Ex situ n and p doping of vertical epitaxial short silicon nanowires by

advertisement
IOP PUBLISHING
NANOTECHNOLOGY
Nanotechnology 20 (2009) 165706 (7pp)
doi:10.1088/0957-4484/20/16/165706
Ex situ n and p doping of vertical
epitaxial short silicon nanowires by
ion implantation
Pratyush Das Kanungo1,3, Reinhard Kögler2 , Kien Nguyen-Duc1,
Nikolai Zakharov1 , Peter Werner1 and Ulrich Gösele1
1
2
Max Planck Institute of Microstructure Physics, Weinberg 2, 06120 Halle, Germany
Forschungszentrum Dresden—Rossendorf, FWIM, 01314 Dresden, Germany
E-mail: kanungo@mpi-halle.de
Received 9 November 2008, in final form 2 January 2009
Published 1 April 2009
Online at stacks.iop.org/Nano/20/165706
Abstract
Vertical epitaxial short (200–300 nm long) silicon nanowires (Si NWs) grown by molecular
beam epitaxy on Si(111) substrates were separately doped p- and n-type ex situ by implanting
with B, P and As ions respectively at room temperature. Multi-energy implantations were used
for each case, with fluences of the order of 1013 –1014 cm−2 , and the NWs were subsequently
annealed by rapid thermal annealing (RTA). Transmission electron microscopy showed no
residual defect in the volume of the NWs. Electrical measurements of single NWs with a Pt/Ir
tip inside a scanning electron microscope (SEM) showed significant increase of electrical
conductivity of the implanted NWs compared to that of a nominally undoped NW. The p-type,
i.e. B-implanted, NWs showed the conductivity expected from the intended doping level.
However, the n-type NWs, i.e. P- and As-implanted ones, showed one to two orders of
magnitude lower conductivity. We think that a stronger surface depletion is mainly responsible
for this behavior of the n-type NWs.
(Some figures in this article are in colour only in the electronic version)
In order to meet the ever increasing demand for miniaturization
of microelectronic devices and circuits, the use of silicon
nanowires (Si NWs) can offer an alternative scaling down
approach [1, 2]. Because of their quasi-one-dimensional
structure and possibilities of quantum confinement [3], Si
NWs have been widely investigated as prototype devices
for many future applications [4–6]. However, in order to
integrate them with the existing ultra-large scale integration
(ULSI) technology, it is essential to demonstrate that they
are compatible with the current semiconductor processing
steps. One of the most important steps in the fabrication of
planar silicon devices and circuits is doping by implantation
with ions of group III and V elements [7, 8] in specific
areas of the devices known as the ‘active regions’, where
the transport as well as recombination and generation of the
charge carriers take place. So far, doping of Si NWs has
mostly been done in situ [9–11] by incorporating the dopant
species in the gas phase along with Si during NW growth.
Although prototype devices have been fabricated with in
situ doped NWs [11, 12] they have often encountered two
problems. Firstly, an amorphization of the NW structures
was observed [10] on doping them with B from the diborane
(B2 H6 ) gas. As a result, instead of a homogeneously doped
crystalline NW a ‘p-type amorphous shell–intrinsic crystalline
core’ kind of structure was formed, which is undesirable for
most of the microelectronics applications. Secondly, an interdiffusion of the dopant species between differently doped
segments in an NW has often deformed the ideal axial p–
n junction and introduced unwanted lateral junctions [13].
This inter-diffusion and transition region between p- and ntype segments in an NW has been theoretically modeled and
fitted to experimental data elsewhere [14]. Doping by ion
implantation on the other hand can vertically confine the
dopants within their stopping ranges, which are determined by
the implantation energy and dose. It is possible that the NWs
are amorphized during implantation. However, by a suitable
3 Author to whom any correspondence should be addressed.
0957-4484/09/165706+07$30.00
1
© 2009 IOP Publishing Ltd Printed in the UK
Nanotechnology 20 (2009) 165706
P D Kanungo et al
Figure 1. (a) SEM image of as-grown vertical NWs. (b) Cross sectional transmission electron microscope (TEM) image of an as-grown NW.
(c) A NW with the Au cap removed before implantation.
3 × 1018 cm−3 of B was grown prior to the Au deposition at
525 ◦ C. This was a necessary step to grow Si NWs for all
the p-type wafers, as the wafer surface plays a more critical
role for MBE [17, 18] than the other VLS growth techniques.
A buffer layer offers a cleaner Si(111)-7 × 7 surface than
the original wafer surface. We observed that for the p-type
wafers introduction of the buffer layer drastically improves the
density of the NWs as compared to the n-type ones. After
the Au deposition, about 270 nm of Si was evaporated, that
resulted in the simultaneous growth of the NWs as well as an
epitaxial Si layer underneath and surrounding them as reported
earlier [17, 18].
A SEM image of a cluster of the as-grown NWs is shown
in figure 1(a). Statistics on such as-grown NWs of each type
showed that the average NW length was around 320 nm and
the diameter around 150 nm. The dark hemispherical objects
on the top of the NWs in figure 1(a) are the Au droplets
that initiated the NW growth. Previous investigations by
transmission electron microscopy revealed [17, 18] a singlecrystalline structure of the NWs oriented along the 111
direction. A cross-sectional TEM image of an as-grown NW
is shown in figure 1(b). It can be seen in figure 1(b) that in
addition to the dark Au cap on top, a decoration of dark spots
is also visible on the surface of the NW. These were reported
as Au nanoparticles earlier [17, 18] and are characteristics of
Si NWs grown by the vapor–liquid–solid (VLS) mechanism.
Because of its heavy mass compared to the implanted species
(B, P and As), the Au droplet at the tip of the NW can act as
an effective ion-stopper, resulting in lower penetration of the
implanted ions into the NW volume than expected. Besides,
subsequent annealing process they can be well recrystallized.
Very recently Colli et al [15] have shown that P and B
implantation on horizontally laid down Si NWs with a dose as
high as 1015 cm−2 does not lead to a high degree of structural
defects in the NWs.
In this paper we demonstrate that short vertical Si NWs
grown by molecular beam epitaxy (MBE) on Si(111) substrates
can be doped with B, P and As by multiple energy/dose
ion implantations with the highest dose of the order of
1014 cm−2 to achieve homogeneous carrier concentrations or
flat concentration profiles along the lengths of the NWs. We
investigated both the structural and the electrical properties
of the implanted NWs. The electrical conductivities of the
implanted NWs were significantly higher than the MBEgrown, nominally undoped NWs reported earlier [16] in the
measured voltage range.
The growth of nominally undoped Si NWs by MBE was
reported earlier [17, 18] by us. In the present work we have
used the previously grown nominally undoped NWs to perform
the ion implantation. In order to avoid any unwanted electrical
junction between the NWs and the substrates, we used p-type
wafers (resistivity 5–10 cm) for the NWs implanted with
B, and n-type wafers (resistivity 0.001–0.006 cm) for the
NWs implanted with P and As respectively. In brief, RCAcleaned 5 Si(111) wafers were annealed at 870 ◦ C inside an
ultra-high vacuum (10−10 mb) MBE chamber to get rid of the
native oxide. Immediately afterward 1–2 nm Au was deposited
in situ at 525 ◦ C, that subsequently broke into droplets acting
as the growth initiator of the Si NWs. For the NWs prepared
for B implantation, a 200 nm thick Si buffer layer doped with
2
Nanotechnology 20 (2009) 165706
P D Kanungo et al
Figure 2. Implantation schemes for the vertical Si NWs. (a) An
intrinsic Si NW grown on p-type Si(111) wafer and B doped (p-type)
buffer layer; the epi-layer is intrinsic as well. (b) B-ion implantation
at an angle of 7◦ w.r.t. the substrate/NW tip to avoid channeling.
(c) The intrinsic Si NW turns p-type. (d) An intrinsic Si NW grown
on n-type Si(111) wafer; the epi-layer is intrinsic as well.
(e) P/As-ion implantation at an angle of 7◦ w.r.t. the substrate/NW tip
to avoid channeling. (f) The intrinsic Si NW turns n-type.
for electrical measurements, a direct Au–Si contact has a more
Schottky type behavior [19] at room temperature than a Pt/Ir–
Si contact, which is part of our measurement system [16]. To
get rid of these two problems, we removed the Au caps from
the NWs by a 5 min dip in an aqueous solution of KI and I2 ,
a standard Au etchant. Figure 1(c) shows the cross-sectional
TEM image of such a NW with the Au cap removed. This
however did not remove the Au nanoparticles on the surface
of the NWs, which is evident from figure 1(c). We assume
that these Au nanoparticles are strongly attached to the NWs
in the form of an Au:Si alloy. As a result of the Au droplet
removal from the NW tip, we found the average length of the
NWs reduced to 270 nm.
The ion implantation was carried out with the goal of
homogeneously doping the NWs along the entire length as
mentioned above. The implantation process and the resulting
change in the NW are schematically depicted in figures 2(a)–
(f). The orientation of the NWs with respect to the impinging
ions is as shown in figures 2(b) and (e). An angle of 7◦
between the ion beam and the NW tip/substrate surface was
maintained in order to avoid ion channeling [8]. In these
specific experiments we did not avoid the implantation of
the surface of the substrate (which was actually the epilayer in between the NWs) during the implantation process.
However, the type (n or p) of the implanted species and
the substrates were always the same which let us avoid the
formation of an unwanted bipolar junction between them. An
implantation with a single energy and dose results in a near
Gaussian concentration profile [7]. In order to achieve a more
homogeneous dopant concentration along the length of the
NWs, i.e. to get a rectangular doping profile, multi-energy
implantation [20] is often used. We applied three to four
Figure 3. Transport of ions in matter (TRIM) simulated
concentration profiles of a multi-energy (a) B implantation, (b) P
implantation, (c) As implantation. The ‘0’ on the x -axis refers to the
NW top. Figure 3(a) and (b) are the resultant profiles of a three-step
implant, and figure 3(b) is the resultant profile of a four-step implant.
Refer to table 1 for details on the implantation energies and doses.
steps of implantations with varying implantation energies and
doses. The resultant profile should flatten over the entire range
of 300 nm. We kept the range higher than the average NW
length so that the epi-layer directly underneath the NWs is also
doped to a significant extent. The intended doping profiles
were simulated by the standard software TRIM [21] code. The
simulations were performed for every individual implantation
first, and then they were added to produce the resultant
concentration profiles, which are shown in figures 3(a), (b)
and (c) respectively. It should be noted that in general the
TRIM model does not differentiate between an NW and a Si
layer. In figure 3 the ‘0’ on the depth scale (x -axis) corresponds
to the top of the NW. Table 1 lists the energies and doses used
for each of the multiple implantations. We denote the resulting
intended dopant concentrations as Nint . For the B implantation
Nint was 3 × 1018 cm−3 (the same as the doping level of the
buffer layer), for the P implantation Nint was 1 × 1019 cm−3 ,
3
Nanotechnology 20 (2009) 165706
P D Kanungo et al
Figure 4. Post-implantation and annealing cross-sectional TEM images of the NWs (upper row) and the corresponding epi-layer (lower row).
B-implanted—figures (a), (b), P-implanted—figures (c), (d), As-implanted—figures (e), (f).
Table 1. Implantation energies and doses for the B, P and As multi-energy implantations of the NWs.
First
implantation
Second
implantation
Third
implantation
Fourth
implantation
Dopant
Energy
(keV)
Dose
(cm−2 )
Energy
(keV)
Dose
(cm−2 )
Energy
(keV)
Dose
(cm−2 )
Energy
(keV)
Dose
(cm−2 )
B
P
As
55
140
300
3.4 × 1013
1.6 × 1014
1.7 × 1013
35
70
140
1.7 × 1013
6.5 × 1013
5.5 × 1012
15
30
80
1.7 × 1013
3.35 × 1013
2.2 × 1012
—
—
30
—
—
2.2 × 1012
and for the As implantation Nint was 1 × 1018 cm−3 . All
the doping concentrations were chosen below the solubility
limit and the following conditions were taken into account. A
high doping level was desired to form an Ohmic contact [19]
during electrical measurements. A high doping level also
reduces the surface depletion [16] of the NWs. By surface
depletion we mean trapping and subsequent deactivation of
dopant atoms by the dangling bonds between Si and the native
SiO2 at the unpassivated NW surface. However, previous
experiments with As implantations showed that if the Nint
value is in the range of 1019 cm−3 (as for the P implantation),
requiring very high energy implants, it leads to a permanent
amorphization of the NWs. Subsequent to the implantations
we performed rapid thermal annealing (RTA) to activate the
dopants and recover the NWs from any implantation-induced
damage. For B implantation, an annealing at 850 ◦ C for 30 s
in Ar atmosphere was tried. For P and As implantations,
the annealing temperature was raised to 1000 ◦ C for the same
duration. As a result of annealing, the concentration profiles
should be flattened even more and extend further into the epilayer underneath the NWs.
After the implantation and RTA, TEM imaging was
performed to investigate any structural defect in the NWs
caused by the implantation process. Typical TEM images of
NWs corresponding to B, P and As implantations are shown
in figures 4(a), (c) and (e) respectively. Figures 4(b), (d)
and (f) show the epi-layers directly underneath the NWs. No
Figure 5. SEM image of a NW contacted with the Pt/Ir tip for I –V
measurement.
residual amorphization was observed in the volume of the
single-crystalline NWs for any of the implantations. However,
we observed some defects in the epi-layers of the As implanted
NW (figure 4(f)). Concerning our TEM analysis we think
these are dislocation loops created during the implantation and
annealing process, since they were not present in the as-grown
NWs.
4
Nanotechnology 20 (2009) 165706
P D Kanungo et al
<2 fA noise) with an internal voltage source. The back contact
was realized with a conductive Ag paste that yielded a contact
resistance in the range of 500–1000 . It is not possible to
evaluate the top contact resistance in a two-point measurement
such as this. However, the change of measured current from
an unimplanted (nominally undoped) to an implanted (n- or ptype) NW will still reflect the effect of doping.
The measured I –V curves of three each from B-, P- and
As-implanted NWs of different diameters and lengths along
with that of an unimplanted NW are plotted in figures 6(a),
(b) and (c) respectively. All the I –V curves were measured
in the range of −0.1 to 0.1 V, which is enough to excite
the majority carriers at room temperature. Higher voltages
often resulted in melting of the NWs. The current at 0.1 V
was between 10 and 40 μA for the B-implanted and 1–5 μA
for the P- and As-implanted NWs. This leads to current
densities of around 105 A cm−2 and 104 A cm−2 respectively.
The thermal conductivity of Si NWs is reported [23] to be
two orders of magnitude lower than bulk Si. The high
current density coupled with low thermal conductivity (low
heat dissipation) may cause the observed burn-out due to
self-heating. As can be seen from figure 6, all the I –V
curves of the doped NWs demonstrate ohmic behavior with
a significantly higher current compared to an unimplanted
(nominally undoped) NW. It clearly indicates a doping induced
increase of the electrical conductivity of the NWs. It was
found earlier [16, 22] that the MBE-grown NWs suffer from
a strong surface depletion. For the in situ B doped NWs [22]
we found that with increased doping the volume conductivity
from the NW dominates, and the surface depletion subsides.
The intended dopant concentrations ( Nint ) in all the implanted
NWs were quite high (1018 –1019 cm−3 ) in order to compensate
the surface depletion as much as possible. However, it has been
predicted [24] that for n-type Si NWs the surface depletion is
much stronger than for the p-type NWs due to larger surface
segregation energy of the n-type dopants. Due to the large
area of the substrate compared to a single NW we can neglect
its resistance compared to a NW. The exact contribution of
the Au nanoparticles on the NW surface caused by the VLS
growth mechanism is still unknown. However, they exist in
both the unimplanted and implanted NWs in almost equal
amounts. Therefore, we think that their contribution, if any,
should be equal for both. From the linear I –V curves of
figures 6, we calculated the resistance values. If we assume
the whole of the calculated resistance comes from the volume
of the NW, we can calculate the resistivity (ρmeas ) taking
into account its cylindrical geometry. We have listed the
resistance and resistivity for each of the measured NWs in
table 2. If we assume the NWs are doped to the level of our
intended concentrations ( Nint ), the dopants are fully activated,
and furthermore the mobility (μ) values are same as the
bulk mobility values for the corresponding Nint [19], we can
calculate the expected resistivity (ρexp ) of the NWs from the
following equation.
Figure 6. The I –V curves of (a) three B-implanted NWs, (b) three
P-implanted NWs and (c) three As-implanted NWs. The I –V curve
of an as-grown, unimplanted NW is included in each case for
comparison. Refer to table 1 for the dimensions and the measured
electrical properties of the NWs.
To confirm the effect of doping on the electrical properties
of the implanted NWs, we measured the current–voltage
characteristics ( I –V curves) of individual NWs. Therefore,
we contacted vertical NWs with a Pt/Ir tip fitted to a micromanipulator inside an SEM. By moving the tip with the
manipulator we could measure different individual NWs.
We have already reported on the electrical measurements of
nominally undoped [16] and in situ B doped Si NWs [22] using
this measurement technique. With this approach we can image
the NW first in the SEM, and then carefully make a contact
between the tip and the top of the NW as shown in figure 5. The
I –V curves were measured by a picoammeter (Keithley 6487,
ρexp =
1
Nint eμ
(1)
where e is the electronic charge. The ρexp values for the B-,
P- and As-implanted NWs are also shown in table 2. It can
5
Nanotechnology 20 (2009) 165706
P D Kanungo et al
Table 2. The measured dimensions (from SEM images) and extracted electrical resistances (from the I –V curves in figure 6) and resistivities
along with the expected resistivities corresponding to the intended doping levels of the NWs mentioned in figure 6.
Nint
(intended
doping level)
Implantation type (cm−3 )
NW number
D (diameter)
(nm)
L (length)
(nm)
R (resistance)
()
ρmeas (measured
resistivity =
π D2 R
) ( cm)
4L
ρexp (expected
resistivity =
1
eμ) ( cm)
Nint
B-implanted
3 × 1018
NW 1
NW 2
NW 3
Unimplanted NW
160
176
191
160
240
308
255
280
3.0k
2.6k
6.8k
1.0G
0.02
0.02
0.07
7180
0.02
P-implanted
1 × 1019
NW 1
NW 2
NW 3
Unimplanted NW
150
140
160
165
270
270
240
295
21k
37k
71k
980M
0.13
0.20
0.50
7103
0.003
As-implanted
1 × 1018
NW 1
NW 2
NW 3
Unimplanted NW
166
175
180
165
240
235
260
295
65k
77k
80k
980M
0.50
0.70
0.80
7103
0.03
be seen from table 2 that for B-implanted NWs ρmeas and ρexp
are of the same order of magnitude. This confirms that the
full activation of our intended doping was achieved in the Bimplanted NWs. For both the P- and As-implanted NWs, the
ρmeas values are higher than ρexp values by one to two orders of
magnitude. Possible reasons for different behavior of n and p
implantations will be discussed in the following sections.
Boron is a light ion which is easily activated electrically
in Si on annealing at 700 ◦ C or above [8]. Therefore, most of
the B atoms are incorporated in the substitutional sites of the
Si lattice at the annealing temperature of 850 ◦ C. This is why
we do not see any defects either in the B-implanted NWs or in
the epi-layer underneath and almost 100% electrical activity of
the B atoms is achieved. As we demonstrated earlier [22], the
surface depletion on B doped NWs is not significant beyond
the doping level of 1018 cm−3 , which is our case.
In the case of heavy P implantations like that of ours,
transient enhanced diffusion (TED) [25–27] during RTA can
increase the diffusivity of P atoms by orders of magnitude
due to interaction with the defects created by the implantation.
Using a value of 1.2 × 10−12 cm2 s−1 [25] for√the effective
diffusion constant ( Deff ), the diffusion length (2 Deff t , where
t = annealing time) reaches 120 nm, which is comparable
to the average diameter of the NWs. Thereby, a significant
number of P atoms can diffuse out of the NWs, especially in
the radial direction, and reach the surface. This will further
enhance the surface depletion, resulting in the higher than
expected values of NW resistivity.
In the case of the As-implanted NWs, TED is not as
significant as in the P-implanted ones. However, it has been
reported [28] that the surface depletion still plays a big role in
increasing the resistivity of the As-doped NWs. Clément et al
[28] reported an increase in resistivity by a factor of 14 for Asimplanted horizontal NWs, which agrees with our result.
In addition, in the case of both the P- and As-implanted
n-type NWs, the top contact resistance between Pt/Ir tip–nSi NW could be much higher than for the B-implanted p-type
NWs due to differences in the barrier heights. This can also
contribute to the measured higher resistivity.
In conclusion, we have demonstrated both p- and ntype doping of intrinsic vertical Si NWs by ion implantation.
We found a significant difference in electrical conductivity
between the p- and n-implanted NWs that can be expected from
previous theoretical predictions.
Acknowledgments
The authors would like to thank Dr O Breitenstein and Mr
J Bauer for scientific discussions, and Mr A Frommfeld, Mr
K U Assmann, Ms S Hopfe and Ms C Muenx for technical
support. The authors would also like to acknowledge the
financial support from the FP6 EU project NODE (Nanowire
based one dimensional electronics).
References
[1] Cui Y and Lieber C M 2001 Science 291 851
[2] Chen L J 2007 J. Mater. Chem. 17 4639
[3] Zhao X, Wie C M, Yang L and Chou M Y 2004 Phys. Rev. Lett.
92 236805
[4] Zhou X T, Hu J Q, Li C P, Ma D D D, Lee C S and
Lee S T 2003 Chem. Phys. Lett. 369 220
[5] Hofheinz M, Jehl X, Sanquer M, Molas G, Vinet M and
Deleonibus S 2006 Appl. Phys. Lett. 89 143504
[6] Tian B, Zheng X, Kempa T J, Fang Y, Yu N, Yu G,
Huang J and Lieber C M 2007 Nature 449 885
[7] Ziegler J F 1984 Inc. Ion Implantation Science and Technology
(New York: Academic)
[8] Gandhi S K 1994 VLSI Fabrication Principles 2nd edn
(New York: Wiley)
[9] Cui Y, Duan X, Hu J and Lieber C M 2000 J. Phys. Chem. B
104 5213
[10] Lew K K, Pan L, Bogart T E, Dilts S M, Dickey E C,
Redwinga J M, Wang Y, Cabassi M, Mayer T S and
Novak S W 2004 Appl. Phys. Lett. 85 3101
[11] Zheng G, Lu W, Jin S and Lieber C M 2004 Adv. Mater.
16 1890
[12] Lauhon L J, Gudiksen M S, Wang D and Lieber C M 2002
Nature 420 57
[13] Tutuc E, Appenzeller J, Reuter C M and Guha S 2006 Nano
Lett. 6 2070
6
Nanotechnology 20 (2009) 165706
P D Kanungo et al
[22] Das Kanungo P, Zakharov N, Bauer J, Breitenstein O,
Werner P and Goesele U 2008 Appl. Phys. Lett. 92 263107
[23] Li D, Wu Y, Kim P, Shi L, Yang P and Majumdar A 2003 Appl.
Phys. Lett. 83 2934
[24] Fernández-Serra M V, Adessi Ch and Blasé X 2006 Phys. Rev.
Lett. 96 166805
[25] Solmi S and Selvidori M 1988 Solid State Phenom. 2 65
[26] Angelucci R, Negrini P and Solmi S 1986 Appl. Phys. Lett.
49 1468
[27] Oehrlein G S, Cohen S A and Sedgwick T O 1984 Appl. Phys.
Lett. 45 417
[28] Clément N, Tonneau D, Dallaportaa H, Bouchiat V,
Fraboulet D, Mariole D, Gautier J and Safarov V 2002
Physica E 13 999
[14] Li N, Tan T Y and Gösele U 2008 Appl. Phys. A 90 591
[15] Colli A, Fasoli A, Ronning C, Pisana S, Piscanec S and
Ferrari C A 2008 Nano Lett. 8 2188
[16] Bauer J, Fleischer F, Breitenstein O, Schubert L, Werner P,
Gösele U and Zacharias M 2007 Appl. Phys. Lett.
90 012105
[17] Zakharov N D, Werner P, Sokolov L and Gösele U 2007
Physica E 37 148
[18] Werner P, Zakharov N D, Gerth G, Schubert L and
Gösele U 2006 Int. J. Mater. Res. 97 1008
[19] Sze S M 1981 Physics of Semiconductor Devices 2nd edn
(New York: Wiley)
[20] Wang D and Wang W 1987 J. Electron. (China) 4 39
[21] www.srim.org
7
Download