Download Damage in hydrogen plasma implanted silicon

JOURNAL OF APPLIED PHYSICS
VOLUME 90, NUMBER 4
15 AUGUST 2001
Damage in hydrogen plasma implanted silicon
Lianwei Wang,a) Ricky K. Y. Fu, Xuchu Zeng, and Paul K. Chub)
Department of Physics and Materials Sciences, City University of Hong Kong, Tat Chee Avenue, Kowlong,
Hong Kong
W. Y. Cheung and S. P. Wong
Department of Electronic Engineering, The Chinese University of Hong Kong, Shatin, New Territories,
Hong Kong
共Received 20 November 2000; accepted for publication 4 June 2001兲
The damage and defects created in silicon by hydrogen plasma immersion ion implantation 共PIII兲
are not the same as those generated by conventional beamline ion implantation due to the difference
in the ion energy distribution and lack of mass selection in PIII. Defect generation must be well
controlled because damage in the implanted and surface zones can easily translate into defects in the
silicon-on-insulator structures synthesized by the PIII/wafer bonding/ion-cut process. The defect
formation and its change with annealing temperature were investigated experimentally employing
channeling Rutherford backscattering spectrometry, secondary ion mass spectrometry, and
atomic-force microscopy. We also calculated the damage energy density of the three dominant
⫹
hydrogen species in the plasma 共H⫹ , H⫹
2 , and H3 兲 as well as displacement of silicon atoms in the
silicon wafer. H⫹
2 creates the most damage because its damage energy density is very close to the
silicon threshold energy. The effects of atmospheric gaseous impurities unavoidably coimplanted
from the overlying plasma are also modeled. Even though their concentration is usually small in the
plasma, our results indicate that these gaseous impurities lead to significant silicon atom
displacement and severe damage in the implanted materials. © 2001 American Institute of Physics.
关DOI: 10.1063/1.1389073兴
I. INTRODUCTION
beam in beamline ion implantation is mass and energy selected to comprise only one ion species with a specific energy. Therefore, the formation and effects of damage and
defects in hydrogen plasma implanted silicon are expected to
be different from those of beamline ion implantation. The
damage in the hydrogen plasma implanted wafer can conceivably translate into defects in the SOI structure produced
by PIII/ion-cut/wafer bonding and impact the yield of the
fabrication process. In order to investigate this problem, the
following factors must be considered:6,7
Silicon on insulator 共SOI兲 is an important technology for
radiation-hardened integrated circuits as well as low-power,
low-voltage, and high-temperature microelectronics. In fact,
with the introduction of consumer microelectronic chips fabricated using SOI, SOI is now regarded to be not just the
material of the future but also the material of the present. 1,2
Compared to the more mature SOI technologies such as
separation by implantation of oxygen 共SIMOX兲3 and
smart-cut,4 synthesis of SOI materials by plasma immersion
ion implantation 共PIII兲 combined with wafer bonding and ion
cut is an attractive approach due to the high efficiency of PIII
and relatively inexpensive instrumentation.5 In fact, the cost
saving is more significant for 300 mm silicon wafers as the
PIII processing time is independent of the wafer dimension.
There are many intrinsic differences between PIII and
conventional beamline ion implantation. For example, in
PIII, all ion species in the overlying plasma are implanted
when a negative voltage pulse is applied to the silicon wafer
and the ion energy distribution depends on factors such as
the sample voltage wave form, plasma sheath propagation,
collision effects, and hardware-related issues such as displacement current and cable capacitance. In contrast, the ion
共1兲 The ion implant dose required for the hydrogen PIII
layer transfer process is quite large 共high 1016 cm⫺2 to low
1017 cm⫺2 兲. Therefore, significant crystal damage results in
spite of the small hydrogen mass.
共2兲 Different hydrogen ion species exist in the plasma,
⫹
mainly, H⫹ , H⫹
2 , and H3 , and all of them are coimplanted
into the silicon wafer. The damage profile is different for
each ion species due to the difference in the net impact energy 共for example, each H atom in the H⫹
3 molecular ion
possesses 1/3 of the kinetic energy of the H atom in the H⫹
atomic ion兲 and dose 共for instance, there are three hydrogen
atoms in each H⫹
3 molecular ion兲.
共3兲 Since PIII machines are typically not designed for
ultra-high-vacuum 共UHV兲 operation, there are residual oxygen, nitrogen, water, and other atmospheric species in the
vacuum chamber. The plasma thus contains some of these
ions and they are unavoidably coimplanted into the silicon
wafer together with hydrogen. The damage caused by these
residual gas species must be taken into account.
a兲
Also affiliated with Shanghai Institute of Metallurgy, Chinese Academy of
Sciences, Shanghai 200050, China and currently on leave at Delft University of Technology, DIMES TC, Feldmannweg 17, P.O. BOX 5053, 2600
GB, Delft, Netherlands.
b兲
Author to whom correspondence should be addressed; electronic mail:
[email protected]
0021-8979/2001/90(4)/1735/5/$18.00
1735
© 2001 American Institute of Physics
1736
Wang et al.
J. Appl. Phys., Vol. 90, No. 4, 15 August 2001
共4兲 The ion energy distribution is broad due to multiple
ion species as well as the low-energy component arising
from the nonzero rise and fall times of the sample voltage
pulse.
In general, a surface dislocation density of less than 50
cm⫺2 is desired in a production environment.1 Hence, damage in the plasma implanted wafer must be carefully controlled in each step during the manufacturing process. In this
work, the damage characteristics of Si after hydrogen PIII
are investigated experimentally using secondary ion mass
spectrometry 共SIMS兲, channeling Rutherford backscattering
spectrometry 共RBS/C兲, and atomic-force microscopy 共AFM兲.
We also use a relatively simple model to derive the damage
energy density of the three hydrogen ions as well as gaseous
impurity ions.
II. EXPERIMENT
Boron-doped p-type Si共100兲 with a resistivity of 14 –21
⍀ cm was hydrogen plasma implanted using our semiconductor PIII instrument.8 The base pressure in the vacuum
chamber was 9.4⫻10⫺7 Torr. Before implantation, highpurity hydrogen gas was bled into the chamber to establish a
working pressure of 8.0⫻10⫺5 Torr. The PIII experiments
were conducted at a bias voltage of 25 kV, current of 7.8 –1.1
A, pulsing frequency of 200 or 300 Hz, and pulse duration of
30 ␮s.
The hydrogen implant dose was calculated based on the
SIMS results using a relative sensitivity factor derived from
standard ion implant materials. A simple wet cleaning process was carried out to remove some surface contaminants
before the SIMS measurements. The damage characteristics
of the as-implanted and annealed samples 共200 and 400 °C兲
were assessed using helium RBS/C. The incident energy was
2 MeV and the backscattering angle was 170°. The surface
morphology of the samples was studied by AFM and it was
conducted using a Park Instrument SPM machine at room
temperature and atmospheric pressure.
III. RESULTS AND DISCUSSION
In PIII, there is no mass selection and this is one of the
reasons why PIII boasts a high ion flux and throughput.
However, it also means that all the ions in the plasma are
implanted into the wafer simultaneously. A typical hydrogen
⫹
plasma consists of three dominant ions, H⫹ , H⫹
2 , and H3 ,
and their implantation into Si can be easily verified by fitting
the SIMS hydrogen depth profile. According to the molecular
ion implantation theory,9 the net ion energy is given by
Em/M , where E is the energy of the molecular ion, m is the
mass of the atom, and M is the total mass of the molecule.
Hence, the hydrogen atoms in these three ion species have
different net implant energies. For example, each hydrogen
atom in the H⫹
3 molecular ion has 1/3 of the implant energy
and, consequently, smaller penetrating depth. Since the
nuclear stopping power is a function of the ion energy, the
damage profile is different for the three hydrogens containing
ions. In our fits, we use the following simple relationship:
FIG. 1. Hydrogen depth profile acquired by SIMS from hydrogen plasma
implanted silicon 共300 Hz pulsing frequency and 30 min implantation time兲
⫹
and the theoretical fit using overlapping H⫹ , H⫹
2 , and H3 Gaussian distributions. Hydrogen diffusion that causes broadening of the half width and
depth shift 共80–100 nm兲 due to the surface treatment has been considered.
The high surface peak in the SIMS profile is a measurement artifact.
n
N共 x 兲⫽
di
兺 冑2 ␲␴
i⫽1
冉
exp ⫺
pi
共 x⫺R pi 兲 2
2 ␴ 2pi
冊
,
共1兲
where N(x) is the density of H ions at distance x from the
surface, d i is the dose of the ith species, R pi is the project
range of the ith species, and ␴ pi is the straggle which can be
obtained from TRIM simulation. However, considering the
heating effects during implantation, hydrogen diffusion
needs to be taken into account as well. Moreover, surface
oxidation during PIII and the surface treatment before SIMS
measurement may cause a small shift in the depth profile
measurement.
Figure 1 depicts the hydrogen SIMS depth profile of the
sample implanted at 300 Hz for 30 min, and the integrated
ion dose was calculated to be 1.74⫻1017 cm⫺2 . This dose is
typical of the ion-cut process by PIII. Using TRIM and Eq.
共1兲, we fit the SIMS data and the doses are determined to be
1.9⫻1016, 4.5⫻1016, and 2.0⫻1016 cm⫺2 for H⫹ , H⫹
2 and
H⫹
3 , respectively. The sum of the ion doses from the fit is
1.69⫻1017 cm⫺2 . The difference between the fitted value
and measured result is due to surface hydrogen that contributes to the SIMS result but not to the modeled value. To
investigate the damage, we derive the damage energy density
using these experimentally determined doses and the following relationship:10
e d共 x 兲 ⫽
D dE d 共 x 兲
,
N 共 Si 兲 dx
共2兲
where e d (x) is the damage energy density in eV/atom, D is
the dose, N共Si兲 is the atomic density of Si (5⫻1022/cm3 兲,
and dE d (x)/dx is the nuclear stopping energy, which can be
directly calculated by TRIM. The calculated damage energy
⫹
densities for H⫹ , H⫹
2 , and H3 are 2.1, 15, and 12.6 eV/
atom, respectively, and the simulated damage 共atomic displacement兲 distribution is exhibited in Fig. 2. Here, we consider that the atomic displacement is mainly caused by
J. Appl. Phys., Vol. 90, No. 4, 15 August 2001
FIG. 2. Calculated composite damage distribution 共dpa⫽displacement per
⫹
atom兲 at an implantation voltage of 25 kV using H⫹ , H⫹
2 , and H3 with
doses of 1.9⫻1016, 4.5⫻1016, and 2⫻1016 cm⫺2 , respectively.
nuclear stopping.11,12 Comparing the composite damage en⫹
ergy density of H⫹ , H⫹
2 , and H3 with the threshold energy
of Si 共about 15 eV兲, our hydrogen PIII process causes significant damage to the crystal structure of silicon and can
render the region in the vicinity of the projection range R p
amorphous. It should also be pointed out that a high concentration of implanted hydrogen will cause high pressure or
stress within the Si crystal, contributing to additional damage
to the crystal structure.13
Figure 3 shows the RBS/C spectra of the sample implanted at 200 Hz and for 30 min, and those of the sample
annealed at 400 °C are displayed in Fig. 4. Comparison between the random and channeled spectra shows that the damage layer in both samples is almost amorphous and the thickness is approximately 170 nm. The damage profile acquired
from a smaller implant dose sample is exhibited in Fig. 5.
However, such a small implantation dose is inadequate for
effective microcavity formation and ion cut, and so there is
no practical need to investigate the latter case. Figure 5
shows that a lower dose implant 共25 kV, 200 Hz, 10 min兲
leads to a thinner damage layer 共about 70 nm in thickness兲.
FIG. 3. Channeling RBS 共RBS/C兲 spectrum acquired from a hydrogen PIII
sample 共300 Hz and 30 min兲.
Wang et al.
1737
FIG. 4. RBS/C spectrum acquired from the hydrogen PIII sample shown in
Fig. 2 after annealing at 400 °C for 2 h.
In this case, the damage caused by residual gaseous impurities is relatively significant, and this issue will be discussed
later in this article.
Piatkowska, Gawlik, and Jagielski14 studied the relationship between the surface morphology and hydrogen implant
dose in beamline ion implantation. They, however, did not
present detailed results on the change of surface morphology
with temperature. Here, we show the surface morphological
change in the as-implanted and annealed samples. Figure 6
depicts the AFM topographic maps of the as-implanted and
annealed hydrogen PIII samples, and the change in the surface roughness as indicated by our root-mean-square 共rms兲
calculation in a 5 ␮m⫻5 ␮m region for different annealing
temperatures is shown in Table I. Based on the observed
increase in the surface roughness with the anneal temperature, the hydrogen movement or coalescence process during
annealing plays an important role in the surface morphology,
even though the change in the RBS channeling behavior after
FIG. 5. RBS/C spectrum acquired from a lower dose sample 共200 Hz and
10 min兲.
1738
Wang et al.
J. Appl. Phys., Vol. 90, No. 4, 15 August 2001
FIG. 6. AFM topographical map of the hydrogen PIII samples 共300 Hz and
30 min兲: 共a兲 as implanted and 共b兲 after annealing at 400 °C for 2 h.
annealing at 400 °C is not obvious. It should be noted that
our work focuses on temperature at or below 400 °C because
in the hydrogen PIII/ion-cut process, an implantation temperature of 400 °C or higher will cause hydrogen bubble formation or blistering and, consequently, premature exfoliation. As an interstitial atom with a dissolution enthalpy of 0.8
eV, hydrogen displays an extraordinary chemical reactivity to
silicon, leading to the formation of point and extended defects irrespective of its atomic or molecular state. Cerofolini
et al.15 investigated the bubble formation in H and He implanted Si and deduced the change of the enthalpy in the Si–
H2 system with respect to temperature, hydrogen concentration, and other factors. One of the important factors influencing the surface morphology is internal stress caused by the
high pressure exerted by hydrogen in the microvoids. This
stress leads to an increase of the surface roughness, and finally, surface blistering at a high enough temperature or under mechanical stress.
Normally, for a Si wafer implanted with a hydrogen dose
higher than 6⫻1016 cm⫺2 , surface blistering will occur and
become visible when the wafer is heated to 450 °C. At a
higher implant dose, the required temperature is lower and
this is also true in the case of boron coimplantation. Figure
6共b兲 confirms the gradual change in the surface morphology
that eventually leads to surface blistering. For the ion-cut
process, the generation of a sufficient amount of bubbles or
microcavities is necessary. However, since the damage and
defects cannot be annealed out at low temperature, a paradox
is created when a high-temperature treatment before wafer
bonding is not feasible. The broad damage zone buried in the
SOI structure will affect recrystallization of the transferred
layer during the subsequent solid-state reaction at high temperature. One can argue that if hydrogen PIII is conducted on
a silicon wafer with a pregrown thin surface oxide, the damage region on the surface can be confined to the oxide that
can be removed prior to wafer bonding. However, our RBS
results 共Figs. 3–5兲 show that the damage zone stretches all
the way from the vicinity of the ion-projected range to close
to the surface. In addition, oxygen recoil may cause other
problems and the damage created in the bulk cannot be circumvented totally.
Figure 5 shows that surface damage is still serious even
when the dose is lower. According to our previous studies,
under typical conditions, contaminants such as oxygen, nitrogen, and carbon in the plasma constitute a few percent of
the total ion current in hydrogen PIII.16 This is because PIII
equipment is usually not designed for UHV conditions. For a
total ion implant dose in the high 1016 to low 1017 cm⫺2 , the
oxygen, nitrogen, and carbon ion doses may be close to or
exceed 1015 cm⫺2 . Based on our experimental results at 25
kV for an oxygen dose of 1⫻1015 cm⫺2 , we calculate damage energy densities of 29.3 and 33.5 eV/atom for O⫹ and
TABLE I. Root-mean-square 共rms兲 surface roughness values of the asimplanted sample and samples annealed at 200, 300, and 400 °C.
Surface roughness rms 共nm兲
As implanted
200 °C
300 °C
400 °C
0.06
0.13
0.203
0.165
FIG. 7. Calculated damage distribution profile of Si atoms 共dpa
⫽displacement per atom兲 in 25 kV oxygen plasma implanted samples.
Wang et al.
J. Appl. Phys., Vol. 90, No. 4, 15 August 2001
TABLE II. Calculated damage energy densities of carbon and nitrogen 共N⫹
and N⫹
2 in Si 共implantation energy⫽25 keV兲.
Ion species
Implantation energy 共keV兲
Implantation dose 共ions cm⫺2 兲
Projection range R p /
longitudinal straggling 共nm兲
Damage energy density
共eV/atom兲
C⫹
N⫹
N⫹
2
25
1⫻1015
75.7/34.2
25
1⫻1015
65.1/29.5
25
0.5⫻1015
33.7/17.6
16.3
22.5
26.5
O⫹
2 , respectively. The simulated displacement of Si atoms is
shown in Fig. 7. The calculated damage energy densities of
nitrogen and carbon are listed in Table II for the same implant conditions 共25 kV兲. In spite of their small percentage,
the presence of gaseous contaminants can significantly increase the damage. These contaminant ions are heavier and
the damage zone is closer to the surface. Hence, the RBS/C
spectrum shown in Fig. 5 makes sense. This damage region
is buried in the SOI structure after layer transfer, that is,
close to the buried oxide, and affects recrystallization even
more severely. This contamination issue must be addressed
properly in experiments, for instance, by using better
vacuum and pumping devices.
IV. CONCLUSION
The damage in hydrogen plasma implanted Si has been
investigated experimentally and theoretically. Our RBS/C results indicate that the damaged layer is quite broad, and different from that observed in single-energy beamline ion implantation in which the damage only occurs near the
projected range of the implanted ions. The broadness of the
damage zone is attributed to different ion species from the
plasma implanted to different depths. Finally, we discuss the
contribution of inevitable coimplanted gaseous contaminants. Even though they are small in percentage, our damage
energy density and silicon atom displacement calculations
1739
reveal the severity of the effects, and care must, therefore, be
exercised in reducing the amount of these gaseous species in
PIII equipment.
ACKNOWLEDGMENTS
The work was jointly supported by the Hong Kong Research Grants Council 共CERG Grant No. 9040412 or CityU
Grant Nos. 1003/99E and 9040498 or CityU Grant No. 1032/
00E兲, the City University of Hong Kong 共SRG Grant No.
7001028兲, and the Chinese NSF 共Grant No. 59982008兲.
1
J.-P. Colinge, Silicon-on-Insulator Technology, Materials to VLSI, 2nd ed.
共Kluwer Academic, Dordrecht, 1997兲.
2
S. Cristoloveanu and S. S. Li, Electrical Characterization of Silicon-onInsulator Materials and Devices 共Kluwer Academic, Dordrecht, 1995兲.
3
L. Alles, P. Dolan. M. J. Anc, P. Allen, F. Cordts, and T. Nakai, Proc. IEEE
Int. SOI Conf. 共1997兲, p. 10 共unpublished兲.
4
M. Bruel, B. Aspar, B. Charlet, C. Maleville, T. Poumeyrol, A. Soubie, A.
J. Auberton-Herve, J. M. Lamure, T. Barge, F. Metral, and S. Trucchi,
Proc. IEEE Int. SOI Conf. 共1995兲, p. 178 共unpublished兲.
5
W. G. En, I. J. Malik, M. A. Bryan, S. Farrens, F. J. Henley, N. W.
Cheung, and C. Chan, Proc. IEEE Int. SOI Conf. 共1998兲, p. 163 共unpublished兲.
6
P. K. Chu, S. Qin, C. Chan, N. W. Cheung, and L. A. Larson, Mater. Sci.
Eng., R. R17, 207 共1996兲.
7
Z. Fan, Q. C. Chen, P. K. Chu, and C. Chan, IEEE Trans. Plasma Sci. 26,
1661 共1998兲.
8
P. K. Chu, S. Qin, C. Chan, N. W. Cheung, and P. K. Ko, IEEE Trans.
Plasma Sci. 26, 79 共1998兲.
9
S. Prussin, in Ion Implantation in Semiconductors, edited by S. Namba
共Plenum, New York, 1975兲, p. 449.
10
J. Ziegler, J. Biersack, and U. Littmar, in The Stopping and Range of Ions
in Solids 共Pergamon, New York, 1985兲.
11
S. Tian, M. F. Morris, S. J. Morris, B. Obradovic, G. Wang, A. F. Tasch,
and C. M. Snell, IEEE Trans. Electron Devices 45, 1226 共1998兲.
12
W. Bohmayr, A. Burenkov, J. Lorenz, H. Ryssel, and S. Selberherr, IEEE
Trans. Comput.-Aided Des. 17, 1236 共1998兲.
13
G. F. Cerofolini, G. Calzolari, F. Corni, C. Nobili, G. Ottaviani, and R.
Tonini, Mater. Sci. Eng., B 71, 196 共2000兲.
14
A. Piatkowska, G. Gawlik, and J. Jagielski, Appl. Surf. Sci. 141, 333
共1999兲.
15
G. F. Ceroflini, F. Corni, S. Frabboni, C. Nobili, G. Ottaviani, and R.
Tonini, Mater. Sci. Eng., R. 27, 1 共2000兲.
16
Z. Fan, X. Zeng, P. K. Chu, C. Chan, and M. Watanabe, Nucl. Instrum.
Methods Phys. Res. B 155, 75 共1999兲.