CHAPTER 5

Optical Properties and Electronic Structure of

Amorphous Semiconductors
J. Tauc
Institute of Solid State Physics of the Czechoslovak Academy of Sciences
Prague, Czechoslovakia

I. INTRODUCTION

Optical properties of amorphous solids are of considerable interest since

glasses of various kinds are very important optical materials. We shall
treat in this chapter an example of the absorption spectrum of amorphous
material which is particularly simple. We chose amorphous germanium
for this study and compared its optical properties with the well-known
properties of its crystalline form. The atomic arrangement in amorphous Ge
was investigated by X-ray diffraction [1]. The results of these studies can
be interpreted in two ways; it is at present impossible to decide between
them. In one model amorphous Ge is pictured as composed of small
crystallites with linear dimensions of the order loA. In the other model
amorphous Ge is considered as a homogeneous medium. The tetrahedra
are still the basic units of the structure but the neighboring tetrahedra are
irregularly rotated so that the long-range order is lost after several atomic
distances. For the interpretation of the optical properties we shall use
the latter model here, but the former model would give similar results.
We shall show what information concerning the changes of the electronic
structure due to the loss of the long-range order can be deduced from the
changes observed in the optical properties. This chapter is based on work
performed jointly in the Institute of Solid State Physics in Prague and the
Institute of Physics in Bucharest and reported elsewhere [1-3].
Let us note that both models of atomic arrangement lead necessarily
to the presence of vacancies; these are either concentrated on the boundaries
of the crystallites or homogeneously distributed in the solid. The studies of
transport properties have shown [2] that these vacancies act as acceptors
and are responsible for the p-type conductivity of amorphous Ge having a
123
S. Nudelman et al. (eds.), Optical Properties of Solids
© Springer Science+Business Media New York 1969
124 J. Tauc

large concentration of holes (1018_10 19cm- 3 ). These vacancies may be

filled with foreign atoms.

n. EXPE~ENTALMETHOD

Ge can be prepared in an amorphous state only in thin layers. The

procedure, carried out in a vacuum (10- 5-10- 6 atm), involved evaporation
of very pure Ge on quartz or KCI substrates which were kept at room
temperature. Some layers were annealed at 250°C; they are still amorphous
but have somewhat different properties. Annealing at 450°C converts the
layers into a polycrystalline state whose optical properties are essentially
those of the crystal. We shall denote nonannealed layers by "a-Ge," the
annealed ones by "at-Ge." The thickness of the layers was between 800 and
4000 A. In this range the optical constants did not depend on the thickness
of the layers and are therefore believed to be properties of the "bulk"
material. The structure of the layers was tested by X-ray diffraction and the
thickness was measured by the method of interference fringes.
The optical constants were determined in the range 0.08 to 1.7 eV by
the measurement of transmission and reflection; it was necessary, of course,
to take proper account of the interference effects in the layers. In regions
where the layers were opaque (hw > 1.7 eV), only the reflection spectrum
to be used in the determination of the optical constants from the dispersion
relations was measured. However, in our case we met with particular diffi-
culties which made the determination of the optical constants in the region
of high absorption difficult. We shall, therefore, return to this problem later
and discuss first the three low-energy regions shown in Fig. 1.

t.,
6.0

I,D
I II
TRANSITIONS IN THE TAIL
VALENCE BAND
2D

o
o 0.2 0.4 0.6 0.8
Fig. I. Absorption in amorphous Ge in the infrared region (after [5]).
Optical Properties and Electronic Structure or Amorphous Semiconductors 125

m. TRANSITIONS IN THE VALENCE BAND

The spectrum between 0.08 and 0.6 eV shown in Fig. 2 is qualitatively

similar to the spectra in p-type crystalline G.e; the peaks in amorphous Ge
are shifted toward lower energies. These peaks in crystalline p-type Ge were
interpreted as direct (that is, k-vector conserving) transitions of holes from
the two upper valence bands 1 and 2 into the deep-lying valence band 3
(cf. Fig. 3) and can be approximately described by the formulas

(5.1)
ro 2 B~3 ,." Ifun-LlEI! exp ( m3 Iiro-LlE)
m3-m2 kT
where LIE is the spin-orbit splitting and ml' m2' and m3 are the (averaged)
effective masses.
As seen in Fig. 4, equation (5.1) describes approximately also the absorp-
tion in amorphous Ge (the agreement is better the farther from LIE). The
appropriate values of the parameters LIE, ml , m2' and m3 for various cases
are listed in Table I.
We may now assume that the observed changes of effective masses are
due to the transition from the crystalline to the amorphous state; the changes
are qualitatively in accord with the relation

~3 =K~l + ~J (5.2)

as they should be. This assumption involves the exact conservation of the k
vector (L1k = 0) during the transitions.

Table I. The Effective Masses in Crystalline

and Amorphous Germanium *

LlE(eV)

Crystalline Ge 0.29 0.30 0.04

a-Ge 0.195 0.49 0.029
at-Ge 0.21 0.34 0.033

• The term ma was assumed to be the same as in the

crystal, i.e., ma = 0.08 m.
 ,
126 J. Tauc

-non-anneoled
20 ~---onneo/ed
1
~16

11

0.8
\ \.
0.1, . \x,
Fig. 2. Absorption in amor-
0 01 01 03 O~ 05 06 phous Ge in the region 0.08
fiw(eY)- to 0.6 eV (after [4]).

o Fig. 3. Valence band of Ge.

In (1\.,1' t,
III., - IJf / J/,
5

Fig. 4. Determination of the effective masses (after [5]).

Optical Properties and Electronic Structure of Amorphous Semlcooductors
'"
Alternatively, we may assume that the observed changes of effective
masses are at least partly fictitious and due to the nonconservation of the k
vector du ring the transitions. If we ascribe the whole observed changes of
effective masses to this effect, we obtain the upper limit for the uncertainty
Jk in the conservation of Ikl.
We shall now deduce equation (5.1) under the assumption that k is not
exactly conserved during the transitions. Let us consider the transitions of
holes between the bands I and 3 which are assumed parabolic. We have
ep .... UI -f) Goint density of states)
where f, and f3 are the probabilities that the states in bands 1 and 3 are
occupied by holes. Band 3 is practically free of holes; therefore, f3 = O.
For fl we have
fl = const. exp (-e"dkT)

where £,,1 is the energy of holes in band I. Energy is conserved during the
transitions:
(5.3)
where
h' , h' ,
tP l = - - k, tp) = - - k, (5.4)
2m, 2m,
From equations (5.3) and (5.4) it follows that

k: )
e,,1 = (-m, 2' - 1 (hw-JE) (5.5)
m3 k,

If kl = k3' then the first of equations (5.1) is obtained. Let us now assume
thatk) = kl( I +J) where ,1 is in the interval (-,10' ,10)' ,10 ~ I.
We shall calculate ·the mean value offl assuming for simplicity that all
J's in the above-mentioned interval have the same probability:

1. - _1_ f" /, (~) dJ (5.6)

2,10 -.to

We find that, if Ihw- .1El/kT'P-l, II has tbe same fonn as in the first
of equations (5.1); however, instead of mj we have
(5.7)
We have ascribed the factor to m l and left mJ unchanged; this (nonessential)
assumption seems plausible in view of the fact that band 3 lies deeper than
band I. A similar reasoning regarding the transition 2 .... 3 gives
(5.8)
128 J. Tauc

Table I shows that the difference between the effective masses m 1a , m2a
and m 1 , m2 observed in amorphous and crystalline Ge, respectively, gives
10- 1 as the order of magnitude for 211 0 , In the spectral region considered
k < 10 7 cm -1 and 11k < 106 cm -1. The volume corresponding to 11k in
the reciprocal space Bo = 4n/3' (11k)3 is 4 X 10 18 cm- 3 •
How can this result be understood? The theories proposed for the
electronic structure of an amorphous solid by Gubanov and others [7,8]
consider the difference between the amorphous structure and the crystalline
structure with the same short-range order as a perturbation. Assuming that
the perturbation of the effective potential in which the electron moves is
sufficiently small, they construct the wave functions of an electron (or hole)
in an amorphous solid as linear combinations of the Bloch functions in the
same band of the corresponding crystal. As during the transitions between
the Bloch functions the k-vector is conserved; Bo determined above gives an
estimate of the spread of the k vectors of the Bloch functions entering
into the linear combinations. If we assume that the wave functions in band 3
are exact Bloch functions, Bo characterized the wave functions in bands I or 2
of amorphous Ge.
Bo is a very small part of the reduced Brillouin zone
B = (2n)3. 4/a 3 = 5.6 x 10 24 cm - 3
of the crystal: B" ,..., 10- 6 B. This result can be stated in an alternative way
using the uncertainty relation Bo Vo ,..., (2n)3 where Vo is the volume in which
the wave functions are localized. In our case we obtain Vo "'" 10- 17 cm 3 •
That means that the valence-band wave functions are localized over many
millions of atoms; they are therefore practically delocalized and reasonably
well described by Bloch functions.

IV. TAIL OF THE ABSORPTION EDGE

The tail near the absorption edge can be separated if an extrapolation

is applied to the absorption edge as shown in Fig. 6. In Fig. 5 we see that
the tail has a Gaussian shape and the integrated absorption JWB 2 dw increases
with temperature; at 300 oK it is 1.45 times Jarger than at 150 oK. The tail is
not a transition from or into the valence or conduction bands because we
would expect in such a case an edge rather than a band and therefore it does
not give us information about the distortion of the band edges. It appears
likely that the tail is related to the presence of localized defects.
We have suggested that this tail is caused by excitons bound to neutral
acceptors which are presumably the vacancies mentioned above. From the
temperature dependence of the absorption we have found that their activation
energy must be larger than [Link] eV [5].
Optical Properties and Electronic Structure of Amorphous Semiconductors 129
w.-----------------------,

01
Fig. 5. Temperature dependence
nco
of the absorption in the tail. J =
0'-------""'=----'----------'-- - - ' - ------"""""- - - ' nCO-nCOrn. where nCOrn is the maxi-
-0.2 -01 o a7 0.2 mum of the band (ncom = Eg-O.02
~ h<.J (eVj - eV) (after [4]).

15
~~
~
3

0
06 0.8 12 U 16 Fig. 6. Absorption edge (after [4]).
'}(ufer)

Imagine that the following process takes place: An acceptor captures an

electron by thermal excitation and the photon excites this electron into a
state in which the acceptor binds an electron and a hole. The extent to which
such a process occurs should be proportional to the product of the concen-
tration of ionized acceptors nA and the concentration of holes p.
Let us consider the simplest possibility. In the material there are NA
acceptors/cm 3 with an activation energy EA and ND donors/cm 3 with energy
levels many kTabove the energy level of acceptors; one has then nA = ND+p.
In such a case the de Boer-van Geel formula gives

pnA = (NA-nA) N v exp (- EA) (5.9)

P kT
where N v is the effective density of states in the valence band, and P is the
130 J. Tauc

spin degeneracy of acceptors. In the case that nAIN A ~ 1 (or p ~ N A- N D)

we obtain from the above, given the ratio 1.45 of integrated absorptions at
300 and ISooK, EA = 0.01 eV. Of course, we have no evidence that only a
small number of acceptors is ionized. It may well be that the compensation
is high and nA/NA may not be ~ L In such a case, EA would be greater than
0.01 eV. Detailed measurements of temperature dependence of the tail are
necessary to obtain definite information about the parameters of the
statistics.
The suggested explanation of this tail as caused by excitons bound to
defects is in principle similar to the explanation of similar tails observed
near the fundamental absorption edges of ionic crystals (IX and Pbands [9]).

V. ABSORPTION EDGE

.J
In Fig. 6, Itw 8 2 is plotted against Itw near tile absorption edge for one
of the a-Ge samples. It is apparent that the absorption in the tail is followed
by an absorption which is represented by a straight line showing that in this
region W 28 2 '" (ltw - Eg)2. If we extrapolate this line, we obtain at 300 OK
Eg = 0.88 eV and Eg = 0.92 eV for a-Ge and at-Ge, respectively.
The dependence of 8 2 on Itw near the absorption edge suggests that
one has here transitions for which only energy but not the k vector is con-
served (indirect transitions). We shall show that such behavior is expected
if the wave functions are linear combinations of the Bloch functions as
discussed in section III. We shall develop a theory of the optical absorption
valid for such kinds of wave functions and compare it with the experimental
results. We shall first state more explicitly our basic assumptions.
The wave functions in the valence and conduction bands of the crystal
are Bloch functions:

(5.10)

normalized in the basic volume V; k, k' are in the first Brillouin zone.
We assume that the basic volume V contains the same number of atoms
in the amorphous as in the crystalline state (for simplicity we neglect the
difference of densities of both states).
It is known from both experiment and theory that the valence and
conduction bands retain their meaning in the amorphous state. We shall
assume that the perturbation describing the change of the crystalline into
the amorphous state is such that the wave functions in the valence band of
the amorphous state are in zero-approximation linear combinations of the
wave functions in the valence band of the crystal; an analogous statement
Optical Properties and Electronic Structure of Amorphous Semiconductors 131

holds for the conduction band. Therefore, the ground state described by a
Slater determinant is the same as in the crystal. If we now produce an elec-
tron-hole pair in the crystal with wave functions Ivx >, lex' > and then
apply the perturbation we obtain the wave functions in the amorphous
solid. We can label these wave functions with indices x, x' and denote them
by Ivx >a' lex' >a' As the perturbation is performed in a continuous
spectrum we can assume that the energies of the states Ivx > a and Ivx >
measured from the top of the valence band are the same (similarly for the
conduction band). The consequence of these assumptions is that the cor-
responding densities of states are the same in both cases.
The calculation of the absorption in an amorphous solid is based on
the general one-electron formula

(5.11)

where the summation is performed over all initial and final states i, f in the
basic volume V. P iJ is the momentum matrix element between the wave
functions of the final and initial states lex' >a and Ivx >a' Expressing them
as linear combinations of Bloch functions in the respective bands, we obtain

P iJ = a<ex' Ipi vX)a = L L a<ex'lek') <vklvx)a <ek' Ipl vk)

k k'

= L a<ex'lek) <vklvx)a Pvc(k)

k
(5.12)

where

(5.13)

is the matrix element in the crystal. The summations are performed over all k
vectors in the first Brillouin zone. Q is the volume of the elementary cell
over which the integration is performed.
From equation (5.4) we obtain

2
G2(W) = ( -
ne)2 2 '\'L., L.,'\' IL.,'\' Pvc(k) a<ex' lek) <vklvx)al
~
2

mw II" ,,' k

• c5 [EcCx')-Ev(x)-hw] (5.14)

The summations are taken over all x, x' in the first Brillouin zone. The
factor 2 respects the doubling of the number of states by spin which is
conserved during the transition.
132 J. Tauc

In the crystal we have

(cx'lck) = Jkx ,

Summing over k and x' and replacing then L by V/(2n)3 . J d 3 x we obtain

" B
the usual formula for direct transitions:

Let us now simplify equation (5.14) by an assumption, usually made in the

theory of the optical properties of crystals, that IPve(k)1 is a constant. We put

f(x, x') = N II a(cx'lck) (vklvx)aI 2 (5.16)

k

where N = v/a is the number of unit cells in the basic volume V. In a

crystalf(x, x') = NCJ"",. As it is seen from equation (5.16), in an amorphous
solid, f(x, x') will have a finite maximum for x = x' and a certain nonzero
width. The double sum ILf(x, x') is independent of the actual form of
" ,,'
Ivx > a,lex' > a if these wave functions are constructed as described above
and IPvel is a constant. In the crystal it gives

I If(x, x') = NI I J",,' = NI.1 = N 2 (5.17)

x x' )it x' )it

Therefore, the following sum rule holds

ta 3 x a3 x'f(x, x') = B2 (5.18)

where B is the volume in the reciprocal space of the Brillouin zone. The
existence of a sum rule for f(x, x') is expected from intuitive reasoning.
This sum rule replaces the CJ function by a function with a finite width which
must obey the same normalization equation.
With the functionf(x, x') equation (5.14) can be written as

82(W) = (2ne)2 2
mw (2n)3

We shall now discuss the particularly simple case that near x = x' = 0
the factor f(x, x') is constant, equal to fo in a certain part of the Brillouin
zone Be, and zero in the rest. From the sum rule equation (5.18) we determine
Optical Properties and Electronic Structure of Amorphous Semiconductors 133

10 to be BJBe. In this case in the range considered equation (5.19) gives

= (~n;y (2~)3 ;e IPvel 2 ffdE d 3 x' 0 [Ee(x')-E] f

d 3 x 0 [E-

f
Ev(x)-hw]

= ( -2ne)2 -3
2 1
-IPvel 2
dEge(E) gv(liw-E) (5.20)
mw (2n) Be

It is apparent that e2 (w) is determined by a convolution of the densities

of states in the valence and conduction bands gv(E) and ge(E) for which
energy is conserved. If the energies are measured from the band extrema
we have
gv(Ep) ...., Et ge(En)"'" E: En+Ep = hw-Eg
We obtain then from equation (5.20)

w 2 e2(w) ...., (hw-Eg)2 (5.21)

in accord with observation.

The suggested approach allows a rough estimate of Be because IPvel and
gv, ge are crystalline quantities and are known in Ge. We assume that the
edge observed in amorphous Ge corresponds to r 2 ,-+r25 , direct transition
in the crystal. We calculate the densities of states with the effective masses
mp = 0.3 m and mn = 0.04 m, and we take Pve = 7 x 10- 20 g-cmJsec PO].
The slope of the straight line in Fig. 6 is dhw.Je2Jdhw = 6.5. With these
values we obtain from equation (5.20) Be ...., 10 21 cm- 3 ...., 2 X 10- 4 B.
As Be ~ Bv, Be gives the reciprocal volume of k vectors from which
linear combinations of the conduction band wave functions in amorphous
Ge are formed. Using the uncertainty relation, we find that these combina-
tions describe wave functions localized in real space in a volume
Ve = (2n)3JBe ...., 2 x 10- 19 cm 3

which contains several thousand atoms.

It is seen that the conduction band wave functions are much more
localized than the valence band functions; in linear dimensions, the latter
extend over at least hundreds of atoms, the former only over a (ew tens of
atoms.
134 J. Tauc

VI. MAIN ABSORPTION BAND

At photon energies larger than 1.8 eV we could measure only the

reflectivity. In Fig. 7 the reflectivity of amorphous Ge is compared with that
of single-crystal Ge and of polycrystalline layers obtained from amorphous
layers by annealing at 450°C. We see that the sharp structure observed in
crystalline material is absent in amorphous Ge. This is in accord with the
explanation of sharp structure as due to the long-range order. The valence
band plasma region (above 8 eV) seems to be the same as in the crystal.
We attempted to calculate the optical constants from this curve by using
the dispersion relations [11 J, and obtained B I (w), B2 (w) curves published
elsewhere [3J. However, it was realized later that the integral

f we2 (w) dw

0.65

0.60

0.'0

0 ,35

0,25

Fig. 7. Reflectivity of amorphous

Ge (curve 3) compared with the
0 .200~.I..-..J..
2----'-'-.l...-.J..
6 ----'----'-8--''-,LO---L.-.J'2 reflectivity of single crystals (curve I)

"W (eV) and polycrystaIIine layers (curve 2)

(after [3]).
Optical Properties and Electronic Structure of Amorphous Semiconductors 135

over the whole spectrum is roughly one-half that for the crystal although one
expects it should be the same. This is due to the absence in amorphous Ge
of the peak near 4.5 eV which is dominant in crystalline Ge. Before entering
into an analysis of the spectrum in the fundamental absorption region it
should be made clear that this strange feature of the spectrum of amorphous
Ge is not due to some experimental error. It is interesting to note that, in a
recent paper of Zeppelfeld and Raether [14] on characteristic energy losses
of electrons, the peak near 4.5 eV is observed in crystalline Ge and Si, but
not in the amorphous state in accordance with preliminary optical results.

VII. CONCLUDING REMARKS

By studying the optical properties of a particularly simple amorphous

material we have obtained information about the wave functions near the
band extrema. It is hoped that the basic features of this simple analysis may
be applicable also to other amorphous materials; however, results reported
by Stuke on amorphous Se and Te [12] and preliminary results of
Grigorovici on amorphous Si [13] indicate that the situation is probably
more complicated. An essential point in our analysis is the assumption that
the perturbation caused by the transition to the amorphous state is small;
this may not always be true.
There are also many problems connected with the electronic structure
which we have not mentioned here. One is the presence of localized states
inside the gap. We can only state that our spectra appear not to reveal any
presence of such states. Absorption at these states (perhaps because of their
small concentration) may be very small and overlapped by other absorptions.
Another interesting problem is the change of the gap during the transition
from the crystalline to the amorphous state; Keller and Stuke [12] discussed
some features of this problem but work on a larger variety of amorphous
semiconductors is needed before the situation can be properly understood.

Vill. REFERENCES

1. H. Richter and G. Breitling, Z. Naturforsch. 13a:988 (1958).

2. R. Grigorovici, N. Croitoru, A. Devenyi, and E. Teieman, Proc. Seventh Coni
Phys. Semicond., Paris, 1964, Dunod (Paris), 1964, p. 423.
3. J. Tauc, A. Abraham, L. Pajasova, R. Grigorovici, and A. Vancu, Proc. Intern.
Coni Phys. of Noncrystalline Solids, Delft, 1964, North-Holland Publishing Com-
pany (Amsterdam), 1965, p. 606.
4. J. Tauc, R. Grigorovici, and A. Vancu, Phys. Statlls Solidi 15: 627 (1966).
5. J. Tauc, R. Grigorovici, and A. Vancu, J. Phys. Soc. Japan (Suppl.J 21: 123 (1966).
6. F. R. Kessler, Phys. Status Solidi 5:3 (1964); 6:3 (1964).
136 J. Tauc

7. A. I. Gubanov, Quantum Electron Theory of Amorphous Conductors, Consultants

Bureau (New York), 1965.
8. K. Moorjami and C. Feldman, Rev. Mod. Phys. 36: 1042 (1964).
9. F. Seitz, Rev. Mod. Phys. 26: 7 (1954).
10. M. Cardona, J. Phys. Chem. Solids 24: 1543 (1963).
11. J. Taue, Progr. Semicond. 9:87 (1965).
12. H. Keller and J. Stuke, Phys. Status Solidi 8:831 (1965).
13. R. Grigoroviei and A. Vaneu (to be published).
14. K. Zeppelfeld and H. Raether, Z. Physik 193:471 (1966).