M. SCEPANOVIC:and M.

JwnC: Mercury Diffusion in HgCdTe during Annealing 289

phys. stat. sol. (a) 143, 289 (1994)

Subject classification: 66.30; 61.80; S8.13

Insiiiute of Physics, Beograd, Zemun ')

Numerical Simulations of Mercury Diffusion in HgCdTe

during Laser Annealing
BY
M. S~EPANOVICand M. J E V T I ~

The effects of pulsed laser irradiation on the Hg concentration in HgCdTe are investigated. A model
for laser induced diffusion of Hg, taking into account the heat flow transient and the melting and
solidification of the surface of HgCdTe under pulsed laser irradiation, is developed. It allows the
analysis of the role of irradiation and of diffusion parameters on Hg behavior. Comparison of the
calculated compositional depth profiles of Hg with the published experimental data enables the
assessment of the values of the diffusion, out-diffusion, and segregation parameters, unknown under
laser annealing conditions.

1. Introduction
The electrical and optical properties of HgCdTe are determined both by electrically active
impurities and by native defects. Mercury vacancies act as acceptors, typically dominating
all other electrically active impurities, whereas Hg interstitials act as donors [ 13.
It is known that the weakness of Hg-Te bonds, which tend to weaken even more due
to the presence of nearby Cd-Te bonds, is often a source for the appearance of acceptor
concentration changes. For this reason, Hg losses observed under conventional or pulsed
thermal annealing techniques, are the critical mechanism of change in the electrical properties
during HgCdTe processing. Short-time annealing techniques, such as laser annealing, offer
the advantage of minimizing the stoichiometric changes and thereby any undesired changes
in the electrical properties [2].
Therefore, investigation of Hg behavior in the laser annealed HgCdTe has great
importance. Here we have proposed a dynamical model for laser induced Hg losses in
HgCdTe. The numerical results are compared with the published experimental data for
Nd :YAG laser irradiation.

2. Mercury Losses under Laser Irradiation of HgCdTe

The behavior of nominal Hg,-,Cd,Te ( X = 0.2) samples under pulsed laser irradiation
was investigated earlier [2]. It was found that the most significant laser induced effect is
the compositional change in the direction perpendicular to the surface of the wafer. This
change is mainly restricted to the irradiated area, emphasizing the local character of laser
treatment. The laser induced process is unambiguously accompanied by loss of Hg. The
first depth profiles of the laser treated HgCdTe area, both by Auger sputter depth profiling
and by Rutherford backscattering, have been reported by Afonso et al. [2]. They have shown
that laser induced effects extend to a depth of 60 to 150 nm, depending on the laser irradiation

') PregCrevica 118, 11080 Zemun, Yugoslavia.

290 M. S ~ E P A N Oand
V I ~M. JEVTIC

conditions. A deficient region of Hg has been found to be located a few tens of monolayers
below the top of the surface. Also, it seems that there might be higher Hg losses due to the
thicker Hg deficient layer rather than to the lower Hg content at the points of maximum
depletion. The minimal measured Hg concentration was found to stay around 50% of the
surface as-grown concentration. An adequate model for simulation of the compositional
change during laser irradiation has not been given until now.
To analyze the laser induced compositional change, we assume that once the irradiation
starts and the surface is heated, the first monolayer becomes Hg depleted. Consequently,
a concentration gradient is generated and a diffusion process of Hg towards the surface is
activated. Two competing processes then take place while laser interaction exists, the
out-diffusion of Hg and the diffusion of Hg towards the surface. The Hg diffusion coefficient
increases with temperature [3], and although no data have been reported for melted HgCdTe,
its value is expected to increase drastically upon melting, similar to what happens in other
semiconductors. Hence, diffusion lengths larger than in the solid are to be expected with
the consequent stimulation of Hg diffusion to the surface. Moreover, under conditions of
an intense thermal gradient of about lo’ K/cm, a thermal diffusion of Hg (Soret effect)
should be superimposed on ordinary diffusion. However, the formation of a thin surface
layer rich in Hg cannot be understood based only on these two diffusion mechanisms.
There has been no adequate explanation until now. We suppose that the segregation during
the solidification process should be taken into account. Unfortunately, the value of the
segregation Coefficient of Hg under nonequilibrium conditions induced by laser irradiation,
when the liquid-solid interface moves at a high velocity, about a few m/s [4], is unknown.

3. Model
The compositional change is modelled by applying the one-dimensional diffusion equation
for the Hg concentration C(x, t ) at depth x, to the surface layers having temperature T and
diffusion coefficient changing with time t ,

D(T) is the usual diffusion coefficient,

D(T)= Do exp (-Q/kBT) 9 (2)
where k, is the Boltzmann constant. The experimental activation energy Q , and the
pre-exponential factor Do for Hg diffusion in HgCdTe systems vary across a broad range
of values when the experimental conditions change [5]. It should be pointed out that the
diffusion coefficients in a liquid phase should be several orders of magnitude larger than
in a solid phase, and we have introduced the multiplying factor zD1,i.e. D,(T) = zD]D,(T).
The term

is called the Soret coefficient [6].Here AH,,, stands for the heat of transport of Hg atoms.
The minus on the right side of (3) indicates that Hg diffuses preferentially to the hotter end
of a specimen, under the thermal gradient. The fact that under such conditions mercury
Numerical Simulations of Mercury Diffusion in HgCdTe during Laser Annealing 29 1

diffuses mostly by the vacancy mechanism [7] is an explanation for the mentioned diffusion
direction.
We assume that the initial concentration of Hg is the uniform bulk concentration C,,

C(x, 0) = c, . (4)
The effects of ordinary and thermal diffusion are included in the model by (1) and the
initial condition (4). The out-diffusion of Hg is considered by the following semiempirical
relation chosen for the first boundary condition:

Here h(T)is the surface mass transfer coefficient, which controls the diffusion efficiency of
Hg through the surface, and strongly depends on temperature and phase. C(0, t ) is the Hg
concentration at the surface.
The second boundary condition refers to the fact that a change of Hg concentration does
not exist in the sample bulk,

C(oo,t) = c,. (6)

The absorbed laser light melts HgCdTe to a depth dependent on energy density, laser
frequency, and pulse duration time [4]. As the liquid cools by the underlying substrate, the
melt front begins to recede toward the surface with velocity v until it begins to encounter
diffusing Hg atoms. At this time the liquid layer is divided into segments of equal width
Ax (typically 2.5 to 5 nm). Diffusion of mercury atoms is then calculated by successive
numerical solutions of the finite difference equation developed from (1).When the receding
melt front begins to encounter diffusing mercury, solidification of segment Axm, at the
position of the front x,, occurs instantaneously, with Hg concentration

Cs(Xmr t o ) = k'(v) CI(xm, to) t (7)

being incorporated into the solid, and remainder

ACl(xm- 1, to) = (1 - k'(v)) CI(xm, t o ) 9 (8)

being rejected into the adjacent liquid layer at x , - ~ . C, is the Hg concentration in the
liquid at the interface prior to solidification, and k' is the nonequilibrium segregation
coefficient [8]. Rejected mercury atoms then diffuse-into the remaining liquid, with the melt
front remaining stationary for a time Atm = Ax/v, and the resulting profile in the liquid
is calculated by n successive solutions, each for a time At = Atm/n, to the finite differ-
ence equation. After a time Atm,the melt front then advantages instantaneously through
the segment AX^-^, with the partition of Hg between the solid and liquid being given
by k'. Calculations of the type discussed above are continued until the melt front recedes
back to within 5 nm of the surface where Hg remaining in a thin surface layer is considered to
be segregated at the surface, taking into account the Hg out-diffusion. In this model k' is
treated as a fitting parameter, and the values for k' are determined by comparing the
calculations of Hg redistribution to the experimental measurements of Hg profiles after
laser annealing [2].

23 physica (a) 143/2

292 M. SCEPANOVIC and M. JEVTIC

The model assumes that the mechanisms for Hg redistribution are ordinary and thermal
diffusion of mercury atoms, and their out-diffusion in the absence of convection in the
liquid. The segregation effect is taken into consideration assuming the nonequilibrium
segregation coefficient, K , to be independent of Hg concentration.

4. Numerical Results and Discussion

Using the proposed model, the numerical calculations of the Hg concentration changes
have been performed for irradiation with frequency doubled Nd :YAG laser single pulses
(2 = 532 nm) having an energy density E = 0.15 J/cm2 and 8 ns duration.

4.1 Concentration profiles of Hg calculated by neglecting the segregation

The results given in Fig. 1 show the influence of the diffusion coefficient D ( T ) and the surface
mass transfer coefficient h( T ) , on the calculated Hg concentration profiles, without
segregation ( K = l),when only diffusion and out-diffusion processes are present. We assume
that the diffusion coefficient D(T)is raised zD, times due to material melting. It is obvious
from Fig. 1 that the slope of Hg concentration is greater, if zDI is greater, whereas its
dependence on activation energy Q is reciprocal. On the other hand, at the beginning we
supposed that the values of mass transfer coefficient h(T) can be obtained by extrapolation
from its experimentally determined dependence under equilibrium conditions at low
temperatures. However, the effect of out-diffusion, arising from the values of h( T ) ,obtained
in this way, is much smaller than the effect of out-diffusion obtained in [2]. Consequently,
we assume that under the nonequilibrium conditions h ( T )is z, times greater in solid state,
and z1times greater in liquid state than under equilibrium conditions, i.e. h,(T) = z,h(T) and

9.0

8.5

8.0

7.5

7.0

6.5
.- ...- .... Do=l 6cm2/s;Q=O.SeV;zoi=4; z,=l O;zl=l 00
. . . . . . . .. Do=l 6Cm2/S;Q=O.9eV;Z~1=4; z s = l O;zl=l 000
6.0
.-. - .-. - D o = l . 6 c m2/s;Q=O.9eV;z~i=4O
Do=0.2c m2/s;Q=0.9eV;z~i=320
5.5 - - - - - - - Do=0.2c m2/s;Q=0.8eV;zol=l 00

0 10 20 30 40 50 60 70
depth (nm)
80 90
-
100 110 1 2 0 130 1 4 0

Fig. 1. Depth dependence of the Hg concentration as a function of diffusion (Do, Q, zD,)and out-diffusion
parameters (zs,z,), without segregation (k' = 1)
Numerical Simulations of Mercury Diffusion in HgCdTe during Laser Annealing 293

11.00 ~

10.55 :

10.10 7

9.65 :

9.20 :

8.75 :

8.30 :

7.85 :

7.40 :

6.95 -

0 10 20 30 40 50 60 70
depth (nm)
80
-
90 100 110 120 130 140

Fig. 2. Dependence of depth profiles of Hg on the segregation coefficient (k'), without out-diffusion
(zs= 0). The curve designated by a solid line takes into account the Soret effect with heat of transport
AH,,, = 3.2eV

h,(T) = z,h(T).One can see from Fig. 1 that the out-diffusion is low for small z,. Therefore,
zI must be greater than 500 if we want to attain a satisfactory agreement between calculated
and measured profiles. Also, we point out that the experimentally observed minimum in
the mercury concentration cannot be explained in this way.

4.2 Concentration profiles of Hg calculated by neglecting out-diffusion

The results in Fig. 2 show the influence of the segregation coefficient, k',on the Hg concentra-
tion when out-diffusion is neglected, h(T) = 0. The curves given in Fig. 2 refer to values of
k' from 0.6 to 0.9. Note that the segregation coefficient under equilibrium conditions is 0.56.
The curve designated by the solid line is a result of taking into account the Soret effect with
the heat of transport AHmequal to 3.2 eV, for k' = 0.6. Its influence on the Hg concentration
profiles is small. It can be explained by the fact that the applied laser energy causes a small
temperature gradient in the liquid phase, and that the Hg diffusion coefficient is much smaller
in solid than in liquid HgCdTe. Neglecting the out-diffusion of mercury atoms at the surface
influences the appearance of the excess surface Hg concentration and the concentration
minimum under the conditions of mercury segregation.

4.3 Concentration profile of Hg in a presence of &flusion, out-&flusion,

and segregation
Fig. 3 shows depth profiles of the Hg concentration calculated for different values of diffusion
parameters and the segregation coefficient. Location of the Hg minimum concentration is
slowly moved toward the surface dependent on Hg losses, and diffusion and segregation
coefficients. It can be seen that the Hg excess at the surface depends both on out-diffusion

23'
294 M. SCEPANOVICand M. JEVTIC

10.5 t

A
t
(/I
.-
-4-

c
Is

+?
0
W

A
-4-

X
v
0

Fig. 3. Depth dependence of the Hg concentration as a function of diffusion (Do, z,,), surface mass
transfer (z,), and segregation (k') coefficients. The experimental data (A) are taken from [2]

and on segregation conditions. As we can see the proposed model gives results that are in quali-
tative agreement with experimentally observed phenomena: the presence of an excess of mer-
cury concentration at the irradiated surface and the concentration minimum near the surface.
According to the fact that the values of some parameters included in the model are not
known in liquid HgCdTe under nonequilibrium conditions, the fitting procedure could be
applied to attain a satisfactory quantitative agreement between calculated and experimental
results. The results presented by curve a (Fig. 3) show the best agreement with the experimental
data (A) given in [2], especially in the first 30 nm. However, the slope of the calculated curve
is somewhat steeper than for the experimental data [2]. It could be a consequence of the
constant segregation coefficient, i.e., the neglect of its dependence on solidification velocity.
The effect of the nonconstant segregation coefficient will be the subject of a separate analysis.

4.4 Influence of the absorbed laser energy density on Hg concentration profile

The results in Fig. 4 show the influence of the absorbed laser energy determined by laser
absorption coenicients for the liquid and solid phases (a,and a,)on the concentration
profiles of Hg. Parameters of diffusion, out-diffusion, and segregation are the same for all
cases, and equal to that shown by a solid line in Fig. 3. It can be seen that the increase of
absorbed energy drastically enlarges the thickness of the Hg depletion layer, and decreases
the surface concentration of Hg. The width of the concentration minimum increases with
increasing absorbed laser energy. The same effect can be obtained by increasing the
irradiation intensity assuming constant values of the absorption coefficients. No sufficient
experimental results are available by which we can verify these results. However, the liquid
convection must be included into the model to examine in detail the effects of intensive
laser irradiation on the mercury concentration profile [9].
Numerical Simulations of Mercury Diffusion in HgCdTe during Laser Annealing 295

9.0

8.5

8.0

7.5

7.0

6.5

6.0
0 10 20 30 40 50 60 70 80

depth (nm)
90

- 100 110 120 130 140

Fig. 4. Influence of absorbed laser energy on depth profiles of Hg assuming the equality of the absorption
coefficients for the liquid (sl,) and solid (01,) phases

5. Conclusion
We proposed a simulation model based on ordinary and thermal diffusion, out-diffusion,
and segregation of Hg atoms during laser annealing of HgCdTe. The results obtained by
the developed model explain the appearance of a minimum of Hg concentration located a
few tens of monolayers below the top of the surface, and the excess of Hg atoms at the
irradiated surface. The observation that the higher Hg losses due to higher absorbed laser
energy cause a thicker Hg deficient layer rather than a lower Hg content at the point of
minimum Hg concentration, is also in agreement with experimental results.
By comparison of the obtained results with the published experimental data one can
estimate the values of some material parameters, such as diffusion, segregation, and mass
surface transfer coefficients, which are unknown under the laser annealing conditions,
especially in the liquid phase.

References
[l] A. L. DAWAR, SAVITAROY,R. P. MALL,and P. C. MATHUR,J. appl. Phys. 70,3516 (1991).
[2] C. N. AFONSO,M. ALONSO,J. L. H. NEIRA,A. D. SEQUEIRA, M. F. DA SILVA,and J. C. SOARFS,
J. Vacuum Sci. Technol. A 7, 3256 (1989).
[3] M. BROWNand A. F. W. WILLOUGHBY, J. Crystal Growth 59,27 (1982).
[4] M. M. J E V T I ~and M. J. S~EPANOVI~:,
Appl. Phys. A 53, 332 (1991).
[S] Yu. L. KHAIT,Semicond. Sci. Technol. 6, C84 (1991).
[6] A. MIOTELU)and L. F. DONADALLE ROSE,Phys. Letters A 87, 317 (1982).
[7] M. F. S. TANGand D. A. STEVENSON, J. Vacuum Sci. Technol. A 7, 544 (1989).
[8] J. M. POATE and J. W. MAYER(Ed.), Laser Annealing of Semiconductors,Academic Press, New York
1982.
[9] M. BERTI,L. F. DONADALLE ROSE,A. V. DRIGO,C. COHEN,J. SIWKA,G. G. BENTINI, and E. JAN-
NITTI, Phys. Rev. B 34, 2346 (1986).
(Received January 28, 1994; in revised,form April 11, 1994)