Content uploaded by Kenneth D Singer
Author content
All content in this area was uploaded by Kenneth D Singer
Content may be subject to copyright.
Space-charge dynamics in photorefractive polymers
Oksana Ostroverkhovaa) and Kenneth D. Singerb)
Department of Physics, Case Western Reserve University, Cleveland, Ohio 44106-7079
共Received 6 May 2002; accepted for publication 9 May 2002兲
The model of space-charge formation in photorefractive polymers due to Schildkraut and Buettner
has been modified to include thermally accessible deep traps as well as shallow traps. The dynamic
equations have been solved semiempirically using independent measurements of photoconductive
properties to predict photorefractive dynamics. Dependencies of the dynamics on charge generation,
mobility, trap density, acceptor density, ionized acceptor density, as well as their associated rates are
examined. The magnitude of the fast time constant of photorefractive development is successfully
predicted. The introduction of deep traps into the model has allowed us to qualitatively predict the
reduction in speed due to deep trap filling and ionized acceptor growth. Experimental studies of
photoconductivity and photorefraction 共PR兲in several polyvinyl carbazole photorefractive
composites are carried out to demonstrate the applicability of the model. By choosing chromophores
with different ionization potentials and by varying the chromophore concentrations, we investigate
the influence of the chromophore ionization potential on the photoelectric and PR properties and
reveal the nature of deep traps in the composites and their contribution to both photoconductivity
and PR dynamics. Effects of plasticizer components are also discussed. © 2002 American Institute
of Physics. 关DOI: 10.1063/1.1491279兴
I. INTRODUCTION
The photorefractive 共PR兲effect involves a change in re-
fractive index in an electro-optical material resulting from
the redistribution of charge carriers created under the influ-
ence of optical beams. This mechanism results in a process,
where the phase shift between the incident intensity pattern
and resulting refractive index pattern due to Poisson’s equa-
tion leads to a number of useful nonlinear optical phenomena
of interest for image and data processing and storage. The
specific processes required for the PR effect include: photo-
generation of charge carriers, transport of mobile carriers,
trapping of these carriers in the regions of destructive inter-
ference and a change of the refractive index in response to
space-charge field. Considerable effort has been applied in
order to understand the influence of each of these processes
on the PR performance in a variety of materials.
Organic and polymeric materials have been the subject
of numerous recent studies. In these materials, several groups
have addressed these photoelectric mechanisms, especially
photogeneration and transport. Photogeneration efficiency is
usually probed using the xerographic discharge technique1–3
or is estimated from dc photoconductivity,4while charge car-
rier mobility is measured using the time-of-flight or holo-
graphic time-of-flight techniques.5,6 Charge trapping mecha-
nisms in PR polymers are still not clearly understood,
although several techniques such as two-beam coupling,7ab-
sorption spectroscopy,8and comparison of external photocur-
rent efficiency to the photogeneration efficiency3have been
used to study the nature of traps. The mechanisms of space-
charge field-induced refractive index change has been inves-
tigated by ellipsometric9–12 and electric field-induced second
harmonic generation13 experiments.
Since polymeric PR materials are potentially useful for a
number of applications that require fast response times, PR
dynamics is an important subject for investigation. Recently,
a number of studies have been aimed at understanding the
relationship between photoconductivity and PR speed4,14,15
and to develop a theory describing grating formation.16,17
However, a systematic theoretical study of photoconductive
mechanisms in PR speed in polymers has not been carried
out, as has been for steady-state PR.3,18 Even for steady-state
PR, only the limiting cases of deep traps or no traps are
usually considered, although the presence of shallow traps
has been confirmed both by the dispersive nature of charge
transport in these disordered media19,20 and by the sublinear
intensity dependence of the PR grating erasure rate.21
In the theoretical part of this article, we modify Schild-
kraut and Buettner’s16 model to take into account both shal-
low and thermally accessible deep traps. We then use a semi-
empirical technique to solve the dynamical and constitutive
equations of the model to predict the dynamic PR response
from independently measured photoelectric properties. In
particular, we obtain the trapping, detrapping, and recombi-
nation properties of the material from the dynamics of dc
photoconductivity, and use them in combination with mea-
surements of charge generation and mobility to reveal the PR
dynamics. By performing numerical simulations, we explore
the dependence of the PR speed on all of these photoconduc-
tive parameters. The introduction of thermally accessible
deep traps into the model has allowed us to fully characterize
the PR dynamic, and to access their consequences. Thus we
are able to describe the observed growth of radical C60
⫺共ion-
ized acceptors兲in various materials8,22 as well as the PR
response time fatigue due to sample illumination prior to PR
a兲Present address: Department of Chemistry, Stanford University, Stanford,
CA 94305-5080.
b兲Electronic mail: kds4@po.cwru.edu
JOURNAL OF APPLIED PHYSICS VOLUME 92, NUMBER 4 15 AUGUST 2002
17270021-8979/2002/92(4)/1727/17/$19.00 © 2002 American Institute of Physics
measurements.22 Our aim is to provide a useful experimental
and numerical modeling process for predicting PR dynamics
from basic material properties in order to guide the develop-
ment of new materials.
In the experimental part of this article, we assess the
applicability of the modified model and examine the contri-
bution of chromophores and plasticizers to photoconductive
and photorefractive performance of PR polymers. We ad-
dress how the PR-relevant photoelectric properties such as
mobility, charge generation efficiency, trapping, detrapping,
and recombination rates are influenced by the degree of dis-
order, trap depths, ionization potentials of the constituents,
presence of ionic impurities, etc. We analyze the photocon-
ductive and photorefractive behavior of both plasticized4,11,14
and unplasticized polyvinyl carbazole 共PVK兲composites in-
cluding the sensitizer (C60) and several chromophores. For
composites of both classes, we determine the quantum effi-
ciency, mobility, trapping, detrapping, and recombination
rates from photoelectric measurements. Then using these
rates we 共i兲calculate the PR speed as determined by four-
wave mixing, 共ii兲compare corresponding rates for different
chromophores and relate them to relevant ionization poten-
tials, 共iii兲study the influence of plasticizer on photoconduc-
tivity and photorefractive performance of the composite, and
finally, 共iv兲study the nature of shallow and deep traps in
composites and investigate their influence on photoconduc-
tive and photorefractive properties of the materials.
II. THEORY
The first theoretical description of PR in polymers was
adapted from Kukhtarev’s23 theory of inorganic crystals by
Schildkraut and Buettner.16 They included the rate equation
for traps in the system of PR dynamical equations and took
into account the field dependence of both photogeneration of
mobile carriers and mobility. The modified model presented
here differs from Schildkraut and Buettner’s by introducing
two kinds of traps—shallow and deep. Here, by the term
‘‘deep,’’ we mean that the rate of thermal detrapping for
these traps is at least an order of magnitude lower than that
of shallow traps, but still having a nonzero probability for
detrapping. The processes taken into account in here are de-
picted in Fig. 1. A sensitizer 共acceptor兲with density NAis
excited and subsequently ionized by light of frequency
with cross-section s. A free hole is injected into the transport
manifold and hops between transport sites until it either be-
comes trapped or recombines with ionized acceptors with
rate
␥
. Although, generally, the energy spectrum of trapping
sites has a continuous distribution,21 for tractability we con-
sider only two kinds of traps with well-defined energy levels
共ionization potential兲—shallow traps MT1and deep traps
MT2. We assume that the trapping rate does not depend on
the trap depth,24,25 so that shallow and deep traps are filled
with the same trapping rate
␥
T. Detrapping proceeds with a
thermal excitation rate

1for shallow traps or

2for deep
traps. Optical detrapping is not considered because the depth
of both shallow and deep traps 共⬍0.5 eV兲is much smaller
than the photon energy ប
共⬃1.96 eV for HeNe 633 nm
light兲. Then, the modified system of nonlinear equations de-
scribing the PR dynamics is given by
t⫽
NA
i
t⫺
M1
t⫺
M2
t⫺1
e
J
x,
M1
t⫽
␥
T共MT1⫺M1兲
⫺

1M1,
M2
t⫽
␥
T共MT2⫺M2兲
⫺

2M2,
共1兲
NA
i
t⫽sI共NA⫺NA
i兲⫺
␥
NA
i
,
E
x⫽e
⑀
0
⑀
共
⫹M1⫹M2⫺NA
i兲,
J⫽e
E⫺e
x.
Here
is the free charge 共hole兲density, NAthe total density
of acceptors 共e.g., C60兲,NA
ithe density of ionized acceptors
共e.g., C60
⫺兲,M1,M2,MT1, and MT2the densities of filled
shallow traps, filled deep traps, and total shallow and deep
trapping sites, respectively, Ethe electric field, and Ithe
incident light intensity. Jis the current density,
the charge
carrier drift mobility, and
is the diffusion coefficient given
by
⫽kBT/e. The quantity sis the cross section of photoge-
neration,
␥
T,

1,

2the trapping rate and detrapping rates
for shallow and deep traps, respectively,
␥
the recombination
rate, and
⑀
the dielectric constant. We consider the param-
eters s,
,
␥
T, and
␥
to be electric field dependent assuming
the following dependencies:18
s⫽s共Eref兲共E/Eref兲p,
⫽
共Eref兲e

共E1/2⫺Eref
1/2兲,
共2兲
␥
⫽
␥
共Eref兲e

共E1/2⫺Eref
1/2兲,
␥
T⫽
␥
T共Eref兲e

␥
共E1/2⫺Eref
1/2兲.
Here Eref is the relevant reference electric field for each of
the parameters. During photorefractive grating formation, the
reference electric field for photogeneration efficiency is the
external applied field, while for mobility, trapping, and re-
FIG. 1. Schematic representation of the modified model for photorefractive
polymers. Symbols are E: electric field,
: frequency of light,
: free charge
density, s: photogeneration cross section,
␥
T: trapping rate,
␥
: recombina-
tion rate, and

1,2 : detrapping rates.
1728 J. Appl. Phys., Vol. 92, No. 4, 15 August 2002 O. Ostroverkhova and K. D. Singer
combination rates the reference electric field is the projection
of the applied field on the grating vector. Parameters p,

␥
,
and

are determined experimentally.
Since the creation of a photorefractive hologram as-
sumes a nonuniform light intensity pattern created by the
interfering beams, the incident light intensity can be ex-
pressed as a periodic function of x:
I⫽I0⫹I1cos kx,共3兲
where kis the grating vector chosen to be parallel to the x
direction. The general spatial dependence of Eq. 共3兲will also
apply to the response of the polymeric material. The exact
electric field and charge density distributions must be deter-
mined by numerically solving Eqs. 共1兲–共3兲. However, as
shown by Schildkraut and Cui18 for the steady state, the nu-
merically determined values for free charge density and the
amplitude and phase of the space-charge field were in a good
agreement with the corresponding values obtained from the
Fourier analysis of the equations analogous to our Eq. 共1兲.
Also, the Fourier decomposition approach was used by Cui
et al. for the analysis of dynamics of the PR grating
erasure.21 So, as we proceed, we use a Fourier decomposition
of all the densities, electric field, and current and separate the
equations on the basis of the spatial dependence of the terms.
We also consider the case of moderate applied electric fields
(Ea⬍50 V/
m), where effects of grating bending and
higher spatial harmonics can be neglected,26 and thus limit
our analysis to the zeroth and first spatial Fourier compo-
nents. We experimentally investigated the dependence of the
PR dynamics on the intensity modulation depth m⫽I1/I0.
We found that in contrast with a steady-state PR perfor-
mance, the PR dynamics observed under our experimental
conditions of short prior-to-experiment light exposure time
of ⬃10 s and total writing beam intensities below 400
mW/cm2depended only on I0rather than the modulation
depth m, as expected from the theory for the first spatial
order Fourier decomposition. Therefore in our range of elec-
tric fields and incident intensities, the PR dynamics did not
depend on the modulation depth, and thus, the unity modu-
lation depth 共a conventional choice of experimental geom-
etry in PR polymer literature兲, is used in this article. Then,
the solution of Eq. 共1兲can be written in the following form:
⫽
0共t兲⫹
1共t兲共C
1cos kx⫹C
2sin kx兲,共4兲
where
⫽
,M1,M2,NA
i,J. Since the mobility and all of
the relevant charge generation, trapping, and recombination
rates are field-dependent, we assume a form similar to Eq.
共4兲for each of these parameters as well as for the electric
field Ewith time-independent
0.18 We substitute Eq. 共4兲in
Eq. 共1兲and separate the spatially independent zeroth order
and spatially varying first order systems of equations. We
start from the zeroth order equations that describe photocon-
ductivity in PR polymers under homogeneous illumination of
intensity I0. As we proceed, we will use our photoconduc-
tivity experiments and theory 共zeroth order兲to predict the PR
dynamics 共first order兲assuming that the influence of Gauss-
ian rather than uniform incident beams is the same for both
zeroth and first order processes.
A. Zeroth order: Photoconductivity in photorefractive
polymers
In this section we explore how the information about
rates relevant for PR grating formation can be extracted from
the photocurrent dynamics. The zeroth order system of equa-
tions derived from Eq. 共1兲is written as follows:
dM10共t兲
dt ⫽
␥
T共E0兲关MT1⫺M10共t兲兴
0共t兲⫺

1M10共t兲,
dM20共t兲
dt ⫽
␥
T共E0兲关MT2⫺M20共t兲兴
0共t兲⫺

2M20共t兲,
共5兲
dNA0
i共t兲
dt ⫽s共E0兲I0关NA⫺NA0
i共t兲兴⫺
␥
共E0兲NA0
i共t兲
0共t兲,
0共t兲⫹M10共t兲⫹M20共t兲⫺NA0
i共t兲⫽0.
One more equation that is not included in Eq. 共5兲but pro-
vides a link to dc photoconductivity experiments is the con-
stitutive equation for the photocurrent J0(t) given by J0(t)
⫽e
(E0)
0(t)E0. Equation 共5兲describes the dynamics of
free charge generation followed by transport, trapping, de-
trapping, and recombination in the photorefractive polymers
under external electric field E0. We consider the electric
field E0to be constant and given by E0⫽V/d, where Vis the
applied voltage and dis the thickness of the polymeric film.
Then, the time evolution of the photocurrent J0(t) probes the
dynamics of free charge density
0(t) that is connected
through Eq. 共5兲to the generation, trapping, detrapping, and
recombination processes in the PR polymers. The rates de-
scribing these processes (s,
␥
T,

1,2 ,
␥
) are intrinsic charac-
teristics of the polymer composites, and our goal is to deter-
mine their influence on both photoconductivity and
photorefractive speed. It should also be mentioned that since
all the rates are electric-field dependent, the photoconductiv-
ity experiments have to be conducted in the range of external
electric field E0that covers both applied electric field in PR
experiment Eaand its projection on the grating vector E
˜
ain
order to proceed with calculating the space-charge field dy-
namics 共first order兲on the basis of parameters determined
from the photoconductivity 共zeroth order兲.
To study the temporal behavior of the free charge
,
trapped charge M1,M2, and ionized acceptor NA
idensities,
it is convenient to transform Eq. 共5兲to a dimensionless form.
The time scale is normalized by the average drift time of the
free carrier before it is trapped by a shallow trap:
⫽t/
0,
where
0⫽1/关
␥
T(E0)MT1兴. The reason for this choice of
time scale will be explained later in the Sec. II A 1. We ex-
press all the densities in terms of total acceptor density NA:
%⫽
/NA,mT1,2⫽MT1,2 /NA,m1,2⫽M1,2 /NA, and nA
i
⫽NA
i/NA. We also introduce the relative photogeneration,
recombination, and detrapping parameters s
˜
I0⫽sI0
0,
␥
˜
⫽
␥
0NA, and

˜
1,2⫽
0

1,2 , respectively. The dimensionless
analog of Eq. 共5兲is then written as follows:
1729J. Appl. Phys., Vol. 92, No. 4, 15 August 2002 O. Ostroverkhova and K. D. Singer
dm10
d
⫽
冉
1⫺m10
mT1
冊
%0⫺

˜
1m10 ,
dm20
d
⫽mT2
mT1
冉
1⫺m20
mT2
冊
%0⫺

˜
2m20 ,
共6兲
dnA0
i
d
⫽s
˜
I0共1⫺nA0
i兲⫺
␥
˜
nA0
i%0,
%0共
兲⫹m10共
兲⫹m20共
兲⫺nA0
i共
兲⫽0.
This is a nonlinear system that cannot be solved analyti-
cally. However, we can consider different cases depending
on the total trap densities with respect to the acceptor density
in the material, leading to simplifications of these equations.
Case 1. Both the total shallow and deep trap densities
are smaller or on the order of the acceptor density: mT1,2
ⱗ1.Case 2. The total deep 共shallow兲trap density is smaller
or on the order of the acceptor density, but the total shallow
共deep兲trap density is much larger than the acceptor density:
mT2ⱗ1, mT1Ⰷ1 or vice versa. In this case, the ratio
m10 /mT1Ⰶ1共or m20 /mT2Ⰶ1兲is always valid, and the first
共second兲equation in Eq. 共6兲is simplified.
Case 3. Both deep and shallow total trap densities are
much larger than the acceptor density: mT1,2Ⰷ1. In this case
both ratios m10 /mT1Ⰶ1 and m20 /mT2Ⰶ1 are always valid,
so that both the first and second equations in Eq. 共6兲become
linear.
Our simulations show that in the trap-limited regime
共Case 1兲photocurrent dynamics is similar to the short-time
scale regime of Cases 2,3 共Sec. II A 1兲.Case 1 does not
describe long-time scale changes in both photocurrent and
space-charge field, and therefore, seems not to be applicable
for most of our materials. Therefore we consider only Case 2
and Case 3, which can be applicable depending on the chro-
mophore ionization potential and concentration.27,28 We start
our analysis from Case 3 and then extend it to Case 2.
1. Trap-unlimited regime
In this section we consider the case when the total den-
sity of both shallow and deep traps is much higher than the
acceptor density mT1,2Ⰷ1共Case 3兲. Then, Eq. 共6兲is simpli-
fied to
dm10
d
⫽%0⫺

˜
1m10 ,
dm20
d
⫽mT2
mT1%0⫺

˜
2m20 ,
共7兲
dnA0
i
d
⫽s
˜
I0共1⫺nA0
i兲⫺
␥
˜
nA0
i%0,
%0共
兲⫹m10共
兲⫹m20共
兲⫺nA0
i共
兲⫽0.
We also assume that the density of total shallow traps is
much larger than the density of total deep traps mT1ⰇmT2
which seems to be relevant for all the composites we studied
共Sec. III A兲. Then, we can separately consider different time
scales at which either shallow 共‘‘short time scale’’兲or deep
共‘‘long time scale’’兲trap dynamics prevails in order to pro-
vide insight into the experimentally observed trapping, de-
trapping, and recombination rates as deduced from dc pho-
toconductivity.
a. Short time scale. On the short time scale, the dynam-
ics of the system is entirely determined by shallow traps. To
probe the behavior of our system, we first consider the initial
rise in photocurrent as the nonlinear term
␥
˜
%0nA0
iin the third
equation of Eq. 共7兲is much smaller than the linear term
s
˜
I0nA0
iin that same equation. This transforms the nonlinear
system of Eq. 共7兲into a linear one that can be solved ana-
lytically. Then, we seek a solution of the form ⫽0e⫺
˜
,
where ⫽m10 ,m20 ,nA0
i, and then solve for the three char-
acteristic rates
˜
that determine the dynamics of dc photo-
conductivity at this time scale:
˜
1⫽1⫹mT2/mT1⬇1;
˜
2
⫽s
˜
I0; and
˜
3⬇

˜
2⫹(mT2/mT1)

˜
1. In these equations we
assume that the detrapping parameters

˜
1and

˜
2are much
smaller than 1 and (mT2/mT1), respectively. This assump-
tion holds when the free charge density is much smaller than
the density of filled traps as observed in a variety of
materials.3,29 We have also confirmed this in our photocon-
ductivity experiment for materials described here. For this
reason and also in keeping with our experimental observa-
tions that the charge generation rate at reasonable experimen-
tal intensities I0⬍1 W/cm2is much smaller than the trapping
parameter 1/
0, i.e., s
˜
I0Ⰶ1共Sec. IV A 1兲, we can assume
that
˜
2Ⰶ
˜
1and
˜
3Ⰶ
˜
1.
Thus the fastest photoconductivity dynamics is given by
unity
˜
1⬇1 in the dimensionless form. This fact explains our
choice of the time scale being normalized with respect to
0⫽1/(
␥
TMT1). Thus in dc photoconductivity experiments
the fastest photocurrent dynamics 关J0(t)⬃
0(t)兴is deter-
mined by the value for the shallow trapping product
␥
TMT1.
In the next longer time regime 共though still in the shal-
low trapping fast limit兲we need to consider that, as we will
see later, the free charge density has reached a maximum and
starts to decrease. Then, we can replace the free charge den-
sity function %0(
) in the nonlinear term of the third equa-
tion in Eq. 共7兲with a quasi-steady value %0. We again obtain
three time constants out of which
˜
2is of relevance and
given by
˜
2⫽s
˜
I0⫹
␥
˜
%0. In this intermediate time regime,
˜
2
contains information regarding the recombination parameter
␥
˜
.Summarizing the dynamics of the photoconductivity on
a short time scale 共for PVK-based materials we studied, this
corresponds to
ⱗ50兲, we obtain the following expressions
for dc photoconductivity rates:
˜
1⬇1;
˜
2⫽s
˜
I0⫹
␥
˜
%0.In
keeping with our experiments, in the low intensity regime
共below 1 W/cm2兲, we can simplify these to
˜
1⬇1;
˜
2
⬇
␥
˜
%0.
Thus in dc photoconductivity experiments, the short time
scale dynamics can be fitted with a biexponential function,
where the faster rate 1yields the shallow trapping product
␥
TMT1, and the slower rate 2is directly related to the
recombination rate
␥
.
To test our approximations, we performed numerical
simulations fixing s
˜
I0⫽10⫺2
0,mT2/mT1⫽0.1,

˜
1
1730 J. Appl. Phys., Vol. 92, No. 4, 15 August 2002 O. Ostroverkhova and K. D. Singer
⫽0.1
0,

˜
2⫽0.001
0, and varying the average carrier life-
time
0and the recombination parameter
␥
˜
. This choice of
parameters used for the simulation was suggested by the cor-
responding typical values observed experimentally in unplas-
ticized composites 共Sec. IV A 1兲. First, at
␥
˜
⫽5⫻104
0we
substituted different values of the parameter
0into Eq. 共7兲,
and numerically solved this system to find the dependence of
%0(
) for
⭐50. At this time scale, the free charge density
grows as the charge is injected into the transport manifold,
reaches the maximum %0,max , and then decreases, due to re-
combination and trapping. We then fit the calculated curve of
the free charge density %0(
) with a biexponential function
%fit⫽A共1⫺Be⫺
˜
10
⫹共B⫺1兲e⫺
˜
20
兲共8兲
to determine
˜
10 共the faster constant兲and compare it to
˜
1
⫽1. At B⬎1 Eq. 共8兲describes the photocurrent rise with the
rate
˜
10 and then the photocurrent decay with the rate
˜
20 .
We found that the faster constant in the biexponential fit
yields values equal to unity within 10%, as anticipated. To
extract the slower rate, we used a fixed
0⫽5ms 共typical
value for an unplasticized sample observed in dc photocon-
ductivity experiment at E0⬃30 V/
m and I0⬃50 mW/cm2兲
for different recombination parameters
␥
˜
, found %0(
), and
fit it using Eq. 共8兲to determine
˜
20 . We found that
˜
20 agrees
with 2⫽
␥
˜
%0(%0⫽%0,max) within a factor of 3. As we show
in Ref. 28 and mention later in Sec. II B 1, an error of this
magnitude in the recombination rate actually has a negligible
effect on the PR speed. This agreement is remarkable since
we dealt so crudely with the nonlinear term containing the
charge density.
As an example, consider the dynamics of the dc photo-
current simulated using the parameters s
˜
I0⫽5⫻10⫺5,
mT2/mT1⫽0.1,

˜
1⫽5⫻10⫺4,

˜
2⫽5⫻10⫺6, and
␥
˜
⫽2.5
⫻102共solid line in Fig. 2兲. The dashed line in the figure is a
biexponential fit of the short time behavior using Eq. 共8兲. The
faster inverse time constant of this fit
˜
10⫽1.1 reflects the
expected value of 1⫹mT2/mT1⫽1.1. The slower speed
˜
20
⫽1.9⫻10⫺2divided by the dimensionless free charge den-
sity in its maximum %0,max⬇4.4⫻10⫺5yields, according to
˜
20⫽
␥
˜
%0,max , the recombination parameter
␥
˜
⬇4.3⫻102,
which is within the expected range of values given the input
value
␥
˜
⫽2.5⫻102. The biexponential of Eq. 共8兲fits the
%0(
) dependence perfectly at
⬍50 共inset of Fig. 2兲, but as
the processes which are not taken into account at short time
scale take over at
Ⰷ50, the short time scale fit does not
follow the photocurrent dynamics correctly 共dashed line in
Fig. 2兲, and thus long time scale analysis should be applied.
We now use a procedure to produce a better time evolu-
tion for %0that will also yield the detrapping parameter for
shallow traps. Although it is not obvious how to analytically
extract the detrapping parameter for shallow traps

˜
1,%0(
)
at short time scales is rather sensitive to changes in

˜
1.28 To
find the detrapping parameter

˜
1and to fine-tune the recom-
bination parameter
␥
˜
, we used the
␥
˜
estimated above as the
initial value in Eq. 共7兲and varied both

˜
1and
␥
˜
. At every
step we calculated the dimensionless free charge density
%0(
) and fit to Eq. 共8兲. Then we transformed the dimension-
less fit parameter
˜
20 to the dimensional form 20⫽
˜
20 /
0
and compared them to the analogous parameters of a biex-
ponential fit of dc photoconductivity data. The comparison
was executed by searching for a minimum of the function
f⫽
冉
20,exp⫺20
20,exp
冊
2
⫹
冉
Bexp⫺B
Bexp
冊
2,共9兲
where 20,exp ,Bexp are the experimental parameters analo-
gous to corresponding parameters 20 ,Bintroduced in Eq.
共8兲.Summarizing the short time scale dc photoconductivity
dynamics for the case mT1ⰇmT2Ⰷ1, we are able to deter-
mine the shallow trapping parameter
␥
TMT1, recombination
rate
␥
, and shallow detrapping rate

1.
b. Long time scale. On the long time scale (
⬎103),
shallow traps have reached quasiequilibrium, and deep traps
determine the dynamics of the dc photoconductivity. The di-
mensionless parameters to be determined here are the ratio of
total deep traps with respect to shallow traps mT2/mT1and
the thermal detrapping parameter of deep traps

˜
2. We per-
formed a numerical simulation fixing the parameters s
˜
I0
⫽5⫻10⫺5,

˜
1⫽5⫻10⫺4,
␥
˜
⫽2.5⫻102, and varying
mT2/mT1共while maintaining the ratio mT2/mT1Ⰶ1兲and

˜
2.
Similar to the short time scale approach, we fit the free
charge density %0(
) with a biexponential similar to Eq. 共8兲
共dotted line in Fig. 2兲. The faster speed
˜
10 was kept fixed
equal to unity. Then, our fit yielded two coefficients—the
slower speed
˜
20,long and the exponential prefactor Blong . Al-
though at this time scale, it is not straightforward to relate
the constants of the fit to the parameters of the material di-
rectly, our simulations show that the deep to shallow trap
density ratio mT2/mT1and deep detrapping parameter

˜
2
can be found from the fit constants in a manner similar to the
short time scale analysis. When dealing with the experimen-
tal data, we constructed the function
FIG. 2. Simulated dynamics of dc photocurrent for the case mT1ⰇmT2Ⰷ1
using Eq. 共7兲and parameters s
˜
I0⫽5⫻10⫺5,mT2/mT1⫽0.1,

˜
1⫽5
⫻10⫺4,

˜
2⫽5⫻10⫺6, and
␥
˜
⫽2.5⫻102. The inset shows the short time
scale part of the photocurrent transient 共data and fit兲.
1731J. Appl. Phys., Vol. 92, No. 4, 15 August 2002 O. Ostroverkhova and K. D. Singer
f1⫽
冉
20,exp,long⫺20,long
20,exp,long
冊
2
⫹
冉
Bexp,long⫺Blong
Bexp,long
冊
2,
where 20,exp,long and Bexp,long are the experimental constants
analogous to 20,long⫽
˜
20,long /
0and Blong . Similar to the
short time scale approach, we sought values for mT2/mT1
and

˜
2that would minimize the function f1.
Summarizing the long time scale dc photoconductivity
dynamics for the case mT1ⰇmT2Ⰷ1, we are able to experi-
mentally determine the product
␥
TMT2and the detrapping
rate

2.
2. Trap-limited regime
In this section, we consider the dc photoconductivity dy-
namics when the total density of traps is on the order or less
than the acceptor density. We will limit our discussion to the
case when the regime is ‘‘trap-limited’’ only with respect to
deep traps 共Case 2兲that appears to be relevant for the mate-
rials we studied.28 So, further in this section we assume that
mT1Ⰷ1 and mT2⭐1. In this case, the condition mT1ⰇmT2is
satisfied automatically and thus the time scale division onto
‘‘short’’ and ‘‘long’’ is still appropriate. Also, since for shal-
low traps the condition mT1Ⰷ1 is the same as in the previous
Sec. II A 1, all the short time scale considerations are valid.
However, long time scale behavior is no longer the same as
in the trap-unlimited regime, mainly because in this case the
condition m20 /mT2Ⰶ1 is no longer valid, so Eq. 共6兲with the
first equation replaced with the first equation of Eq. 共7兲has
to be solved. Modified in this way, the system of Eq. 共6兲
contains one more parameter than Eq. 共7兲, so that it is not
enough to determine the ratio mT2/mT1because of mT2in
the term m20 /mT2. Our simulations show that if mT2⬃1
then the use of simplified Eq. 共7兲instead of Eq. 共6兲is still
possible which allows us to determine the ratio mT2/mT1
within 10% error. However, if mT2⬍1, then the modified Eq.
共6兲must be solved since the error becomes ⬎100%. This
complicates the analysis of the long time scale behavior.
Also, when the density of total deep traps becomes of the
order of 1% or less of the acceptor density (mT2⭐0.01), it
appears to be impossible to detect deep traps in the material
using dc photoconductivity. We simulated the long time scale
evolution of the dimensionless free charge density %0(
)
with the fixed parameters s
˜
I0⫽5⫻10⫺5,

˜
1⫽5⫻10⫺4,

˜
2
⫽5⫻10⫺6,
␥
˜
⫽2.5⫻102,mT1⫽10, and varying mT2from
0.01 to 1. Figure 3共a兲shows the deep-trap limited behavior
of the photoconductivity for various total deep trap densities.
As seen from Fig. 3共a兲, when mT2⫽0.01, the decay of %0(
)
is less than 3% over the time scale corresponding to the
experimental run of duration ⭓103s for the PVK-based
composites we studied 共Sec. III兲, so that it would be hard to
obtain a reliable fit to such data and thus the trap densities
below 0.01 cannot be detected by this method. As mentioned
before, in the case when mT2⫽0共no deep traps兲, the photo-
conductivity degradation during continuous illumination is
not observed.
3. Time evolution of ionized acceptor density n
A
0
i
According to the last equation in Eq. 共5兲, the free charge,
filled traps, and ionized acceptor densities are constrained by
the charge neutrality equation, thus the photoconductivity de-
cay, trap filling, and the growth of the density of ionized
acceptors are directly connected to each other. Figure 3共b兲
shows the simulated time growth of the ionized acceptor
density nA0
i(
) using the fixed parameters listed in the pre-
vious Sec. II A 2 and varying the total deep trap density mT2
from 0.01 共deep trap-limited regime兲to 10 共deep trap-
unlimited regime兲. As seen in Fig. 3共b兲, the more deep traps
are available in the material, the more pronounced is the
ionized acceptor density growth. Also, the trap depth is the
factor that affects the time evolution of nA0
i. Figure 3共c兲
shows how the depth of the traps 共thermal detrapping rate兲
affects the formation of ionized acceptors. For this simula-
tion we used the same parameters as for the simulation
shown in Fig. 3共b兲, but with fixed mT2⫽1 and varied

˜
2
from 5⫻10⫺7to 5⫻10⫺5.
Our simulations show that the steady-state number den-
sity of ionized acceptors (NA0
i) in the material depends on all
FIG. 3. Long time scale dynamics of 共a兲free charge density 共dc photocur-
rent兲共b兲ionized acceptor number density at B
˜
2⫽5⫻10⫺6at various deep
trap number densities; and 共c兲ionized acceptor number density at mT2⫽1at
various deep trap detrapping rates, as calculated from Eq. 共6兲using param-
eters s
˜
I0⫽5⫻10⫺5,

˜
1⫽5⫻10⫺4,
␥
˜
⫽2.5⫻102, and mT1⫽10.
1732 J. Appl. Phys., Vol. 92, No. 4, 15 August 2002 O. Ostroverkhova and K. D. Singer
the rates participating in the photoconductivity–
photogeneration cross-section s, trapping rate
␥
T, recombi-
nation rate
␥
, and detrapping rates

1,2 . In particular, Fig. 4
shows the simulated ionized acceptor density achieved in a
typical unplasticized material as a function of photogenera-
tion cross section 共simulated dependence is fitted with power
law NA0
i⬃sb, where bwas determined to be ⬇0.5兲. This
dependence explains differences in the steady-state PR per-
formance, which depends on the density of filled traps re-
lated to ionized acceptor density by the electric neutrality
equation 关the last equation of Eq. 共5兲兴, observed in the same
material sensitized with different sensitizers.30
In summary, the density and depth of available traps as
well as other photoconductivity parameters are directly re-
sponsible for the variations in the growth of ionized accep-
tors nA0
i共e.g., C60
⫺兲experimentally observed by several
groups.8,22
We have determined many of the parameters of Fig. 1,
and, in the next section, will show how these may be used to
predict the PR dynamics. Here we provide a numerical simu-
lation of the dynamics of the free charge and ionized accep-
tors for materials with deep traps, and in Sec. III we consider
experimental data for different PVK-based composites and
discuss the nature of traps in these materials.
B. First order: Photorefraction
In this section, we consider the formation of the first
spatial Fourier component of the free charge, filled traps, and
ionized acceptors densities as well as space charge field and
hence PR dynamics. Here it appears more convenient to use
dimensional equations rather than dimensionless for easier
comparison of the simulated PR dynamics with the experi-
ment. After substitution of Eq. 共4兲into Eq. 共1兲, and sepa-
rately collecting terms with sin kx and cos kx, we obtain a
system of ten equations—eight differential equations 共with
respect to time兲and two equations relating the space-charge
field to free charge, filled traps, and ionized acceptor densi-
ties. We consider that the photogeneration cross section, mo-
bility, recombination, and trapping rates are electric field de-
pendent and assume the dependencies given in Eq. 共2兲.As
mentioned before, due to these field dependencies all the
parameters acquire a spatially varying part upon formation of
the space-charge field. Using Eq. 共2兲and the approach sug-
gested by Schildkraut and Cui,18 we can express all the rates
in terms of space-charge field as follows:
s⫽s共Ea兲关1⫹p共E1/Ea兲兴,
⫽
共E
˜
a兲关1⫹共1/2兲

E
˜
a
⫺1/2E1兴,
共10兲
␥
⫽
␥
共E
˜
a兲关1⫹共1/2兲

E
˜
a
⫺1/2E1兴,
␥
T⫽
␥
T共E
˜
a兲关1⫹共1/2兲

␥
E
˜
a
⫺1/2E1兴,
where Eais the applied electric field, E
˜
ais the projection of
the applied field on the grating vector, and E1is the space-
charge field given in accordance with Eq. 共4兲by the relation
E1⫽E11(t)cos kx⫹E12(t)sin kx. Then, the first order system
of equations describing PR dynamics is
dE11,12
dt ⫽⫺
e
⑀
0
⑀
冋
冉
1⫹

E
˜
a
1/2
2
冊
0共t兲E11,12共t兲
⫺E
˜
a
11,12共t兲⫿k
11,12共t兲
册
,
dM11,12
dt ⫽
␥
TMT1
冉
11,12共t兲⫺

␥
E
˜
a
⫺1/2
2E11,12共t兲
0共t兲
冊
⫺

1M11,12共t兲,
dM21,22
dt ⫽
␥
TMT2
冉
11,12共t兲⫺

␥
E
˜
a
⫺1/2
2E11,12共t兲
0共t兲
冊
⫺

2M21,22共t兲,
dNA11
i
dt ⫽sNAI1⫺关sI0⫹
␥
0共t兲兴NA11
i共t兲⫺
␥
NA0
i共t兲
11共t兲
⫺
冉
sNAI0
p
Ea
⫺
␥
E
˜
a
⫺1/2
2
0共t兲NA0
i共t兲
冊
E11共t兲,
共11兲
dNA12
i
dt ⫽⫺关sI0⫹
␥
0共t兲兴NA12
i共t兲⫺
␥
NA0
i共t兲
12共t兲
⫺
冉
sNAI0
p
Ea
⫺
␥
E
˜
a
⫺1/2
2
0共t兲NA0
i共t兲
冊
E12共t兲,
E11共t兲⫽e
⑀
0
⑀
k关
12共t兲⫹M12共t兲⫹M22共t兲⫺NA12
i共t兲兴,
E12共t兲⫽⫺ e
⑀
0
⑀
k关
11共t兲⫹M11共t兲⫹M21共t兲⫺NA11
i共t兲兴.
Here E11 ,M11 ,M21 , and NA11
iare the time-dependent am-
plitudes of space-charge field, filled shallow traps, filled deep
traps, and ionized acceptors, respectively. These have a spa-
tial dependence cos kx 共in-phase with the incident light illu-
mination兲. The quantities E12 ,M12 ,M22 , and NA12
iare the
corresponding amplitudes of functions with a spatial depen-
dence sin kx 共90° out-of-phase with the intensity of incident
light兲.
It is conventional16,21,29 to assume that the PR dynamics
is much slower than the photoconductive dynamics, which is
FIG. 4. Ionized acceptor number density as a function of photogeneration
cross section as simulated using Eq. 共6兲with parameters

˜
1⫽5⫻10⫺4,

˜
2
⫽5⫻10⫺6,
␥
˜
⫽2.5⫻102,mT1⫽10, and mT2⫽1 transformed into dimen-
sional form using
0⫽5msandNA⫽5⫻1024 m⫺3, and fitted with a power
law function NA0
i⬃sb.
1733J. Appl. Phys., Vol. 92, No. 4, 15 August 2002 O. Ostroverkhova and K. D. Singer
analogous to setting the zeroth order 关
0(t),NA0
i(t)兴func-
tions to be constant in time in Eq. 共11兲. In this case, Eq. 共11兲
can be solved analytically as linear equations with constant
coefficients. This approach can be applied on the short time
scale. However, longer times require accounting for deep
traps, which result in slowly changing components of both
0(t) and NA0
i(t), requiring Eq. 共11兲to be solved numeri-
cally. In the next section we will concentrate on obtaining the
four wave mixing 共FWM兲dynamics from the theory and
photoconductivity parameters.
1. Photorefractive rise
In this section we examine the factors that affect the PR
rise time. First of all, it is important to perform a simulation
using conditions as close as possible to a real experiment. In
our case, a real experiment 共FWM兲is performed as follows:
we turn on the electric field with one writing beam on, then
in 10 s we turn on the other writing beam of the same inten-
sity and monitor the space-charge field formation with a
probe beam. In Eq. 共11兲that describes the PR dynamics, we
need to define the behavior of zeroth order functions
0(t)
and NA0
i(t), so first we simulate the time evolution of these
functions under the experimental conditions described in
Sec. III. The dependence of these zeroth order functions on
the experimental conditions is responsible for the history de-
pendence of the PR performance, as we will explore later in
this section. So, as we determined
0(t) and NA0
i(t) for the
time span of our PR experiment, we use them in Eq. 共11兲to
simulate the evolution of the space charge field. We then
calculate the diffraction efficiency,
共t兲⬃E11
2共t兲⫹E12
2共t兲共12兲
and fit it to a single exponential,
⫽
0共1⫺e⫺
t兲2共13兲
where the parameter
is PR speed. Usually in the literature,
the experimentally measured diffraction efficiency is fit with
a biexponential function. In this case, the initial rise is attrib-
uted to photoconductivity, and the slower one to chro-
mophore reorientation. Thus a single exponential fit 关Eq.
共13兲兴 to describe the initial space-charge field formation is
adequate. This case corresponds to PR dynamics observed in
our unplasticized composites. When both faster and slower
experimental time constants are due to photoconductivity,
which is the case for our plasticized composites, the simu-
lated data should be fit with a biexponential function. Here,
for simplicity, we consider the former case and use single
exponential fits 关Eq. 共13兲兴 to describe the diffraction effi-
ciency rise due to space-charge field formation.
To explore the factors that affect the speed of space-
charge field formation, we modeled the PR experiment by
varying the photogeneration rate s, thermal detrapping rate

1, recombination rate
␥
, total density of shallow trapping
sites MT1, and mobility
. For each set of parameters we
calculated PR speed
from Eqs. 共11兲–共13兲. Although some
of these parameters depend on each other and, strictly speak-
ing, cannot be varied independently, this simulation still can
provide some insight into factors determining the PR speed.
Our simulation shows that the PR speed is nearly insensitive
to the shallow traps release rate: the change in parameter

1
over four orders of magnitude barely changed the PR speed
by a factor of 2.28 Larger effects were observed when chang-
ing other parameters: a decrease in the total density of shal-
low trapping sites of a factor of 500 led to a sixfold increase
in PR speed, and a four order of magnitude decrease in the
recombination rate yielded a 30-fold increase in PR speed.28
It should be noted that a decrease in total trap density may
decrease the diffraction efficiency. Thus in this case there is a
trade-off between PR speed and steady-state diffraction effi-
ciency. As shown in Fig. 5共a兲, the changes in mobility and
photogeneration cross section had the largest impact on a PR
speed. However, it should be mentioned that an independent
variation of the photogeneration cross section is more justi-
fied than an independent variation of mobility, because the
recombination rate is mobility-dependent and affects the PR
speed the opposite way to the mobility itself. Thus, in a real
system, the mobility dependence will be smaller than that
shown in Fig. 5共a兲.
We now explore the history dependence of the PR speed.
As an example, we consider a composite with deep traps that
led to decay of the dc photoconductivity and the slow growth
of the ionized acceptor density. The parameters used in this
simulation were experimentally observed typical values for
the unplasticized composite PVK/AODCST/C60 共Sec. IV兲at
an electric field of 40 V/
m and total light intensity of 1
W/cm2.
We simulated a FWM experiment using a fresh sample
and both beams being turned on simultaneously, then the
same experiment after illuminating the sample with one
beam for 10 s and then turning on another one and so on up
to homogeneous illumination with one beam for 5000 s prior
to the PR experiment. The results of the simulation are pre-
sented in Fig. 5共b兲and show that there is a substantial history
FIG. 5. Dependence of PR speed on 共a兲photogeneration cross section and
charge carrier mobility and 共b兲the time of the homogeneous illumination
prior to PR experiment, as calculated from Eqs. 共11兲–共13兲.
1734 J. Appl. Phys., Vol. 92, No. 4, 15 August 2002 O. Ostroverkhova and K. D. Singer
dependence of the PR speed, in particular the response time
degradation in materials with deep traps, as reported in pre-
vious measurements.31,32 The material will relax back to its
initial 共‘‘fresh’’兲state in the dark when the traps empty, and
the released charge recombines with ionized acceptors. The
relaxation time is mostly determined by the trap depth.
III. EXPERIMENT
A. Materials
For our experiments we chose PVK as a photoconduc-
tive polymer, C60 as a sensitizer, BBP as a plasticizer, and
the chromophores AODCST, PDCST, and 5CB.33 One class
of composites under investigation included the molar con-
centrations of the following: PVK(99%)/C60(1%) 共compos-
ite 1, Tg⬃230 °C兲and PVK(89%)/C60(1%)/NLO(10%)
where NLO is a nonlinear chromophore which in our case
was represented by AODCST 共composite 2, Tg⬃133 °C兲,
PDCST 共composite 3, Tg⬃137 °C兲,or5CB共composite 4,
Tg⬃130 °C兲. Another class had the plasticizer at a loading
complementary to the chromophore molar concentrations:
PVK(49%)/C60(1%)/BBP(50%– x%)/AODCST(x%) and
PVK(49%)/C60(1%)/BBP(50%– x%)/5CB(x%), where x
was varied from 0% to 40%. Studying the chromophore con-
centration dependence of such composites where the chro-
mophore is substituted with the plasticizer rather than just
being added provides for consistent orientational effects
since the glass transition temperature (Tg) was near room
temperature of ⬃21 °C for all the concentrations. In particu-
lar, Tgof AODCST-containing plasticized composites re-
mained constant within 1 °C as xchanged from 2% to 40%
(Tg⬃20 to 21 °C兲, and glass transition temperature of 5CB-
containing plasticized composites increased from ⬃21 °C at
x⫽5% to ⬃24°C at x⫽40%. Thus orientational effects are
similar for all the composites, as we proved by an orienta-
tional dynamics study using electric field induced second
harmonic generation 共EFISHG兲.13 Also, the photoelectric
properties of the materials connected to charge transport
共mobility, trapping, and recombination rates兲change with
temperature relative to Tg,⌬T⫽Texp⫺Tg, where Texp is the
temperature at which the experiment is conducted.3,34,35
Therefore we ensured that ⌬Tis almost the same for all our
composites. We used purified materials and freshly made
samples for all our experiments since we found that both
chemical impurities and sample aging over a period of sev-
eral months led to deep trap formation, which would con-
found the data.
Sample preparation included the following steps. First,
PVK was dissolved in a mixture of toluene and cyclohex-
anone wt. 4:1. Then, C60 was dissolved in toluene and added
to a solution of PVK. Finally, the dye and the plasticizer
were added to a solution of PVK and C60 . The volume con-
centration of C60 was calculated to be NA⬇3.8⫻1024 m⫺3.
For mobility and photogeneration efficiency measurements,
we prepared spin-coated samples on an Al substrate, while
another Al electrode was deposited directly on top of the
films. For photorefractive measurements we first prepared
the films on the indium tin oxide 共ITO兲slides and then sand-
wiched them together and baked them in a vacuum oven at
120 °C for 1 h.
The thickness of the samples for mobility and photoge-
neration efficiency measurements 共Al electrodes兲was
⬃5–10
m, and ⬃30–70
m for the photorefractive
samples 共ITO electrodes兲. We used both types of electrodes
for dc photoconductivity measurements to ensure that the
parameters calculated from the photocurrent transients are
not influenced by the type of the electrode or sample thick-
ness. The dielectric constant was measured at a frequency of
1 Hz using a capacitance bridge.
B. Photoconductive measurements
Equation 共1兲introduced in Sec. II is valid for the bulk
material of infinite extent. In real experiments, the external
circuit and electrodes may influence the observed
behavior.19,36,37 Because of the high degree of disorder and
field dependence of all of the photoelectric parameters in
polymers, a systematic analysis of current–voltage character-
istics for different electrodes and their combinations is nec-
essary to fully elucidate the electrical characteristics.38–41
Schildkraut and Cui18 found good agreement between
the steady state values for free carrier density, amplitude, and
phase of space-charge field calculated from the bulk dynamic
equations 关e.g., Eq. 共1兲兴 for no boundary conditions 共infinite
bulk兲, ohmic boundary conditions 共‘‘infinite’’ supply of
charge from the electrodes兲, and blocking boundary condi-
tions 共Schottky barrier兲. We performed the dc photoconduc-
tivity experiments with two types of electrodes, Al and ITO,
and for our experimental conditions 共range of electric fields
and intensities兲did not find differences between the param-
eters calculated from the photocurrent transients. Thus we
assume Eq. 共1兲approximates our samples well.
1. dc photoconductivity (short time scale)
For the short time scale measurements 共t⬍50
0, where
0is the average lifetime of a free carrier兲, we applied an
electric field to the sample and waited until all the transient
processes disappeared, then opened a shutter 共switching time
below 40
s兲and recorded the sample current under 633 nm
illumination with an oscilloscope. For unplasticized compos-
ites the time span of this short time scale experiment was ⱗ4
s, while for the composites with plasticizer it was ⱗ40 s
depending on the applied electric field. We performed this
experiment for various electric fields and incident intensities.
The photocurrent transients were then fit to a biexponential
function
fit⫽A关1⫺Be⫺10t⫹共B⫺1兲e⫺20t兴共14兲
and the product of trapping rate and density of available
shallow traps
␥
TMT1, the recombination rate
␥
, and the shal-
low trap detrapping rate

1were determined as functions of
intensity and electric field in accordance with the procedure
described in Sec. II A 1 a. From the electric field dependence
of the trapping rate, we calculated the parameter

␥
关defined
in Eq. 共2兲兴.
1735J. Appl. Phys., Vol. 92, No. 4, 15 August 2002 O. Ostroverkhova and K. D. Singer
2. Dark conductivity and dc photoconductivity (long
time scale)
For dark conductivity and the long time scale (t
⬎103
0) photoconductivity measurements, we monitored
the current through the sample for ⬃20 min using a Keithley
6517 electrometer. A typical experimental run for plasticized
samples included the following: the electric field was turned
on, and the dark current (jdark) was recorded for ⬃20 min.
Then, the electric field was turned off and then in 2 min
turned on again, and dark current was monitored again for
⬃20 min. If the first two dark current runs reproduced, then
the light was turned on, and the current under illumination
(jlight) was monitored for another ⬃20 min. Then, the pho-
tocurrent jphoto was calculated using jphoto⫽jlight⫺jdark .If
the first two dark current runs did not reproduce, a third dark
current run was executed, and for all the samples under study
the third run reproduced the second.
In unplasticized composites the dark current is due to
injection from the electrodes. It reaches steady state in sev-
eral seconds and is at least an order of magnitude smaller
than the photocurrent at the incident intensity 20 mW/cm2at
an applied field of E0⬃40 V/
m. In plasticized samples the
dark current is due to both injection from the electrodes and
to impurity ions moving towards opposite electrodes. It
reaches a steady state only after 20–30 min and at low elec-
tric fields (E0⬃10 V/
m) is ⬃30% of the photocurrent at
the incident intensity I0⬃40 mW/cm2, so it is important to
take careful measurements of dark current to reliably deter-
mine the photocurrent.28
The long time scale photoconductivity measurement was
performed as a function of applied field and incident inten-
sity. The photocurrent transients were fit to Eq. 共14兲, and the
product of trapping rate and density of available deep traps
␥
TMT2and the detrapping rate

2for deep traps were deter-
mined as described in Sec. II A 1 b.
3. Mobility
Mobility was measured using the time-of-flight 共TOF兲
technique. We used a wavelength of 320 nm that is the third
anti-Stokes of H2—stimulated Raman-shifted 532 nm of a
Nd:YAG laser with a 3.5 ns pulse of ⬃5
J/pulse. The
sample current was transformed to a voltage, amplified, and
monitored with an oscilloscope. Then, mobility was calcu-
lated using a log-log plot for determining a transient time tT
and the relation
⫽d2/(tTV), where dis the thickness of the
sample and Vis the voltage applied. We measured mobility
as a function of electric field to determine the parameter

defined in Eq. 共2兲that describes the mobility field depen-
dence. For unplasticized samples and plasticized samples
with low chromophore content 共⬍10% concentration兲the
TOF transients could be easily resolved at electric fields E
⬎20 V/
m. The transients for high chromophore content
plasticized samples 共⬎10% chromophore concentration兲at
all electric fields and low chromophore content plasticized
samples at electric fields E⬍10 V/
m yielded TOF tran-
sients that were too dispersive for reliable determination of
the mobility from TOF experiment. In this case we estimated
the mobility from the dc photoconductivity measurements
using the formula4
⫽d2(
␥
TMT1)/(2 V). This estimate
showed reasonable agreement with TOF results for unplasti-
cized and low chromophore content plasticized samples at
higher electric fields (E⬎20 V/
m), thus we used it to cal-
culate the mobility for the samples which could not be accu-
rately measured by the TOF technique.
4. Photogeneration efficiency
The xerographic discharge technique was used for this
measurement for unplasticized samples and plasticized
samples at higher electric fields (E⬎20 V/
m). The sample
was charged to an initial voltage and then disconnected from
the power supply. The decay under HeNe illumination of 633
nm was monitored using a static voltmeter and an oscillo-
scope. The cross section of photogeneration swas calculated
from the slope of the discharge rate versus illumination in-
tensity:
兩
(dV/dt)
兩
light⫽关sd2eNA/(
⑀
0
⑀
)兴I0, where I0is the
intensity of light, dis a thickness of the sample,
⑀
is a di-
electric constant, and NAis a number density of C60 .We
also estimated the photogeneration cross section from the dc
photoconductivity as described in Ref. 4 using the formula
s⫽(
0
␥
TMT1)/(I0NA) where
0is the maximum free
charge density. We ensured consistency of this method by
comparing the photogeneration cross sections determined
from dc photoconductivity and xerographic discharge for un-
plasticized and plasticized samples at higher fields where the
xerographic discharge technique provided reliable data. For
plasticized samples at low electric fields (E⬍10 V/
m) re-
liable measurements of xerographic discharge could not be
obtained because of relatively large dark current. In this case,
we calculated the photogeneration cross section sfrom the dc
photoconductivity.Analyzing the electric field dependence of
s, we obtained the photogeneration field-dependence param-
eter pof Eq. 共2兲for each composite.
C. Photorefractive measurements
The diffraction efficiency was measured in a degenerate
four-wave mixing geometry. The grating was written with
two s-polarized HeNe 633 nm beams of the same internal
intensity with a total intensity varying from 25 to 400
mW/cm2. The probe beam was p-polarized with intensity 5
mW/cm2. The external angle between the crossing beams
was 28°, and the external angle between the sample normal
and the bisector of two writing beams was 50°. This experi-
mental geometry along with the index of refraction n⫽1.637
yielded a diffraction grating of period ⌳⬇1.8
m. A typical
experiment for unplasticized samples included the following
steps: first, we applied the electric field 共⬃30 V/
m兲with
one writing beam and the probe beam on, then in 10 s when
all the transients disappear, we opened the other writing
beam with a shutter and recorded the diffracted signal with a
photodetector, lock-in amplifier, and computer. After the dif-
fracted signal reached the quasi-steady state 共⬃20 s兲, the
light was blocked and the field was turned off. The samples
were kept in the dark for ⬃30 min before the next measure-
ment to assure complete decay of the space charge field and
the absence of ionized acceptors and filled traps. For plasti-
cized samples, the experimental run was similar, with the
1736 J. Appl. Phys., Vol. 92, No. 4, 15 August 2002 O. Ostroverkhova and K. D. Singer
difference being that we used a lower electric field 共⬃10
V/
m兲and waited for at least 60 s after turning the electric
field on to ensure that most of the impurity ions have accu-
mulated at the electrodes.42 Also, for plasticized samples, we
monitored the diffracted signal following homogeneous illu-
mination of various durations 共up to 1 h兲to study the influ-
ence of deep traps on the photorefractive signal for different
chromophores and chromophore concentrations.
We fit the PR grating formation dynamics with a biex-
ponential
⫽
0关1⫺ae⫺
1t⫺共1⫺a兲e⫺
2t兴2.共15兲
In unplasticized composites the faster speed
1was attrib-
uted to photoconductivity and dominated with weight a
⬇0.7 to 0.8 共inset of Fig. 7兲. The slower speed
2in these
composites is due to orientational enhancement which was
verified experimentally by comparing the ratio of p-polarized
and s-polarized diffraction efficiencies
p/
sto this ratio
calculated from geometry of the experiment and the electro-
optic effect.43 The slower speed was intensity independent
and on the order of 0.5–1 s⫺1, which is consistent with our
EFISHG measurements of the dynamics of chromophore re-
orientation in these unplasticized materials.13
In plasticized composites, both the faster (
1) and
slower (
2) speeds were attributed to photoconductivity and
varied from 0.1 to 10 s⫺1and from 0.01 to 1 s⫺1, respec-
tively, depending on the chromophore concentration, applied
electric field, and incident light intensity. Thus the chro-
mophore reorientation time constant of about 50 ms 共as de-
termined by EFISHG兲which is faster than both
1and
2
does not contribute.
IV. RESULTS
A. Unplasticized composites
In this section we determine the photoconductivity pa-
rameters for composites 1– 4 and discuss the differences in
the parameters depending on the chromophores. Then based
on these parameters we model the dynamics of photorefrac-
tive grating formation and compare it to experiment.
1. Photoelectric properties
The photoconductivity parameters and rates introduced
in Sec. II and calculated from the TOF, xerographic dis-
charge, and dc photoconductivity 共at incident intensity I0
⫽100 mW/cm2兲experiments described in Sec. III at the
electric field E0⫽40 V/
m are summarized in Table I. All
the trends we describe in this section are applicable for com-
posites 1– 4 in the studied electric field range of E
⬇20– 80 V/
m. Diagrams describing charge generation,
transport, and trapping for different composites are shown in
Fig. 6. Here we used the relative highest occupied molecular
orbital 共HOMO兲level energies for PVK 共0eV兲,C60 共⫺0.17
eV兲, AODCST 共0.03 eV兲, PDCST 共0.1 eV兲, and 5CB
共⬍⫺0.4 eV兲provided in Ref. 8.
Photogeneration cross-section sis a measure of quantum
efficiency and for low absorption is given by the relation s
⫽
␣
/(ប
NA) where
␣
is the absorption coefficient,
quantum efficiency, and
is the light frequency. Photoge-
neration cross section microscopically depends on the
donor–acceptor charge transfer and electron–hole dissocia-
tion rates.44 According to their HOMO levels, chromophores
AODCST and PDCST as well as PVK are donors with re-
spect to C60 , and thus could participate in photogeneration.
If we take into account the dependence of the charge transfer
rate kCT on the energy difference ⌬EDA between HOMO
levels of donor and acceptor44,45 kCT⬃exp关⫺(⌬EDA
⫺)2/(4kBT)兴where is the reorganization energy, then
in the noninverted regime, the photogeneration efficiency
would be highest for composite 3, followed by composites 2,
1, and 4. The noninverted regime refers to the case ⌬EDA
⬍, which seems to describe our composites.44 However,
the donor–acceptor charge transfer is not the only factor that
contributes to charge generation. The other factor is the
electron–hole dissociation,44 which proceeds more strongly
as the mobility increases. Based on our results for photoge-
neration cross sections and mobilities for composites 1– 4
共Table I兲, we conclude that the mobility differences that af-
fect dissociation in addition to the charge transfer rate kCT
could account for the differences in quantum efficiencies for
the composites 1– 4.
The hole mobility of all of the composites 2– 4 is smaller
than that for the PVK/C60 共composite 1兲. This is expected
when adding polar chromophores to the system due to the
FIG. 6. Schematic representation of the intrinsic states and processes of a
PR composite with various chromophores.
TABLE I. Photoelectric parameters relevant for photorefraction for composites 1–4. No plasticizers were added.
Composite s,10
⫺5
m2/J
,10
⫺11
m2/共Vs兲
␥
TMT1,
s⫺1
␥
TMT2,
s⫺1

1,
s⫺1

2,
s⫺1
␥
,10
⫺19
m3/s

,10
⫺4
共m/V兲1/2

␥
,10
⫺4
共m/V兲1/2 p
⑀
1 Neat 1.7 5.3 850 2.7 2.5 4.9⫾0.2 8.3⫾0.5 2.04⫾0.06 3.1
2 AODCST 1.6 3.9 300 3 0.75 8.5⫻10⫺40.09 5.8⫾0.5 3.3⫾0.4 2.04⫾0.04 4.3
3 PDCST 0.84 2.5 770 23 0.3 1.5⫻10⫺40.08 4.9⫾0.2 6.5⫾0.3 2.04⫾0.04 3.8
4 5CB 0.83 1.7 550 1.5 0.09 6.0⫾0.3 4.6⫾2.0 2.1⫾0.2 4.5
1737J. Appl. Phys., Vol. 92, No. 4, 15 August 2002 O. Ostroverkhova and K. D. Singer
increase in energetic disorder.20,46 Based only on the dipole
moment of the chromophore, composites 2 and 3 would be
expected to yield similar mobilities due to almost equal di-
pole moments 共6.9 D for AODCST and 6.6 D for PDCST兲,
and composite 4 would yield the higher mobility than the
composites 2 or 3 since the dipole moment of 5CB 共4.1 D兲is
lower than that of both AODCST and PDCST. Our results
show that the composite 2 possesses the largest value for
mobility out of three composites with chromophores 共2–4兲.
We attribute this to the fact that the HOMO level of
AODCST is situated inside the transport manifold of PVK,
so that AODCST molecules participate in transport by in-
creasing the density of transport states.
All three chromophores influence the PVK/C60 system
modifying the position and depth of shallow traps that are
intrinsic to PVK.3Here we need to make a distinction be-
tween shallow traps whose release 共detrapping兲time is much
smaller than the transit time, in other words shallow traps
that broaden the tail of the current transients observed in our
TOF experiment, and traps whose release time is much larger
than the transit time. The latter traps are relevant for photo-
refractive performance of the polymer composites considered
here.
The parameters describing shallow traps are the trapping
rate
␥
T, the total number density of available shallow traps
MT1, and the detrapping 共releasing兲rate

1. Comparing de-
trapping rates

1for composites 1– 4 共Table I兲, we observe
that the addition of all chromophores leads to a decrease in
detrapping rate

1that may arise from a decrease in the
overlap integral 共intersite distance兲, an increase in the trap
energy depth ⌬ET共Fig. 6兲,24,47 or both. Both AODCST and
PDCST have a larger impact on

1than 5CB since they can
provide sites that can serve as deeper traps. Since in our
experiments we cannot measure the trapping rate
␥
Tand
total trap density MT1, but only the product
␥
TMT1, we can
only speculate about possible contributions of each. The
trapping rate
␥
Tat a given electric field is expected to de-
pend on the free hole mobility and the neutral trap capture
cross section.24 In this case we should expect a smaller trap-
ping rate
␥
Tvalue for the composites 2– 4 in comparison to
composite 1 due to decreased mobility. In composite 3 the
product
␥
TMT1is larger than that of composites 2 and 4
which could indicate that PDCST actually adds shallow traps
to the system in addition to just changing the average depth
of existing shallow traps in pure PVK. On the contrary, com-
posite 2 has the smallest product
␥
TMT1which could mean
that AODCST reduces the relative density of relevant shal-
low traps intrinsic to PVK by providing extra transport sites.
Similar to shallow traps, we characterize deep traps by their
number density MT2and detrapping rate

2. We could not
detect deep traps in the PVK/C60 system and in composite 4
共with 5CB as chromophore兲which means that the density of
available deep traps (MT2) in these composites is on the
order of 1% or less of the acceptor density NA. Both
AODCST 共composite 2兲and PDCST 共composite 3兲create
deep traps, with the trap depth larger for PDCST than for
AODCST which is consistent with experimental studies per-
formed with these chromophores in Ref. 8 and with our nu-
merical simulations. From the comparison between
␥
TMT2
values for composites 2 and 3 we conclude that PDCST cre-
ates more deep traps than AODCST as would be expected by
their HOMO levels.
The recombination rate
␥
describes interaction of the
free hole with the ionized acceptor (C60
⫺) and usually is
treated as Langevin bimolecular recombination
␥
⫽e
/(
⑀⑀
0).24 According to this relation, the ratio
␥
/
de-
pends only on dielectric constant
⑀
of the material. In poly-
mers, due to disorder, deviations from the Langevin form are
observed.24 Our measured recombination rate for PVK/C60
shows good agreement with the value obtained using the
Langevin form. However, the measured values for compos-
ites 2– 4 are considerably lower than the corresponding val-
ues calculated from the Langevin form.
2. Photorefractive properties
In this section we applied all the photoelectric param-
eters we determined above to Eq. 共11兲. When considering the
space-charge field formation, we used the values for mobil-
ity, trapping, and recombination rates calculated using the
value of the projection of the electric field on grating vector
while the photogeneration cross section was calculated using
the applied electric field. We solved Eq. 共11兲to determine the
dynamics of photorefractive grating formation 关E1(t)兴for
the composites 2– 4. Then, we calculated the diffraction ef-
ficiency signal time evolution as it appears in the four wave
mixing 共FWM兲experiment48
⬃E1(t)2and fit with the
function of Eq. 共13兲to predict the photoconductive part of
photorefractive speed
. The calculated speed
is to be com-
pared with the measured
1introduced in Sec. III C. The
anticipated speed
as a function of total internal intensity of
two beams for the composites 2– 4 is shown in Fig. 7 共lines
with symbols兲. The measured 共as described in Sec. III C兲
faster component of the photorefractive speed
1for com-
posites 2– 4 is also shown in Fig. 7 共symbols兲and is in a
reasonable agreement with the speed
predicted using ex-
FIG. 7. Photorefractive speed as a function of intensity for composites 2–4:
line with symbols represents a theoretically predicted speed for this com-
posite 共no adjustable parameters兲using Eqs. 共11兲–共13兲; symbols represent
actual PR speed determined from the FWM experiment. The inset shows a
typical transient measured in composite 2 at a total intensity of 125
mW/cm2.
1738 J. Appl. Phys., Vol. 92, No. 4, 15 August 2002 O. Ostroverkhova and K. D. Singer
perimentally determined photoelectric parameters for these
composites. Both calculated and experimentally measured
PR speed are sublinear functions of light intensity 共
,
1
⬃Ia, where a⫽0.6 to 0.7 depending on a composite兲, simi-
lar to intensity dependencies of simulated and measured pho-
toconductivity 共taken at maximal photocurrent value, see
Fig. 2兲in these composites.
B. Plasticized composites
In this section we consider the dependence of the PR
properties of composites on the chromophore and plasticizer
concentration. First, we consider how substitution of the
chromophore molecule by the plasticizer 关increase in param-
eter x(%) introduced in Sec. III A兴affects the photoelectric
parameters such as mobility, photogeneration efficiency, re-
combination, trapping, and detrapping rates. We studied this
for two chromophores, 5CB and AODCST, to determine the
influence of the chromophore ionization potential on these
characteristics. Second, we used the parameters determined
from photoconductivity to model the time evolution of a
photorefractive grating 共diffracted signal observed in FWM
experiment兲and compare it with experimental data for dif-
ferent concentrations. Finally, we discuss the dependence of
the diffracted signal on the illumination history for different
concentrations of AODCST and 5CB.
1. Photoelectric properties
The most distinct feature of the plasticized composites in
comparison to unplasticized ones is the presence of large
dark current observed in ‘‘fresh’’ samples that were not ex-
posed to either electric field or illumination. In this case the
dark current is caused by both charge injection from the elec-
trodes and native ionic impurities49 that become mobile un-
der the electric field due to the conformational freedom of
the plasticized polymer chains.24 As the sample is kept under
electric field, the mobile ions move towards the oppositely
charged electrodes and either neutralize42 or build up, reduc-
ing the electric field inside the polymeric film. The samples
of all concentrations xfor both 5CB and AODCST showed
similar behavior. After the transient, the current gradually
decreases until it reaches a quasi-steady level, which in our
materials occurs at time ⬃20–30 min after the electric field
is turned on.28 We performed all the measurements after the
samples were electrically cleansed to avoid dynamic effects
directly induced by moving impurity ions. Although we tried
to maintain exactly the same experimental conditions for the
samples at all concentrations, our measurements of both pho-
toconductivity and diffraction efficiency at different times
after turning on the electric field without any prior illumina-
tion show that internal electric fields are different for differ-
ent concentrations of the chromophore. However, these ef-
fects were minor in comparison with the direct concentration
effects that we investigated. The energy diagram illustrating
different composites studied is shown in Fig. 8. Based on the
ionization potentials of the plasticizer and chromophores, we
expect that an increase in concentration of 5CB 共relative to
the concentration of plasticizer BBP兲should not change the
mobility, photogeneration efficiency, trapping, and other
photoelectric parameters for reasons other than an increase in
energetic disorder due to the difference in dipole moments of
5CB (4.1D) and BBP (1.1D). In the case of AODCST, how-
ever, in addition to the change in energetic disorder, we also
expect changes due to the contribution of AODCST sites in
charge photogeneration and transport. Indeed, our results
show that both mobility and photogeneration cross section
increase with concentration of AODCST and stay almost
constant for all concentrations of 5CB. Although these trends
were observed for the whole range of electric fields studied
共1–50 V/
m兲, the most pronounced concentration depen-
dence was found at low electric fields 共⬍15 V/
m兲. This
could be due to the smaller influence of energetic disorder at
low fields, so that the largest impact on concentration depen-
dence is attributed to the effects only due to ionization po-
tential of the chromophores relative to transport states. The
concentration dependence of mobility and photogeneration
cross section for 5CB and AODCST at electric field E0
⫽10 V/
m is shown in Fig. 9. As determined from dc pho-
toconductivity at electric field E0⫽10 V/
m and intensity
I0⫽40 mW/cm2, the product
␥
TMT1increased monotoni-
cally in a similar to sand
manner from 3.5 s⫺1for x
⫽0% 关PVK(49%)/BBP(50%)/C60(1%)兴to 20 s⫺1for x
⫽40% in the case of AODCST and did not change in the
case of 5CB. This change reflects the increase in mobility
and intersite distance that affect the trapping rate for
AODCST and no changes in these for 5CB.28 The detrapping
rate

1increased from ⬃0.05 to ⬃0.1 s⫺1for AODCST,
reflecting increase in the overlap integral 共decrease in inter-
site distance兲. The recombination rate was ⬃2⫻10⫺21 m3/s
and did not change appreciably with concentration, probably
because the increase in mobility in the case of AODCST was
partially compensated by an increase in dielectric constant as
the concentration of AODCST increased.
The presence of deep traps in the composites was stud-
ied by monitoring dc photoconductivity on a long time scale
共Sec. II A 1 b兲.28 The composites with x⫽0 and with any
FIG. 8. Illustration of chromophore and plasticizer roles in charge genera-
tion, transport, and trapping: 共a兲high x%, AODCST; 共b兲low x%,
AODCST; and 共c兲any x%, 5CB.
1739J. Appl. Phys., Vol. 92, No. 4, 15 August 2002 O. Ostroverkhova and K. D. Singer
concentration of 5CB showed no photodegradation and thus
the deep trap density in these composites was below our
detection limit of MT2⭐0.01NA. Our numerical simulations
of PR dynamics show that when the available density of
deep traps is on the level MT2⬃0.01NAor less 共provided
shallow trap-unlimited regime MT1ⰇNA兲, the PR grating
time evolution is not influenced by deep traps. In the
AODCST-containing composites, the degradation of the pho-
tocurrent increased as the concentration of AODCST in-
creased, so that the product
␥
TMT2describing deep trapping
increased from ⬃0.02 s⫺1for x⫽2% to ⬃1.6 s⫺1for x
⫽40%, and the detrapping rate

2correspondingly changed
from 1.1⫻10⫺5to 8.2⫻10⫺4s⫺1, although the values for
low concentrations of the chromophore may contain a large
error due to the small concentration of deep traps MT2in
these composites. Similar to the method described in Sec.
IV A 2, we substitute all the calculated values into the equa-
tions describing photorefractive grating formation 关Eq. 共11兲兴
and model the PR performance of the composites. We fit the
calculated diffraction efficiency
(t) with a biexponential
function 关Eq. 共15兲兴 since, as we mentioned above for these
samples, both the slow and fast components of the photore-
fractive speed are due to photoconductivity. The calculated
faster and slower PR speed 共
1and
2, respectively兲as a
function of concentration of the chromophore with no adjust-
able parameters is presented in Fig. 10共a兲共line with symbols
corresponding to concentrations for which we determined
photoelectric parameters on whose basis the PR speed was
calculated兲.
2. Photorefractive properties
The faster and slower PR speed 共
1and
2兲obtained
from the biexponential fit of Eq. 共15兲to the experimentally
measured diffraction efficiency for different concentrations is
plotted in Fig. 10共a兲共symbols兲along with a characteristic fit
关Fig. 10共b兲兴. Faster PR speed in both 5CB and AODCST-
containing composites shows reasonable agreement with the
values of
1calculated from photoconductivity 共line with
symbols兲. The experimentally measured slower speed
2
shows good agreement with calculated values for 5CB-
containing composites. Although we could not find mono-
tonic dependence of slower PR speed on various parameters,
our simulations show that at low photogeneration efficiencies
and charge carrier mobilities electric field dependence of
photoelectric rates 共i.e., nonzero parameters p,

,

␥
兲leads
to nonsingle exponential behavior of space-charge field in
materials without deep traps, such as our 5CB-containing
composites.
In materials with deep traps, the slower component is
influenced by trapping rate
␥
T, deep trap density MT2, and
detrapping rate

2. The slower PR speed in AODCST-
containing composites shown in Fig. 10共a兲is calculated us-
ing shallow and deep trap unlimited approximation 共Case 3兲.
As seen from Fig. 10共a兲,Case 3 describes well composites
with high chromophore content. However, at low chro-
mophore concentrations, it seems that the deep trap limited
Case 2 needs to be considered.27,28 In these materials, it is
not enough to know the trapping parameter
␥
TMT2that we
are able to determine from the photoconductivity, but knowl-
edge of the trapping rate
␥
Tand trap density MT2separately
is required to predict the slower PR speed.
According to our simulations, the materials with deep
traps are expected to show a change in both rise and decay
FIG. 9. Concentration dependence of 共a兲photogeneration cross section and
共b兲mobility. The lines provide visual guidance.
FIG. 10. 共a兲Concentration dependence of faster and slower PR speed for
5CB- and AODCST-containing composites. Lines with symbols correspond
to the photoelectric parameter values for PR speed calculated using Eqs.
共11兲,共12兲, and 共15兲; symbols correspond to FWM experimental data at ap-
plied field Ea⫽10 V/
m and total internal intensity I⫽300 mW/cm2.共b兲
Data and fit with Eq. 共15兲to two composites.
1740 J. Appl. Phys., Vol. 92, No. 4, 15 August 2002 O. Ostroverkhova and K. D. Singer
transients due to the duration of uniform illumination prior to
the experiment, which is supported by our experimental data.
Figure 11 illustrates the diffraction efficiency time evolution
in ‘‘fresh’’ samples and preilluminated samples of the com-
posites containing x⫽40% of 5CB 关Fig. 11共a兲兴 and x
⫽40% of AODCST 关Fig. 11共b兲兴. The experimental run con-
sisted of the following steps. First, the electric field of 10
V/
m was turned on and kept on for 60 min without any
illumination. Then, one of the writing beams of intensity 150
mW/cm2was turned on, and then after ⬃20 s the second
writing beam of the same intensity was turned on, and the
diffraction grating formation was monitored for 60 s. Then,
one writing beam was turned off, and the grating decay was
monitored. The second writing beam 共with the applied field
on兲was illuminating the sample for 5–15 min, and then
again the second beam was turned on and grating formation
monitored. To ensure that the effects we observe in this ex-
periment are due to illumination and not due to internal fields
formed by uncompensated traps and impurity ions, we per-
formed a similar experiment but without any illumination
between the grating decay and formation measurements. Our
results show that all the dependencies described here and
shown in Fig. 11 are due to illumination only. The 5CB-
containing composite showed no substantial history depen-
dence because the dynamics involves only shallow traps, and
the equilibrium between photogeneration, trapping, detrap-
ping, and recombination processes in the system is reached
within several seconds and then does not change over a long
time scale. In such systems, the long time scale illumination
does not change the density of ionized acceptors and filled
traps, and so the initial conditions for the onset of diffraction
are the same at any time. Thus there is no illumination his-
tory dependence 共Fig. 11兲. The behavior is different for the
composites with high AODCST concentration. After the
sample illuminated for 5 min, the grating formation and de-
cay are substantially different in comparison with those ob-
served in the fresh sample. This is because after illumination
the initial conditions for the grating formation change dra-
matically due to deep trap filling and ionized acceptor
growth, so that deep traps noticeably contribute to the grat-
ing dynamics, as predicted by numerical simulations.27,28
The illumination history dependence can be undesirable
for applications requiring a long time period grating
formation/decay repetition in the presence of illumination
and, thus materials containing deep traps are not suitable for
these applications.
V. CONCLUDING REMARKS
We have carried out a detailed investigation of the pro-
cesses that affect the PR speed. Although it has been previ-
ously experimentally demonstrated that photogeneration
cross section and mobility have a large impact on PR speed,
our calculations confirm this. However, we also demonstrate
the role of other parameters such as recombination rate, trap-
ping rate, and density of traps. We have shown the subtle role
that traps play in photorefractive dynamics. We have also
shown how the PR dynamics have a rather complicated
form, but have nonetheless demonstrated how biexponential
fits to the data can be useful when carried out over the ap-
propriate time scales.
We considered two kinds of thermally accessible traps,
shallow and deep. We attribute the shallow traps to structural
defects and conformational traps of the carbazole units50 in
the photoconductor itself 共PVK兲, which can be affected by
the presence of chromophores. Although there could be some
deep traps in PVK itself, the essential contribution in the
density of deep traps that may affect the space-charge field
formation is determined by the chromophores. Deep traps
lead to slow growth of ionized acceptor density, as previ-
ously observed,8which leads to complicated long time scale
PR dynamics and illumination history dependence of both
photoconductive and photorefractive properties.
Ionized acceptor density (NA0
i) was identified earlier in
the literature8,22 as ‘‘deep photoexcitable photorefractive trap
density,’’ since it turned out to be approximately equal to
filled trap density (M10⫹M20) measured in the two-beam
coupling experiment.8We believe that photorefractive traps
in polymer composites are not photoexcitable and are the
ones we discussed above, which are due to structural defects
in PVK and to the position of the ionization potential of a
chromophore with respect to the photoconductor. The density
of such traps in the material and their depth, together with
other photoelectric rates intrinsic to the material, determines
the dynamics of growth of the ionized acceptors. However,
ionized acceptor density is indeed a good measure for filled
trap density, since the relation NAO
i⬇M10⫹M20 is always
true according to the charge neutrality equation 关last equation
in Eq. 共5兲兴 under the condition of free charge density (
0)
much smaller than filled trap densities (M10,20), which seems
to be the case in most PR polymer composites. For example,
if the same material is sensitized with different sensitizers
共meaning that the photogeneration cross section sis the only
parameter that changes in the system兲, the different filled trap
densities 共as manifested through different gain coefficients兲
FIG. 11. Influence of homogeneous illumination prior to FWM experiment
on diffracted signals for 共a兲x⫽40%, 5CB and 共b兲x⫽40% , AODCST at
applied electric field Ea⫽10 V/
m and total incident intensity I0
⫽300 mW/cm2.
1741J. Appl. Phys., Vol. 92, No. 4, 15 August 2002 O. Ostroverkhova and K. D. Singer
observed in the two-beam coupling experiment30 are due to
the difference in charge photogeneration efficiency, which
leads to the difference in ionized acceptor density and, there-
fore, filled trap density.
Experimentally, we studied a set of unplasticized com-
posites with 5CB, AODCST, and PDCST as chromophores
and chromophore concentration dependence of plasticized
composites with 5CB and AODCST as chromophores. We
applied the modified Schildkraut and Buettner’s model to
compare both photoconductive and photorefractive proper-
ties of the composites under investigation. We were able to
predict with no adjustable parameters the faster photorefrac-
tive speed for a variety of composites from the experimen-
tally measured values of the relevant photoelectric rates. The
slower PR dynamics was predicted with no adjustable pa-
rameters for composites with no deep traps 共composites with
5CB as a chromophore兲and with a high concentration of
deep traps 共high AODCST content composites兲. We also
studied the influence of deep traps on the illumination history
dependence of the photorefractive performance. The modi-
fied model seems to describe qualitatively the experimentally
observed trends.
Figure 10 points out the complexities of assessing pho-
torefractive speed and designing high speed materials. As
can be seen in that figure, the slow time constant in
AODCST composites is always slower than that for 5CB
composites due to more deep trapping, while the fast time
constant for AODCST composites can be faster than 5CB
composites due to a higher density of states created by
AODCST states within the transport manifold. The complete
fit in Fig. 10共b兲indicates the difficulties in assessing the PR
speed. Diffraction builds more quickly in AODCST compos-
ites, but stabilizes more slowly compared with 5CB compos-
ites. An optimum chromophore would possess a fast compo-
nent like the AODCST composite, but suppress a slow
component like the 5CB composite. Such a chromophore
would possess an ionization potential high enough and/or a
density of states narrow enough so that deep traps are not
formed even at high concentration to suppress the long time
constant, but an ionization potential low enough so that its
density of states overlaps with the transport manifold in-
creasing the initial speed of grating formation. The ionization
potential of 5CB is too high for this, while that of AODCST
may be too low to avoid deep trapping. We are not aware of
an optimized composite that exhibits both a fast short time
constant and a short or suppressed long time constant. As is
well-known, the presence of deeper traps slows the dynam-
ics, but increases the diffraction efficiency. This trade-off
must also be addressed depending on the requirements of the
application. We believe that this study provides a method to
assess the impact of trapping on dynamics in a quantitative
manner.
Despite some successful predictions we were able to
make using the modified model, there are some outstanding
issues. First of all, the composites with trap densities on the
level of MT⬃0.01⫺1NAcannot be treated the way elabo-
rated for the case MT1,2ⰇNA, and thus in many composites
the longer scale behavior of space-charge field cannot be
predicted as well as the shorter time scale behavior. Second,
the trapping, detrapping, and recombination rates (
␥
T,

,
␥
)
were found to depend sublinearly on light intensity28 which
is not explicitly included or predicted by the model, thus it
seems that some process has not been taken into account.
Third, it might be helpful to include the effect of the forma-
tion of the internal electric field inside the sample due to the
filling of uncompensated traps as well as non-neutralized im-
purity ions. Also, both photoconductivity and photorefractive
signals are sensitive to the presence of deep traps in the
composite, so that impurities that could serve as deep traps
can obscure the performance of a purer system. A final re-
mark is that Eq. 共1兲is written for an infinite bulk material
and does not take into account the possible effects of elec-
trodes. These issues require further study.
ACKNOWLEDGMENTS
The authors thank Professor R. Twieg for AODCST. Dr.
V. Ostroverkhov is acknowledged for assistance in the nu-
merical simulations. Dr. I. Shiyanovskaya and Dr. J. An-
drews are acknowledged for helpful discussions. The re-
search described in this article was made possible by the
support of the National Science Foundation through the AL-
COM Science and Technology Center 共DMR 89-20147兲, and
by the Air Force Office of Scientific Research, Air Force
Material Command, USAF 共F49620-99-1-0018兲.
1J. C. Scott, L. T. Pautmeier, and W. E. Moerner, J. Opt. Soc. Am. B 9,
2059 共1992兲.
2S. Schloter, A. Schreiber, M. Grasruck, A. Leopold, M. Kol’chenko, J.
Pan, C. Hohle, P. Strohriegl, S. J. Zilker, and D. Haarer, Appl. Phys. B:
Lasers Opt. 68, 899 共1999兲.
3T. K. Daubler, R. Bittner, K. Meerholz, V. Cimrova, and D. Neher, Phys.
Rev. B 61, 13515 共2000兲.
4E. Hendrickx, Y. Zhang, K. B. Ferrio, J. A. Herlocker, J. Anderson, N. R.
Armstrong, E. A. Mash,A. Persoons, N. Peyghambarian, and B. Kippelen,
J. Mater. Chem. 9, 2251 共1999兲.
5M. Grasruck, A. Schreiber, U. Hofmann, S. J. Zilker, A. Leopold, S.
Schloter, C. Hohle, P. Strohriegl, and D. Haarer, Phys. Rev. B 60, 16543
共1999兲.
6S. J. Zilker et al., Chem. Phys. Lett. 306, 285 共1999兲.
7A. Grunnet-Jepsen, C. L. Thompson, and W. E. Moerner, J. Opt. Soc.Am.
B15, 905 共1998兲.
8A. Grunnet-Jepsen, D. Wright, B. Smith, M. S. Bratcher, M. S. DeClue, J.
S. Siegel, and W. E. Moerner, Chem. Phys. Lett. 291, 553 共1998兲.
9B. Kippelen, Sandalphon, K. Meerholz, and N. Peyghambarian, Appl.
Phys. Lett. 68, 1748 共1996兲.
10 K. Hoechstetter, S. Schloter, U. Hoffmann, and D. Haarer, J. Chem. Phys.
110, 4944 共1998兲.
11 E. Mecher, C. Brauchle, H. H. Horhold, J. C. Hummelen, and K. Meer-
holz, Phys. Chem. Chem. Phys. 1, 1749 共1999兲.
12 K. S. West, D. P. West, M. D. Rahn, J. D. Shakos, F. A. Wade, K. Khand,
and T. A. King, J. Appl. Phys. 84, 5893 共1998兲.
13 O. Ostroverkhova, A. Stickrath, and K. D. Singer, J. Appl. Phys. 91,9481
共2002兲.
14 D. Wright, M. A. Diaz-Garcia, J. D. Casperson, M. DeClue, W. E. Mo-
erner, and R. J. Twieg, Appl. Phys. Lett. 73,1490共1998兲.
15 G. Bauml, S. Schloter, U. Hoffmann, and D. Haarer, Opt. Commun. 154,
75 共1998兲.
16 J. S. Schildkraut and A. V. Buettner, J. Appl. Phys. 72, 1888 共1992兲.
17 B. Yuan, X. Sun, C. Hou, Y. Li, Z. Zhou, Y. Jiang, and C. Li, J. Appl.
Phys. 88, 5562 共2000兲.
18 J. Schildkraut and Y. Cui, J. Appl. Phys. 72, 5055 共1992兲.
19 J. Mort and D. Pai, Photoconductivity and Related Phenomena 共Elsevier,
New York, 1976兲.
20 A. Goonesekera and S. Ducharme, J. Appl. Phys. 85, 6506 共1999兲.
21 Y. Cui, B. Swedek, N. Cheng, J. Zieba, and P. N. Prasad, J. Appl. Phys. 85,
38 共1999兲.
1742 J. Appl. Phys., Vol. 92, No. 4, 15 August 2002 O. Ostroverkhova and K. D. Singer
22 J. A. Herlocker, C. Fuentes-Hernandez, K. B. Ferrio, E. Hendrickx, P.-A.
Blanche, N. Peyghambarian, and B. Kippelen, Appl. Phys. Lett. 77, 2292
共2000兲.
23 N. V. Kukhtarev, V. B. Markov, M. Soskin, and V. L. Vinetskii, Ferroelec-
trics 22,961共1979兲.
24 M. Pope and C. Swenberg, Electronic Processes in Organic Crystals and
Polymers 共Oxford University, New York, 1999兲.
25 B. Movaghar, Philos. Mag. B 65,811共1992兲.
26 K. Meerholz, E. Mecher, R. Bittner, andY. Nardin, J. Opt. Soc.Am. B 15,
2114 共1998兲.
27 O. Ostroverkhova and K. D. Singer, Proc. SPIE 4462,163共2001兲.
28 O. Ostroverkhova, Ph.D. thesis, Case Western Reserve University, 2001.
29 Photorefractive Materials and Their Applications I, edited by P. Gunter
and J.-P. Huignard 共Springer-Verlag, Berlin, 1988兲.
30 O. Ostroverkhova, U. Gubler, D. Wright, W. E. Moerner, M. He, A.
Sastre-Santos, and R. Twieg, Adv. Fun. Mat. 共to be published兲.
31 S. M. Silence, G. C. Bjorklund, and W. E. Moerner, Opt. Lett. 19, 1822
共1994兲.
32 J. A. Herlocker, K. B. Ferrio, E. Hendrickx, B. D. Guenther, S. Mery, B.
Kippelen, and N. Peyghambarian, Appl. Phys. Lett. 74, 2253 共1999兲.
33 M. A. Diaz-Garcia, D. Wright, J. D. Casperson, B. Smith, E. Glazer, W. E.
Moerner, L. I. Sukhomlinova, and R. J. Twieg, Chem. Mater. 11, 1784
共1999兲.
34 M. Abkowitz, M. Stolka, and M. Morgan, J. Appl. Phys. 52, 3453 共1981兲.
35 M. Abkowitz, H. Bassler, and M. Stolka, Philos. Mag. B 63,201共1991兲.
36 R. Bube, Photoconductivity of Solids 共Wiley, New York, 1960兲.
37 Photoconductivity in Polymers, edited by A. Patsis and D. Seanor 共Tech-
nomic Westport, CT, 1976兲.
38 R. Blum, M. Sprave, J. Sablotny, and M. Eich, J. Opt. Soc.Am. B 15,318
共1998兲.
39 P. Blom, M. de Jong, and M. van Munster, Phys. Rev. B 55, R656 共1997兲.
40 L. Bozano, S. Carter, J. Scott, G. Malliaras, and P. Brock, Appl. Phys.
Lett. 74, 1132 共1999兲.
41 T. Daubler, D. Neher, H. Rost, and H. Horhold, Phys. Rev. B 59, 1964
共1999兲.
42 M. Pauley, H. Guan, and C. Wang, J. Chem. Phys. 104,6834共1996兲.
43 W. E. Moerner, S. M. Silence, F. Hache, and G. C. Bjorklund, J. Opt. Soc.
Am. B 11, 320 共1994兲.
44 E. Hendrickx, B. Kippelen, S. Thayumanavan, S. Marder, A. Persoons,
and N. Peyghambarian, J. Chem. Phys. 112,9557共2000兲.
45 P. Marcus and P. Siders, J. Phys. Chem. 86,622共1982兲.
46 R. Young, J. Chem. Phys. 103, 6749 共1995兲.
47 P. Borsenberger and D. Weiss, Organic Photoreceptors for Imaging Sys-
tems 共Marcel Dekker, New York, 1993兲.
48 W. E. Moerner, A. Grunnet-Jepsen, and C. Thompson, Annu. Rev. Mater.
Sci. 27, 585 共1997兲.
49 A. Blythe, Electrical Properties of Polymers 共Cambridge University,
Cambridge, England, 1979兲.
50 J. Slowik and I. Chen, J. Appl. Phys. 54, 4467 共1983兲.
1743J. Appl. Phys., Vol. 92, No. 4, 15 August 2002 O. Ostroverkhova and K. D. Singer