Home Search Collections Journals About Contact us My IOPscience

Electrical conduction by interface states in semiconductor heterojunctions

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2000 Semicond. Sci. Technol. 15 341

(http://iopscience.iop.org/0268-1242/15/4/307)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 128.59.222.12
This content was downloaded on 16/02/2015 at 21:39

Please note that terms and conditions apply.

Semicond. Sci. Technol. 15 (2000) 341–348. Printed in the UK PII: S0268-1242(00)09107-0

Electrical conduction by interface

states in semiconductor
heterojunctions
M El Yacoubi, R Evrard, N D Nguyen and M Schmeits
Institute of Physics, B5 University of Liège, B4000 Sart Tilman/Liège 1, Belgium

Received 1 November 1999, accepted for publication 17 January 2000

Abstract. Electrical conduction in semiconductor heterojunctions containing defect states in

the interface region is studied. As the classical drift-diffusion mechanism cannot in any case
explain electrical conduction in semiconductor heterojunctions, tunnelling involving interface
states is often considered as a possible conduction path. A theoretical treatment is made
where defect states in the interface region with a continuous energy distribution are included.
Electrical conduction through this defect band then allows the transit of electrons from the
conduction band of one semiconductor to the valence band of the second component. The
analysis is initiated by electrical measurements on n-CdS/p-CdTe heterojunctions obtained by
chemical vapour deposition of CdS on (111) oriented CdTe single crystals, for which
current–voltage and capacitance–frequency results are shown. The theoretical analysis is
based on the numerical resolution of Poisson’s equation and the continuity equations of
electrons, holes and defect states, where a current component corresponding to the defect
band conduction is explicitly included. Comparison with the experimental curves shows that
this formalism yields an efficient tool to model the conduction process through the interface
region. It also allows us to determine critical values of the physical parameters when a
particular step in the conduction mechanism becomes dominant.

1. Introduction of CdTe. Cadmium sulphide with a bandgap of 2.42 eV is a

useful window material for heterojunction solar cells.
The understanding of electrical conduction through semi- The first part of the present paper is devoted to the
conductor heterojunctions is of basic interest as many de- experimental analysis of the electrical characteristics of
vices such as heterojunction bipolar transistors, heterojunc- p-CdTe/n-CdS heterojunctions obtained by chemical vapour
tion LEDs and lasers as well as photodetectors or solar cells deposition (CVD) of CdS onto single-crystal CdTe substrates.
contain such junctions between two different semiconduc- In such crystalline systems, effects such as recombination at
tors. The efficiency of the conduction mechanism determines grain boundaries are avoided. Nevertheless, the difference
the performance of any device [1].
in crystalline parameters of about 10% is such that a
The electrical conduction in these systems cannot be in
high concentration of defects in the interface region is to
all cases interpreted by a classical drift-diffusion mechanism.
be expected. In particular, dislocations extending over a
Tunnelling through defect states in the interface region is
thousand Ångstroms or more are possible. Atomic diffusion
often suggested as a possible conduction path. The object of
from one side of the heterointerface to the other is also
the present paper is to propose a theoretical modelling of the
conduction mechanism by including defects in the interface possible.
region with a continuous energy distribution. The charge Results of electrical characterization of CdTe/CdS
carriers are then allowed to transit through this defect band heterojunctions have been reported in the literature [3–12].
from the conduction band of one component to the valence In all cases, it does not seem possible to explain the observed
band of the other component. characteristics by drift and diffusion as in p–n homojunctions.
The study is initiated by electrical measurements made Tunnelling of electrons from the CdS conduction band to the
on p-CdTe/n-CdS heterojunctions. Cadmium sulphide– CdTe valence band via interface defect states has often been
cadmium telluride heterojunctions have received much suggested to explain qualitatively the electrical conduction.
attention as they are potential candidates for low-cost, highly In the experimental part of this paper, we report results of
efficient solar cells. The cells are generally made by measurements of dc current–voltage (I –V ) characteristics, as
successive deposition of thin films of CdS and CdTe. Despite well as of capacitance–frequency curves taken on samples of
their polycrystalline structure, high efficiencies, up to 15%, p-CdTe/n-CdS heterojunctions. These measurements were
can be obtained [2]. These performances benefit from the performed at different temperatures, from room temperature
optimum gap of 1.45 eV and the high absorption coefficient down to 200 K.

0268-1242/00/040341+08$30.00 © 2000 IOP Publishing Ltd 341

M El Yacoubi et al

The interpretation of the results is made through a

theoretical analysis based on the assumption of the existence
in the interface region of defects with energy states in the
common bandgap. If their concentration is large, these states
may form a quasi-continuous energy distribution leading to
a defect-state band. This is a way to represent an imperfect
CdS overlayer grown on a CdTe substrate, with defect states
present on both sides of the interface. In the model used for
the calculations, the main contribution to the charge transport
in the interface region under forward-bias conditions is due
to electrons of the CdS conduction band which recombine on
defect states in the CdS bandgap; from there they transit by
conduction in the defect-state band to the CdTe side of the
heterointerface where they recombine with holes from the
p-CdTe valence band.
As will be shown, this numerical treatment constitutes Figure 1. Experimental current–voltage curves for a typical
an efficient way to simulate the various recombination, sample at five different temperatures.
tunnelling and hopping mechanisms, which are most
probably present. Due to the complexity of the mechanisms behaviour. The convention of sign is such that positive
occurring at a real heterojunction a complete description voltages correspond to forward biased heterojunctions.
including all intermediate steps of the current transport is Figure 4, which will be discussed later on, shows the band
not possible. Within the numerical modelling it is also structure constructed in the framework of the usual model
possible to simulate various configurations corresponding to based on the difference of electron affinities and on the
different parameter sets and to determine critical values of hypothesis of a neutral CdTe/CdS interface. Due to the
the parameters which determine basic macroscopic quantities relatively high energy barrier between the CdTe and CdS
such as those resulting from the electrical measurements. conduction bands, and the even higher barrier between the
The comparison between experimental and theoretical valence bands, the injection of minority carriers predicted
dc and ac characteristics shows that the model proposed in within this model should remain negligible at weak bias.
this paper can be useful for the study of electrical conduction This is in contradiction with our observations. In the case
in heterojunction systems. of the sample discussed here for instance, we measure at
room temperature a current of about 0.3 µA for a bias as
weak as 0.1 V. This shows that the current transport through
2. Experimental results
the heterojunction is certainly not due to thermal injection of
Without discussing all the experimental details on the minority carriers. This is confirmed by the examination of the
preparation conditions, let us just say that the CdS layers, behaviour of the I –V characteristics in terms of temperature.
about 0.2 µm thick, were grown by a simple chemical-vapour As shown in figure 1, the slope of the characteristic curves
deposition technique on single crystals of p-type As-doped is almost temperature independent up to 0.3 V of applied
CdTe; the substrates are oriented parallel to (111) planes. To voltage, in the linear region. This leads us to suggest, for this
improve the n character of the CdS layers, vapour of CdCl2 region, an I –V relation of the form
was introduced into the reactor during growth. Finally, we I = I0 exp(BV ). (1)
deposited In contacts onto the CdS surface of the samples
by thermal evaporation in a vacuum of 2 × 10−6 Torr. We We obtain that B is about 21 V−1 and almost completely
prepared ohmic contacts on the CdTe substrates by first independent of temperature. Therefore, the temperature
treating the samples for 10 minutes in a solution of lithium dependence is mainly contained in the pre-exponential factor
nitrate in water, with a concentration of 10−4 mol ml−1 . Then I0 .
we used electroless deposition of gold to obtain contacts of All this suggests that conduction mechanisms other than
relatively low resistance. These contacts have a surface of thermal injection of minority carriers must exist. Moreover,
about 1 mm2 . Measurements between two such contacts on the reverse voltage characteristics show no saturation,
the same substrate show that they are ohmic with a resistance indicating also that defect-assisted generation or tunnelling
of 100  at most. mechanisms occur. Above 0.3 V of applied voltage, the I –V
Acceptor concentrations in CdTe substrates, as curves deviate from the characteristic exponential behaviour,
determined by Hall measurements, are about 4 × 1016 cm−3 . which is indicative of a series resistance effect.
The donor concentration in the CdS layers, determined from In figure 2, we show for values of V equal to 0.1, 0.2 and
C–V measurements, is about 2 × 1016 cm−3 . 0.3 V, the current as a function of the inverse thermal voltage
We first measured the dc current–voltage characteristics q/kT . The results indicate that the conduction process is
of these samples between room temperature and 210 K. The thermally activated with an activation energy which has only
current was limited to 0.1 mA to avoid sample breakdown. a weak dependence on the applied voltage. For the three
In figure 1 we show these current–voltage characteristics values of V considered here, we find activation energies
at five temperatures ranging from 210 to 300 K, for one εa of respectively 0.32, 0.29 and 0.25 eV. This proves that
sample. The other samples show the same characteristic changing the forward bias does not appreciably modify the

342
 Electrical conduction by interface states

occupation can be described by a Fermi function ft (x, Et )

with a quasi-Fermi energy Ft for non-equilibrium situations.
The detailed formal developments can be found in [13] and
[14]. The occupation function for defect states at a given
energy Et is given by
1
ft = (2)
1 + g exp[(Et − Ft )/kT ]
where g is the degeneracy factor. The local concentration of
occupied states at a position x is given by
Z Ec
nt (x) = ft (Et )Dt (Et ) dEt (3)
Ev
Figure 2. Experimental current–inverse thermal voltage curve for
same sample as in figure 1, for values of the applied voltage equal
where Ev and Ec are the valence and conduction band edges.
to 0.1, 0.2 and 0.3 V. Integration over the bandgap of Dt (x) yields the local total
concentration of defect states Nt (x).
We assume the existence of a physical mechanism
allowing for a rapid spatial exchange of electrons between
neighbouring defect sites. In our model this is accounted for
by introducing a current in the defect band. We assume that
the current in the defect band is given by the expression

JEt = nt µt ∇F
E t (4)

where µt is the mobility of the electrons in the defect band.

This expression leads to a current proportional to the gradient
of the quasi-Fermi level Ft and is of course zero at thermal
equilibrium. The form is similar to that used for the electron
and hole currents in non-degenerate semiconductors [15].
Due to the conduction in this defect-state region, the
electrons come to the interface where they recombine with
Figure 3. Experimental capacitance–frequency curve for zero holes from the CdTe valence band. The detailed mechanism
bias and T = 300 K. of the electrical conduction by these defect states is not known
at the atomic scale. It may be hopping motion or tunnelling
band bending either in CdTe or in CdS. Therefore, the change between adjacent defects. What is important for our purpose
in voltage due to an external source is almost entirely applied is that equation (4) yields a tractable expression allowing us
to a relatively thin interface layer. to simulate the effect of the conduction by the defect states
We also have determined the ac-current characteristics in the numerical computation.
for various values of temperature and applied steady-state The electron mobility in the defect band µt is of course
voltage. As a typical result, we show in figure 3 for T = a parameter difficult to estimate, as it is not a directly
300 K and V = 0 V the capacitance–frequency (C–f ) measurable quantity. It will be shown that the results of the
curve of the CdTe–CdS junction. The C–f curve shows calculations are not very sensitive to the value chosen for
a decrease of the capacitance from 650 pF to 250 pF in the µt , provided that it is not too small. One reason is that the
102 –104 Hz range and from 250 to 150 pF in the 104 to 106 Hz conduction paths in the defect band are restricted to the near
frequency range. A first analysis suggests the presence of interface region, and thus are quite short. Another reason
a continuous distribution of defect states, with a resulting is that the capture and emission of charge carriers by the
continuous distribution of cutoff frequencies, leading to a traps are the dominant mechanisms limiting the current in
progressive decrease of the response of the defect states to the junction.
the applied ac voltage. The theoretical analysis of the junction is based on the
numerical resolution of a set of four equations coupling the
3. Theoretical analysis electrical potential ψ, the electron and hole concentrations n
and p and the concentration of occupied states nt . These are
The theoretical study is based on the numerical resolution of Poisson’s equation and the continuity equations for electrons,
Poisson’s equation and the continuity equations for electrons, holes and occupied defect states.
holes and occupied defect levels [13, 14]. The experimental This latter equation is given by
results lead us to propose the following model to explain the ∂nt 1E E
charge transport through the junction. Defect states located in = Rnt − Rpt + ∇ · Jt (5)
∂t q
the interface region form a continuous distribution of energy
levels, with a density of states per unit energy Dt (x, Et ). The where Rnt is the transition rate for exchange of electrons
defect states are supposed to interact strongly, so that their with the conduction band and Rpt that for exchange of holes

343
M El Yacoubi et al

with the valence band. We assume that these exchanges

can be described by the usual expressions resulting from
capture and emission according to the Shockley–Read–Hall
recombination scheme [14]. Then Rnt is the transition rate
resulting from the difference between the integrated electron
capture rate and the integrated electron emission rate
Z Ec
Rnt (x) = rnt (Et , x) dEt (6)
Ev

with
rnt (Et ) = cn n(1 − ft )Dt − en ft Dt . (7)
Similarly, the defect to valence band transition rate Rpt (x)
can be written with rpt (Et ) given by
Figure 4. Energy band diagram for CdTe-p/CdS-n heterojunction
rpt (Et ) = cp pft Dt − ep (1 − ft )Dt . (8) at V = 0.2 V applied voltage. The interface defect band extends
from x = −0.01 to 0.2 µm. The total defect concentration is
In the above relations cn and cp are the electron and Nt = 1017 cm−3 for x = −0.01 µm to 0.1 µm, and goes linearly
hole capture coefficients, related to the electron and hole to zero from x = 0.1 to 0.2 µm. Nt = 0 outside that region. The
p
capture cross sections by cn = vth n
σn and cp = vth σp , defect density of states Dt (x, Et ) is constant as a function of Et ,
where vth are the respective thermal velocities. The relations over a width 1Et = 0.7 eV.
between cn , cp and the electron and hole thermal emission
rates en and ep are given by the following equations, Figure 4 gives the result of our calculations for the
expressing microreversibility between individual states at energy-band diagram in the interface region, and a steady-
thermal equilibrium. state applied voltage of V = 0.2 V. In the range of values
of x represented in the figure, i.e. from −0.2 to 0.4 µm, the
en = gcn Nc exp[(Et − Ec )/kT ] (9) electron and hole quasi-Fermi levels Fn and Fp are nearly
cp constant and separated by 0.2 eV. They join for x-values
ep = Nv exp[(Ev − Et )/kT ]. (10) outside the figure as usual. The defect quasi-Fermi level
g
Ft is nearly flat in the whole defect region, with a position
In the numerical study of the p-CdTe/n-CdS heterojunc- intermediate between Fn and Fp , indicating that, for the
tions we have taken the dopant concentrations respectively chosen set of parameters, the defect states are in equilibrium
equal to NA = 4 × 1016 cm−3 and ND = 2 × 1016 cm−3 , in with neither of the two bands.
accordance with the experimental values. As the system has In figure 5 we show, for the same value of the applied
a planar symmetry, we use a one-dimensional analysis, where voltage, the recombination rates Rnt and Rpt as functions of
the distances from the interface are measured by a variable x, the position x, alongside with the different current densities
with x = 0 at the position of the heterointerface. The x-axis Jn , Jp and Jt . These figures clearly reveal the conduction
is oriented in the direction of the CdS epitaxial film. mechanism of the junction. When the electrons injected from
The first results we show are obtained with the following the n-CdS side reach the defect region, they are captured by
choice of parameters. We take the conduction band the defect traps. These capture transitions take place for x
discontinuity 1Ec equal to 0.22 eV, the value estimated from between 0.15 µm and 0.20 µm, i.e. in the region where the
the difference in electron affinities of CdTe and CdS [11]. concentration of traps has not yet reached its maximum value.
All other parameters describing bulk properties correspond As figure 5(a) shows, the capture rate of electrons by defects
to common values of CdTe and CdS [15]. The defect band Rnt is maximum near the boundary of the region with defect
mobility is taken equal to 10−3 times the electron mobility in states in the CdS layer and decreases rapidly when going
CdTe. For the electron and hole capture cross sections, we towards the interface with CdTe. After capture, the electrons
use energy-independent values σn = σp = 5 × 10−15 cm2 . are swept through the defect band up to the heterointerface at
The defect states are supposed to be acceptor-like, in the sense x = 0, where they recombine with the holes on the p-CdTe
that they are neutral when empty and negatively charged when side of the junction. As figure 5(b) shows, below 0.15 µm
occupied. the current is almost entirely due to the transport of electrons
For the defect distribution function Dt , we choose a via trap states. Indeed the total current density J is such that
constant value in a band of width 1Et = 0.7 eV located J ∼ = Jt for this CdS region close to the interface. In the
0.21 eV below the CdS conduction band edge, 0.54 eV below small layer of 0.01 µm where the defect layer is allowed to
the CdTe conduction band edge. The x-dependence of the extend on the CdTe side in our model, the current transforms
density of states is such that Nt (x) is constant and equal from a defect dominated contribution Jt to a hole current Jp ,
to 1 × 1017 cm−3 for values of x between −0.01 µm and due to the strong hole transition rate Rpt which appears in
+0.1 µm. Beyond this last value, Nt decreases linearly to figure 5(a). Beyond this x value, the current is transported
zero between +0.1 and +0.2 µm. This simulates a defect by holes exclusively.
band existing on both sides of the interface, with a larger This conduction mechanism is of course much more
extent in the CdS layer, as can be expected in the case of efficient than a thermally activated emission over the energy
films grown on substrates with a large lattice mismatch. barrier. Indeed, for an electron, the energy barrier is of the

344
 Electrical conduction by interface states

Figure 6. Theoretical current density J as function of inverse

thermal voltage q/kT for the system described in figure 4, for
three values of applied voltage.

Figure 5. (a) Conduction band to defect state transition rate Rnt

and defect state to valence band transition rate Rpt , as a function of
x, for forward biased junction, with V = 0.2 V. (b) Current
densities for electrons Jn , holes Jp and defect band Jt , and total
current density J for applied voltage V = 0.2 V.

order of the built-in potential plus 1Ec , the conduction band Figure 7. Current density at T = 300 K, V = 0.2 V applied dc
voltage for defect concentration Nt as described in figure 4, for
discontinuity, i.e. about 1 eV. The barrier is even higher different values of the parameter sets εµt = µt /µt0 , with
for the holes. In contrast, within our model, the part of µt0 = µn (CdTe) × 10−3 , εn = σn (CdS)/σn0 , σn0 = 1 × 10−15 cm2
the process that is thermally activated is the transport of and εp = σh (CdTe)/σp0 , with σp0 = 10−15 cm2 . When one
electrons and holes from their respective ohmic contact to parameter is modified, the two others are kept equal to their
the recombination region, which is close to the limits of standard values.
the defect region. This requires little thermal activation,
typically a few hundreds of meV, as will appear from 1 mm2 , leading to current densities of the same order of
the analysis of the current–voltage values for different magnitude as those of figure 6.
temperatures. The conduction process depends on a series of
Figure 6 shows the logarithm of the calculated total parameters which describe the different steps and which are
current densities J as functions of the inverse thermal voltage difficult to estimate from basic principles or to obtain from
q/kT , for three values of the applied voltage V = 0.1, 0.2 direct experimental measurements. Therefore we show the
and 0.3 V. Notice that, for this temperature range, log J is a effect of some of these parameters on basic dc characteristics.
linear function of q/kT . The activation energies which can In figure 7, we show the resulting current at 300 K and
be deduced from the slope of the functions are respectively 0.2 V of applied voltage as a function of the defect band
0.34 eV for V = 0.1 V, 0.30 eV for V = 0.2 V and mobility µt , the electron capture cross section σn of CdS and
0.26 eV for V = 0.3 V. As can be inferred from figure 4, the hole capture cross section of CdTe σp . The abscissa are
these activation energies correspond to mean values of the normalized values εµt , εn and εp of the three parameters with
barrier felt by the electrons on the n-CdS side of the junction. the set of parameters of figure 4 taken as reference values.
The thermal activation of electrons is the transport-limiting Modifying one of the three values εµt , εn or εp , keeping
mechanism as the valence band edge on the CdTe side is all others fixed, yields the three curves J (ε). They show
nearly flat. that the modification of the defect band mobility µt only
Comparing with the experimental results shown in leads to a sizeable effect when µt is reduced by at least
figure 2, one retrieves the linear slope of the log I –q/kT a factor 10−4 . Decreasing the electron and hole capture
curves, with similar values of the activation energies. Even cross sections σn and σp relative to the reference value
the absolute values of the currents are comparable. Indeed produces important modifications of the resulting current, as
the junctions studied experimentally have sections of about expected. Increasing σp with respect to the reference value

345
M El Yacoubi et al

(a)

Figure 8. Activation energy εa at V = 0.2 V applied dc voltage,

T = 300 K as a function of x 2 , the width of the defect region on
the CdS side of the junction. The three curves correspond to the
same value of the integrated defect density Nt , but with different
width of the defect density of states function 1Et . The
experimental activation energy is εa = 0.29 eV.

modifies only slightly the current value, whereas increasing

the electron capture cross section σn of CdS produces a
further increase of the current. This illustrates that within this
parameter region, the filling of the defect states is effectively
the current-controlling part of the conduction mechanism.
Other important parameters describing the defect states
(b)
are those related to the defect distribution function Dt , with
its energy dependence and its spatial extension. The effect Figure 9. (a) Electron, hole and occupied defect state densities at
of the extension x2 of the defect region on the CdS side is steady state for V = 0 V (dotted curves) and real part of small
signal amplitude of low-frequency ac component of the same
illustrated in figure 8, where three different values of the quantities, as a function of position x. (b) Real and imaginary part
defect density of states width 1Et are considered. The of electron, hole, displacement and defect current densities as a
integrated value Nt (x) and all other parameters are kept function of position x.
equal to those of the reference system. For V = 0.2 V
of applied voltage and 1Et = 0.5, 0.7 and 0.7 eV, we concentration for example can be written
have determined the thermal activation energy resulting from
the computed values of the current as a function of 1/T , n(x, t) = n0 (x) + ñ(x) eiωt . (11)
modifying the value of x2 between 0.1 µm and 0.3 µm. The
In figure 9(a) we show for V = 0, as a function of
curves show that in all cases, the activation energies decreases
position x, the steady-state concentrations of electrons, holes
when x2 increases, which is a result of the modification
and occupied defect levels nt , together with the real part
of the shape of the conduction band edge on the CdS side
of the low-frequency response of the same quantities. At
of the junction. The value of 0.3 eV suggested by the
this frequency, the imaginary part is orders of magnitude
experiment, is best reproduced with x2 = 0.2 µm and a
weaker. The figure shows that the modification of the electron
defect band width of 1Et = 0.7 eV. Considering other and hole concentration at the depletion layer edges occurs
values of the applied voltage yields similar agreement for while the defect concentration is modified, but with opposite
the corresponding activation energies. These results justify sign. The x-integration of the real part of the ac-value
the choice of parameters we have used for figure 4 and the of n(x) − nt (x) − p(x) yields zero, thus ensuring charge
resulting curves. Due to the large number of parameters, conservation and electrical neutrality.
several combinations of parameters could yield valuable Figure 9(b) shows the real and imaginary parts of the
comparison with experiment. As shown by figure 8, taking different current components, namely the electron and hole
1Et = 0.9 eV instead of 0.7 eV could be another plausible currents Jn and Jp , the defect band current Jt , as well as
solution. the displacement current JD . This latter makes that the
In order to determine the capacitance–frequency curves, total current is constant as a function of x. The real parts
we have performed a small-signal ac analysis which consists correspond to the resistive part of the current and have the
in solving the same set of basic equations in response to an same slope as the dc components already shown in figure 5.
applied voltage consisting of a steady-state component V0 The imaginary parts correspond to the capacitive part of the
plus a harmonic component of amplitude ṽ and frequency current and illustrate well the respective filling and emptying
ω = 2πf . Within the small-signal approximation, all of the conduction band, valence band and defect states
quantities can be written as the sum of a steady-state dc as appears in figure 9(a). This already indicates that the
component and a sinusoidal ac component [14]. The electron capacitance, which is determined by the modification of the

346
 Electrical conduction by interface states

voltage. The figure shows that below the reference set of

parameters, it is the hole capture cross section σp in CdTe
which allows us to modify the cutoff frequency. Above the
reference value, both electron and hole capture cross sections
allow us to increase the cutoff frequency. The defect band
mobility must be modified by at least a factor of 103 to
produce a reduction of the response of the hole defect state
system.
Comparison with the experimental C(f ) curve requires
great care. It is effectively difficult to reproduce
experimentally the steplike shape with a constant value of
the C(f ) function beyond the cutoff frequency [16]. In our
case, the fact that the experimental curve of figure 3 is not
reproduced by one of the shown theoretical curves seems
to indicate that the real system is even more complicated
than we assumed. Beside the states that we have included
Figure 10. Theoretical capacitance–frequency curves for V = 0,
to explain the dc-current behaviour, there must exist another
T = 300 K for CdTe-p/CdS-n heterojunction as described in distribution of defect states whose population is modified
figure 4 (curve a), values of the relative defect mobility εµt = 0 when an ac voltage is applied to the system. They should
(curve a), 10−6 (b), 10−4 (c), 3 × 10−4 (d), 10−3 (e) and 1 (f). yield a contribution to the capacitance, without participating
in the dc conduction mechanism. This could for example
be described by a mechanism where one of the three steps
occurring in our model of the current transport is forbidden.
The C(f ) curve a in figure 10 shows such a behaviour.

4. Conclusion

We have presented a way of modelling electrical conduc-

tion mediated by interface states in a semiconductor hetero-
junction. The analysis starts from an experimental study of
electrical characteristics of CdTe/CdS heterojunctions. The
interpretation was made by a numerical analysis based on a
model including conduction through a band of defect states
in the interface region. In a forward biased junction, the elec-
trons are injected from the CdS-n side of the junction into the
Figure 11. Theoretically determined cutoff-frequency ωc for
defect band, which performs the current transport towards
different values of the relative value of the parameters εµt , εn , εp
as defined in figure 7. the CdTe-p valence band.
This model correctly explains conduction in the dc
space charge with applied voltage will depend on the way the regime, at least for the low-voltage linear region. The analysis
charge on the defect states is modified. Usually this occurs of our model shows that the electrons are captured by the
by only local exchange of electrons with the conduction band defect states relatively far from the interface with CdTe.
or of holes with the valence band, but here displacement of This explains the small value of the activation energy in
mobile electrons inside the defect band contributes too. the mechanism of charge transport, as well as the fact that
This is further illustrated in figure 10, where we show the slopes of the curves giving log I against V are almost
the calculated C(f ) curves for different values of the defect- independent of temperature.
band mobility µt . It is seen that when µt is set equal to zero, The experimentally obtained C–f behaviour suggests
the C(f ) curve shows a continuous decrease up to 105 Hz. the presence of a distribution of defect states with varying
Increasing the mobility µt produces an increasing value of energy position inside the gap. The theoretical study allows
the low-frequency capacitance and a steplike shape with a us to analyse a defect dependent behaviour of the ac response
cutoff frequency moving towards higher frequencies when µt of the junction. It suggests that the model should be refined
increases. In figure 11 we show in a way similar to figure 7, by adding a second group of interface states which would
the cutoff frequencies as a function of the relative values of interact with the bands without participating in the defect
the defect band mobility εµt , the CdS electron capture cross band conduction.
section σn and the CdTe hole capture cross section σp . The The results show that the inclusion of a current in the
cutoff frequency is defined as the frequency for which the defect state band represents an efficient tool in the study of the
capacitance takes the mean value between the low-frequency electrical characteristics of semiconductor heterojunctions,
capacitance and the high-frequency capacitance. In the case where a full treatment taking into account the presence of
of discrete bulk defects, it corresponds to the frequency above interface states is otherwise nearly impossible, if one wants
which the defect states no longer respond to the applied to describe a real system.

347
M El Yacoubi et al

Acknowledgment [6] Ercelebi C, Brinkman A W, Furlong T S and Woods J 1990

J. Cryst. Growth 101 162
Financial support of the Belgian Fonds National de la [7] Ringel S A, Smith A W, MacDougal and Rohatgi A 1991
Recherche Scientifique (contract 9.4565.96F) is gratefully J. Appl. Phys. 70 881
acknowledged. [8] Morris R G and Das S K 1992 Int. J. Solar Energy 12 95
[9] Brinkman A W, Al Allak H M, Arwan G R, Brown P D,
Durose K, Ercelebi C, Simmons M V and Woods J 1992
References Int. J. Solar Energy 12 233
[10] Das S K 1993 Solar Energy Mater. Solar Cells 29 227
[1] Morgan V and Williams R H (eds) 1991 Physics and [11] Tomita Y, Kawai T and Hatanaka Y 1993 Japan. J. Appl.
Technology of Heterojunction Devices (IEE Materials and Phys. 32 1923
Devices Series 8) (London: Peregrinus) [12] Bayhan H and Ercelebi C 1997 Semicond. Sci. Technol. 12
[2] Partain L D1995 Solar Cells and their Applications (New 600
York: Wiley) [13] Sakhaf M and Schmeits M 1996 J. Appl. Phys. 80 6893
[3] Ou S S, Stafsudd O M and Basol B M 1984 Solid State [14] Schmeits M 1997 Solid State Commun. 12 1217
Electron. 27 21 [15] Sze S M 1981 Physics of Semiconductor Devices (New York:
[4] Anthony T C, Fahrenbuch A L, Peters M G and Bube R H Wiley)
1985 J. Appl. Phys. 57 400 [16] Chen J F, Chen N C, Wang P Y and Tsai M H 1996 J. Appl.
[5] Chu T L, Chu S S and Ang S T 1988 J. Appl. Phys. 64 1233 Phys. 81 1255

348