Optical-constants
model
for
semiconductors
and
insulators
Y.
F.
Chen andC. M.
KweiDepartment
of
Eiectronics Engineering, National Chiao Tung University, Hsinchu, Taiwan, Republicof
ChinaC.
J.
Tung*Institute
of
Nuclear Science, Nationa! Tsing Hua University, Hsinchu, Taiwan, Republicof
China (Received 5April 1993)The refractive index and the extinction coefficient, i.e., the real and imaginary parts ofthe complex in-dex ofrefraction, for semiconductors and insulators are derived as afunction ofphoton energy. These derivations apply f-sum rules and symmetry relations tocritical-point transitions from the valence band to the conduction band. Comparison with measured optical data reveals that present formulations are valid over awide range ofphoton energies immediately above the band gap tothe first ionization thresh-old for inner shells. The present work shows an improvement and extension over the theory ofForouhi and Bloomer, which applies for anarrow range ofphoton energies above the absorption edge.
I.
INTRODUCTIONOptical properties
of
solids can be characterized by quantities such as the complex indexof
refraction n( )co= i(i)co+i (tr)coand the complex dielectric functions(co) a=i(co)+isa(co) T.hese quantities are photon energy
codependent and provide information on the propagation
of
radiation and on the electronic structureof
solids.'
Optical quantities are useful in the design
of
mirrors, prisms, multilayers,etc.
The refractive index il(co) and the extinction coefficient tc(co) are commonly determined by
rejective
and absorp-tive spectroscopies. Utilizing modern synchrotronra-diation sources, these quantities can be measured over a wide range
of
photon energies, from the far infrared to the hard x-ray region. Theoretical formulationsof
g and~ can be obtained by the energy-dependent dielectric
function through the relations
il=
[[(E,
+
sz)'~+E, ]/2j'~
andtc=
j[(E,
+sz)'~
—
E,]/2j'~
. The real and imaginary partsof
the dielectric function, c.& and c.2,can be derived by the quantum theory for the interaction
between radiation and matter. A widely accepted model
for the derivation
of
dielectric functions is the extendedDrude model ' for the description
of
collectiveoscilla-tions and interband transitions. Parameters in this model
are determined by a fit
of
the imaginary partof
thedielectric function tocorresponding experimental data.
However, it is the refractive index and the extinction
coefficient rather than the dielectric function which are
experimentally accessible. Therefore, analytical expres-sions for the former two quantities are useful in the deter-mination
of
fitting parameters and in the analysisof
ex-perirnental data. Previously, such expressions forsemi-conductors and insulators were established' ' only for anarrow range
of
photon energies immediately above the band gap. Later, Forouhi and Bloomer(FB)
(Refs. 14 and 15) derived a formula for the extinction coefficient toextend the validity to photon energies in the vicinity
of
interband transitions. TheFB
formula sufFers at least two deficiencies,i.
e.
, the extinction coefficient does notII.
THEORYA. f-sum rules
f-sum rules are important constraints for the analysis
of
energy-dependent optical quantities. These rules may be expressed in different forms involving the imaginarypart
of
the dielectric function, E2(co), the extinctioncoefficient a.(co), and the energy-loss function
Im[
—
1/E(co)]. They are explicitly expressed as'coE (co )dco
=
cooi K(co )dio
=
co (2)where co =Pi(4mNZe
/m)'
is the free-electron plasma energy, N is the atomic (or molecular) density,Z
is the satisfy the symmetry relation and it does not comply withf-sum rules. As a consequence, their formula cannot be applied
to
photon energies around and above the plasmon resonant energies.In the present work, we construct an analytical expres-sion for the extinction coefficient which satisfies the sym-metry relation and sum rules. We also derive an expres-sion for the refractive index by applying Krarners-Kronig
dispersion relations. These expressions are then applied
to the critical-point transitions for electrons from the
valence band to the conduction band
of
insulators andsemiconductors. Comparing the results between this work and experimental data on il(co), tc(co),
s,
(co),E2(co), and the energy-loss function Im[—
1/s(co)],
it isdemon-strated that present formulas for the optical quantities
are valid for photon energies over a wide range from the
band gap to the edge
of
inner shells.4374
Y.
F.
CHEN, C.M.KWEI,AND C.J.
TUNG 48 total numberof
electrons per atom (or molecule), and eand m are, respectively, electron charge and mass. Note that these sum rules apply to optical quantities contribut-ed by all absorption processes including valence-band
ex-citations and inner-shell ionizations over the entire ener-gyinterval:
co=0
—
+~.
A typical spectrum
of
the optical quantity usually re-veals well-separated peaks, with little overlapping, due todifferent absorption processes. In such a case, one can set the upper limit
of
integration in Eqs. (1)—
(3)equal to an energy cutoff co;,large compared to the ith peak energy but small compared to other peak energies. Thus the above sum rules may be separated into individualcontri-bution from each isolated absorption process by the ap-proximate relations
f
,
(')d
0 2
4me
I
W(co)=
i(k;„,
ixik;„,
)
i 5(co—
cop),3Ac
where W(cp) is the energy absorbed by the solid in the range co to co+dco per unit time,
I
isthe incident photon intensity, cisthe speedof
light, and the Dirac 5function isowing to the conservationof
energy.For
the moment, we consider a single critical-point transition from the valence band to the conduction band, with transitionma-trix element
(
k;„,
~x
~k,
'„,
)
and the transition energyct)0=E
E .
The extinction coefficient isgiven by A'cp(co)W(co)26)I
where p(cp) is the number
of
possible transitions in the range co toco+de
per unit volume. Defining the dipoleoscillator strength
of
the transition asand
2
f
coK(co )dcp=
0 4 r]b
fo=
and substituting it into Eqs. (10)and
(11),
we obtain (12) 2 i—
1 m ~P,Ico'Im,
dco'=—
o E(co') (6) neff(co)=
z ~ zf
cp'E2(co')dco'2~'e'A'N
(7) andn,
ff(co)=
f
c'o(Kco)d c'o,~e
ANwhere
co,
is the plasmon energy associated with the ith absorption process and qb is the background refractive index. Note that the above partial f-sum rules for K(co) and Im[—
1/s(co)] yield eff'ective oscillator strengths that are reduced by qb and gb due to the shieldingof
other absorption processes. A detailed discussion about the background refractive index will begiven below.It
is useful to define an effective numberof
electrons per atom (or molecule) corresponding to a given opticalquantity according to
p
K(CO)
—
5(co cop)4cop (13)
Equation (13)does not satisfy the symmetry relation, '
i.
e.
,K(co)=
—
K(—
co).To
rectify this situation, we let~fo
K(co)
=
[5(co—
cop)—
5(co+cop)].
4ct)p (14)
(15) The extinction coefficient
of Eq.
(14)also satisfies the sum ruleof
Eq. (2). In this case, however, the lower limitof
integration must be extended from0 to
—
~
with the right-hand sideof
this equation multiplying by2.
Derivations so far assume zero energy breadth or infinite lifetime for the excited state in the critical-pointtransi-tion. This assumption leads to the 5-function dependence
of
the extinction coefficient on photon energy. In reality, the spontaneous emission produces the dampingof
excit-ed states.
To
accommodate the damping effect, one may replace the 5 functions inEq.
(14) by the Lorentzian line-shape function according tom CO
n,
ff(co)=
co'Im,
dco' .2me AN E Co
For
each absorption process, we expect that(9) where
I
0 is the damping parameter relating to the full
width at half maximum
of
the K(co)spectrum. Therefore,the extinction coefficient becomes
n,
ff(co)~,:n,
ff(co)~,:n,
ff(co)~,
=
1:g&'.
ribf
plpco
K(CP)
=
2 2[(cp
—
cop)+I
p][(co+coo)+I
p](16) in the limit
co~co;.
Equations (7)—(9) provide guidancefor the determination
of
the numberof
valence electrons participating in critical-point transitions.B.
Index ofrefractionConsider the forward propagation
of
photons in asolid. ' The first-order time-dependent perturbation theory under the electric dipole approximation gives'
g(co)
=qb+
—
1 Pf,
K(co')dco',
7T —~CO CO
(17)
where P denotes the Cauchy's principal value.
Substitut-The refractive index is connected to the extinction coefficient by Kramers-Kronig
(KK)
dispersion relations.These relations are direct consequences
of
the causality principle. Based on the Hilbert transform inKK
rela-tions, itgivesing
Eq.
(16)intoEq.
(17),we obtainf
o(m'—
mo—
r(')2
[(~
—
~,
)'+
r,
'][(~+~,}'+
r2]
(18)3
1,2The above formulas for the refractive index and the
ex-tinction coefficient are approximate since they are de-rived by first neglecting the damping effect and then resuming this effect using a Lorentzian line-shape
func-tion. Corresponding approximate formulas for the real and imaginary parts
of
the dielectric function can also be derived using Eqs. (16) and (18) and the relationse&(co)=q) (co)
—
Ic(co) and e2(co)=2qI(co)a(co). Alterna-tively, the quantum theory yields a Drude-type dielectricfunction as ' 0.8 0.4
1.
41.
0—
0.2 I I I I I I I IFc
(co coo ) ,()=
b-(c& coo) +co l'0
Fpgpcs E,
(~)
=
( 2~2)2+~2y2
(20) -0.20.
8 0.60.
40.
2 0.0 I I«
i I 8&eCOI-f
=
3 'l(k:.
I-Ik:"& I'P(-
) (21)be the oscillator strength for the ith critical-point
transi-tion. One can generalize Eqs. (16)and (18)into
where c.& is the background dielectric function
of
the solid, I'p is related tothe osci11ator strength, and yp isthe damping parameter. Comparing the approximatedielec-tric function with the Drude dielectric function, we find that they are equal under conditions E&=q)z,
Fc=fogb,
ye=21
c,I
o«coo,
andfo
«coo.
To
illustrate this com-parison, inFig.
1 we plot results in E,(co), E2(co), andIm[
—
I/e(co)]
calculated using the approximate (solid curves) and the Drude (dashed curves) formulas under conditionse&=qlb=l, In=coo/10,
andfo=coo/10. It
is seen that fairly good agreement isfound for all functionsplotted. The imposed conditions
I
0«coo
andfc«coo
make Eqs. (16) and (18) valid only for insulators and semiconductors where cop are large enough to satisfythese conditions.
The above formulations are for a single critical-point
transition
of
electrons from the valence tothe conductionbands. The generalization
of
these formulations to a gen-eral situation for several critical-point transitions can be made in a straightforward way.It
isnot difficult to showthat each critical-point transition makes an independent contribution to the extinction coefficient. Thus, one can
divide the oscillator strength into fractions
f;
with transi-tion energy co;and damping coefficientI;.
LetI I I I I I I I I I I I I I I I i i I
0Q
0.0 0.5
1.
01.
5 2.0Flax. 1. Aplot ofe&(co},Ez(co},and Im[
—
I/E(c0}]calculated using Eqs. (16) and (18) and the relations c&=
g—
a andc2=2qa (solid curves) and using Eqs. (19) and (20) (dashed curves) bytaking Eb=r)~
=1,
I
0=coo/10,and fo=c002/10.C
f,
(co colI
.)—
—
1.
=(
[(co—
co,)+I
1.][(co+co.
)+I
)(24) where the first and second summation terms indicate,
re-spectively, contributions from the valence band (having q critical-point transitions) and from inner shells (with c
ex-citation groups). Since we are interested in the energy
re-gion co
«
cuj (j
=
1toc),
Eq.
(24) reduces toEq.
(23) withwhere q is the total number
of
critical-point transitions from the valence to the conduction bands.To
show that gb is indeed the background refractive index due to the contribution from core electrons in inner shells, one can rewriteEq.
(23) asq
f
(~2 ~2~ F2~)';
=)[(
co—
co; )+
I";
][(co+
co, )+
I,
]f;
I,
co;=i
[(co—
co;)+I;][(co+co;)
+I;]
(22)C
rl„=
1+
—,'g
co+I
(25) and qf
(~2 ~2 P2};
=
i[(
co—
co;)+
I;
][(co+
co; )+
I,
]
It
is seen that core electrons contribute to theback-ground refractive index by an amount directly
propor-tional to their oscillator strengths but inversely
propor-tional to the square binding energies. Equation (25}will be applied toafew examples discussed below.
4376
Y.
F.
CHEN, C.M.KWEI,AND C.J.
TUNGIII.
RESULTSAND DISCUSSIONWe now fit
Eq.
(22) to optical data' for several in-sulators and semiconductors. TablesI
—
IVlist the fitting parameters for MgO, SiOz, Si, and GaAs. In these fittings, we check the accuracyof
optical quantities in-cluding ri, a, and Im(—
1/E),
critical-point energies tu;, and sum rulesof
Eqs. (4)—
(6) by comparing fitted valuesto experimental data.
It
is confirmed that the totaloscil-lator strength
of
critical-point transitions satisfies the sum rule,i.
e.,gq=,
f;
=co~,
/rib, with co~, being the plasmon energyof
valence electrons. Critical-point ener-gies measured by the optical reAectometry ' andcalcu-lated by the pseudopotential band theory for Si and
GaAs are also listed in Tables
III
and IV for comparison.For
MgO andSi02,
our fits extend to~=300
eV, which exceeds the L-shell threshold energiesof
Mg and Si. Inthese cases, the first eleven terms
(i
=1
—11)
in TablesI
andII
correspondto
critical-point transitionsof
the valence band, whereas the last three terms (i=
12,13, 14) are associated with L-shell excitationsof
Mg and Si. Neglecting the K-shell contribution, we find that the re-fractive index at the region well below the L-shellexcita-tion energy may be given by
Eq.
(24) with 14 terms orbyEq.
(23)with 11 terms. In the latter equation we find thatqb=1.
035 and1.
004 for MgO and Si02, respectively. Thus the background refractive index is indeed deter-mined byEq.
(25).Figure 2 shows a comparison
of
q(cu), a(co), andIm[
—
1/E(co)] for semiconducting Si calculated presently (solid curves) using Eqs. (22) and (23) and the relationsc&
=
g—
~ ande2=2g~,
measured experimentally(dashed curves) and determined by
FB
(chain curves).It
is seen that close agreement is found between present re-sults and experimental data for all quantities plotted.
It
is also seen that prominent interband transitions, represented by the strong resonant peaks in g and ~spec-tra, occur in the region
of
critical-point energies below—
10 eV. Above this energy, the refractive index gradual-ly approaches the asymptotic value gb, and the extinction coefficient decreases with co.
The correspondingFB
re-1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 2 3 4 5 6 7 8 9 10
TABLE
II.
Parameters in Eqs.(22)and (23) forSiO&.gb=1
(1.004 if i=1
to 11 are used)f;
(eV')I;
(eV) 6 15 10 38 51 110 45 41 33 70 63 16 58 170 0.16 0.60 0.69 1.31 1.70 3.96 4.97 5.00 4.98 8.00 12.94 5.95 14.90 59.82 10.2 11.6 12.4 14.0 16.9 20.0 25.0 30.0 40.0 51.0 72.0 108.0 134.6 189.5 0.2 0.4 11.0 16.0 10.0 21.0 23.0 70.0 70.0 50.0 0.01 0.01 0.36 0.25 0.30 0.50 0.70 1.60 2.30 2.80 3.43 3.48 3.72 4.35 4.75 5.45 6.40 7.80 10.60 14.00 3.40 3.45 3.66 4.30 4.57 5.48 3.42 3.48 4.47 4.60 5.56 TABLEIII.
Parameters in Eqs.(22)and (23) for Si.Experimental Theoretical
gb
=
1.015 reAectivity critical-pointf;
(eV )1;
(eV) co; (eV) structure (eV) analysis (eV)TABLE
I.
Parameters in Eqs.(22) and (23) forMgO.gb=1
(1.035 if i=1
to 11 are used)f,
(eV')I;
(eV) co; (eV) TABLEIV. Parameters inEqs.(22)and (23) forGaAs.1 2 3 4 5 6 7 8 9 10 11 12 13 14 2 24 35 35 24 84 38 28 29 131 143 18 95 500 0.10 1.30 0.90 0.59 0.91 1.40 1.00 0.79 1.58 7.47 11.92 1.96 8.95 29.64 7.8 9.1 11.0 13.3 14.8 17.5 19.4 21.2 24.1 34.0 49.0 58.0 72.0 99.0 1 2 3 4 5 6 7 8 9 10 11 0.8 3.4 5.8 13.0 20.0 13.0 20.0 44.0 59.0 49.0 13.0 0.10 0.20 0.40 0.50 0.40 0.50 0.79 1.40 2.20 2.20 1.70 3.03 3.25 3.70 4.60 5.05 5.76 6.70 8.00 10.50 13.50 21.50 3.02 3.25 4.64 5.11 5.64 6.60 3.03 3.25 4.54 5.07 5.76 6.67 9.87 12.55 Experimental Theoretical gb
=1.
01 reAectivity critical-point8 I I I I I I I II I 1 I II I 'I I I I I I I i I I I I I I I 5 6 I
Si
3
23
03
3
03
2 0i2
18 2430
—— 036
0 10 15 20 250FIG.
2. A cornomparison ofg(co), a(co), and Im—
1Si calculated using
E
(22) & ann c,&= g~(solid curves), measured e
(.
.
.
h.
.
.
.
„.
.
)(=.
f
2)
.
.
".
..
'(Ref.15).
e
.
and determined by FB(chain curves) es d 2 determined byFB
(chainsuits, on the other hand, approach constant both optical quantities. Th
tinction coefficient
t
1~ ~
iies. econstant value
of
theFB
ex-the integr 1 th
n a arge co is clearl in
a in e sum
ruleofE
.
2y 'ncorrect because q. ( ) will then diverge. ring wit the experimental data it '
th t th 1'd't
of
FB
1b lo
-7
Vo results is restricted to h energy-loss function, where
B
resule
.
is is true es eciallf
a energy. Figure 3 shows a lot
of
defined in Eqs. (7)—
(9),foror
SiSi calculated presently (solidcurves), measured experimentall determined uusing'
FB
optical uantin a y (dashed curves), and
It
is seen that thquantities (chain curves). a t e e6'ective number
of
valencormulas lead toto unphysical results for n z(co
3 I I 1 I I I I I I I I ' ' ' ' I ' ' ' ' 4 I I I I I II I I I I I MgO Z
=4
V v~b v~b 0 20 40(e
60 80 0 15 I s s i s I30
45 60 750FKx.3. The results ofn,n,s(~)la.
co,
,n,a(col~„,and n,a.(co)~,
fori calculated presentl ( 1
d,
'
ex
y soi curves), usin ex
(dashed curves) (Re
f
.21)and usingFB
o tica'
g experimental data curves) (Ref. 15). The
.
.
edotted line atZ=4
re r'
g ptical quantities (chain ration value for the number
represents the satu-enum erofvalence electrons per atom.
~ p n ofg(co), a(co), and Im
—
1/cFIG.
5. A corn arisongO calculated using
E
s. (22)an c2
=
g~(solid curves) and measured4378
Y.
F.
CHEN, C.M.KWEI,AND C.J.
TUNG 48I I
20 s s s &
( &» &
( '' & & &
( s I I I I I ( I I I ( I I I ( I I
2.
0 Z +Z=16
151.
5 103
3
03
—1.
0 50 ( & I I I I I I I I I I I I I I 100 150 200 250300
(eV) —0.
5FIG.
6. The results ofn,ff(co)~, ,n,z(co)~,and n, (acyl~, for M0
l lated resently (solid curves) and using experimental data (dashed curves) (Ref. 19). The dotted line at Z&represents the saturation value for the number ofelectrons con-tributing toanisolated absorpting group.
I ( I I I I i I I
20 40 60 80
100
0.
0FIG.
7. A comparison ofg(co), ir(co), and Im[—
I/E(col] for Si02glass calculated using Eqs. (22) and (23)and the relationsEl g K and E2=2gK (solid curves) and measured
experimen-tally (dashed curves) (Ref. 20). and
,
n(sue)~, at large co. TheFB
results onn,
fr(co)~2
also deviate greatly compared to experimental data at co&15eV.
Similarly, in Fig. 4 we plot the results
of
q(co), ir(co,and Im[
—
I/E(ru)] for semiconducting GaAs calculated presently (solid curves), measured experimentally (dashed curves), and determined byFB
(chain curves).It
isagain seen that close agreement is found between present re-sults and experimental data for all quantities plotted. The small deviations occurring at cu&22 eV are due tothe contribution from d-band excitations. The onset
of
these excitations is-25
eV, above which optical data areunavailable.
It
is also seen that theFB
results are in agreement with experimental data only forco(6
eV. Thefailure
of
these results around and above plasmon ener-gies is clearly demonstrated.As an example
of
insulators, inFig.
5 we plotg(co,
ir(ro), and Im[—
I/E(co)] for MgO. Here the agreement between present results (solid curves) and experimentaldata (dashed curves) is quite good for all quantities
plot-ted. Figure 6 shows a plot
of n,
ff(co) calculated presently( 1'd curves) and measured experimentally (dashed
f
curves) for MgO. Note that the saturation value o
( )~ is near 16instead
of
8, the numberof
valencenet 6) ~ Is nea
electrons. This is due to the onset
of
Mg L-shellexcita-tions near 55 eV; there only about
75% of
the oscillatorstrength
of
valence electrons is exhausted. Thus a strong overlappingof
oscillator strengths between the valence band and the MgL
shell exists above 55 eV. Also, the next filled0
X
shell lies-500
eV above the valence ban.
d Thereforee the isolated absorption group shouldin-2
.
2 2 4elude eight valence electrons
(3s of
Mg; 2s and p o0)
and eight Mg L-shell electrons (2s and 2pof
Mg). The mergeof
saturation valuesof
all threen,
( a)scupectraf
indicates that gb
=1.
It
means that absorption peaks oMg and
0
K
shells are far away from critical-point tran-sition peaksof
the valence band.Finally, in
Fig.
7we plot optical quantities for another insulator, Si02 glass. Good agreement is found between present results (solid curves) and experimental data(dashed curves). In this case, the isolated absorption grou~roup involves 16 valence electrons
(3s
an 3p oof
Sii; 2s and 2ppof
o0)
and 8Si L-shell electrons (2s2and 2p6=
300of
Si). The saturationof n,
fr(co) to 24 occurs atco=
eV.IV. CONCLUSIONS
We have constructed analytical expressions for the ex-tinction coefficient and the refractive index
of
semicon-ductors and insulators. Our approach involved the
appli-cation
of
f-sum rules and symmetry relations. Theseex-' ~
pressions contain parameters charactenzing the oscillator
strength, the damping effect, and the transition energy
as-sociated with each critical-point transition from the valence to the conduction bands. Contributions from inner shells to the refractive index in the vicinity o critical-point energies were included as a background re-fractive index. Applications
of
present formulations were made for MgO, Si02, Si,and GaAs over a wide rangeof
photon energies. Results were in very good agreement with experimental data for all optical quantities studied.ACKNO%'LED GMENT
This work was supported by the National Science
Council
of
the Republicof
China under Contract No. NSC82-0208-M-009-013.*Present address: Departmerit of Nuclear Engineering, Texas A8cM University, College Station, TX77843.
F.
Bassani and G. P.Parravicini, Electronic States and Optical Transitions in Solids (Pergamon, Oxford, 1975).2F.Wooten, Optical Properties ofSolids (Academic, New York, 1972).
R.
M. A.Azzam and N. M. Bashara, Ellipsometry and Polar-ized Light (North-Holland, Amsterdam, 1977).4P.S.Hauge, Surf. Sci. 96, 108(1979). 5R. H.Muller, Surf. Sci. 56,19(1976).
A. V.Rzhanov and
K. K.
Svitashev, Adv. Electron. Electron Phys. 49,1(1979).7R.Loudon, The Quantum Theory ofLight (Clarendon, Oxford, 1973).
sH. Raether, Excitations
of
Plasmons and Interband Transitionsby Electrons, Springer Tracts in Modern Physics Vol. 88 (Springer, Berlin, 1980).
C.M. Kwei,
T.
L.Lin, and C.J.
Tung,J.
Phys. B21, 2901 (1988);see also C.M.Kwei,Y.
F.
Chen, C.J.
Tung, andJ.
P. %'ang, Surf.Sci.(to be published).D.
L. Greenaway and G. Harbeke, Optical Properties and Band Structureof
Semiconductors (Pergamon, Oxford, 1968i. ~iJ.I.
Pankove, Optical Processes in Semiconductors (Dover,New York, 1971).
J.
C.Phillips, Bonds and Bands in Semiconductors (Academic,New York, 1973).
' Handbook
of
Optical Constants ofSolids, edited by E. D.Palik (Academic, New York, 1985).i4A.
R.
Forouhi andI.
Bloomer, Phys. Rev.B34, 7018 (1986). i5A.R.
Forouhi andI.
Bloomer, Phys. Rev.B38,1865(1988).D. Y.Smith and
E.
Shiles, Phys. Rev.B17, 4689 (1978).'7J.
J.
Sakurai, Advanced Quantum Mechanics (Addison-Wesley, Reading, MA, 1967).H. M. Nussenzveig, Causality and Dispersion Relations (Academic, New York, 1972).
'9D. M. Roessler and D.
R.
Hu(oman, in Handbook ofOptical Constantsof
SolidsII,
edited byE.
D.Palik (Academic, NewYork, 1991),p.919.
2oH.
R.
Philipp, in Handbookof
Optical Constants ofSolids I, edited byE.
D.Palik (Academic, New York, 1985),p. 749.2'D.
F.
Edwards, in Handbookof
Optical Constantsof
SolidsI,
edited by
E.
D.Palik (Academic, New York, 1985),p. 547.D.
E.
Aspnes and A. A.Studna, Phys. Rev.B27,985(1983).C.V.Festenberg, Z.Phys. 227, 453(1969).
24R.
R.
L.Zucca,J.
P.Walter, Y.R.
Shen, and M. L.Cohen, Solid State Commun. 8, 627(1970).25R.