Spontaneous vacancy array formation on FeSi2 and CoSi 2 formed on Si(100)2×n surface

Article abstract of DOI:10.1063/1.1461904

Atomic structure of FeSi₂ or CoSi₂ grown on the Si(100)2×n surface has been investigated by scanning tunneling microscopy. After annealing the Fe or Co covered Si(100)2×n substrate at ～800°C, an ordered adatom vacancy array appears on the nominal 1×1 surface of the formed FeSi₂ or CoSi₂ islands, which has not been observed for silicide on the Si(100)-2×1. Upon further annealing to ～1100°C, the vacancies coalesce into striped domains along one of the 〈011〉 directions. These nanostructured features are a result of the Ni impurities, and can be a promising template for fabricating nanodot arrays.

Full text of DOI:10.1063/1.1461904

Spontaneous vacancy array formation on FeSi 2 and CoSi 2 formed on Si (100)2×n
surface
Jun-Zhong Wang, Jin-Feng Jia, Hong Liu, Jian-Long Li, Xi Liu, and Qi-Kun Xue
Citation: Applied Physics Letters 80, 1990 (2002); doi: 10.1063/1.1461904
View online: http://dx.doi.org/10.1063/1.1461904
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/80/11?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Erratum: “Layer-by-layer growth of Ag on a GaN(0001) surface” [Appl. Phys. Lett. 82, 1389 (2003)]
Appl. Phys. Lett. 82, 3991 (2003); 10.1063/1.1577218
Layer-by-layer growth of Ag on a GaN(0001) surface
Appl. Phys. Lett. 82, 1389 (2003); 10.1063/1.1556572
Spontaneous formation of ordered indium nanowire array on Si(001)
Appl. Phys. Lett. 79, 2826 (2001); 10.1063/1.1413722
Silicon on GaN(0001) and (0001) surfaces
J. Vac. Sci. Technol. B 19, 1619 (2001); 10.1116/1.1383074
Self-organization in Si/CoSi 2 (111) heteroepitaxy
J. Vac. Sci. Technol. B 17, 1848 (1999); 10.1116/1.590837
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.248.155.225 On: Sun, 23 Nov 2014 05:13:01

APPLIED PHYSICS LETTERS
VOLUME 80, NUMBER 11
18 MARCH 2002
Spontaneous vacancy array formation on FeSi2 and CoSi2 formed
on Si„100… 2Ãn surface
Jun-Zhong Wang, Jin-Feng Jia, Hong Liu, Jian-Long Li, Xi Liu, and Qi-Kun Xuea)
State Key Laboratory for Surface Physics, and International Center for Quantum Structures, Institute of
Physics, the Chinese Academy of Sciences, Beijing 100080, People’s Republic of China
͑Received 12 December 2001; accepted for publication 28 January 2002͒
Atomic structure of FeSi2 or CoSi2 grown on the Si͑100͒ 2ϫn surface has been investigated by
scanning tunneling microscopy. After annealing the Fe or Co covered Si͑100͒ 2ϫn substrate at
ϳ800 °C, an ordered adatom vacancy array appears on the nominal 1ϫ1 surface of the formed FeSi2
or CoSi2 islands, which has not been observed for silicide on the Si͑100͒–2ϫ1. Upon further
annealing to ϳ1100 °C, the vacancies coalesce into striped domains along one of the ͗011͘
directions. These nanostructured features are a result of the Ni impurities, and can be a promising
template for fabricating nanodot arrays. © 2002 American Institute of Physics.
͓DOI: 10.1063/1.1461904͔
Ferromagnetic metal silicides grown epitaxially on Si
have attracted great interest in the past decade due to their
importance in microelectronics. Among them CoSi2 and
NiSi2 are by far the most studied systems.1 CoSi2 is attrac-
tive as a self-aligned silicide for high-performance ultralarge
scale integrated devices because of its low lattice mismatch
with Si ͑Ϫ1.2%͒ and low electrical reistivity ͑14 ␮⍀ cm͒ at
room temperature ͑RT͒.2,3 Compared with other silicides,
FeSi2 is more interesting because it exists in four different
phases with distinct properties denoted as ␣, ␤, pseudomor-
phic, and ␥-FeSi2 , respectively. ␥-FeSi2 has a CaF2 struc-
ture of metallic character with a lattice constant ϳaSiϭ5.43
Å.4,5
In the growth of various epitaxial layers, defects play
very important roles in determining qualities of the grown
layers. The dimer vacancy lines ͑DVLs͒ giving rise to the
2ϫn structure on Si͑100͒ are one of the well known defects
in this category.6–10 A tiny amount of metal contaminations,
like Ni, can lead to dimer vacancies to coalesce into
DVLs.11,12 The ordering of dimer vacancies is driven by a
short-range attractive interaction between vacancies in adja-
cent dimer rows, and by a long-range repulsive interaction
between vacancies in the same row.13 Zandvliet et al.14 have
shown that the long-range repulsive interaction is related to
strain relaxation.
reported previously,16–18 it is difficult to determine whether
such an array is an intrinsic feature of the Ni silicide. This
work shows that vacancy array is a common phenomenon for
the ferromagnetic metal silicides grown on Si͑100͒ 2ϫn,
and its formation can be controlled easily by a small amount
of certain dopants such as Ni.
The experiments were performed on a commercial Omi-
cron ultrahigh vacuum ͑UHV͒ variable temperature scanning
tunneling microscope ͑STM͒, equipped with a low energy
electron diffraction ͑LEED͒/Auger spectrometer with a base
pressure of 5.0ϫ10Ϫ11 mbar. The STM images were ob-
tained in the constant-current mode with electrochemically
etched W tips. The sample used was a heavily P-doped Si
͑001͒ wafer with a resistivity of ϳ0.001 ⍀ cm, which was
intentionally contacted to a stainless steel tweezer before
loaded into the UHV system. The small amount of Ni con-
tamination from the contact was found to be enough to pro-
duce the Si͑100͒ 2ϫn structure as described in Ref. 19. To
prepare a clean surface, the sample was flashed up to
1150 °C in the UHV chamber for several seconds, maintain-
ing the pressure better than 10Ϫ9 mbar. A well ordered 2
ϫn structure can be achieved with 15–20 cycles of flashing.
A water-cooled e-beam bombardment evaporator was
used to deposit Fe, while Co was deposited by direct current
heating of a W wire coated with Co film by electrodeposi-
tion. Fe or Co silicides were prepared by subsequent anneal-
ing to specified temperatures, which were measured by an
infrared pyrometer.
Owing to a big difference in defect density between the
Si͑100͒ 2ϫn and 2ϫ1 surfaces, it is expected that growth
and structure of silicides on two surfaces should differ sig-
nificantly. Recently, Goldfarb and Briggs15 have found two
different reconstructions for the CoSi2 islands on the 2ϫn
and vicinal 2ϫ1 substrates: 3D islands with a c(2ϫ4) struc-
ture on the former, while 2D platelets with a p(2ϫ2)
ϩc(2ϫ2) mixed phases on the latter.
Figure 1͑a͒ shows the STM image of the as-prepared 2
ϫn surface, which is characterized by the DVLs perpendicu-
lar to the dimer rows. The DVLs pervade the whole surface
with fairly good spatial periodicity. For most samples we
studied, n varies from 8 to 12.20
In this letter, we report ordered adatom vacancy array on
the FeSi2 or CoSi2 islands that are grown on the Si͑100͒ 2
ϫn substrate by molecular beam epitaxy ͑MBE͒. Such an
array does not appear on the silicides formed on the Si͑100͒
2ϫ1 substrate. Although similar vacancy array on NiSi2 was
ϳ2 Å Fe was then deposited on the above 2ϫn surface
at RT. After subsequent annealing at 800 °C for 5 min, Fe
silicide islands with an orientation parallel to the ͗011͘ direc-
tions were observed. The typical morphology includes both
3D islands with flat tops and faceted huts, similar to the
observation by Goldfarb and Briggs.15 According to Behar
et al.,4 the Fe silicide formed at 800 °C can be identified as
a͒
Electronic mail: qkxue@aphy.iphy.ac.cn
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
0003-6951/2002/80(11)/1990/3/$19.00 1990 © 2002 American Institute of Physics
128.248.155.225 On: Sun, 23 Nov 2014 05:13:01

Appl. Phys. Lett., Vol. 80, No. 11, 18 March 2002
Wang et al.
1991
FIG. 1. ͑a͒ Typical STM image of as-prepared Si͑100͒ 2ϫn surface ͑100
ϫ100 nm2, VsϭϪ2.0 V, Iϭ0.03 nA͒; ͑b͒ atomic resolution image of the
FeSi2 islands after annealing at 800 °C ͑14ϫ14 nm2, VsϭϪ0.67 V, Iϭ1.0
nA͒; ͑c͒ FeSi2 island after annealing at 1100 °C ͑20ϫ20 nm2, VsϭϪ0.55 V,
Iϭ0.5 nA͒; ͑d͒ FeSi2 island formed on Si͑100͒ 2ϫ1 after annealing at
800 °C for 5 min ͑8ϫ8 nm2, Vsϭ3.2 V, Iϭ0.028 nA͒.
FIG. 2. ͑a͒,͑b͒ Atomically resolved STM images of the CoSi2 island on
Si͑100͒ 2ϫn surface after annealing at 800 °C, ͑a͒ 12ϫ12 nm2 ͑VsϭϪ1.14
V, Iϭ0.036 nA͒; and ͑b͒ 17ϫ17 nm2 ͑VsϭϪ0.44 V, Iϭ0.04 nA͒; ͑c͒ CoSi2
island grown on Si͑100͒ 2ϫ1 ͑10.5ϫ10.5 nm2, VsϭϪ1.0 V, Iϭ0.04 nA͒;
͑d͒ top view of the structure model proposed for the vacancy array. The
hatched and solid circles denote Si adatoms and Fe ͑Co͒ atoms, respectively.
The open circles correspond to the second layer Si.
the FeSi2 phase. The atomic resolution image of the 3D is-
lands is shown in Fig. 1͑b͒, where the adatom vacancy array
appears on the islands with vacancy sizes of a0ϫa0 , 2a0
ϫ2a0 , 2a0ϫ3a0 , 3a0ϫ3a0 , etc. Within the vacancies,
stronger than that on the Si͑100͒ 2ϫn. This feature can be
understood based on the fact that the long-range interaction
decays as LϪ2 14 For the striped vacancies, L equals to 3a0 ,
.
there are some abnormal adatoms, which are shifted by 1.9 Å
much less than na0 of the Si͑100͒ 2ϫn ͑nу8 in our case͒.
As a consequence, the striped vacancies become more or-
dered than the 2ϫn.
We also deposited the same amount of Fe on the Si͑100͒
2ϫ1 surface with similar annealing procedure. In this case,
no vacancy array, but only a 3ϫ3 reconstruction is observed,
as shown in Fig. 1͑d͒. Since the difference between the 2
ϫn and 2ϫ1 is in the DVLs, the formation of the vacancy
array must be a result of the defects induced by the Ni con-
tamination.
¯
͑the half unit cell͒ along the ͓011͔ and 011 directions with
͓
͔
respect to the surrounding normal adatoms. Hereafter we re-
fer to the former as ‘‘center atoms,’’ and the latter as ‘‘frame
atoms.’’18 The vacancies usually contain four center atoms,
and vacancies with 0, 1, or 2 protrusions are also observed.
Interestingly, there are no more than two adjacent adatoms to
¯
align simultaneously along the ͓011͔ and 011 directions.
͓
͔
Unlike the case on 2ϫn surface, the vacancies are not con-
nected together to form lines. This suggests that, at the an-
nealing temperature, although some adatom vacancies are
mobile, the short-range attractive interaction between the ad-
jacent vacancies is not strong enough to overcome the ther-
mal fluctuation. In addition, both the center and frame atoms
have a nearest-neighboring distance of 3.8 Å, the same as the
lattice constant of ␥-FeSi2 ͑100͒ so this Fe disilicide can be
recognized as the ␥-FeSi2 phase.
To understand the formation of the vacancy array, we
carried out a similar experiment with Co. The Co silicide
islands with heights of 3.5–4.0 nm are formed, and most of
them have a flat top. As CoSi2 phase appears around 500 °C,
the observed silicides can be identified as the CoSi2 phase.21
The high resolution images under different biases are shown
in Figs. 2͑a͒ and 2͑b͒, in which the vacancy features are very
similar to that of ␥-FeSi2 . Both the frame and center atoms
have a nearest-neighboring distance of 3.8 Å, the lattice con-
stant of the CoSi2(100) surface. Some vacancies with three
protrusions can be observed occasionally. For comparison,
the same amount of Co were deposited on the Si͑100͒ 2ϫ1
surface. After annealing at 800 °C for 5 min, CoSi2 forms
and the high resolution STM image is shown in Fig. 2͑c͒.
Instead of the ordered vacancy array, only small point defects
without center atoms appear randomly on the CoSi2 surface.
These results are similar to the observation by Sirringhaus,
After further annealing the sample up to ϳ1100 °C, the
island number significantly decreases at an expense of in-
creasing island sizes. The atomic resolution image is shown
in Fig. 1͑c͒, where the vacancy array becomes much ordered.
More interestingly, some vacancy patches connect each other
along one of the ͗011͘ directions, forming a straight stripe
with a width of 3a0 . In term of the above two annealing
temperatures, we can infer that the barrier height for the
coalesce of adatom vacancy ranges from ϳ70 to ϳ90 meV.
This striped domain is very similar to the DVLs on the
Si͑100͒ 2ϫn surface. The difference lies in a monotonic
width of 3a0 and the embedded center atoms for the former.
As the striped vacancy domains are straight, we speculate the
Lee, and Kanel,22,23 where the &ϫ&R45° reconstruction of
¨
S surface was also observed. The experiment suggests that
the formation of vacancy array on CoSi2 is also related to the
dimer defects on the 2ϫn substrate.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
long-range repulsive interaction between vacancies to be
128.248.155.225 On: Sun, 23 Nov 2014 05:13:01

1992
Appl. Phys. Lett., Vol. 80, No. 11, 18 March 2002
Wang et al.
Three structure models were previously proposed to ac-
1 S. Mantl, Mater. Sci. Rep. 8, 1 ͑1992͒.
2 S. P. Murarka, Silicides for VLSI Applications ͑Academic, New York,
1983͒.
count for the vacancy arrays on NiSi2 islands.16–18 In the
model by Ono et al.,18 the NiSi2 surface was supposed to be
similar to the C surface of CoSi2 , i.e., the Si adatoms are
located at the bridge sites of the second layer Ni. Khang and
Kuk17 suggested the stacking fault to be the main mechanism
for the vacancy array formation, and assigned the first layer
to be Co atoms. However, the AES measurement indicated
that the surface of NiSi2 is Si-rich compared to bulk NiSi2 .16
As ␥-FeSi2 , CoSi2 , and NiSi2 all have the same CaF2
structure with a lattice constant of 3.8 Å on the ͑100͒ plane,
the structure model of the vacancy array should be very simi-
lar. We cannot rule out the possibility of the stacking fault
model by Khang and Kuk, but we find that the model by
Becker et al.16 can best explain our observations. As illus-
trated in Fig. 2͑d͒, there are 1.5 ML of Si on top of the upper
Fe or Co layer, with the frame atoms located at the T4 site
and the center atoms at H4 site.
3 R. T. Tung and K. Inoue, in Microscopy of Semiconducting Materials
͑Institute of Physics, London, 1997͒, Vol. 157, p. 487.
4 M. Behar, H. Bernas, J. Desimmoni, X. W. Lin, and R. L. Maltez, J. Appl.
Phys. 79, 752 ͑1996͒.
5 A. L. Vazquez de Parga, J. de la Figuera, C. Ocal, and R. Miranda, Euro-
phys. Lett. 18, 595 ͑1992͒.
6 J. A. Martin, D. E. Savage, W. Moritz, and M. G. Lagally, Phys. Rev. Lett.
56, 1936 ͑1986͒.
7 X. Chen, F. Wu, Z. Zhang, and M. G. Lagally, Phys. Rev. Lett. 73, 850
͑1994͒.
8 F.-K. Men, A. R. Smith, K.-J. Chao, Z. Zhang, and C.-K. Shih, Phys. Rev.
B 52, R8650 ͑1995͒.
9 M. H. Tsai, Y. S. Tsai, C. S. Chang, Y. Wei, and I. S. T. Tsong, Phys. Rev.
B 56, 7435 ͑1997͒.
10 K. Muller, E. Lang, L. Hammer, W. Grimm, P. Heilman, and K. Heinz, in
Determination of Surface Structure by LEED, edited by P. M. Marcus and
F. Jona ͑Plenum, New York, 1984͒, p. 483.
11
¨
H. Niehus, U. K. Kohhler, M. Copel, and J. E. Demuth, J. Microsc. 152,
735 ͑1988͒.
12 K. Kato, T. Ide, S. Miura, A. Tamura, and T. Ichinokawa, Surf. Sci. 194,
L87 ͑1988͒.
In summary, we have studied the surface structures of Fe
and Co silicides grown on the Si͑100͒ 2ϫn and 2ϫ1 sur-
faces, respectively. Ordered vacancy array forms on both
FeSi2 and CoSi2 surfaces when the Si(100)–2ϫn substrate
is used. In contrast, vacancy array is absent for the silicides
grown on the Si͑100͒ 2ϫ1. Such an array is found to be a
common feature for ferromagnetic metal silicides formed on
Si͑100͒ 2ϫn, and is closely related to the DVLs induced by
Ni contamination. Upon annealing further to 1100 °C, some
vacancies coalesce into stripes along one ͗011͘ direction.
There are no more than two adjacent adatoms to align simul-
taneously along both the principal directions. We emphasize
that upon further optimization, this array can be a useful
template for growing nanodot arrays.
13 P. C. Weakliem, Z. Zhang, and H. Metiu, Surf. Sci. 336, 303 ͑1995͒.
14 H. J. W. Zandvliet, H. K. Louwsma, P. E. Hegeman, and B. Poelsema,
Phys. Rev. Lett. 75, 3890 ͑1995͒.
15 I. Goldfarb and G. A. D. Briggs, Phys. Rev. B 60, 4800 ͑1999͒.
16 R. S. Becker, A. J. Becker, J. Sullivan, and R. T. Tung, J. Vac. Sci. Tech-
nol. B 11, 752 ͑1993͒.
17 Y. Khang and Y. Kuk, Phys. Rev. B 53, 10775 ͑1996͒.
18 I. Ono, M. Yoshimura, and K. Ueda, J. Vac. Sci. Technol. B 16, 2947
͑1998͒.
19 J.-Y. Koo, J.-Y. Yi, C. Huang, D. Kim, and S. Lee, Phys. Rev. B 54, 10308
͑1996͒.
20 J.-L. Li, X.-J. Liang, J.-F. Jia, X. Liu, J.-Z. Wang, E.-G. Wang, and Q.-K.
Xue, Appl. Phys. Lett. 79, 2826 ͑2001͒.
21
¨
¨
B. Voigtander, V. Scheuch, H. P. Bonzel, S. Heinze, and S. Blugel, Phys.
Rev. B 55, R13444 ͑1997͒.
H. Sirringhaus, E. Y. Lee, and H. von Kanel, Phys. Rev. Lett. 74, 3999
͑1995͒.
22
23
¨
This work is supported by the Natural Science Founda-
tion of China under Grant Nos. 69625608 and 60076009.
¨
R. Stalder, C. Schwarz, H. Sirringhaus, and H. von Kanel, Surf. Sci. 271,
355 ͑1992͒.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.248.155.225 On: Sun, 23 Nov 2014 05:13:01