PHYSICAL REVIE% B VOLUME 34, NUMBER 8 15 OCTOBER 1986

Optical bowing in zinc chalcogenide semiconductor alloys

James E. Bernard and Alex Zunger

Solar Energy Research Institute, Golden, Colorado, 80401
(Received 9 June 1986)
Alloys of zinc chalcogenides exhibit both some of the smallest (for ZnS Se|-, ) and the largest
(for ZnS„Tei-„)optical bowing observed in isovalent semiconductor systems. A theoretical
analysis of this effect, by use of self-consistent band-structure techniques for ordered 509o-50% al-
loys in the CuAu-I structure, predicts correct bo~ing parameters and chemical trends, yet sug-
gests a physical model which is altogether different from the virtual-crystal model.

The smooth variation with composition x of the lattice (AB&C virtual alloy is assumed to have the following: (i) a
parameter a(x) and the optical band gap Es(x) of iso- single type of nearest-neighbor bond with length R~n&c,
structural, isovalent semiconductor alloys' A,
long formed the basis for their extensive utilization in
B|,
C has (ii) the crystal structure common to AC and BC, and con-
comitantly, (iii) a linearly averaged crystal potential (and
semiconductor technologies, which often require a's and crystal screening); hence symmetries of the band structure
Es's intermediate between those of pure binary compounds and charge densities identical to those of the underlying
AC and BC. Whereas a(x) often can be estimated binaries. These assumptions constitute an enormous com-
accurately from a linear interpolation' a (x ) xa~c ~ putational simplification, for the alloy band structure can
+(1 —x )age between the lattice parameters a~c and age be treated precisely with the same computational tools
of the constituents (Vegard's rule), the optical band gapa used to describe each of its binary constituents. This sim-
of alloys show quadratic nonlinearities. 'z plicity is largely responsible for the great popularity this
approach has enjoyed over the yearss 9 within diverse
Es(x) [xagc+(I —x)age] -bx(1 —x) . computational schemes such as the empirical pseudopoten-
tial method, the dielectric two-band model, s and tight
Perhaps the most striking example of this "optical bow- binding. s
ing" effect (b &0) is offered by solid solutions of zinc chal- Recent experimental'o '2 and theoretical'3 studies have
cogenides, 2 5 exhibiting (Table I) both one of the smallest
bo~ing parameters (b ~0 -0.4 eV for ZnS Sei
~
the largest (b 3 eV ins ZnS„Tet-„)
„)
and
ever to be observed
been critical of the VCA, demonstrating that assumptions
(i)-(iii) above are inappropriate to an accurate model of
semiconductor alloys. These studies have found (i) bi-
in isovalent pseudobinary semiconductor alloys. modal (RqcsaRnc) bond-length distributions, o'3 (ii) low-
Current understanding of the origin of the experimental er symmetry (partially ordered) alloy space groups with
(expt) optical bowing parameters rests on a separation distinct A and 8 sublattices, '3's and (iii) the identification
of b, „atinto a contribution bt present in an ideal hypotheti- —
of distinct A C and 8— C features in the alloy
cal alloy (either perfectly substitutionally random or per- photoemission, "(') reflectivity, "(b) and nuclear magnetic
fectly ordered), modeled within the virtual-crystal approx- resonance chemical shift spectra. 'z In addition, the wide
imation (VCA), and a contribution btt due to alloy range of degrees of successs 9 enjoyed by VCA-based ap-
disorder. (') 9 In VCA models of an 2~8~ —, C alloy the proaches to the optical bowing problem (depending on the
atoms A and 8 are replaced by some average atom (AB) method used to interpolate the band parameters of the al-
whose properties are modeled as a linear average of those loy from those of the constituents) reduces one's confi-
of A and 8. Hence, unlike a true ternary system, the dence in the UCA itself. In cases where the VCA contri-
bution btvc" fell substantially short of the experimental

TABLE I. Observed (Refs. 2-5) and calculated bowing pa-

value b, „~„ the remainder was ascribed to disorder ef-
fects. 5s However, such large disorder contributions ap-
rameters (in eV) of the fundamental direct band gap in zinc pear to be inconsistent with experimental evidence in the
chalcogenide alloys. The observed lattice parameters a 5.409,
5.668, and 6.089 A were used for ZnS, ZnSe, and ZnTe, respec-
tively. The calculated u~ are 0.236, 0.229, and 0.215 for
case of size-mismatched alloys, which show remarkably
sharp Raman lines, sharp reflectivity spectra,
electron mobilities,
"
etc. , and with coherent potential
high

Zn2SSe, Zn2SeTe, and Zn2STe, respectively. calculations, "(') which show little additional bowing (bn)
due to disorder.
Alloy b vcA bcxpt bf'" bs bca For these reasons, we approach the optical bowing prob-
lem from an altogether different viewpoint, using a
ZnSSe 0. 14' 043b 0 39 045 003 009
ZnSe Te —0.04' 1.23' 1.32 —0.36 nonempirical, self-consistent approach free of the VCA
1.96 1.00
ZnSTe 0.28' —3.0" 3.83 2.68 1.68 —0.53 and motivated as follows. Alloys can be described as col-
lections of local atomic arrangements (clusters), each of
'Reference 6. 'Reference 4. which occurs with a statistical weight appropriate to the
bReference 3. dReferences 2 and 5. composition. ' '5 These same local arrangements occur in

34 5992 1986 The American Physical Society

 OPTICAL BONING IN ZINC CHALCOGENIDE. ..

nearest-neighbor clusters of four A and 8 atoms lying at formation potentials of AC and BC, we calculate it directly
the vertices of a tetrahedron, are Aq, 238, A28z, 883, and from the volume-dependent band structures. Second,
84 which occur also in the ordered phases AC, A38C4, bring together n units of AC and 4 —n units of BC, both
ABC2 A83C4 and BC, respectively. Since zinc chal- with lattice constant a(x), to form the compound alloy
cogenide alloys exhibit the face centered cubic (fcc) struc- A„84 „C4,without relaxing the A C and 8 C bond — —
ture for all compositions, a superposition of these five lengths. The charge exchange (CE) between the units
structures A„84—„C4for O~n ~4 (all fcc sublattices) forming the alloy contributes (again for n 2)
correctly describes the Bravais lattice of the alloy. ' We
consider equilibrium structures by allowing the common 2[eye(a )+ ettc(a )] —4egec(a, u '
—, ) . (3)
atom C (in our case Zn) to relax to its minimum energy
position (generally producing nonequal A — —
C and 8 C Third, allow the A — —
C and 8 C bond lengths to relax to
their equilibrium (eq) total energy minimizing value
bond lengths'0'3) for each structure n at its appropriate
(hence, u~W 4 ). This structural (S ) contribution is
equilibrium lattice constant at"1. We Hmit further the
number of structures to be considered to AC, ABCz, and bs -4e~ac(a, u - '
—, ) —4e„,
c(a, u~) . (4)
BC by noting that (i) the bowing b does not depend on
composition x to a very good approximation [Eq. (1)], All terms in Eqs. (2)-(4) are calculated separately via
hence one can select any value of x, (ii) choosing x self-consistent band-structure calculations for the ap-
the ABC2 structure occurs with overwhelmingly larger propriate structues. The sum bt bvD+bcE+bs of Eqs.
concentration than A38C4 and ABsC4 at this composi- (2)-(4) produces just the total bowing of Eq. (1) for
tion. '3 Hence we neglect any shift in the band gap contri- x &. Although we could have calculated this sum
buted by minority species at x —, '
(i.e., disorder effects). directly as bt 2e~c(a~c) + 2ettc(attc) 4e~ac(a, uo~)
Among the benefits of this approach are the following: with just three band-structure calculations per alloy, we
(i) the calculation can be done by a first-principles method believe that the three-step process outlined above serves to
(unlike VCA which requires a potential averaging which is clarify the physical origins of bt in terms of (i) a hydro-
not amenable to a first-principles description), (ii) the static pressure effect (bvD), (ii) a charge-transfer effect
unique identities of the A and 8 atoms are retained (hence (bcE), and (iii) structural bond-length and bond-angle
A -8 charge transfer is permitted), and (iii) distinct bond relaxation (bs ).
lengths R~c and Rec are permitted. The reasonable Equations (2)-(4) require the knowledge of the lattice
agreement obtained with experiment (Table I) implies the constants a~c and attc of the endpoint compounds (taken
relative unimportance of disorder effects for these size- here from experiment, Table I), as well as a(x) (from
mismatched alloys, contrary to previous conclusions, sug- Vegard's rule) and u~ for the alloy at x —, For the '.
gests a physical origin of bowing effects completely dif- latter, we follow the method of Martins and Zunger, '7 who
ferent from that implied by the VCA, and indicates a cru- obtained alloy bond lengths in excellent agreement with
cial experiment which can distinguish clearly the validity experiment by minimizing the bond-bending and bond-
of the two approaches. stretching deformation energy, using Keating's valence
We use the first-principles self-consistent all-electron force field. 's Qur results are summarized in Table I.
mixed-basis band-structure method's within the local- A few observations are apparent from the results. First,
density formalism. Convergence tests's have assured a the calculated total bowing bf"' reproduces well the ob-
precision of-0. 1 eV or better for band energies. We have served values and their trends (the observed value of
used a simple tetragonal ABCz structure derived from that ZnS„Te~ „
is the least certain since this sample was
of the CuAu-I alloy'3'4 by the inclusion of an additional metastably grown inside the miscibility gap). This leaves
simple tetragonal sublattice occupied by a third type of smaller discrepancies b, „~t bP" t— o be accounted for by
atom (Zn here). It is characterized by the space group disorder contributions btt relative to those suggested by the
P4m2, a tetragonal rt c/a ratio (assumed unity, in accor- VCAs (compare b, „~t bt" in T—"able I). Second, the
dance with the observedz s cubic lattice parameters of the structural contribution bs, neglected altogether in the
alloys), and an internal displacement parameter u measur- VCA, 6 9 dominates bP" This i.s so because the band
ing a possible difference between the two bond lengths structure is extremely sensitive to bond deformations'
Rgc-(rt u + g )' a and Rec~[rt (u —— + — a If ') ']' (measured by u ——, '
), where u~ values are given in the
',
~

u —, then R~c Rec, as in VCA. This structure has caption of Table I), and since in turn, substantial bond de-
been observed previously' and used theoretically'3 for formations are required to relax the elastic strain energy'
III-V alloys. imposed by packing of binary components ~ith a large lat-
We follow the evolution of the bands of the ordered al- tice mismatch ha a~c —aec ) (hence, bs ~ha ).
~

loy model A„84 „C4from those of its constituents AC and Third, the VD contributions are negative, since the in-
8C in three steps. ' First, dilate the AC lattice and crease in eec due to compression of aec exceeds the de-
compress the BC lattice to a common lattice constant crease in e~c due to dilation of a~c. Fourth, the CE con-
a a(x) appropriate to the aHoy. This volume deforma- tributions are positive, and as can be expected, increase
tion (VD) contributes (for the 50%-50% alloy)

b» -2[e~c(a~c) e~c(a )]+2[ettc(—

aec) eec(a )];—
(2)
monotonically
bX (X~ —
~ith the electronegativity difference
Ztt ( (0. 1, 0.3, and 0.4, on Pauling's scale, for
S-Se, Se-Te, and Te-S, respectively). The significance of
the dual scaling with both ha and dX is exemplified by the
while it could be calculated from the known band gap de- fact that Ga AI~, As has a small bowing t'~ despite hav-
ing 4X 0. 1, just as does ZnS, Se1
bowing b ~0.
which has a larger
5 eV; the difference is traceable to the form-
„, tinent to the binary compounds are symmetric about the
vertical plane containing the Zn site (dashed vertical lines
er alloy's having ha ~0,
while the latter has a substantial in Fig. 1). Were the VCA description accurate, the con-
mismatch, ha 0.26 A, giving it a significant contribution tours for the ABC' alloy (center column of Fig. 1) would
from bq. be symmetric too. This is obviously not the case: The VB
The departure from the VCA in these systems is evident min, I I„(1 I„),
of STeZnq [Fig. 1(b)] has a strong sulfur
most clearly in the calculated band-by-band charge densi- s-like character with little amplitude on Te; the next band,
ties depicted in Fig. 1. The two columns at the extreme I ~. folded in from zinc blende X~„is complementary in
left and extreme right depict, for ZnS and ZnTe, respec- having most of its amplitude around Te [Fig. 1(e)]. Hence
tively, the calculated densities in the (110) plane, includ- we predict a splitting of 2.6 eV between these two lowest
ing those for the valence-band minimum (VB min) I ~„ valence bands at I to be obseved in photoemission (zero in
[Figs. 1(a) and 1(c)], the XI, band [Figs. 1(d) and 1(f)l, VCA). A similar effect has been observed in photoemis-
the valence-band maximum (VB max) I I5„[Figs. 1(g) sion in the valence band of Hg, Cd1- Te alloys. In "'
and 1(i)], and the conduction-band minimum (CB min) addition to the splitting due to folding, some states which
I"~, [Figs. 1(j) and 1(1)]. As expected, the contours per- would be degenerate due to folding alone (e.g. , at M, de-

zn s '(]) (k) g'. . ~ -, ','.'X 0-.7//, znTO

fog~ C 1C
(
' .ia'

zns (I) zn Te (1)

r$5y ~15v
A $
»0=. „.
= -la.W&jj, ] '
"~ 0.I &i 4j*Q -= =-IN~i'&I

~a HO ";0 ~i .&". ~o
""

IP
.,

f
I
I
I
I
I
)O-&

1 i
zns (&) 'STezns
,
'(I) z Te (f)
X1v - ',
i, '.
Rv(XIv) Xqy
,

~, i

(s I
! I

8 Te2ng
()
0"3 -' ' ~/
10.v ''' - '10 «2)' t1» {Rv}
--, ,=.-. ~A~+
—. .W Cw

/-'~L
FIG. 1. Electronic charge&ensity contours (logarithmically spaced, in units of e/a, u. } of a few electronic states in ZnS (left),
ZnTe (right), and their STeZnq alloy in the CuAu-I type structure (center). For the latter, we show both the band labels in P4rn2
symmetry and, in parentheses, the zinc-blende labels. Solid circles around atoms indicate core regions; straight solid lines denote
bonds. The vertical dashed line is the symmetry plane for pure ZnS or Zu Te; this symmetry is broken in the alloy (central panels).
 OPTICAL BO%ING IN ZINC CHALCOGENIDE. .. 5995

completely ordered structures. For example, the possible character is evident in our alloy calculation [Figs. 1(h) and
rived from two zinc-blende X points, and at 8, derived l(k)], this being precisely what is required to explain the
from two zinc-blende L points) split in the alloy due to observed bowingz' of the spin-orbit splitting of the VB
non-zinc-blende components of the potential. This split- max. The significant departures from VCA noted in this
ting differs qualitatively in the case of mixed anions (e.g., and other work suggest a reexamination of applications of
ZnzSTe) from that in the case of mixed cations [e.g. , VCA to semiconductor alloys.
CdMnTez (Ref. 20) or GaA1Asz]. For example, the Note added in proof. Recently, J. C. Mikkelson and
lowest two valence states at M, being anion states, are split J. B. Boyce [paper presented at the Seventh International
in the materials we consider here (with a splitting of 2. 7 Conference on Ternary and Multinary Compounds,
eV in ZnzSTe), whereas they are degenerate~ in Snowmass, Colorado, Sept. I986, edited by S. K. Deb and
CdMnTez. The lowest two valence bands and lowest two A. Zunger [Mater. Res. Soc. Symp. Proc. (to be pub-
conduction bands at R (all Ri) are split by 2.5 and 1.6 eV, —
lished)]] have measured the Zn Se and Ze Te bond —
respectively, in ZnzSTe. Similarly there are states which lengths in ZnSe„Tei „alloys using the extended x-ray ab-
split in structures with mixed cations but are degenerate in sorption fine structure (EXAFS) method, finding at
the case of mixed anions. All such splittings must vanish x 0.5 Rz, a, 2.472 A and Rz, s, 2.617 A, in excellent
in zinc blende (hence in VCA), which has neither mixed agreement with the values Rz„s, 2.472 A, and Rz, y,
anions nor mixed cations, so photoemission measurements 2.616 A, calculated here from minimization of the elas-
of these splittings would provide a critical test of the VCA tic energies (corresponding to u~ 0.229, caption to Table
versus the present physical model of bowing. Further- I).
more, in assuming the alloy to share with its constitutents
the same space group symmetry, the VCA would predict This work was supported by Office of Energy
the VB max, I is„,to be a pure p+d state [as in Figs. 1(g) Research-Basic Energy Sciences through Materials

I i„
and 1(i) for ZnS and ZnTe, respectively), and the CB min,
to be a pure s state [as in Figs. 1(j) and 1(1) for ZnS
and ZnTe, respectively]. However, substantial intermix-
Research Division, Grant No. DE-AC02-77-CH00178.
One of us (J.E.B.) wishes to thank J. Jaffe for consulta-
tions regarding the use of the "potential-variation-mixed-
ing of valence- (I4„+I5„) and conduction- (I i, ) band basis" band structure program.

'J. C. W'ooley, in Compound Semiconductors, edited by R. K. 'OJ. C, Mikkelsen and J. B. Boyce, Phys. Rev. Lett. 49, 1412
Willardson and H. L. Goering (Reinhold, New York, 1962), (1982).
p. 3; A. N. Pikhtin, Fiz. Tekh. Poluprovodn. 11, 425 (1971) ii(a) W. E. Spicer, J. A. Silberman, J. Morgen, 1. Lindau, J. A.
[Sov. Phys. Semicond. 11, 245 (1977)l. Wilson, A. 8. Chen, and A. Sher, Phys. Rev. Lett. 49, 948
2S. Larach, R. E. Shrader, and C. F. Stocker, Phys. Rev. I.OS, (1982); (b) K. E. Kirschfeld, N. Nelkowski, and T. S.
587 (1957). Wagner, ibid 29, 66 (1.972).
3A. A. El-Shazly, M. M. H. El-Naby, M. A. Kenny, M. M. 2A. Willig, 8. Sapoval, K. Leibler, and C. Verie, J. Phys. C 9,
El-Nahass, H. T. El-Shair, and A. M. Ebrahim, Appl. Phys. A 1981 (1976).
36, 51 (1985); R. Mach, P. Fl gel, L. G. Suslina, A. G. Aresh- G. P. Srivastava, J. L. Martins, and A. Zunger, Phys. Rev. 8
kin, J. Maege, and G. Voigt, Phys. Status Solidi 8 109, 607 31, 2561 (1985); A. Zunger and J. Jaffe, Phys. Rev. Lett. 51,
(1982). 662 (1983); J. L. Martins and Z. Zunger, ibid 56, 1400.
4A. Ebina, M. Yamamoto, and T. Takahashi, Phys. Rev. 8 6, (19&6).
3786 (1972). '4T. S. Kuan, T. F. Kuech, W. I. Wang, and E. L. Wilkie, Phys.
sR. Hill and D. Richardson, J. Phys. C 6, L115 (1973). Rev. Lett. 54, 201 (1985); H. R. Jen, M. J. Cherng, and G. 8.
6J. A. Van Vechten and T. K. Bergstresser, Phys. Rev. 8 1, 3351 Stringfellow, Appl. Phys. Lett. 4$, 1603 (1986).
(1970). '50. de Fontaine, in Solid State Physics, edited by H. Ehren-
7(a) A. Baldereschi, E. Hess, K. Maschke, H. Neumann, K. R. reich, F. Seitz, and D. Turnbull (Academic, New York, 1979),
Schulze, and K. Unger, J. Phys. C 10, 4709 (1977); (b) Vol. 34, p. 73.
R. HBI and D. Richardson, ibid 4, L289 (197.1); 4, L339 'sP. Bendt and A. Zunger, Phys. Rev. 8 26, 3114 (1982).
(1971). i7J. L. Martius and A. Zunger, Phys. Rev. 8 30, 6217 (1984).
S(a) K. E. Newman and J. D. Dow, Phys. Rev. 8 2'7, 7495 'sP. N. Keating, Phys. Rev. 145, 637 (1966).
(1983); (b) R. E. Allen, T. J. Humphreys, J. D. Dow, and '9J. Jaffe and A. Zunger, Phys. Rev. 8 29, 1882 (1984).
O. F. Sankey, J. Vac. Sci. Technol. 8 2, 449 (1984). zoS. H. Wei and A. Zunger, Phys. Rev. Lett. 56, 528 (1986).
9A. 8. Chen and A. Sher, Phys. Rev. 8 23, 5360 (1981); 22, z'D. J. Chadi, Phys. Rev. 8 16, 790 (1977).
3886 (1980).
View publication stats