1) Evaporated and sputtered InSb films have different crystallization properties. Evaporated films have a lower activation energy of 1.39 eV and crystallize by three-dimensional growth on existing nuclei.
2) Sputtered films have a higher activation energy of 2.7 eV and crystallize by nucleation and subsequent growth on the nuclei. The Avrami exponent indicates sputtered films require crystalline nuclei to form before crystallization can begin.
3) Transmission electron microscopy images show evaporated films have existing nuclei while sputtered films have no visible crystallites in the as-deposited state, requiring nucleation.
2. 234
the crystallization. The dimensions of the probe
beam and the film thickness are small enough to 550 500 450 400
consider the temperature uniform over the
2 InSb
volume probed. 20 ~
The samples that we have investigated consist
of a single InSb layer on a thick (1.2 mm) glass oo
/
substrate. The heat diffusivity D and heat con-
ductivity K of the substrate are D = 4.90 x 10 ~ /
m 2 s 1 and K = 1.1 W K t, respectively. From /
/
these thermal properties it can be deduced that /
the rise time of the temperature is approximately -2 /
200 /~s [2]. The temperature distribution has
/
-3
/
been calculated using the expression of Pittaway
for a surface heat source on a semi-finite sub- 1.8 2.0 2.2 2.4 2.6
strate [3]. The InSb layers were prepared by
IO00/T [K1] ......
sputtering or by flash evaporation. The composi-
tion of the film was determined by X-ray fluores- Fig. I. Arrhenius plot of the transformation time r', as
cence (XRF) to be InSb within 2 at.%. The deduced from the transmittivity, vs. the temperature. The
slope of the line gives the activation energy. The sample was a
non-crystallinity of the samples was checked with 20 nm InSb film prepared by evaporation.
X-ray diffraction (XRD) by the absence of sharp
peaks. It appeared that evaporated films thicker
than 100 nm were at least partially microcrystal-
line. In our experiments we used evaporated films 30 t • oo
with thicknesses between 20 and 90 nm, and I 2.5 •
sputtered films between 20 and 160 nm.
2.0F
> i
3. Results uZ 1.5~ :: s
o
At constant temperature T the crystalline frac-
tion x(t) during the transformation is usually
described by the Avrami equation [4] k k
0 4~0 8~0 120 160 200
x(t) = 1 -exp{ -(t/r)'"} d (nm) ---,,.
where m is the Avrami exponent. For an activated Fig. 2. Activation energy vs. film thickness for a number of
lnSb films. The full circles represent samples prepared by
process the temperature dependence of the sputtering, the open circles films prepared by flash evapora-
characteristic transformation time r can be tion.
written as
r = r 0 exp(EacJkT ) Transformation times longer than 1 ms are
where East is the activation energy. included only, since for shorter times the non-
This activation energy can be deduced directly isothermal part of the process cannot be neglec-
from measurements of the transmittivity. In our ted. The crystallization time varies rapidly over a
samples the transmittivity increases during the temperature range less than 200 K. Also, the
amorphous-to-crystalline transition. For this experimental points can be fitted with a straight
purpose let us define the transformation time r' line. This indicates that the crystallization process
as the time needed for the transmittivity to cross a can be described with one activation energy only.
level that is 1.15 times the start level. This r' is The same conclusions can be made for the other
proportional to the transformation time r. How- samples that have been measured.
ever, it should be noted that the proportionality From the slope of the line in Fig. 1 the activa-
constant is different for different samples. From a tion energy can be calculated. In Fig. 2 the activa-
plot of log r' against l/T, the activation energy tion energies for a number of evaporated and
Eac t c a n be calculated. In Fig. 1 the result for an sputtered samples are shown. The activation
evaporated sample of 20 nm of InSb is shown. energy is independent of the film tl-fickness. There
3. 235
is a large difference, however, between sputtered
I
and evaporated films. The average activation
energy for an evaporated film is 1.39(5) eV
atom-~, while for sputtered films it is 2.7(1) eV
g120
140
u~ 100
/
/
atom ~.
To investigate this difference further, we have
80 /"
60
determined the Avrami exponent m for one sput- d c = 1.08 d a
40
tered and one evaporated sample. From the
20
Avrami exponent it can be deduced whether the
i i
crystallization starts on existing nuclei or not and 0 2'0 4'0 6'0 dO 100 120 140
what the dimensionality of the growth process of da (nm) --
the crystallites is (see e.g. ref. 5). However, the Fig. 3. Film thickness after crystallization, d~, vs. film thick-
measurement of rn is less trivial than that of E~,,t. ness of the amorphous phase, d,, for films prepared by
To obtain accurate values of m, the crystalline evaporation. The solid line corresponds to d = 1.08(4)d..
.fraction x(t) has to be calculated from the experi-
mental reflectivity and transmittivity. For this we
need to know the values of the optical constants
of the amorphous and crystalline phases at the
temperature at which the transmittivity and
reflectivity have been measured. Since accurate
values for thin films are available only at room
0 1 2 3 4 5 6
temperature [6], the following procedure was
In(t/tp)
applied. The sample is heated with a short laser
pulse from the heating laser. After the sample has Fig. 4. Avrami plot of the cryslallinc fraction x(t) for an
evaporated sample of 70 nm; tr is the pulse time of the
cooled down, the transmittivity and reflectivity heating pulse.
are measured. This is repeated a large number of
times on the same area of the sample. In this way
the sample is crystallized gradually while the denced by XRF and Rutherford backscattering
transmittivity and reflectivity are measured at measurements.
room temperature. This allows us to use the opti- In Fig. 4 an Avrami plot is shown of the crys-
cal constants that have been determined for sput- talline fraction x(t) resulting from an experiment
tered InSb at room temperature: for amorphous on a 70 nm InSb layer produced by flash evapor-
InSb n = 4 . 8 2 - 1.95i and for crystalline InSb ation. If the crystallization process could be
n = 4 . 0 7 - 0 . 7 5 5 i [6]. The dielectric constant of described completely by the Avrami equation,
the mixture of the amorphous and crystalline this would be a straight line with slope m. How-
phase is calculated using an effective medium ever, the curve deviates strongly from the simple
theory [2, 7]. This is valid if the crystalline phase Avrami behaviour. This is typical for all samples
appears homogeneously distributed throughout considered, both sputtered and evaporated. This
the volume of the film. deviation is not unexpected for a thin film. In the
The effective optical constants are used to derivation of the Avrami expression the im-
extract the crystalline fraction x(t). The calcula- pingement of crystallites is taken into account.
tion includes volume changes during the transi- However, it is assumed that the material extends
tion. For sputtered films Holtslag and Scholte [6] infinitely in all directions. This is obviously not
observed that the film thickness d changes slightly true for a thin film. Therefore one may expect the
during crystallization: d c = 1.015 d a. In flash- crystallization to become lower dimensional
evaporated films the volume change is much when the crystallites reach the interfaces of the
larger (see Fig. 3): d~=l.08d~. The volume film [9]. Also, one cannot neglect the effect of the
change upon fusion of lnSb is reported to be stress induced by the volume expansion during
11.4%-13.7% [8]. Therefore the density of the the crystallization. This may become especially
evaporated sample is comparable to the density important when the crystals start to impinge.
of the liquid. The sputtered sample is less dense. From the slope at the beginning of the curve,
This can be attributed to the incorporation the Avrami exponent at the start of the crystalliza-
of a considerable amount of argon, as evi- tion process can be calculated. For an evaporated
4. 236
film of 70 nm m = 1.5(2) was found. For a sput- Combined with the high vaiues ot the Avrami
tered film of 92 nm rn = 3(1). The value for the exponent and the activation energy, this indicates
sputtered film is rather inaccurate owing to the that crystalline nuclei first have to be formed
high activation energy of the film. To obtain before crystallization can start. The value of the
enough points at the start of the crystallization, a Avrami exponent is in accordance with t ; ; - 2.5
short pulse time has to be used: 0.1 ms vs. 10 ms This value corresponds to a crystallization pro-
for an evaporated film. Consequently, the non- cess that is similar to that in the evaporated layer
isothermal part of the crystallization may have plus an extra nucleation step. Also, as one might
influenced the experimental value of the Avrami expect, the value of the activation energy in the
exponent. However, within the given limits the
experimental value for the Avrami exponent is
correct.
4. Discussion and conclusions
In Table 1 the results for sputtered and evap-
orated InSb are summarized. It is clear that the
crystallization properties depend strongly on the
preparation method. Unfortunately, there is no
unique relationship between the value of the
Avrami exponent and the characteristics of the
crystallization process. Different processes may
give the same value for the Avrami exponent [5].
To understand these differences we have to com-
bine the results of the optical measurements with
the results of T E M experiments on a sputtered
and an evaporated layer of InSb on Si3N 4 sub-
strates [10].
First consider the evaporated layer. From the
Fig. 5. TEM pholograph of an amorphous lnSb layer
low value of the Avrami exponent m = 1.5(2) it is prepared by evaporation. The number of crystalline nuclei
highly unlikely that the crystallization starts with visible corresponds to approximately I() ~ crystatlites m ~'
the formation of crystalline nuclei. This is evi-
denced by Fig. 5, in which a section of an evap-
orated layer is shown. The number of crystallites
visible corresponds to 1 0 ~~ crystalhtes m - .-
- "
Therefore it can be concluded that in evaporated
InSb the crystallization proceeds by three-dimen-
sional diffusion-limited growth on already exist-
!
ing nuclei. In sputtered InSb the situation is
different. As can be seen from Fig. 6, no visible
crystallites are present in the as-deposited layer.
TABLE 1 Comparison between some parameters deter-
mining the crystallization process in thin amorphous InSb
films prepared by sputtering and by evaporation. E~t is the
activation energy for crystallization, m is the Avrami
exponent and d~ and d~ are the film thicknesses of the
crystalline and amorphous layers respectively
t~rameter E vapora~d Aput~red
E~,~, 1.39(5)eV atom t 2.7(l)eV atom
m 1.5(2) 3(1)
Fig. 6. TEM photograph of an amorphous InSb laver
d~/d. 1.08(4) 1.015(5)
prepared by sputtering.
5. 237
sputtered layer is higher than in the evaporated during sputtering. Before crystallization can start,
layer. argon first has to be removed.
Until now, transient nucleation effects have In conclusion, it has been shown that the
been neglected, since the Avrami analysis crystallization process in evaporated InSb pro-
neglects the incubation time that is needed to ceeds by three-dimensional diffusion-limited
reach a steady state nucleation rate. This is justi- growth on existing nuclei. In sputtered InSb
fied for evaporated InSb, since nuclei are already the crystallization proceeds by nucleation and
present. For sputtered InSb one has to take care. subsequent three-dimensional diffusion-limited
However, Gravesteijn has shown that it is pos- growth. The high activation energy in sputtered
sible to crystallize a thin InSb film near the melt- films is mainly due to the high barrier against
ing point in 15 ns [1]. The incubation time at that nucleation in these layers.
temperature will be even shorter. Also, a non-
negligible incubation time would have turned up
in the Arrhenius plot of the crystallization time as Acknowledgments
a deviation from linear behaviour. No such devia- The author gratefully acknowledges Mr. P. van
tion has been observed for crystallization times der Werf and Mr. N. Dreesen for the preparation
larger than 1 ms. Therefore it is not unreasonable of the samples. Dr. J. Coombs, Dr. A. Holtslag
to neglect the incubation time due to transient and Dr. G. Thomas are acknowledged for
nucleation in sputtered InSb also. critically reading the manuscript.
Now we can estimate the activation energies
for the nucleation step and the growth separately.
It can be deduced straightforwardly that the acti- References
vation energy for a sputtered film can be written
as [3, 1 1] D. J. Gravesteijn, Appl. Opt., 27(1988) 736.
C. J. van der Poel, J. Mater. Res., 3 (1988) 126.
Eac, spur= ( E n ~- E~)/rn~p~ L. G. Pittaway, Br. J. AppL Phys., 15 ( 19641967.
M. Avrami, J. ('hem. Phys., 9 (1941 ) 177.
and for an evaporated film as J. W. Christian, in R. W. Cahn (ed.), l~hysical Metallur~w,
North-Holland, Amsterdam, 1971t, p. 47 I.
E act evap = Eg/t~lcv.lp A. H. M. Holtslag and P. M. L. O. Scholte, to be
published.
where E n and Eg are the activation energies for J. C. Maxwell Garnett, l'hil. 7)ans. R. 3oc. Lond., 203
nucleation and growth respectively and rn~p~ and 119114) 385.
rn~v,~ are the Avrami exponents of the sputtered
p N. A. Goryunova, The ('hemist O, ~1 Diamond-like Semi-
and evaporated films respectively. From this we conductors, Chapman and Halk London. 1965, p. 114.
9 M. C. Weinberg, J. Non-('ryst. Solids, 70 1985) 253.
find Eg = 2.113) eV atom i and E n = 612) eV
10 F. J. A. M. Greidanus, B. A. J. Jacobs. F. J. A. den
atom 1. The barrier against nucleation in the Broeder, J. H. M. Spruit and M. Rosenkranz, Appl. f'hvs.
sputtered film is very high. This is most likely due Lett., 54 (1989) 963.
to the argon that is incorporated in the film 11 E. A. Marseglia, .I. Non-(rvst. Solids, 41 1981)) 3 I.