafer
⫻ 1019
Park et al.
B: boron, 0.01– 0.02 ⍀ cm, 3.1⫻ 1018 – 8.4
cm−3.
C: boron, 0.001– 0.009 ⍀ cm, 9.7⫻ 1018 – 1.3
cm−3.
D: antimony, 0.005– 0.020 ⍀ cm, 1.2⫻ 1018 – 1.1
cm−3.
The thicknesses of the native oxide layers were measured by ellipsometry. Within experimental error, all wafers
were found to contain a 2 nm thick oxide. A laser power of
400 mW corresponds to an 8.8 GW/ cm2 peak intensity on
the sample surface.
Only the highly boron-doped wafers B and C showed an
initial marked decrease in our TD-SHG measurements. However, we believe that wafer A would also show the initial
SHG decrease under appropriate experimental conditions.
Because the time-dependency in SHG comes from the electric field at the interface of Si/ SiO2, we suggest that the
source of this initial marked decrease is the presence of a
substantial electric field at the Si/ SiO2 interface prior to the
initiation of an optical measurement. We attribute this field to
a concentration of B− in the silicon and a concentration of B+
in the oxide, forming a near capacitorlike field at the interface. The initial SHG signal at t = 0 comes from the combination of ␹共2兲 and ␹共3兲E共t = 0兲 as represented in Eq. 共1兲, where
E共t = 0兲 is the initial electric field. This initial electric field
arises from boron charge traps as indicated by ① in Fig.
1共b兲. The direction of this initial field is from the oxide to the
silicon substrate. During laser illumination, photoexcited
electrons may both fill the boron B+ charge traps and be
captured by ambient oxygen traps on the oxide surface. In
the first case, the electrons captured by the B+ charge traps
decrease the magnitude of the initial field. In the second case
involving the oxygen traps, a TD capacitorlike field is created opposite to the direction of the initial boron-induced
electric field. Thus the net electric field decreases as indicated by the decreasing SHG signal, designated by ② in Fig.
1共b兲. At the minimum of the SHG signal, the net electric field
becomes zero, though the two separate electric fields may
not be zero, as indicated by ③ in Fig. 1共b兲. After reaching a
minimum, the SHG signal increases as more electrons are
transferred onto the oxygen molecules on the oxide surface
and onto the B+ charge traps until both are saturated, as
shown in ④ and ⑤ in Fig. 1共b兲.
Because the oxide layer is relatively thin, oxygen charge
traps can easily release electrons. When the laser beam is
blocked, the electrons captured by the oxygen molecules can
tunnel back to empty states in the valence band of the silicon
substrate, which results in a decrease in the electric field at
the interface.3,4 An important question is what happens to the
boron charge traps after they are filled. Will they also spontaneously lose the trapped electrons? If so, one would expect
that after blocking the beam, the sample would tend toward
its original configuration. However, when the beam was unblocked after being blocked for a short period, no initial
SHG decrease was observed as shown in Fig. 1共a兲. After a
blocking time of approximately 1 h, no significant recovery
of the initial SHG decrease was observed. However, if the
beam was blocked for more than several hours, a discernable
recovery of the initial SHG decrease was observed. Thus we
infer that the detrapping rate for the boron charge traps is
much less than the detrapping rate for the surface oxygen
Appl. Phys. Lett. 95, 062102 共2009兲
traps. Quantum mechanical tunneling can be determined by
barrier thickness, barrier height, and the availability of unoccupied levels on the other side of the barrier. Our interpretation is supported by a density functional calculation, which
shows that neutral boron is more stable in amorphous SiO2
than positively ionized boron.7 Prior to this theoretical work,
it was suggested that B atoms diffuse into SiO2 as B+ ions.8
More recently, another group showed that boron has various
stable and metastable forms in crystalline and defective
SiO2, and they also showed that B+ is one of these stable
forms,9,10 which showed similar results from the previous
research7 in amorphous SiO2.
It is a well-known experimental result that when oxide
grows on silicon, boron atoms are incorporated into the oxide near the interface. The ratio of the concentration of a
dopant in silicon to the concentration of the dopant in the
oxide at the interface is called the segregation coefficient m.
Boron is the only typical silicon dopant that has a segregation coefficient less than one.11 This means that at the interface, the concentration of boron in the oxide is greater than
that in the silicon substrate. Based on our measurements and
the theoretical calculations involving the segregation coefficient of boron, we suggest a plausible mechanism for creating the boron positive charge traps in the oxide. At room
temperature, a boron atom may be easily ionized by acquiring an electron from a silicon atom, leaving a hole state in
the valence band. During the growth of silicon oxide on the
silicon substrate, boron atoms near the interface are incorporated into the oxide as a neutral state near the interface. Some
of the boron atoms in the oxide can be ionized as positive
ions leaving an electron in silicon substrate. The electron can
occupy the hole state of the silicon or an acceptor state of the
boron atom in silicon. Therefore, across the interface B− and
B+ ions are formed in the silicon and the oxide, respectively
关Fig. 2共a兲兴. These charge traps are responsible for the initial
electric field and can be detected by the TD-SHG measurement.
Because electron injection to the oxygen molecules is
carried out by three-photon absorption, it requires a highly
intense beam.3,4,12 When our incoming laser beam intensity
is below 200 mW, there is no time-dependency in the SHG
from wafers A and D, which means that the electron injection
to the oxygen molecules is negligible. In this manner, we
could fill the boron-induced traps first and then fill the oxygen charge traps. As shown in Fig. 2共b兲 共red lower graph兲, at
the beginning, when the laser power is 173 mW, the photoinjected electrons fill up the boron-induced charge traps, and
then, during the blocking, the laser beam intensity is increased to 436 mW. When the laser beam is unblocked again,
the photoinjected electrons fill oxygen charge traps, inducing
the E1 electric field in Fig. 2共a兲 共dark arrow兲. The fact that
we were able to fill the B+ traps at relatively low intensity is
evidence that it require less than a three-photon absorption
process to fill up the B+ traps. The determination of the energy level of these traps is a subject of ongoing experimental
and theoretical study.
Our measurements show that the magnitude of the initial
decrease depends on doping concentrations 共Fig. 3兲. When
the laser power is relatively high, the initial SHG decrease
cannot be observed for wafer B since the oxygen traps are
occupied very quickly 关Fig. 3共a兲: 396 mW兴. However, at low
laser power, the oxygen traps have a very low probability of
occupation 关Fig. 3共a兲: 154 mW and Fig. 3共b兲: 155 mW兴.
Downloaded 10 Aug 2009 to 129.59.73.55. Redistribution subject to AIP license or copyright; see http://apl.aip.org/apl/copyright.jsp
062102-3
Appl. Phys. Lett. 95, 062102 共2009兲
Park et al.
FIG. 3. 共Color online兲 共a兲 SHG results from B-doped silicon
共0.01– 0.02 ⍀ cm兲 collected using different laser powers. 共b兲 SHG results
B-doped silicon 共0.001– 0.009 ⍀ cm兲 using different laser powers. The data
were taken under identical experimental conditions.
FIG. 2. 共Color online兲 共a兲 Depiction of boron charge traps formed in borondoped silicon during oxidation. White arrows: the electric field from the
boron charge traps and photoinjected electron-induced oxygen charge traps.
Dark arrow: the electric field from the oxygen traps. 共b兲 Blue 共upper兲 graph:
filling both traps simultaneously between 10 and 178 s and then filling only
the oxygen charge traps after 266 s. Red 共lower兲 graph: filling only the
boron-induced charge traps first between 10 and 178 s and then filling only
the oxygen charge traps after 266 s. The sample is B-doped silicon
共0.001– 0.009 ⍀ cm兲.
Thus only the decrease in the boron-induced electric field is
observed.
Equation 共1兲 was numerically fit to the data to obtain
relative information on the susceptibilities. For an intrinsic
silicon wafer, ␹共2兲 and ␹共3兲 can be considered as real numbers
because both ␹共2兲 and ␹共3兲 effectively come from a few
atomic layers at the interface and absorption is very low at
800 nm wavelength. However, when the wafer is highly
doped, these parameters cannot be assumed to be real numbers. If the susceptibilities are real numbers, the minimum
SHG in Eq. 共1兲 is zero, and therefore it will not fit the data.
In general, ␹共2兲 and ␹共3兲 are complex numbers and Eq. 共1兲 can
be expressed as
I2␻ = 兩兩␹共2兲兩 + 兩␹共3兲兩ei␪E共t兲兩2共I␻兲2 ,
where ␪ is a relative phase between the ␹
electric field E共t兲 can be expressed by
E共t兲 = E0e−t/t1 − E1共1 − e−t/t2兲,
共2兲
共2兲
共3兲
and ␹ . The TD
共3兲
where E0 is the magnitude of the electric field from boron
charge traps, E1 is the magnitude of the electric field from
oxygen charge traps, t1 is the detrapping time constant of the
boron-induced charge trap, and t2 is the trapping time constant of the oxygen charge traps. For accurate fitting, electron
detrapping from the oxygen traps should also be considered.
However, if we need to know only the effective value of
兩␹共2兲兩, introducing a detrapping constant 共t3兲 does not change
the value of 兩␹共2兲兩. The value of 兩␹共3兲兩 cannot be separated
from the fitting because it is combined with the electric fields
E0 and E1. In Figs. 3共a兲 and 3共b兲, Eqs. 共2兲 and 共3兲 are fit to
the two data sets 共307 and 305 mW兲. The fitting result indicates that 兩␹共2兲兩 increases with increasing doping concentrations. We used van der Pauw four-point probe measurements
to determine the resistivities of wafers B and C to be 0.0165
and 0.00519 ⍀ cm, respectively. We observed that the
effective 兩␹共2兲兩 increased nonlinearly with respect to the
boron concentration by comparing the relative ratios
共兩␹共2兲兩wafer C / 兩␹共2兲兩wafer B = 1.55 and nwafer C / nwafer B = 4.72兲.
It is important to note that there is a band-bending shift
as the field direction is changed. It is possible that this band
bending may affect our SHG measurements, especially the
filling rate due to a change in the barrier height in two ways.
共1兲 The change in barrier height may influence the number of
lower-energy electrons that make it over the barrier, and 共2兲
if the barrier is high enough, the order number of photons 共a
multiphoton process兲 required to scale the barrier may
change. Previous research carried out in our group showed
that even at intensities of 100 GW/ cm2, well above the intensities used in the present experiments, there is no significant change in the trap-filling rate as the field changed, nor
was there any indication of a change in the multiphoton order number.13 Therefore, we can conclude that band bending
does not significantly alter our results or our interpretation.
This work was supported by the Department of Energy
共DOE兲 under Grant No. DE-FGO2-99ER45781. We are also
grateful to Dr. John Rozen for carrying out the four-pointprobe measurement.
1
C. H. Lee, R. K. Chang, and N. Bloembergen, Phys. Rev. Lett. 18, 167
共1967兲.
2
O. A. Aktsipetrov, A. A. Fedyanin, E. D. Mishina, A. N. Rubtsov, C. W.
van Hasselt, M. A. C. Devillers, and Th. Rasing, Phys. Rev. B 54, 1825
共1996兲.
3
W. Wang, G. Lüpke, M. Gi Ventra, S. T. Pantelides, J. M. Gilligan, N. H.
Tolk, I. C. Kizilyalli, P. K. Roy, G. Margaritondo, and G. Lucovsky, Phys.
Rev. Lett. 81, 4224 共1998兲.
4
T. Scheidt, E. G. Rohwer, H. M. von Bergmann, and H. Stafast, Phys. Rev.
B 69, 165314 共2004兲.
5
J. G. Mihaychuk, J. Bloch, Y. Liu, and H. M. van Driel, Opt. Lett. 20,
2063 共1995兲.
6
J. Bloch, J. G. Mihaychuk, and H. M. van Driel, Phys. Rev. Lett. 77, 920
共1996兲.
7
V. Zubkov, J. P. Senosiain, S. Aronowitz, V. Sukharev, and C. B. Musgrave, Mater. Res. Soc. Symp. Proc. 610, B5.11.1 共2000兲.
8
Y. Wu, H. Niimi, H. Yang, G. Lucovsky, and R. B. Fair, J. Vac. Sci.
Technol. B 17, 1813 共1999兲.
9
M. Otani, K. Shiraishi, and A. Oshyyama, Phys. Rev. Lett. 90, 075901
共2003兲.
10
M. Otani, K. Shiraishi, and A. Oshyyama, Phys. Rev. B 68, 184112
共2003兲.
11
S. A. Campbell, The Science and Engineering of Microelectronic Fabrication, 2nd ed. 共Oxford University Press, New York, 2001兲.
12
J. G. Mihaychuk, N. Shamir, and H. M. van Driel, Phys. Rev. B 59, 2164
共1999兲.
13
Z. Marka, R. Pasternak, S. N. Rashkeev, Y. Jiang, S. T. Pantelides, and N.
H. Tolk, Phys. Rev. B 67, 045302 共2003兲.
Downloaded 10 Aug 2009 to 129.59.73.55. Redistribution subject to AIP license or copyright; see http://apl.aip.org/apl/copyright.jsp