Optuca~
properties
of
semuconducting iron dismcudethin tums
M.
c.
Bost
and.
J. E. Mahan
Department
of
ElectricalEngineering
and
Condensed
Matter
Sciences Laboratory, Colorado State University,
Fort Collins, Colorado 80523
(Received 14 January 1985; accepted for publication 13June 1985)
Irondisilicide thin films were preparedby furnace reaction
ofion
beam sputtered iron layers with
single-crystal silicon wafers and with low-pressure chemical vapor deposition (LPCYD)
polycrystalline silicon thin films. X-ray diffraction indicates the films are single-phase,
orthorhombic, ,8-FeSi
2
•
Impurity levels are below the detection limit of Auger spectroscopy.
Normal incidence spectral transmittance and reflectance dataindicate a minimum, direct energy
gap of0.87 eV. The apparent thermal activation energy of the resistivity in the intrinsic regime is
about halfofthis minimum optical gap. With such a direct band gap, the material may be suitable
for the development of both light-sensitive and light-emitting thin-film devices within the silicon
microelectronics technology.
TABLE
l. Semiconducting transition metal silicides,
SAMPLE PREPARATION AND MATERIALS
CHARACTERIZATION
ed metal.filmwith a silicon substrate. Because of the obvious
opportunity for developing, within the planar silicon tech-
nology, new optoelectronic devices
that
could operate in the
infrared region, a fundamental investigation of the optical
and photoelectronic properties seems long overdue. To that
end, we have prepared FeSi
2
thin films and have investigated
their optical properties, which are described below.
Forbidden energy gap
Reference
0.35 eV
0.30
2
1.38 3
0.9-1.0
4
0.85
5
0.7-D.8
6
0.8 7
0.19
7
0.12
7
0.48
7
1.8
8
1.9
7
0.75 7
LaSi,
ReSi,
BaSi,
OsSi
2
Ca
2
Si
Mg,Si
Material
Iron
disilicide may assume two crystal structures: a
l.ow
temperature, orthorhombic, semiconducting,8 phase, and a
high temperature, tetragonal, metallic
a phase. There is
some uncertainty regarding the transition temperature. Val-
ues given are
915,14.15
937,5
950,15
and 960
0C.
6
Thecomposi-
tion range of the
a phase is thought to be
rather
broad, ex-
tending from 69.6 to 72.1 at.
% silicon (the excess silicon
being due to vacancies in the iron sublattice). The,8phase has
not been found to deviate detectably from the stoichiometric
composition, when carefully prepared.
16
Thin films of iron disilicide were prepared by furnace
reaction (in
99.995% pure argon) of ion beam sputtered iron
INTRODUCTION
The
prospects of fabricating narrow-band-gap, semi-
conducting transition metal silicides within the planar sili-
con technology are intriguing from both practical and fun-
damental viewpoints. One may envision the development
of
new optoelectronic devices, and find a fertile fieldofresearch
leading to a better fundamental understanding of the sili-
cides as a class of materials.
The
semiconducting nature of several transition metal
silicides has been established in past investigations of the
transport properties of bulk samples. Table I givesprevious-
ly published values of the forbidden energy gaps of the semi-
conducting phases thought to
be
of
essentially stoichiome-
tric composition. These values, the simplest interpretation
of
them
being the 0
OK
energy gaps, were inferred from the
temperature dependence
of
the electrical resistivity or in
some cases of the
Han
coefficient. There are other apparent-
ly semiconducting phases whose compositions are less well
defined. These include MnSi
17
[having reported band-gap
values of
0.4,9
0.702,10
and 0.9 eV,7and IrSi
175
(1.2
eV)].1I
Thereare a variety of ways in which the semiconducting
nature of a substance may
be revealed. The concept of
the
minimum metallic conductivity was introduced by Mott,
who
put
an upper limit ofabout 3,000
flO
em on the resistiv-
ity
of
a metallic
phase."
The semiconducting manganese
silicide has been reported to exhibita resistivity as high as0.1
o
em;'?
and the iridium silicide was prepared in thin-film
form having a room temperature resistivity of 6.
70 cm.
The
thermally activated character
of
the resistivity usually is a
reliable indicator of the semiconducting state; it has been
the
basis for classifying almost all
of
the previously mentioned
silicides as semiconductors. Perhaps the most direct indica-
tor
of
semiconduction is the observation of an optical-ab-
sorptionedge. The semiconducting nature
of
the iridium sili-
cide was demonstrated in this way, the data being consistent
with a direct transition.
The optical properties of most of the semiconducting
silicides are unknown, because most property studies of
these have been done on bulk samples which absorb very
strongly above the fundamental edge. However, these mate-
rials may be prepared in thin-film form by codeposition of
metal/silicon mixtures or by furnace reaction of the deposit-
2696
J. Appl. Phys. 58 (7), 1
October
1985
0021-8979/85/192696-08$02.40
© 1985 American Institute of Physics
2696
TABLE
II. X-ray diffraction peaks used in phase identification.
layers (99.99% purity sputtering target) with polished sili-
con wafers [n type,
(100),
20-50 n ern], and with similar
wafers that (before iron deposition) were thermally oxidized
to 1000
Aand coated with 5000 Aof polycrystalline silicon
[polysilicon)by low-pressure chemical vapor deposition. In
the search for suitable furnace annealing conditions, FcSi as
well as the two disilicide phases were observed. The furnace
cool down, though requiring several hours, was sufficiently
rapid to obtain
the
metastable a phase.
It
was found
that
a
fewprominent and unique x-ray diffraction peaks associated
with each phase could be utilized for phase identification in
the furnace-reacted films. The two-theta values of these
peaks for copper
K-a radiation, and their Miller indices, are
given in Table
II.17
For
1000and 1050·C anneals (all anneals were for 120
min except as specificallymentioned below), the only detect-
able phase formed on either substrate type was a-FeSi
z.
For
850 and 900 ·C, x-ray diffraction revealed
,B-FeSi
z,
with one
wafer out of twelve exhibiting a single prominent FeSi peak
(a
3D-min
anneal at 900 ·C yielded samples exhibiting several
FeSi peaks and some from
,B-FeSi
z).
It was found
that
at
950·C
a mixture
of
the
a-
and ,B-disilicide phases was ob-
tained, with the,Bphase enhanced for the polysilicon-coated
substrate as compared to the bare wafer. A
"thick"
iron lay-
er (over 1500
A.)
favored
,B-FeSi
z
formation, while a
"thin"
"The
most intense peak.
Phase
FeSi
,B-FeSi
2
a-FeSi
2
Two theta
28.2'
34.6'
45.3"
50.1"
29.2"
46.0"
46.6"
49.5"
17.4"
37.8"
49.1"
53.6"
Miller indices
(110)
1111)
(210)"
(211)
(220)
and
(202)"
(313)
and
(331)
(040)
(422)
(001)"
(101)
(102)
(113)
layer (under 1000
A)
favored the a phase. Indeed, a radial
variation in the relative amounts of the
a and,B phases over
the 4-in.-diam silicon substrates corresponds in this way to
the radial variation in the sputtered iron layer thickness.
This is consistent with the previously reported tendency of
the
a phase to be deficient in iron.
An annealing temperature
of
900 ·C was selected to pre-
pare,B-FeSizlayers of the highest possible quality on single-
crystal substrates for the optical measurements. A represen-
tative x-ray diffraction pattern is shown in Fig. 1 for a
6700-A.film.The peaks are labeled with the Miller indices of
,B-FeSi
z.
A so-called "forbidden" silicon peak, due to multi-
ple reflections in the single-crystal substrate, is also seen.
Compositional analysis by Auger spectroscopy is given in
Fig. 2.
The
Augerspectra, obtained after ion milling for sev-
eral minutes into the film, exhibit strong
Fe(LMM)
and
Si(KLL) peaks. The
O(KLL)
and
C(KLL)
signals, which
were strong on the film surface, were below the detection
limit of the intstrument within
the
interior of the films
and
are not seen in the data. The
Ar(LMM)
peak isan artifactdue
to the ion milling procedure; there were no other detectable
impurities.
A scannning electron microscope (SEM) fracture cross
section of a typical sample is shown in Fig. 3. Clearly visible
are the
,B-FeSi
z
filmand the underlying single-crystal silicon
substrate. The film thickness is approximately 8000
A;
sur-
face roughness appears to be negligible for the optical mea-
surements to be discussed below.
OPTICAL PROPERTIES
Transmittance and reflectance spectra for a series of
film thicknesses are shown in Fig. 4. The light was incident
on the silicide layer. The transmittance data are referenced
to an identical bare silicon substrate. With increasing film
thickness, the transmission for energies above that of the
absorption edge and the interference fringe spacing below
the edge decrease, as expected. Thesedatawere used, togeth-
er with the optical model given in the Appendix, to obtain by
a damped Newton fitting
procedure" the complex index
of
refraction (n
z
)of the silicide layer in the region of the funda-
mental absorption edge:
n
2
= n
z
+ ik
z
'
(1)
n
z
and k
z
are the real and imaginary parts, respectively. The
model includes reflections at the front surface of the film,
:d
I
...
.s
Si
>-
l-
ii;
z
(220)
ILl
I-
and
~
(422)
(313)
(202)
~
(224)
(204)
(04n
(~~1)
I
~
one!
(042)
(33)1
(511)1
(04011
(600)
(312)
If
1
1
11A
~
I
IA.
- "'-
s
60
50
40
30
28 (deorees)
FIG.
I. X-ray diffraction pattern for a
,B-FeSi
2
film prepared on a single-crystal substrate.
2697
J. Appl. Phys., Vol. 58,
No.7,
1
October
1985 M. C. Bost andJ. E.
Mahan
2697
FIG. 2. Augerelectron spectra in the
appropriate energy windows for (a)
argon
and
carbon, (b) oxygen, (c)
iron, and (d) silicon, obtained after
ion milling for several minutes into a
fJ-FeSi
2
film.
-2
(c)
Fe(LMM)
-8
4
2
Ar(LMM)
~
0
C(KLL.)
O~
f.
-2
- I
!-
~
(0
)
-21-
:d
-4
~
200
220
240
260
280
300
450
470
0
ILl
"U
......
ILl
......
Z
"U
8
4
- 3
L-_
.......
__
L-_-L._--JL.--~
650
670 690
710
730
750
1580 1595
1610
1625 1640
ENERGY
(ev
)
multiple internal reflections in both
the
film and the sub-
strate,
and
absorption within the silicide film but not in the
substrate. Values for the index of refraction of the silicon
substrate were obtained from Briggs.
19 The
data
were
smoothed by averaging the value of a
data
point
and
each
neighbor, with the center point being weighted twice as
much as either neighbor.i?
The
optical-absorption coefficient (K
2
),
calculated from
K
2
=
41Tk
2!A
(2)
(A
being
the
wavelength)
and
the
data
of
Fig. 4, is shown in
Fig. 5.
The
onset of absorption
just
below 0.5 eV is due to
extrinsic transitions involving defect states within the for-
bidden energy gap, while
the
second rise above 0.8 eV indi-
cates
the
onset
of
the fundamental interband transitions.
The
data
have been replotted in Fig. 6 in a manner
that
would
yield a linear plot for direct allowed transitions, with the
intercept indicating the minimum
photon
energy (E
g
)
re-
quired?':
(3)
h is Planck'sconstant, v is
the
frequency,
and
C isa constant
depending on
the
details of
the
band structure.
It
does ap-
pear that
the
transition is direct, with a value of 0.87 eV for
the
minimum
photon energy.
The
solid line in the figure was fitted to the
data
using
the method
of
least squares.
Data
below 0.88 eV were not
used for
the
fit. An analogous plot of (K2
hv
)
1/2
versus
photon
energy, which would be appropriate for an indirect
gap,21.22
is not linear over any appreciable energy range.
Our
belief
that
the absorption below the fundamental
edge is due to defects is consistent with the fact
that
the
material is in
the
form of a fine-grained polycrystalline thin
film.
A similar absorption tail has been observed. for grain
boundary defect states
in fine-grained polysilicon
.23
The
de-
fect
part
of the absorption coefficient was obtained by sub-
tracting from the
data
the
part due to interband transitions;
the results
are
shownin Fig. 7 together with the fundamental
edge as obtained from
the
fitting procedure of Fig. 6.
The
defect absorption spectrumindicates a peakin
the
density of
states at 0.6 eV from either the valence- or the conduction-
band edge.
If
the constant of proportionality were known, it
would
be possible to estimate the total defect state density
from the magnitude of the absorption constant as has been
done for amorphous
silicon."
Alternatively, assuming an
optical cross section of
10-
15
ern? gives a density on the
order
of 10
19
cm-
3.
This cross section is typical of impurity
levels in silicon materials used as extrinsic photoconductive
detectors.
25
Electrical measurements on the films formed on bare
silicon wafers are not meaningful due to the shorting effect of
the substrate, but conductivity data have been obtained for
films made
by an identical procedure on thermally oxidized
and polysilicon-coated substrates as described above.
The
behavior of a representative sample is shown in Fig. 8. The
temperature dependence indicates the films are extrinsic at
room temperature, where
the
value is
-1
(0
cm)-I.
Above
- 500 K the conductivity is thermally activated with an ap-
parentactivation energy
of
- 0.43 eV.
The
simplest interpre-
tation is
that
the
material
has
become intrinsic with
the
acti-
2698
J. Appl. Phys., Vol. 58,
No.7,
1 October 1985
M. C, Bost and
J. E. Mahan 2698
FIG. 3. SEM fracture cross section showing 8-FeSi
2
prepared on a bare silicon wafer.
2699
J.
Appl.
Phys., Vol. 58,
No.7,
1 October 1985
M. C. Bost and
J. E. Mahan 2699
60
_0
o
o
o
o
60
0.2 0.4 0.6
0.8
1.0
1.2
0.2
0.4
0.6 0.8 1.0 1.2
1I00A
o
o
_0
o
o
o
o
00
0.2 0.4 0.6 0.8 1.0 1.2
ENERGY
(eV)
0.2 0.4 0.6 0.8
1.0
1.2
ENERGY
(eV)
FIG.
4. Spectral transmittance and reflectance data for films of four different thicknesses.
vation energy equal to half the forbidden energy gap, which
corresponds fairly well to the optically determined value of
0.87 eV.
Room-temperature Hall-effect measurements indicated
the films are
p type, with an apparent hole concentration of
- 2X
10
18
em
-3
and a Hall mobility
00
cm
2
/V
s. At room
temperature the resistivity in Fig.
8 is thermally activated
but with a smaller activation energy
(0.13
eV) than in the
intrinsic regime. Thus, the room temperature carrier con-
centration is determined by acceptor states separated from
o
FIG. 6. Absorption coefficientdata showing the existence of a direct transi-
tion.
1.20.6
0.8
1.0
ENERGY
(eV)
7
0
6
N
E
u
5
<,
N
>
4t
~
Q
4
N
,..--,
;:,
3
s:
N
..:s
2
°
0
'E
°
...
10
5
0
.£
°
-
e
~
z
5
°
-
IJJ
()
IJ..
3
IJ..
IJJ
°
0
()
z
0
Q
0
I-
cO°cP
a.
10
4
0::
ctJ
O
°0
e
0
(J)
m
0
<t
--l
5
SOD
<t
()
0
j::::
3
0
a.
°
0
0
°
10
3
0.4 0.6
0.8 1.0
1.2
ENERGY
(eV)
FIG.
5. Optical-absorption coefficient vs photon energy.
2700
J. Appl. Phys., Vol. 58.
No.7.
1 October 1965
M. C. Bost and J. E. Mahan 2700