Evidence for band-to-band impact ionization in evaporated
alternating-current
thin-film electroluminescent
devices
ZnS:Mn
W. M. Ang, S. Pennathur, L. Pham, J. F. Wager, and S. M. Goodnick
Deparhnent of Electrical and Computer Engineering, Centerjk Advanced Materials Research,
Oregon State University, Cowallis, Oregon 97331-3211
A. A. Douglas
Planar Systems,inc., Beaverton, Oregon 97006
(Received 25 August 1994; accepted for publication 22 November 1994)
Evidence is presented that the normal operation of evaporated ZnS:Mn alternating-current thin-film
electroluminescent (ACTFEL) devices involves electron-hole pair generation by band-to-band
impact
ionization. Four observations are offered to support this assertion. These observations
involve: (i) empirical field-clamping trends, (ii) experimental and simulated trends in charge transfer
characteristics, (iii) experimental attempts to assess the interface distribution using a field-control
circuit, and (iv) Monte Carlo simulation trends. Furthermore, the absence of overshoot in measured
capacitance-voltage and internal charge-phosphor field curves indicates that a majority of the holes
created by impact ionization are trapped at or near the phosphor/insulator interface. The
multiplication factor (i.e., the total number of electrons transferred across the phosphor divided by
the number of electrons injected from the phosphor/insulator cathode interface) is estimated, from
device physics simulation of experimental trends, to be of the order 4-8 for evaporated ZnS:Mn
ACTFEL devices operating under normal conditions. 0 1995 American Institute of Physics.
I. INTRODUCTION
The issue of whether electron-hole generation by bandto-band impact ionization occurs in ZnS:Mn alternatingcurrent thin-film electroluminescent (ACTFEL) devices has
been a subject of some debate for almost as long as the
device physics of modern metal-insulator-semiconductor
(i.e., phosphor)-insulator-metal (MISIM) devices has been
investigated.lm5 Smith considered’ the possibility of band-toband impact ionization existing in ZnS:Mn ACTFEL devices
because of the presence of the so-called “blue flash” known
to occur during the portion of the applied voltage waveform
in which conduction current flows. However, he felt that,
because the threshold voltage of the ACTFEL could be modified from l-2 MVlcm by the type of insulator or insulator
deposition employed, since the quantum efficiency of the
blue flash did not exhibit any increase with increasing threshold voltage, and because the rate of tunnel emission exhibits
such an abrupt threshold, there was no compelling reason to
believe that band-to-band impact ionization plays an important role in the operation of ZnS:Mn ACTFEL devices. In
contrast, Howard et al. interpretzT3 the blue flash as strong
evidence for electron-hole creation via band-to-band impact
ionization and assert that hole creation is necessary for explaining hysteresis in heavily Mn-doped ZnS ACTFEL devices. Yang et al. point out4,5 that the origin of hysteresis
does not mandate the existence of holes; in fact, Yang et al.
prefer to invoke hot electron impact ionization of an unspecified deep level as a more likely process leading to hysteresis
and provide experimental evidence for their preference. Furthermore, they comment that the blue flash is quite broad
band (actually, “white flash” would be more accurate) and
that the origin of this emission is not necessarily associated
with band-to-band impact ionization of electron-hole pairs
(one possibility is that it is due to intervalley radiative reJ. Appl. Phys. 77 (6), 15 March 1995
combination exclusively within the conduction band#). All
of the discussion up to this point is with regard to rather
early work which primarily focused on heavily Mn-doped,
hysteretic ACTFEL devices.
More recently, Thompson and Allen assessed7the bandto-band impact ionization rate in single-crystal ZnS from dc
photocurrent measurements obtained using Schottky barrier
devices. They found a large increase in the photocurrent at
-2 MVlcm which they attributed to band-to-band impact
ionization and deduced the following expression for the impact ionization coefficient
an=aoexp
i
--
b
FP 1
,
(1)
where b = - 13 MV/cm and Fp is the phosphor field. Since
the measured threshold field for actual ZnS:Mn ACTFEL
devices is found to be in a typical range of - 1.5-1.75 MVI
cm, Thompson and Allen’s analysis would seem to be inconsistent with band-to-band impact ionization playing an important role in the operation of ZnS:Mn ACTFEL devices.
However, this conclusion must be tempered by two considerations. First, in the measurements of Thompson and Allen
the onset of band-to-band impact ionization is obscured by
the presence of an initial rise in the photocurrent at lower
fields, which they attribute to deep level impact ionization.
Thus, the threshold for band-to-band impact ionization may
be indeed less than that estimated by Thompson and Allen
because of the obscuration of the deep level response. Second, the measured threshold field is actually the average field
across the phosphor layer and, if space charge is present due
to any type of impact ionization in the ACTFEL phosphor
region, this measured, average threshold field represents a
lower estimate of what the actual field is in the phosphor at
the cathode side; the cathode field is the maximum field in
OOh-8979/95/77(6)/2719/6/$6.00
Q 1995 American Institute of Physics
2719
Downloaded 07 Mar 2001 to 128.193.48.142. Redistribution subject to AIP copyright, see http://ojps.aip.org/japo/japcpyrts.html
the phosphor and is the relevant field for assessing the
threshold for interface state emission and for determining the
actual threshold for any type of impact ionization that occurs
in an actual ACTFEL device.
Cattell et al. also deduced’ an expression for the impact
ionization coefficient in single crystal ZnS samples from dc
measurements and found a form for the impact ionization
coefficient which differs somewhat from that found7 by
Thompson and Allen
’
where their estimates for era and b are 1.25X105 cm-’ and
1.96 MV/cm.
Bringuier wrote Eqs. (1) and (2) in a generalized
manner’ and provided a more explicit form for the impact
ionization coefficient prefactor as follows:
-2
-1
0
Phosphor
(4
1
2
Field [W/cm]
T2
3241
where Ei is an effective ionization energy. Bringuier then
used Eq. (3) with E, equal to 1.5 of the ZnS band gap,
b= 1.7X 106 V/cm, and p2= 1 or 2 in his device physics
based simulation of the operation of a ZnS ACTFEL device
when electron multiplication is present. He considered two
situations in which electron multiplication can occur, distinguished by whether the impact-ionized holes recombine
quickly or slowly. He predicted that an anomalously steep
charge-voltage (Q , ,2- V,, ; note that what we denote Q,,,
in this article is equivalent to what Bringuier refers to as
Q1,2r as discussed later) curve (with a slope greater than the
insulator capacitance) may be observed under a situation of
high multiplication rates and slow hole recombination. We
are aware of no reports in the literature which confirm this
prediction.
Mueller estimated” that for each electron tunnel injected
from the cathode interface of a ZnS:Mn ACTFEL device,
7.6X106 electron-hole pairs are generated by band-to-band
impact ionization in a lOOO-nm-thick phosphor layer. This
estimate is based on the assumption that electrons travel ballistically across the phosphor layer.” Singh et al. have
providedI a more conservative estimate of 2000 for the carrier multiplication factor in such a phosphor layer if it is
assumed that the electrons transit the phosphor layer ballistically. Furthermore, Singh et al. dismiss the possibility of
having such large carrier multiplication factors in real
ZnS:Mn ACTFEL devices since such a large increase in the
carrier density would be readily observable in the electrical
characteristics of such devices. Singh et al. could find no
experimental evidence for a large carrier multiplication factor and, thus, concluded that the carrier multiplication factors
are relatively small in ZnS:Mn ACTFEL devices, if present
at aLl.
The purpose of the present article is to provide evidence
that band-to-band impact ionization indeed does occur in
evaporated ZnS:Mn ACTFEL devices. The evidence presented arises from electrical characterization and modeling
observations and Monte Carlo high field transport modeling.
Considering all of the evidence offered herein, as well as
2720
J. Appt. Phys., Vol. 77, No. 6, 15 March 1995
$0
0”
z
-,
--------.
\
E,
L
220
-2
----,240\
-260\
-3
-4
-3
-2
-1
Phosphor
(W
0
1
2
3
4
Field [W//cm]
FIG. 1. Q-FP curves for (a) a yellow, evaporated ZnS:Mn ACTFEX device
and (b) a blue, CaGa& :Ce ACTFEL device. V,,,, is approximately 20, 40,
and 60 V above threshold for the Q-F, curves shown.
evidence from previous researchers, we believe that there are
compelling reasons to believe that electron-hole pair generation by band-to-band impact ionization occurs in evaporated
ZnS:Mn ACTFEL devices and that the carrier multiplication
factor is significantly greater than one.
II. EVIDENCE
IONIZATION
A. Q-F,
FOR BAND-TO-BAND
IMPACT
field-clamping
From an electrical characterization point-of-view, we
have found evaporated ZnS:Mn ACTFEL devices to constitute a prototypical (almost ideal) ACTFEL device.‘3-*5 In
particular, evaporated ZnS:Mn ACTFEL devices exhibit
strong field-clamping, symmetrical electrical characteristics
with respect to the applied voltage polarity, and capacitancevoltage ( C-V) characteristics in which the insulator capacitance is accurately determined and the C-V transition is
abrupt (indicating that the interface state density is large and
abrupt).
Focusing on the field-clamping property of these devices, Fig. 1 illustrates a comparison of internal charge versus phosphor field (Q-F,) curves for an evaporated ZnS:Mn
(yellow) and a CaGa,S,:Ce (blue) ACTFEL device.16 Note
the dramatic difference in these curves in the steady-state
Ang et
Downloaded 07 Mar 2001 to 128.193.48.142. Redistribution subject to AIP copyright, see http://ojps.aip.org/japo/japcpyrts.html
al.
field (F,,) regions (where field clamping is manifest, if field
clamping indeed does occur) which are the portions of the
Q-Fp curve labeled BC and GH in Fig. 1. The turn-on transition for the CaGa&:Ce ACTFEL device is rather soft and
the field increases rather gradually in the B to C and G to H
portions of the Q-Fp curve, presumably because progressively deeper and deeper interface state electrons are emitted.
In contrast, the ZnS:Mn ACTFEL device exhibits an exceedingly abrupt Q-F, turn-on transition and nearly vertical BC
and GH steady-state-field regions which indicates that strong
field-clamping occurs.
Previously,16 we have interpreted the abrupt Q-F,
turn-on transition and the nearly vertical steady-state field
region observed in evaporated ZnS:Mn ACTFEL devices as
evidence for a large and abrupt density of interface states at
the phosphor/insulator interface. In contrast to this previous
interpretation, we now believe that the strong field clamping
observed in evaporated ZnS:Mn ACTFEL devices arises as a
consequence of electron multiplication via band-to-band impact ionization. Also, we now believe that the Q-F, curve
shown in Fig. l(b) for the CaGa#,:Ce ACTFEL device is
typical of an ACTFEL device in which electron emission
occurs from an interface with a moderate to large density of
interface states; our present perspective is that field clamping
is not exhibited in this CaGa&:Ce ACTFEL device primarily because of the absence of electron multiplication concomitant with impact ionization. Thus, we believe that the
exceedingly strong field-clamping characteristics that we observe in evaporated ZnS:Mn ACTFEL devices are one piece
of evidence that band-to-band impact ionization occurs in
this device. Further evidence for this conclusion is the subject of’the remainder of this article.
B. Slope
of the O,,,-V,,
characteristics
As noted in the introduction, Bringuier predicts9 that an
anomalously steep charge-voltage (Q,,-V,,)
curve (with a
slope greater than the insulator capacitance) may be observed
under a’situation of high multiplication rates due to band-toband impact ionization in which the subsequent hole recombination occurs slowly. We have measured Q,,-V,,
curves
for evaporated ZnS:Mn ACTFEL devices and also have
simulated such curves using Bringuier’s device physics
ACTFEL models.“,r7
Before discussing the experimental results, we must
clarify what is meant by a Q,,-V,,,
curve. Q,,, is defined
in the Q-F, curve shown in Fig. l(a) and corresponds to the
net charge located at the interface with respect to the flat
band or neutral charge (i.e., the amount of charge in the
interface states when the interface is charge neutral).17 Note
that what we call Q,, corresponds to what Bringuier9 denotes Qt,z. Now that Q,, is defined, it is clear that a
Qmx-Vmax curve may be obtained by measuring a set of
Q-F, curves at various IJ’,,,~‘s, assessing emax’s from these
Q-F, curves, and plotting Q,, vs V,, .
Experimentally measured Q,,-V,,
curves for a
ZnS:Mn ACTFEL device at 77 and 300 K temperatures are
shown in Fig. 2. Curves (a) and (b) are obtained at a temperature of 77 K and correspond to applied voltage waveform frequencies of 1000 and 100 Hz, respectively. Also inJ. Appl. Phys., Vol. 77, No. 6, 15 March 1995
0.8
%0 0.6
3
iii
k!
0.4
0
0.2
0
160
170
180
160
Vmax M
FIG. 2. Q,,,-V,,
curvesfor an evaporated
ZnS:Mn ACTFEL device at (a)
77 K and 1000 Hz, (b) 77 K and 100 Hz, and (c) 300 K and 1000 Hz.
dicated in Fig. 2 are the slopes of each of these curves (or the
straight line portions of these curves), which correspond to
capacitances, so that the slopes are expressed in farads, the
.~
unit of capacitance.
First, focus on the 77 K, 100 Hz curve of Fig. 2 [i.e.,
curve (b)] which has a slope of 2.7 nF. The physical insulator
capacitance of this device is 1.4 nF, as assessedfrom known
thicknesses and dielectric constants of the insulating layers
and also, independently from C-V measurements.r3 Thus,
he Qmx-Vmax slope of the 77 K, 100 Hz curve is substantiahy greater than the physical capacitance of the insulator.
According to the theory of Bringuier,9 the slope of a
Q,,,-~TIIl, curve may be only less than or equal to the insulator physical capacitance unless electron multiplication due
to impact ionization occurs during device operation. We consider this deviation between the measured Qmax-Vmaxslope
and the physical capacitance to be our strongest evidence for
impact ionization. Furthermore, Bringuier contends” that a
large deviation between the insulator physical capacitance
and the experimental Q,,-V,,
slopes can be accounted for
only by postulating both that impact ionization occurs and
that the impact-induced holes recombine very slowly with
electrons when they reach the interface; if the hole recombination is fast, the Q,,-IJm, slope is not substantially greater
than the insulator physical capacitance.
Next, consider the 77 K, 1000 Hz Q,,,-V,,,
curve
shown in Fig. 2 [i.e., curve (a)]. In addition to the 2.5 nF
slope of the 1000 Hz curve (which is very similar to the 2.7
nF slope of the 100 Hz curve), another very large slope of
magnitude 20.7 nF is evident over a very small (i.e., -2 V)
range of V,, near threshold. Furthermore, notice that the
300 K, 1000 Hz curve [i.e., curve (cj] exhibits a similar
two-slope behavior as the 77 K, 1000 Hz curve [i.e., curve
(a)], but that the 300 K curve is more washed out than the 77
K curve.
The slopes of extremely large magnitude (i.e., -20 I-$)
observed in the 1000 Hz Q,,,-V,,
curves shown in Fig. 2
are attributed to the very slow rate of recombination of holes
with electrons when they reach the cathode interface. SupAng et
al.
Downloaded 07 Mar 2001 to 128.193.48.142. Redistribution subject to AIP copyright, see http://ojps.aip.org/japo/japcpyrts.html
2721
FIG. 3. Simulated Q--V,,
curves: (a) with impact ionization, 1000 Hz,
(b) with impact ionization, 100 Hz, and (c) without impact ionization,
1000 Hz.
port for this assertion, that slow hole recombination yields a
very large measured slope for the 1000 Hz curves near
threshold, is provided in Fig. 3.
Figure 3 is a series of Qmax-Vmaxsimulations obtained
using Bringuier’s device physics ACTFEL models.9”7 The
three curves simulated in Fig. 3 are obtained assuming a
uniform distribution of interface states of density 5 X lOI
eV-’ cm-? and a flat-band energy of 1 eV. The impact ionization parameters used for curves (a) and (b) .of Fig. 3 are
identical to those used by Bringuier’ as follows: effective
ionization energy, Ei = 3 Egl 2; characteristic phosphor field,
F. = 1.7 MVlcm; and exponent, n = 1.
Curve (c) in Fig. 3 with the smallest slope (i.e., 0.8 nF),
which is less than that of the insulator physical capacitance
(i.e., 1.4 nF), is obtained assuming no impact ionization. According to the theory of Bringuier,’ for a situation in which
impact ionization does not occur, such a Q,,-V,,,
slope
should be always less than, or at most equal to, the insulator
physical capacitance; this is consistent with the no impact
ionization curve [i.e., curve (c)] shown in Fig. 3. A different
value for the simulated Q,,,=- V,,, slope may be obtained by
assuming a different interface state distribution and/or energy
depth.
Now consider curves (a) and (b) shown in Fig. 3. These
curves are simulated for frequencies 1000 and 100 Hz assuming that the hole recombination rate at the cathode interface is very slow. The kinetics of hole recombination is modeled in this simulation as
4J
dt=
-- P
7’
where p is the hole concentration at the cathode and his the
electron-hole recombination time constant. For simulation of
the 100 and 1000 Hz curves shown in Fig. 3, 7 is assumed to
be 1.2 ms. The important point to note from the simulated
100 and 1000 Hz curves is how qualitatively similar they are
to the corresponding experimental curves shown in Fig. 2.
This good qualitative agreement between the experimental
and simulated Q!,,-V,,
curves is strong evidence that
2722
J. Appl. Phys., Vol. 77, No. 6, 15 March 1995
holes, presumably generated by band-to-band impact ionization, play an important role in the operation of evaporated
ZnS:Mn ACTFEL devices. It is important to realize that we
have found it impossible to simulate, no matter how freely
we vary the simulation parameters, the very large slope (i.e.,
-20 nF) observed near threshold for the 1000 Hz Q,,-V,,
curves unless impact ionization with slow recombination is
assumed.
Note that our present simulations do not quantitatively
reproduce the experimental results of Fig. 2. We believe that
better quantitative agreement may be obtained by optimization of the simulation parameters (i.e., interface state density
and depth, electron-hole recombination time, and the impact
ionization threshold field and energy). However, the uniqueness of these simulation parameters is an issue that we have
not yet resolved. For example, the magnitude of the electron
impact ionization coefficient, CY,, varies significantly depending on whether Eqs. (2) or (3) parameters are employed
in the simulation. It is not clear which equation and set of
parameters more accurately represents the impact ionization
trends. Therefore, the goal of obtaining a close quantitative
agreement between the experimental and simulated results
was not pursued.
A crude estimate of the electron multiplication factor
(i.e., the total number of electrons transferred across the
phosphor divided by the number of electrons injected from
the phosphor/insulator cathode interface) for the evaporated
ZnS:Mn ACTFEL device of Fig. 2 may be estimated from
the simulation results of Fig. 3. The Fig. 3 impact ionization
simulation indicates the electron multiplication factor to be
approximately 4-8 for V,,=40
V above threshold. Thus,
we conclude that the vast majority of the electrons which are
stored in interface states for an ACTFEL device in steady
state are generated by impact ionization in the bulk of the
phosphor. We believe that the exceedingly strong fieldclamping character exhibited by these evaporated ZnS:Mn
ACTFEL devices arises primarily from this electron multiplication due to impact ionization. Finally, if our estimate of
the multiplication factor is accurate, the role played by holes
in the device physics operation of evaporated ZnS:Mn
ACTFEL devices needs to be reassessed.
C. Q- Fp and C- V overshoot
Q-F, and C-V characteristics are monitored during the
Qmax-Vmx measurements in order to assess for the presence
of overshoot in the Q-Fp or C-V curves. Previously, we
ha\r_enoted’5*‘s-zo that the presence of Q-F, and C-V overshoot are evidence for the existence of space charge within
the phosphor. However, in the evaporated ZnS:Mn ACTFEL
devices studied herein, we do not find any appreciable Q-F,
or C-V overshoot.
We believe that overshoot is not observed, even though
electron multiplication via band-to-band impact ionization
occurs in these devices, because the positive hole charge is
mobile charge and is thus free to drift to the cathode interface, where it subsequently recombines. Such a situation is
illustrated in Fig. 4(a) which shows a hole being created near
the middle of the phosphor layer and drifting to the cathode,
An9 et
Downloaded 07 Mar 2001 to 128.193.48.142. Redistribution subject to AIP copyright, see http://ojps.aip.org/japo/japcpyrts.html
al.
EF
h+
EF
(a)
I
EF
FIG. 4. Energy band diagrams illustrating (a) a band-to-band impact ionization event and (b) a deep level ionization event.
where it recombines. For such a situation, no Q-F, or C-V
.overshoot is expected since such overshoot is observed only
when positive space charge is both created and trapped in the
bulk phosphor, A situation in which overshoot is expected to
be observed is illustrated in Fig. 4(b) which shows an event
involving impact ionization of a deep level near the center of
the phosphor. In this case, positive space charge is both created and remains trapped in the bulk phosphor.
Note that the absence of Q-F, or C-V overshoot is consistent with band-to-band but not deep level impact ionization. Moreover, the absence of Q-F, or C-V overshoot also
implies that the majority of holes generated by band-to-band
impact ionization in these evaporated ZnS:Mn ACTFEL devices are trapped near the interface and not extensively in the
phosphor bulk.
D. Interface
state
measurements
Another observation that is consistent with our assertion
that band-to-band impact ionization occurs in ZnS:Mn
ACTFEL devices involves our attempts to estimate the interface state density of such devices.*l We have developed a
technique for estimating the interface state density, based on
the use of a field-control circuit. Employing this technique,
we have attempted to measure the interface state densities for
ZnS:Mn ACTFEL devices when the phosphor is grown by
evaporation and the insulators are sputter-deposited silicon
J. Appl. Phys., Vol. 77, No. 6, 15 March 1995
oxynitride (SiON) or barium tantalate (BTO) and when the
phosphor and insulators are grown by atomic layer epitaxy
(ALE) with aluminum-titanium oxide (ATO) insulators. Our
finding is that, to first order, the measured “interface state
densities” are independent of how the phosphor is deposited
(i.e., by evaporation or ALE), how the insulator is deposited
(i.e., sputtering or ALE), or the nature of the insulator (i.e.,
SiON, BTO, or ATO). These findings are consistent with the
fact that ZnS:Mn ACTFEL devices fabricated by various researchers throughout the world, using a variety of fabrication
techniques and different kinds of insulators, invariably obtain
approximately the same brightness-voltage (B-V) threshold
voltage; the magnitude of the B- V threshold is believed to be
determined by the depth of the interface states.
It is very difficult to believe that the interface state density is independent of the phosphor and insulator deposition
technology employed and the nature of the insulators. This is
strong evidence that the B-V threshold voltage, and other
properties which are attributed to interface states, are more
accurately associated with a bulk property of the ZnS;
namely, band-to-band impact ionization of the bulk ZnS.
Thus, we believe that our measured interface state densities
are not true interface state densities; rather, the true nature of
the interface state distribution is obscured by carrier multiplication in our experiment. Consequently, in order to obtain
an accurate assessment of the interface state density of an
ACTFEL device, it is necessary to avoid operating the
ACTFEL device under conditions in which carrier multiplication occurs.
E. Monte
Carlo
high-field
transport
simulation
Our final reason for believing that band-to-band impact
ionization occurs in ZnS:Mn ACTFEL devices involves our
investigations of high field transport in bulk ZnS via Monte
Carlo simulation. Our initial Monte Carlo simulations” employed a three valley (i.e., IY, X, and L valleys) nonparabolic
model for the ZnS conduction band. Subsequent hot electron
luminescence experiments indicated’w924that the high energy
tail of the hot electron distribution is more energetic than that
predicted using the three nonparabolic valley model. Subsequently, we refined the Monte Carlo simulation” so that the
conduction band is modeled using a realistic density of
states. Additionally, full dispersion for the first two conduction bands is included in the refined, full-band structure
Monte Carlo simulation.
The refined, full-band Monte Carlo simulation is found
to exhibit runaway for phosphor fields in excess of -1 MVI
cm, in which case a fraction of the electrons gain energy
from the electric field at a faster rate than energy is dissipated to the lattice. One example of how electron runaway is
manifest in a Monte Carlo simulation is shown in Fig. 5 in
which the average energy of electrons transiting the phosphor is plotted as a function of time after the phosphor field
is turned on at t =O. Note that for fields less than - 1 MV/
cm, the average electron energy reaches steady-state after a
brief transient period; in contrast, no such steady-state average electron energy is obtained for larger fields in which case
the average energy exhibits runaway (i.e., it monotonically
increases with increasing time).
Ang
et al.
Downloaded 07 Mar 2001 to 128.193.48.142. Redistribution subject to AIP copyright, see http://ojps.aip.org/japo/japcpyrts.html
2723
~‘~~~
-------’
without
impact ionization
2
P
m 1.6
e
s
a
1
0
0.1
0.2
0.3
0.4
0.8
0.6
Time (ps)
FIG. 5. Average energy of an electron transiting the phosphor as a function
of time subsequent to the application of a phosphor field step which is
applied at t=O for various phosphor fields. These curves are obtained from
Monte Carlo simulation.
phosphor/insulator cathode interface) to be of order 4-8 for
evaporated ZnS:Mn ACTFEL devices operating under normal conditions of 40 V above threshold; this estimate is obtamed from device physics simulation of experimental
trends.
There are several important implications of our conclusion that band-to-band impact ionization occurs in evaporated ZnS:Mn ACTFEL devices. Fist, since we regard an
evaporated ZnS:Mn ACTFEL device as a prototypical
ACTFEL device, this suggests that impact ionization is a
desirable characteristic for other ACTFEL phosphors. As
shown, field-clamping seems to be intimately related to impact ionization. Second, the fact that impact ionization occurs underscores the potential importance of holes in contributing to the device physics of ACTFEL operation.
ACKNOWLEDGMENTS
Such electron runaway is unphysical and arises from not
properly accounting for all of the scattering mechanisms operative in a real material. As all of the other appropriate
phonon and impurity scattering mechanisms are already incorporated into the Monte Carlo simulation5 our solution to
this runaway problem is to incorporate band-to-band impact
ionization into the simulation using the standard Keldysh
formulation. This leads to physically meaningful results (i.e.,
electron runaway is no longer observed for phosphor fields
up to at least 2 MV/cm). Thus, we believe that the necessity
of incorporating band-to-band impact ionization into the
Monte Carlo simulation, in order to cool the electron distribution so that physically realistic results may be obtained in
the simulation, is one further reason for believing that bandto-band impact ionization occurs in ZnS ACTFEL devices.
III. CONCLUSIONS
Four observations involving
;;)
(iii)
(iv)
experimentally observed field-clamping trends,
experimental and simuIated trends in Q,,-V,,
characteristics,
experimental attempts to assess the interface distribution using a field-control circuit, and
the necessity of having to include impact ionization
into
full-band
Monte
Carlo
simulations
in order
to
alleviate unphysical electron runaway,
are offered to support the assertion that electron-hole generation by band-to-band impact ionization occurs during the
normal operation of evaporated ZnS:Mn ACTFEL devices.
trends are the strongest eviWe believe the Q,,-V,,
dence for the occurrence of band-to-band impact ionization;
the other three observations support this conclusion but are
not convincing without inclusion of the Q,,-V,,
trends.
Additionally, the absence of overshoot in measured Q-Fp
and C-V curves indicates that a majority of the holes created
by impact ionization are trapped at or near the phosphor/
insulator interface. We estimate the multiplication factor (i.e.,
the total number of electrons transferred across the phosphor
divided by the number of electrons injected from the
2724
J. Appl. Phys., Vol. 77, No. 6, 15 March 1995
We wish to thank Ron Khormaei and Sey-Shing Sun for
providing the samples used in this study and Eric Bringuier
for clarification of several subtle aspects of his ACTFEL device physics models. This work was supported by the U.S.
Army Research Office under Contract No. DAAH04-940324 and by the Advanced Research Projects Agency under
the Phosphor Technology Center of Excellence, Grant No.
MDA 972-93-l -0030.
‘D. H. Smith, J. Lumin. 23, 209 (1981).
“W. E. Howard, J. Lumin. 23, 155 (1981).
“W. E. Howard, 0. Sabni, and P M. Ah, J. Appl. Phys. 53, 639 (1982).
“K. W. Yang, Ph. D. thesis, Oregon State University, 1981.
“K. W. Yang and S. J. T. Owen, IEEE Trans. Electron. Devices ED-30,452
(1983).
6N. E. Rigby and J. W. Allen, J. Phys. C 21, 3483 (1988).
‘T. D. Thompson and J. W. Allen, J. Phys. C 20, L499 (1987).
s A. F. Cattell, J. C. Inkson, and J. Kirton, J. Appl. Phys. 61, 722 (1987).
9E. Bringuier, J. Appl. Phys. 67, 7040 (1990).
“G. 0. Mueller, Acta Polytechnica Scandinavica 170, 15 (1990).
” R. Mach and G. 0. Mueller, J. Cryst. Growth 101, 967 (1990).
‘*V. P. Singh, Q. Xu, S. Krishna, and D. C. Morton, Electroluminescence,
edited by V. l? Singh and J. C. MC Clure (Cinco Puntos, El Paso, TX,
1992), p. 47.
I3 J. D. Davidson, J. F. Wager, I. Khormaei, C. N. King, and D. Williams,
IEEE Trans. Electron Devices ED-39, 1122 (1992).
“A. A. Douglas and J. F. Wager, SID 92 Digest, 365 (1992).
‘$A Abu-Dayab, S. Kobayasbi, and J. F. Wager, Appl. Phys. Lett. 62, 744
(1993).
16L. V . Pham, J. E Wager, S. S. Sun, E. Dickey, R. T. Tuenge, and C. N.
King, in Advanced Flat Panel Display Technologies, SPIE Proceedings
2174, edited by P. S. Friedman @PIE, Bellingham, WA, 1994), pp. 190199.
17E. Bringuier, J. Appl. Phys. 66, 1314 (1989).
18A Abu-Dayah, S. Kobayashi, and J. F. Wager, 5. Appl. Phys. 74, 5575
(1$93).
“A Abu-Dayah, M. S. thesis, Oregon State University, 1993.
*‘A. Abu-Dayab and J. F. Wager, 5. Appl. Phys. 75, 3593 (1994).
‘t A. A. Douglas, M. S. thesis, Oregon State University, 1993.
*?K. Bhattachatyya, S. M. Goodnick, and J. E Wager, J. Appl. Phys. 73,
3390 (1993).
=A. A. Douglas, I. F. Wager, D. C Morton, J. B. Koh, and C. P. Hogh,
Appl. Phys. Lett. 63, 231 (1993).
24A. A. Douglas, J. F. Wager, K. Bhattacharyya, S. M. Goodnick, D. C.
Morton, J. B. Koh, and C. P. Hogh, SlD 93 Digest, 851 (1993).
“S. Pennathur, K. Bhattacharyya, S. M. Goodnick, and J. F. Wager, Proceedings of the Third International Workshop on Computational Electronics, edited by S. M. Goodnick (Oregon State University Press, Corvallis,
OR, 1994), p. 288.
An9 et al.
Downloaded 07 Mar 2001 to 128.193.48.142. Redistribution subject to AIP copyright, see http://ojps.aip.org/japo/japcpyrts.html