Appl. Phys. A 74, 659–664 (2002) / Digital Object Identifier (DOI) 10.

1007/s003390100931
Applied Physics A
Materials
Science & Processing

Optical functions of chalcopyrite CuGax In1−xSe2 alloys

M.I. Alonso1∗ , M. Garriga1 , C.A. Durante Rincón2 , E. Hernández2 , M. León3
1 Institut de Ciència de Materials de Barcelona, CSIC, Campus de la Universitat Autònoma de Barcelona, 08193 Bellaterra, Spain
2 Departamento de Fı́sica, Facultad Experimental de Ciencias, Universidad del Zulia, OPT Galerı́as, Maracaibo, Venezuela
3 Departamento de Fı́sica Aplicada, C-XII, Universidad Autònoma de Madrid, 28049 Madrid, Spain

Received: 20 December 2000/Accepted: 30 May 2001/Published online: 25 July 2001 –  Springer-Verlag 2001

Abstract. We report pseudodielectric functions ε of the terial are obtained, and the features observed in the dielectric
quaternary semiconductor alloys CuGax In1−x Se2 . Measure- function are related to the electronic band structures of these
ments were done in polycrystalline samples from 0.7 to chalcopyrites.
5.2 eV at room temperature by spectral ellipsometry. Accu-
rate values of refractive indices n and extinction coefficients
k representative of the bulk materials are obtained from the 1 Experimental details
data. The structures observed in ε are analyzed by fitting
the numerically differentiated spectra of d2 ε(ω)/dω2 to an- The samples used in this study were cut from polycrystalline
alytic line shapes. The obtained energies are related to given ingots of well-characterized alloys. These ingots were pre-
inter-band transitions by taking into account the electronic pared by direct fusion of the constituent elements in stoi-
band structures of the ternary end-point compounds. chiometric proportions. The compositions of the alloys were
checked by atomic emission spectroscopy to be close to
PACS: [Link]; [Link]; [Link] those of the starting mixtures, and a full characterization
by X-ray diffraction was done. More detailed results are re-
ported elsewhere [9]. The alloys showed the chalcopyrite-
Polycrystalline thin films of chalcopyrite CuGa x In1−x Se2 type structure in the whole composition range with no sec-
alloys are currently used as an absorber layer for highly ondary phases present in the diffractograms. The diffraction
efficient photovoltaic energy generation in terrestrial applica- peaks were rather sharp for all samples except for the sample
tions [1, 2]. A key factor that determines the final efficiency with x = 0.6, which displayed broader peaks.
of those devices are the optical properties of the absorber ma- The optical data were obtained at room temperature in
terial [3]. In spite of the importance of this issue and of the the range from 0.7 to 5.2 eV with a rotating polarizer ellip-
considerable amount of research devoted to these quaternary someter. The light source was a 75-W high-pressure Xe arc
alloys, knowledge of their optical functions is rather incom- lamp. As detectors we used a multi-alkali photomultiplier
plete. In fact, reliable values of the optical functions of the tube for the visible–UV range, and a GaInAs photodiode be-
ternary end compounds CuGaSe2 (CGS) and CuInSe2 (CIS) low 1.4 eV. The angle of incidence was ϕ = 70◦ .
were not well established until recently [4, 5]. In order to obtain dielectric function values representative
In previous optical studies of these alloys, attention has of the bulk materials it is necessary to prepare and main-
been limited to the fundamental absorption edge. The energy tain optically abrupt surfaces stable throughout the meas-
of the lowest band gap as a function of composition has been urements [10]. Previous work on single crystals of CIS and
measured by a number of groups by optical absorption or CGS [4, 5] has shown that the best spectra were obtained
photoluminescence (see [6] and references therein). Depen- in samples that were chemomechanically polished using an
dence of band gap on temperature [7] and pressure [8] has alkaline colloidal silica suspension. Therefore, we have fol-
also been investigated. As far as we know, no data of opti- lowed this procedure immediately before ellipsometric meas-
cal functions or electronic transitions above the band gap are urements, and kept the samples under dry N2 flux while col-
available in the literature. In this work, we have measured el- lecting data.
lipsometric spectra of these materials at room temperature, in
the energy range 0.7 to 5.2 eV. Accurate values of refractive
indices n and extinction coefficients k of polycrystalline ma- 2 Optical functions

∗ Correspondingauthor. The complex refractive index spectra calculated from our data
(Fax: +34-93/5805729, E-mail: [Link]@[Link]) using an isotropic two-phase model [11] are plotted in Fig. 1.
660

Table 1. Real and imaginary parts of the complex refractive index n + ik of

three CuGax In1−x Se2 alloys. Values of k that were lower than 0.1 in our
measurements are considered inaccurate and are left blank

x
E 0.2 0.4 0.75
(eV) n k n k n k

0.7 2.92 ... 2.88 ... 2.84 ...

0.8 2.93 ... 2.89 ... 2.85 ...
0.9 2.96 ... 2.91 ... 2.86 ...
1.0 3.02 ... 2.95 ... 2.87 ...
1.1 3.10 ... 3.00 ... 2.89 ...
1.2 3.08 0.23 3.09 0.18 2.92 ...
1.3 3.07 0.27 3.05 0.27 2.97 ...
1.4 3.04 0.34 3.03 0.31 3.04 ...
1.5 3.03 0.37 3.02 0.37 3.10 0.18
1.6 3.01 0.40 3.00 0.40 3.07 0.26
1.7 3.00 0.42 2.99 0.41 3.06 0.31
1.8 2.99 0.43 2.99 0.43 3.05 0.35
1.9 3.00 0.44 3.00 0.45 3.04 0.37
2.0 3.01 0.46 3.02 0.46 3.05 0.38
2.1 3.03 0.49 3.03 0.49 3.07 0.41
2.2 3.06 0.53 3.06 0.53 3.10 0.44
2.3 3.08 0.58 3.09 0.57 3.13 0.49
2.4 3.10 0.63 3.11 0.64 3.15 0.54
2.5 3.11 0.71 3.14 0.72 3.18 0.61
Fig. 1a,b. Real (a) and imaginary (b) parts of the complex refractive index 2.6 3.11 0.79 3.15 0.82 3.20 0.69
n + ik obtained from our measurements. The spectra that correspond to 2.7 3.10 0.88 3.15 0.93 3.21 0.78
the ternary end compounds are plotted with thicker lines 2.8 3.06 0.98 3.09 1.07 3.20 0.89
2.9 2.98 1.08 2.96 1.18 3.16 1.02
3.0 2.87 1.13 2.81 1.20 3.06 1.13
This model assumes a mathematically sharp sample–ambient 3.1 2.77 1.12 2.70 1.16 2.94 1.18
3.2 2.71 1.10 2.65 1.10 2.83 1.19
interface. To recognize the inevitable presence of a surface- 3.3 2.66 1.08 2.64 1.06 2.76 1.16
transition region, the obtained optical functions are termed 3.4 2.65 1.05 2.66 1.05 2.72 1.13
‘pseudodielectric’ and written in brackets [10]. For complete- 3.5 2.64 1.05 2.68 1.06 2.70 1.11
ness, numerical values of three alloy samples are listed in 3.6 2.63 1.05 2.68 1.09 2.70 1.11
Table 1. Values for x = 0 and x = 1 can be found in [5, 12]. In 3.7 2.60 1.05 2.68 1.13 2.70 1.12
3.8 2.59 1.03 2.65 1.16 2.70 1.15
the spectral region of low absorption, accuracy of k measure- 3.9 2.60 1.02 2.64 1.18 2.68 1.17
ments with the ellipsometer used (without a compensator) is
4.0 2.63 1.02 2.64 1.22 2.67 1.19
known to be rather poor: we consider that measured values of 4.1 2.66 1.03 2.63 1.26 2.66 1.21
k < 0.1 are too imprecise and are not listed. Since chalcopy- 4.2 2.71 1.07 2.61 1.32 2.65 1.24
rites are uniaxial crystals, the measured n and k represent 4.3 2.74 1.13 2.57 1.37 2.65 1.27
effective values with contributions of the ordinary and ex- 4.4 2.77 1.22 2.53 1.41 2.65 1.32
4.5 2.76 1.31 2.48 1.45 2.64 1.36
traordinary components of the dielectric tensor [11]. If the 4.6 2.74 1.41 2.43 1.49 2.62 1.43
sample is formed by randomly oriented crystallites of sizes 4.7 2.67 1.51 2.38 1.51 2.58 1.49
typically smaller than the wavelengths of measurement, the 4.8 2.58 1.58 2.34 1.54 2.53 1.55
effective optical response is isotropic and close to the ordi- 4.9 2.50 1.63 2.28 1.55 2.47 1.61
nary component [12]. Except for the CGS specimen, all other 5.0 2.41 1.64 2.23 1.58 2.39 1.64
samples studied here displayed an isotropic behavior. For the 5.1 2.34 1.65 2.19 1.57 2.32 1.66
CGS sample a weak anisotropy was present, which indicates 5.2 2.29 1.66 2.18 1.56 2.24 1.62
either larger crystallites and/or some preferential orientation
of the grains. Accordingly, if we compare the spectra of the
end compositions shown in Fig. 1 to the known dielectric and the absorption coefficient:
tensor components [4, 5], the CIS spectrum resembles the or-
4πk
dinary part whereas that of CGS contains more extraordinary α = , (2)
contribution. However, the absolute values of both compo- λ
nents are not very different; in particular ∆n, ∆k < 1% below where λ denotes the wavelength of light in vacuum. The in-
∼ 3 eV and therefore the obtained spectra are useful for most verse of the absorption coefficient α−1 gives the so-called
applications. From n and k, optical magnitudes of interest penetration depth, indicative of the thickness of material
are calculated easily. The normal-incidence reflectivity R is which is relevant in optical processes, i.e. absorption, trans-
given by: mission, reflection, and light scattering.
As in other alloy systems [12, 13] the dependence of
the refractive index on composition below the band gap is
(n − 1)2 + k2
R = , (1) a smooth curve. This property can be used as an optical
(n + 1)2 + k2 method to evaluate the composition of these alloys. However,
 661

precision is better using R because the reflectivity is less af- (∆cf ) and the spin–orbit (∆so ) parameters, using the quasicu-
fected by thin surface overlayers than n [13, 14]. A suitable bic model [17]. The polarization selection rules can be under-
working energy for these alloys is that of the near-infrared stood in relationship to their well-known binary analogues,
He–Ne laser (λ = 1.523 µm). The values R(x) calculated as described in [18] and depicted in Fig. 3 for our particu-
from our data are well fitted by R = 0.2510 − 0.0422x + lar compounds. In short, starting from a threefold-degenerate
0.0167x 2. Only the x = 0.6 sample deviates from this behav- Γ15 state in zinc blende, the introduction of a tetragonal crys-
ior, possibly due to its lower crystalline quality. tal field gives rise to one Γ4 plus a twofold-degenerate Γ5
level. The possible dipolar inter-band transitions to the Γ1
conduction-band minimum are Γ4 → Γ1 , allowed in E
c,
3 Inter-band critical points and Γ5 → Γ1 , allowed in E ⊥ c. Spin–orbit interaction further
splits the Γ5 levels and the selection rules are somewhat re-
The dielectric function of a semiconductor is closely linked laxed, as given in Fig. 3. Although only one of the transitions
to its electronic band structure. The features observed in ε(ω) (Γ6 → Γ6 in the double group) has a pure E ⊥ c polariza-
in the optical range are related to inter-band transitions orig- tion and the other two are mixed, it is approximately true that
inated by large or singular values of the joint valence and each transition keeps its original polarization, i.e. before in-
conduction density of states. These points of the band struc- troducing the spin–orbit interaction. Also, even if all three
ture are also known as van Hove singularities or critical transitions result from two simultaneous interactions, in the
points [15]. Several critical points are observed in the spectra present compounds where ∆cf < ∆so , it is useful to make
of Fig. 1. They can be better resolved in their numerical sec- a simplification. In this simplified view, E 0 (A) and E 0 (B) can
ond derivatives, as shown in Fig. 2 for a representative alloy be regarded as a crystal-field-split gap, where each of these
sample. We calculated derivatives using smoothing polyno- transitions has a different polarization as determined by the
mials and obtained critical-point energies from fitting to stan- sign of ∆cf . The most usual situation in chalcopyrite com-
dard analytic line shapes [15], using two-dimensional line pounds is that ∆cf < 0 and then E 0 (A) is allowed in E
c
shapes for all observed structures. The fundamental band- and E 0 (B) in E ⊥ c. This is the case of CGS. However, in
gap transitions were fitted only to the spectra of the real part CIS ∆cf > 0 and the situation is reversed: E 0 (A) is allowed
to avoid possible errors in the region of low absorption. For in E ⊥ c and E 0 (B) in E
c. In this simplified schema, the
the rest of structures we considered both real and imaginary higher energy E 0 (C) transition corresponds to the spin–orbit-
measurements. split gap.
Except in the CGS sample where the three peaks are re-
solved, in the rest of the polycrystalline samples used in this
3.1 Fundamental absorption edge work we only observe two components. In principle, the first
structure must be composed by the two closer peaks E 0 (A, B)
The structure of the fundamental absorption edge of chal- and the second one must be ascribed to E 0 (C) as indicated in
copyrites is well understood [16]. These crystals are semicon- Fig. 2. In polycrystals, the ordinary polarization E ⊥ c is ex-
ductors with a direct fundamental gap at the Brillouin-zone pected to dominate the spectra [12]. Thus, the component ( A
center Γ . The threefold valence-band maximum is composed or B) with E ⊥ c will give the main contribution to the first
of three nondegenerate states, giving rise to three transitions observed peak. As sketched in Fig. 3 this polarization corres-
called E 0 (A), E 0 (B), and E 0 (C) from smaller to larger en- ponds to E 0 (A) in a CIS-like and E 0 (B) in a CGS-like band
ergy. Experimental measurements of these transition energies arrangement. Notice that in CIS, with a small distortion from
and of their selection rules allow us to determine the symme- cubic symmetry, the A and B peaks are so close that they
try of the three valence states and to calculate the crystal-field are never resolved at room temperature. The alloy composi-
tions studied here are all expected to have CGS-like gaps, that
is, the first structure seen in our ellipsometric spectra should

E0(A,B) Cu Ga 0.4 In 0.6 Se 2

×3 CIS-like zinc blende CGS-like
100
E0(C) E1(B)
E '(ΓX) Γ6 Γ1 Γ1 Γ1 Γ6
d 2 ε / d ω2 (eV − 2 )

Γ6 Γ7
E(ΓX) E2(A) Γ4
E(XΓ) E(∆X)
Γ7 Γ5
-100 E2(B)
Γ6
Γ4 Γ15
E1(A) real Γ7 Γ5
imag.
Γ7
-200
1 2 3 4 5 ∆so ∆cf > 0 ∆cf < 0 ∆so
energy (eV)
Fig. 3. Schematic evolution of the energy states at the band gap of chalcopy-
Fig. 2. Second-derivative spectra of a representative sample (x = 0.4). Ex- rite starting from a zinc blende-like gap without spin–orbit coupling. Solid
perimental points are plotted using symbols and our best fits are given by (dashed) arrows represent transitions allowed by symmetry in E ⊥ c (E
c)
lines. The arrows mark the fitted critical-point energies, which are labeled polarization, respectively. Note that the ordering of the topmost valence
according to our assignments bands is different in CIS and CGS
662

2 inter-band transitions show a smooth evolution with compo-

sition. All fitted energies are collected in Fig. 5.
C Based on the band-structure calculations of Jaffe and
gap energy (eV)

Zunger [19] for the ternary end compounds and relating the
B observed transitions to those of binary zinc blende analogues,
1.5
the main transitions that contribute to ε(ω) have been as-
A signed to critical points at the Brillouin-zone center Γ and
edge N and T points. The labeling of the transitions is de-
scribed in [5] and is related to the standard zinc blende no-
1 tation. The different observed transitions and their assign-
a ments within calculated band structure are listed in Table 2. In
0.24
the range accessible to our experimental setup there are sev-
0
eral features without corresponding transitions in zinc blende
which are activated in chalcopyrite by folding of the Bril-

∆so
∆cf

0.23 louin zone and a partial d-orbital character of the electronic

states. At Γ we observe transitions from states originated
both at Γ and at X points of zinc blende, as given in Table 2.
-0.1 Concerning the E 1 transition in zinc blende, notice that the
0.22 four equivalent L points fold in pairs into the two equivalent
0 0.5 1
N points in chalcopyrite, giving rise to transitions allowed
b x in CuGa xIn1-xSe2
in both polarizations. In contrast, the E 2 transition of zinc
Fig. 4a,b. Characteristic energies at the fundamental absorption edge. blende is mapped onto different points, originating both the
a Transition energies measured in the different samples. Solid squares are transition E  (ΓX) at Γ , only allowed in perpendicular po-
obtained from transmission measurements and open circles are determined
from the fits of d2 ε/dE 2 , where fitting errors are smaller than or equal to larization, and a pair of transitions at T allowed in either
the size of the circles. b Parameters ∆cf and ∆so calculated by inserting the polarization.
fitted E 0 ( A), E 0 (B), and E 0 (C) into the equations of the quasicubic model From the assignments suggested in Table 2 several rela-
(see text) tive positions of the involved energy states can be de-
duced and compared to available calculations. We consider
the band-structure calculations of Jaffe and Zunger [19],
correspond mainly to the second gap component E 0 (B). This performed within the local-density-functional formalism,
idea is corroborated by transmission measurements done with which is known to give optical band gaps smaller than
samples from the same ingots. The observed absorption edge experiment. However, this mismatch is easily adjusted by
is always the smallest gap E 0 (A) irrespective of its polariza- a rigid upwards shift of the whole conduction band (some-
tion and, actually, by transmission we always obtain transition times called ‘scissors’ correction). All calculated num-
energies smaller than or equal to by ellipsometry. With these
assignments, we obtain the following best-fit dependences of
the three energies:
E2(B)
E 0 (A) = 1.010 + 0.626x − 0.167x(1 − x) , 5
E 0 (B) = 1.006 + 0.702x − 0.166x(1 − x) ,
E 0 (C) = 1.236 + 0.674x − 0.151x(1 − x) . (3) E2(A)
E'(ΓX)
All three energies vary quadratically with a small downward-
transition energy (eV)

E(∆X)
bowing parameter. In particular, the quadratic coefficient E1(B)
4
0.167 obtained for E 0 (A) is in the range of the most repro-
ducible numbers found in bulk CuGax In1−x Se2 alloys [6].
The obtained energies and their fits are displayed in Fig. 4a.
Using the dependences given by (3) in the framework of the E(XΓ)
quasicubic model [17], we have plotted in Fig. 4b the param- E1(A)
eters ∆cf and ∆so as a function of composition. The former 3
has an almost linear variation with x with a slight downward
bowing, whereas the latter is almost constant around a value
∆so = 0.23 eV and shows a small upward bowing. More pre- E(ΓX)
cise results of all these fundamental gap parameters should
be obtained from a detailed study of single-crystal samples,
where polarization selection rules can be experimentally well 2
0 0.2 0.4 0.6 0.8 1
established.
x in CuGa xIn1-xSe2
Fig. 5. Transition energies in CuGax In1−x Se2 alloys determined from the
3.2 Higher inter-band transitions fits of d2 ε/dE 2 . Squares denote critical-point energies ascribed to the
Γ point, circles at N, and triangles at T . Transitions at T not observed
The spectra ε(ω) of CIS and CGS above the fundamental here but known from single-crystal CIS and CGS are represented by solid
gap have a similar appearance. Accordingly, the observed triangles. Lines are drawn as a guide to the eye
 663

Table 2. Possible inter-band dipolar transitions calculated a in the inves- The transitions observed around ∼ 4 eV in two of the
tigated energy range, their polarizations and the assignment to observed alloys are tentatively identified with E(∆X), the gap at the
structures. Two possibilities for E 2 (B) are given, from smaller to larger T point, by similarity with the spectra of ternary single crys-
calculated energy, as discussed in the text. Spin–orbit interaction is not
considered tals CIS and CGS. Structures at the higher energy part of the
spectra around ∼ 5 eV correspond to T -point transitions in
k Label Polarization Equivalence single crystals. Two split transitions are observed in several
samples. The splitting observed for x = 1 is 0.2 eV, coincid-
Γ ing with that measured in single-crystal CGS. Two possible
E 0 ( A)
Γ4v → Γ1c
(1)
origins have been considered for these transitions, which are
E 0 (B, C) ⊥ Γ5v → Γ1c consistent with the observed selection rules [5]. In the first
(1)
E(ΓX) ⊥ Γ5v → Γ3c possibility, E 2 (A) and E 2 (B) would originate directly from
(2)
E(XΓ) ⊥ Γ5v → Γ1c the E 2 transition in zinc blende. The observed energy separa-
(1) tion would correspond to the splitting of the two valence-band
E  (ΓX) ⊥ Γ5v → Γ2c
E  (ΓX) ⊥ (2)
Γ5v → Γ3c states with symmetries T5v and T3v + T4v coming from X 5v
N in zinc blende. However, the calculated splitting is too large
(1) (1) (∼ 1 eV) compared with the measured 0.15–0.20 eV, because
E 1 ( A)
, ⊥ N1v → N1c
(2) (1)
the state T5v belongs to the upper manifold of antibonding
E 1 (B)
, ⊥ → N1c (2) (2)
N1v bands while the state T3v + T4v is situated down at the
T
(1) (1)
p − d bonding energy region. The second possibility, also
E(∆X)
T3v + T4v → T1c + T2c given in Table 2, is energetically more probable. In this case,
E 2 ( A) ⊥ T5v → T1c + T2c
(1) only E 2 (A) is derived from the zinc blende E 2 transition and
E 2 (B)
T5v → T5c
(2) (2) E 2 (B) has no equivalence in the binary compounds. Then, the
E 2 (B)
T3v + T4v → T1c + T2c
observed splitting must be identified with that of the first two
a [19] conduction states at the T point, calculated as 0.33 eV in CIS
and 0.03 eV in CGS, to be compared with 0.15 and 0.20 eV
respectively.
bers that are given below are taken from the results of
Jaffe and Zunger [19] after having done this correction
to reproduce our experimental values of fundamental gaps 4 Conclusion
at Γ .
First, we compare the relative positions of N and T We have reported pseudodielectric functions measured at
band extrema, and we obtain reasonable agreements be- room temperature on six CuGax In1−x Se2 polycrystalline bulk
tween experiments and theory. Both for CIS and CGS, the samples across the whole composition range from x = 0 to 1.
measured gaps at N [E 1 (A)] (CIS: 2.89 eV, CGS: 3.11 eV) We have paid special attention to the problem of minimizing
are about 0.12 eV smaller than calculated (CIS: 3.00 eV, surface effects on the measurements of n and k, so that the
CGS: 3.24 eV), whereas the gaps found at T [E(∆X)] (CIS: reported values are characteristic of the bulk.
4.07 eV, CGS: 4.49 eV) are about 0.13 eV larger than calcu- From the analysis of the numerical second derivatives of
lated (CIS: 3.92 eV, CGS: 4.38 eV). the ε(ω) spectra, we have obtained the energies of several
With our assignments, it is deduced that the energy dif- critical points as a function of composition. These have been
ference E(Γ3c − Γ1c ) is approximately constant around 1 eV assigned to certain electronic inter-band transitions by anal-
for all compositions. This value is smaller than the calculated ogy with the previous assignments in the ternary end-point
1.9 eV (CIS) and 1.5 eV (CGS). In contrast, the distance to the compounds.
next conduction-band state is larger than the calculated 0.6 eV Both the spectral dependences of the optical functions and
(CIS) and 0.5 eV (CGS), although it is only observed up to the critical-point analysis are expected to be useful in stud-
x = 0.4, giving also a constant value E(Γ2c − Γ3c ) ∼ 2 eV. Ex- ies of solar cells and other heterostructures that contain these
trapolating this to CGS, it becomes clear that the transition compounds.
E  (ΓX) is not observed because it is masked by stronger E 2
transitions in the same energy range. The last transition at Γ
listed in Table 2, E  (ΓX), is a counterpart of T -point transi- References
tions and should have an energy around E 2 (A). According to
(2) 1. R.W. Birkmire, E. Eser: Annu. Rev. Mater. Sci. 27, 625 (1997)
this, the Γ5v level is located deeper below the valence-band 2. U. Rau, H.W. Schock: Appl. Phys. A 69, 131 (1999)
maximum in CIS than in CGS. Our result gives energies of 3. J.J. Loferski: J. Appl. Phys. 27, 777 (1956)
− 2.4 eV (CIS) and − 2.0 eV (CGS), quite close to the calcu- 4. T. Kawashima, S. Adachi, H. Miyake, K. Sugiyama: J. Appl. Phys. 84,
lated −2.3 and − 2.2 eV in [19]. 5202 (1998)
5. M.I. Alonso, K. Wakita, J. Pascual, M. Garriga, N. Yamamoto: Phys.
The two prominent E 1 transitions display similar variation Rev. B 63, 075203 (2001)
with composition, E 1 (B) having a slightly larger dispersion. 6. S.-H. Wei, A. Zunger: J. Appl. Phys. 78, 3846 (1995)
The energy separation between E 1 (A) and E 1 (B) corres- 7. C. Lárez, C. Bellabarba, C. Rincón: Appl. Phys. Lett 65, 1650
ponds to the crystal-field splitting of the valence band at the (1994)
N point. It amounts to 0.75 eV in CIS and 0.95 eV in CGS, 8. I. Choi, P.Y. Yu: Phys. Status Solidi B 211, 51 (1999)
9. C.A. Durante Rincón, E. Hernández, M.I. Alonso, M. Garriga,
larger than the calculated 0.4 eV in both compounds. Unlike S.M. Wasim, C. Rincón, M. León: Mater. Chem. Phys. 70, 300
in CuAlSe2 [20], both in CIS and CGS the first transition, (2001)
E 1 (A), dominates the spectrum. 10. D.E. Aspnes, A.A. Studna: Phys. Rev. B 27, 985 (1983)
664

11. R.M.A. Azzam, N.M. Bashara: Ellipsometry and Polarized Light 16. J.L. Shay, J.H. Wernick: Ternary Chalcopyrite Semiconductors:
(North-Holland, Amsterdam 1977) Growth, Electronic Properties, and Applications (Pergamon, Oxford
12. M.I. Alonso, M. Garriga, C.A. Durante Rincón, M. León: J. Appl. 1975)
Phys. 88, 5796 (2000) 17. J.J. Hopfield: J. Phys. Chem. Solids 15, 97 (1960)
13. F. Lukeš, E. Schmidt, J. Humlíček, M.K. Kekoua, E. Khoutsichvili: 18. J.L. Shay, E. Buehler, J.H. Wernick: Phys. Rev. Lett. 24, 1301
Phys. Status Solidi A 53, 321 (1979) (1970)
14. M.K. Kelly, S. Zollner, M. Cardona: Surf. Sci. 285, 282 (1993) 19. J.E. Jaffe, A. Zunger: Phys. Rev. B 28, 5822 (1983)
15. P. Lautenschlager, M. Garriga, M. Cardona: Phys. Rev. B 36, 4813 20. M.I. Alonso, J. Pascual, M. Garriga, Y. Kikuno, N. Yamamoto,
(1987) K. Wakita: J. Appl. Phys. 88, 1923 (2000)