Sunteți pe pagina 1din 4

PROC.

21st INTERNATIONAL CONFERENCE ON MICROELECTRONICS (MIEL'97), NIŠ, SERBIA, 14-17 SEPTEMBER, 1997

Dispersion of Refractive Index


in Degenerate Mercury Cadmium Telluride
Zoran Jakšić, Olga Jakšić

Abstract – The work presents an approximate expression for In this work we propose a simple empirical relationship
refractive index of mercury cadmium telluride (Hg1-xCdxTe) for refractive index dispersion in dependence on bandgap
applicable to both intrinsic and degenerate material. The energy, temperature, infrared radiation wavelength and car-
relation is applicable for cadmium mole fractions x=0.165-1 at rier concentration. The expression is valid both in nonde-
300 K and 0.165-0.54 at 77 K and electron concentrations up
generate and in degenerate case.
to 1017cm–3. The results are compared to available experi-
mental data and to the results of calculation using Kramers-
II. THEORY
Kronig relation. A unified expression for interband absorp-
tion coefficient based on the Kane three-band model was used There are several mechanisms determining the optical
in the calculation, whilst the Urbach band-tailing and the behaviour of mercury cadmium telluride single crystals.
electron-electron interaction were also taken into account.
The whole wavelength region is usually divided into three
I. INTRODUCTION regions, depending on the prevailing mechanism of optical
processes [5]. The short-wavelength part corresponds to
Due to the possibility to engineer its bandgap, mercury the absorption transitions in near ultraviolet (UV); the
cadmium telluride (Hg1-xCdxTe, MCT) is today surely the middle part, which is the most important for our
most important semiconductor material for detection in calculation, stems from interband transitions near the cutoff
long wavelength infrared range [1]. The dispersion of energy (CO), while the low-frequency part, in the far
refractive index is one of its most important properties infrared region, is connected with the free-carrier
determining behavior of optoelectronic devices fabricated absorption (FC). It is known that the contributions to the
in it. It is especially important to know the refractive index absorption coefficient from various mechanisms are
near the fundamental absorption edge. additive. Therefore the total absorption coefficient (αtot)
A number of authors handled the dispersion of refractive may be written as follows:
index in mercury cadmium telluride single crystals [2-4].
Its value was usually determined by the interference in αtot=αUV+αCO+αFC (1)
transmission spectroscopy, by the multioscillator model or
The term corresponding to the short-wavelength range
using the Kramers-Kronig relation. The most exhaustive
(UV) is usually determined by simple extrapolation of the
model based on quantum mechanical approach was pro-
absorption coefficient in the (CO) range within the
posed by Jensen and Torabi [4]. The authors of [3] point
out that the refractive index calculated according to [4] log(λ)~λ coordinate system [3, 6].
decreases much quicker than observed in recent experi- The well-known Kramers-Kronig relation [6] may be
mental results, the probable reason for this being the used to furnish a dependence between the real part of the
chosen set of calculation parameters. Therefore various refractive index and the absorption coefficient
empirical relations are used today [1-3]. ∞
α (ω )

The most accurate expression to date for the dispersion c
n(ω 0 ) = 1 + dω (2)
relation of refractive index of MCT was reported by Liu, 2 ω 2 − ω 02
0
Chu and Tang [3]. They used the Kramers-Kronig relation
to handle their enormous body of experimental data Because of the linearity of the integrand in the Kramers-
obtained by intrinsic absorption spectroscopy and to Kronig relation, the above expression can be regarded as a
furnish an empirical formula. Their expression is sum of two independent contributions, one off them caused
applicable in the range 0.276<x<0.443 and at temperatures by the interband transitions and comprising the terms (CO)
from 4.2 to 300 K. At room temperature it is valid for and (UV), while the other is the result of the free-carrier
0.205<x<1. absorption (FC),
To the best of the authors knowledge, none of the empi-
rical expressions on refractive index of MCT reported until n=ninterband+nFC (3)
now was applicable for the case of degenerate MCT.
Let us consider first the interband part of the refractive
Authors are with IHTM – Institute of Microelectronic Techno-
coefficient. We use our modification of the empirical ex-
logies and Single Crystals, Njegoševa 12, Belgrade University, pression for refractive index in the interband range, which
11000 Belgrade, Yugoslavia, fax (++381 11 182 995), e-mail was obtained in [3] by applying the Kramers-Kronig
jaksa@hf01.chem.bg.ac.yu relation to experimental absorption data

0-7803-3664-X/97/$10.00 © 1997 IEEE

95
A2 validity, typically about 1017 cm–3. However, as shown by
ninterband = A1 + 2 2 2 experiments, the expression is often used beyond its
1 − A3 h ω validity range [9,11] without a marked loss of the accuracy.
(4)
The second mechanism of the bandgap decrease is the
where the coefficients are Coulomb interaction between electrons and impurities in
~ degenerate semiconductor [9]
A1 = 5.37941 − 0.00343T + (6.65569 + 0.00492T )e
Eg

⎛ π 4 / 3h 2 ⎞⎛ 1 ⎞⎛ Φ ⎞ 2/3
⎟⎜ ⎟n
( ) E imp = ⎜⎜ 1 / 3 ⎟⎜
⎟⎜ m * ⎟⎜ 1 − 2 E
~ (9)
−4 −4 − Eg
A2 = 0.8915 − 8.5 ⋅ 10 T + 0.14027 + 10 T e (5) ⎟
⎝ 3 ⎠⎝ 0 ⎠⎝ g ⎠
~
A3 = −3.712 − 0.00956T + (1.8334 + 0.004463T ) E g where the third bracket takes into account the influence of
~
−Eg nonparabolicity of the conduction band for strong
+ (6.797 + 0.0115T )e
degeneracy of carriers. We calculated the term Φ/Eg by
~ applying a formalism similar to that in [9], but using the
where E g =Eg[eV]/eV and sample temperature T is in K.
Moss-Burstein bandfilling factor instead of the Fermi level.
The influence of mechanisms appearing in degenerate In this manner we avoided the introduction of an additional
semiconductor was taken into account by introducing an fitting parameter into (9) which was used in [9]. The final
effective bandgap energy in a manner which represents a expression for the nonparabolicity correction is
generalization of that described in [7]. The effective
bandgap is obtained as a sum of the fundamental bandgap Φ 4∆E MB m0*
= −1 (10)
energy Eg and the terms corresponding to the bandgap shift
due to the Moss-Burstein bandfilling, electron-electron
Eg 3π 2 ( )
2/3 2 2/3
h n
interaction and electron-impurities interaction.
In (9) and (10) m0* is the effective mass at the bottom of
The term corresponding to the Moss-Burstein bandfilling
the conduction band and may be written as [1]
was calculated according to Scacham and Finkman [2] as
the photon energy for which the intrinsic absorption * Eg + ∆
coefficient of MCT reached 500 cm-1 (the optical me = 0.7168 m0 (11)
bandgap). To this purpose we utilized our unified model of 3 + 2∆ / E g
interband absorption coefficient in mercury cadmium
telluride [8] which takes into account the nonparabolicity where we used the spin split-off energy value ∆=0.9 eV
of bands and effects of carrier degeneration, as well as the and the Kane’s matrix element P=8⋅10-10 eVm. The value
Urbach band tailing for energies below the cutoff. We m0 denotes the free-electron mass.
obtained the following approximate dependence of the It should be noted that in (8, 9) and in our model of the
Moss-Burstein bandfilling factor on cadmium molar absorption coefficient we applied the expression for the
fraction x, temperature T (in K) and electron concentration high-frequency dielectric permittivity in nondegenerate
ne (in cm–3) [7] semiconductor, as reported in [1]. It is given as ε∞=19–11x
and we used it as a fitting parameter independent on
ξ temperature and carrier concentration in order to avoid a
1− x ⎛ ⎞
∆E MB (eV ) =
ne
⎜ ⎟ , (6) self-consistent numerical procedure.
2 ⎝ 7.85 ⋅ 1018 e 0.002T ⎠
We used Hansen’s expression for bandgap energy [12]
−4 −4
ξ = 0.63x + 3.8 ⋅10 T + 0.294 , (7) E g = −0.302 + 193
. x + 5.35 ⋅ 10 T (1 − 2 x )
(12)
2 3
Injected carriers decrease the bandgap due to two main − 0.81x + 0.832 x
mechanisms. One of them is electron-electron interaction
(exchange energy). We used the following expression where T is temperature in K and x is Cd mole fraction in
[9,10] Hg1–xCdxTe. The bandgap energy is obtained in eV.
The effective bandgap is finally calculated as
⎛ e ⎞ ⎛ 3 ⎞ 1/ 3 1/ 3
∆Eel ( eV ) = ⎜ ⎟ ⎜ ⎟ ne . (8) E g eff = E g − ∆E MB − ∆Eel − ∆Eimp (13)
⎝ 2πε 0 ε r ⎠ ⎝ π ⎠

where the exchange energy is obtained in eV if all the By introducing (13) instead of Eg into (4, 5) we obtain an
factors are given in SI units. approximate expression for the dispersion of refractive
Strictly speaking, the above expression is applicable only index due to interband transitions which takes into account
if the ratio between the mean distance between carriers and bandgap shift caused by increased carrier concentration,
the Bohr radius for electrons in the crystal is larger than 1, i. e. valid both in nondegenerate and in degenerate case:
i.e. there is an upper electron concentration limit for its

96
{
ninterband = 5.379 − 0.00343T + [6.655 + 0.00492T ]e
~
− E geff The Cauchy principal value of the Kramers-Kronig inte-
gral (2) is determined according to [15]. For carrier
( )
~
+ ⎡0.8915 − 8.5 ⋅ 10 − 4 T + 0.14 + 10 − 4 T e ⎤ concentrations below 1017 cm–3 the FC absorption causes a
− E geff
⎢⎣ ⎥⎦ faster decrease of refractive index with wavelength far in
[ ( ~
× 1 − h 2ω 2 − 3.71 − 0.0095T + (1.833 + 0.00446T ) E geff (14) the longwave region, while the changes in the CO part
remain negligible. Thus we conclude that for most of the
1/ 2
−1 ⎫ MCT photonic devices the expression (14) can be regarded
⎞⎟ ⎤ ⎪
~2
− E geff
+ (6.797 + 0.0115T )e ⎥ ⎬ as a relatively valid approximation for the total refractive
⎠ ⎦ ⎪ index in the limited range near the absorption edge and for

~ moderate dopings.
where E geff is dimensionless and equal to the value of Egeff
in eV. III. RESULTS AND DISCUSSION
The next mechanism of influence to be considered is the
free carrier absorption. The wavelength dependence of the Figure 1 shows the spectral dependence of refractive in-
absorption coefficient in the FC region may be written as dex at room temperature for different values of x in the
[13] case of nondegenerate semiconductor at room temperature.
The peaks on the left side of the diagram correspond to the
α FC = (δ v1v 2 + δ v1 )P + δ nn N (15)
energies equal to the optical bandgaps for given
where δv1v2 is the cross-section for the intervalence band compositions. The approximation (14) is compared to the
transitions between heavy and light hole band, calculated experimental data published in [3] and to the results of
Kramers-Kronig analysis presented in the same paper.
according to the Kane model. δv1 is the cross-section for
intraband hole transitions [13, 14]. P, N are hole and 4.0
electron concentrations, respectively. The term δnn
describes free-electron absorption by various scattering
mechanisms, and is equal to the sum of the cross-sections 0.165
for scattering on ionized impurities (I), polar optical 3.5
phonons (OP) and deformational potential (DEF) of the
Refractive index

0.205
0.33
acoustical and optical phonons [14]:
0.416
6
e NI 7/ 2
δI = λ (16) 3.0 0.7
4 *3 9
hε ∞ ε s m0 c

δ OP =
(
e 4ω LO χ OP ε ∞−1 − ε s−1 )λ 5/ 2
(17) 2.5
1

hε s m0*π 3 c 7 0 5 10 15 20
Wavelength, µm

e 2 m0* ⎡ Ξ u2 D 2 h 2 ⎤ 3 / 2 Fig. 1 The refractive index dispersion near the bandgap for dif-
δ DEF = ⎢ + ⎥λ (18)
ρ ε ∞π 3 c 5 h 5 ⎢⎣ v II ω TO ⎥⎦ ferent cadmium molar fractions x (0.165; 0.2; 0.33; 0.416;
0.7; 1) of mercury cadmium telluride at 300 K in nonde-
generate case (solid lines). Circles denote experimental data
where εs=20.5–15.5x+5.7x2 [16] and ε∞ (given above) are according to [3] and triangles according to [17], both for
static and high-frequency dielectric permittivity of undoped x=0.205 and 1. Dashed lines are the results of application of
material, NI is the concentration of the scattering ions, Kramers-Kronig relation for x=0.205 and 1 [3]. The x-
ρ=8.0756-2.2571x g/cm3 [16] is MCT density at 300 K, marks denote our experimental values for x=0.205.
D=14.8 eV [14] is the constant of deformation potential for
optical phonons, Ξu=14 eV [13] is the constant of To check the validity of the expression (14) we compared
deformation potential, vII=3000 m/s [13] is the longitudinal it with the results of calculation using the Kramers-Kronig
sound velocity in MCT. The values of the other parameters relation published in [3], with experiments in [3] and [17]
are taken from [14]. and with our own experimental results. Our samples were
In the 10-25µm range the FC absorption coefficient has grown by open tube isothermal vapor-phase epitaxy and
the characteristic shape α∼λr, where r≈2 for p-type MCT at their transmission characteristics were measured on the
300 K, and 0 for p-type at 77 K. The main mechanisms in Specord IR-75 spectrophotometer in the range 2.5 to 25
this case are the intervalence (light hole-heavy hole) band µm at 293 K. The total absorption coefficient for an
transitions and intraband (heavy holes) absorption. For n- epilayer was determined as the mean value through the
type semiconductor r≈3, and the prevailing mechanism is whole layer, i. e. the integral of α(x)dx divided it by the
the scattering of electrons on optical phonons. total layer thickness. In our calculation of α from
transmission we took into account the multiple reflections

97
at the sample interfaces as described in [13] and used the REFERENCES
value R=31% for reflection coefficient of MCT for
composition x=0.205 [14]. The carrier concentration was [1] J. Piotrowski, W. Galus, M. Grudzien, “Near Room-Tempe-
rature Infrared Photo-Detectors”, Infrared Phys. 31, 1, pp.
determined by the Hall measurement. We used two
1-48, 1991
samples 30 µm and 33 µm thick, for both of which we
[2] S. E. Schacham and E. Finkman, “Recombination Mecha-
obtained x=0.198 using (13). Their hole concentration was nisms in p-type HgCdTe: Freezout and Background Flux
P=1016 cm–3. The longwave part of both the absorption Effects”, J. Appl. Phys. 57 (6), pp. 2001-2009, 1985.
characteristics had the typical λr shape, with r=2, which is [3] K. Liu, J. Chu, D. Tang, “Composition and Temperature
in agreement with [13]. Our values of refractive index were Dependence of the Refractive Index in Hg1-XCdXTe”, J.
slightly higher than 3.5, as shown by x marks in Fig. 1. Appl. Phys. 75, 8, pp. 4176-4179, 1994
Thus we conclude that the use of (14) was justified for our [4] B. Jensen, A. Torabi, “Linear and Nonlinear Intensity
samples. Dependent Refractive Index of Hg1-xCdxTe”, J. Appl. Phys.
54, 10, pp. 5945-5949, 1983
[5] C. Tanguy, “Temperature Dependence of the Refractive
4.0
Index of Direct Band Gap Semiconductors near the
3.9
1013 cm–3 Absorption Threshold: Application to GaAs”, J. Appl. Phys
3.8 80, 8, pp. 4626-4631, 1996
Refractive Index

3.7
[6] V. V. Sobolev, S. A. Alekseeva, V. I. Doneckih, Calculation
1015 cm–3 of Optical Functions of Semiconductors According to the
3.6 Krammers-Kronig Relations (in Russian), Shtiinca, Kishi-
3.5 nev, Moldavia, 1976
1017 cm–3 [7] Z. Jakšić, O. Jakšić, “Simple Approximation for Absorption
3.4 10 cm–3
18
Coefficient in Degenerate Mercury Cadmium Telluride",
3.3 IEEE Proc. 20th International Conference on Microelec-
1019 cm–3
3.2 tronics MIEL 95, Vol 1, Niš, Serbia, pp. 117-120, 1995
0 5 10 15 20 [8] Z. Djurić, Z. Jakšić, Z. Djinović, M. Matić, Ž. Lazić, “So-
Wavelength, µm me Theoretical and Technological Aspects of Uncooled
HgCdTe Detectors: A Review”, Microelectronics Journal
25, 2, pp. 99-114, 1994
Fig. 2 Influence of degeneration to the refractive index dispersion
around the bandgap for mercury cadmium telluride. T=77K, [9] D. Hahn, O. Jaschinski, H. H. Wehmann, A. Schlachetzki,
x=0.2. M. von Ortenberg, “Electron-Concentration Dependence of
Absorption and Refraction in n-In0.53Ga0.47As Near the
The refractive index calculated for different electron Band-Edge”, J. Electronic Materials, 24, 10, pp. 1357-1361,
1995
concentrations according to (11) at T=77K and for x=0.2 is
[10] J. Camassel, D. Auvergne, H. Mathieu, “Temperature De-
shown in Fig. 2. As expected, for larger degenerations the
pendence of the Band Gap and Comparison with the
refractive index peak moves toward shorter wavelengths, threshold Frequency of Pure GaAs Lasers”, J. Appl. Phys.
while the decrease is faster in the whole range compared to 46, 6, pp. 2683-2689, 1975
the undoped case. [11] M. Bugajski, W. Lewandowski, “Concentration-Dependent
Absorption and Photoluminescence of n-type InP”, J. Appl.
IV. CONCLUSION Phys. 57, 2, pp. 521-530, 1985
We investigated the behavior of refractive index in [12] G. Hansen, J. Schmit, T. Casselman, “Energy Gap Versus
mercury cadmium telluride taking into account the effects Alloy Composition and Temperature in HgCdTe”, J. Appl.
Phys. 53, 10, pp. 7099-7101, 1982
of increased doping. We proposed an approximation taking
into account the effects of temperature, composition and [13] B. Li, J. Chu, Z. Chen, Y. Chang, H. Ji, D. Tang, "Free-
Carrier Absorption of Hg1-xCdxTe Epitaxial Films", J. Appl.
carrier concentration influence. The relation is applicable
Phys. 79, 10, pp. 7738-7742, 1996
for cadmium mole fractions x=0.165-1 at 300 K and 0.165-
[14] P. Baranskii, O. Gorodnichii, N. Shevchenko, "IR Optical
0.54 at 77 K and electron concentrations up to 1017cm–3. Absorption and Reflection Spectra in Narrow Gap Semicon-
We compared the approximation with experimental and ductors", Infrared Phys. 30, 3, pp. 259-263, 1990
theoretical values. [15] L. Schoenfeld, “On Integrals of the Type ∫f(x)log|x+1|/|x-1|
dx”, SIAM Review, 1, pp. 154-157, 1959
ACKNOWLEDGEMENTS
[16] A. Rogalski, J. Piotrowski, “Intrinsic Infrared Detectors”,
The authors wish to thank Prof. Z. Djuric for stimulating Prog. Quant. Electr. 2, 12, pp. 87-289, 1988
discussions and Dr. Z. Djinovic for doped HgCdTe [17] E. Finkman, S. Schacham, "The Exponential Optical Ab-
samples. sorption Band Tail of Hg1-xCdxTe", J. Appl. Phys. 56, 10,
pp. 2896-2900, 1984

98

S-ar putea să vă placă și