Double Extraction of Uniformly Generated ElectronHole Pairs from Insulators with

Noninjecting Contacts
Alvin M. Goodman and Albert Rose
Citation: Journal of Applied Physics 42, 2823 (1971); doi: 10.1063/1.1660633
View online: http://dx.doi.org/10.1063/1.1660633
View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/42/7?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Electron-hole pair generation in SiC high-temperature alpha particle detectors
Appl. Phys. Lett. 103, 152108 (2013); 10.1063/1.4824774
Enhanced photoluminescence from condensed electron-hole pairs in trenched Si
J. Appl. Phys. 108, 124305 (2010); 10.1063/1.3520663
Vibrational relaxation of NO on Au(111) via electron-hole pair generation
J. Chem. Phys. 125, 154703 (2006); 10.1063/1.2357740
Minimum probability of electronhole pair generation by blackbody radiation
J. Appl. Phys. 58, 575 (1985); 10.1063/1.335665
The generation of electronhole pairs in very thin solar cells possessing an optical backsurface reflector
J. Appl. Phys. 51, 6426 (1980); 10.1063/1.327593

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 210.212.192.132 On: Thu, 05
May 2016 09:08:25

VOLUME 42, NUMBER 7

JOURNAL OF APPLIED PHYSICS

JUNE 1971

Double Extraction of Uniformly Generated Electron-Hole Pairs from Insulators with

N oninjecting Contacts
ALVIN

M.

GoODMAN AND ALBERT ROSE

RCA Laboratories, Princeton, New Jersey 08540

(Received 5 June 1970; in final form 12 October 1970)

A simple physical model has been used to derive an approximate theory of the double extraction of
uniformly generated electron-hole pairs from a photoconductor layer with noninjecting contacts. The
theory predicts 4 regimes of current versus applied voltage behavior:
(1)
(2)
(3)
(4)

At low voltage, I ex: V.

At higher voltage, a transition region between lex: V and Iex:P/2.
At still higher voltage, I ex: V1/2.
At very high voltage (above a saturation value), I = constant.

The model relates the extent of these regimes to the physical parameters of the system, viz., J.lpTp (mobility
X lifetime product for holes), J.lnTn (mobility X lifetime product for electrons), and the layer thickness.
Good agreement is obtained between the theory and measurements on lead oxide vidicons. The theory
also provides some insight into the nature of the transient phenomena "red fade" and "after image" sometimes observed in the operation of lead oxide vidicons. At high light excitation levels, space-charge-limited
currents are expected. In this case, two regimes of current versus applied voltage behavior can be predicted:
(1) Below a saturation value of voltage, I ex: Vl/2.
(2) Above a saturation value of voltage, I = constant.
The reasons for the one-half power current-voltage relationships are distinctly different in the J.lT-limited
and the space-charge-limited cases. In addition, the dependence of current on light intensity in the spacecharge-limited case is a three-quarter power relation whereas in the J.lT-limited case the dependence is a
linear one. Also, the saturation voltage varies as the one-half power of the light intensity in the spacecharge-limited case and is independent of the light intensity in the J.lT-limited case.

has an insulator with a 2-V bandgap and excellent

blocking contacts. The contacts inject less than a
A. Background
nanoampere of current in the dark up to voltages of
An extensive literature already exists, both analytical 10-100 V across a 20-JL-thick layer. Weakly absorbed
and experimental, on the double injection of electrons red light can readily supply photocurrents over a
and holes into solids using Ohmic contacts.! If the hundred times larger.
voltage across these contacts is inverted, the same
The vidicon structure is effectively a sandwich-type
contacts are now blocking (noninjecting) and one is geometry in which the light enters normal to one face.
faced with the problem of double extraction of electrons It has been known and understood, both for the earlier
and holes from the volume of the solid. Only one or two vidicons made of amorphous selenium as well as for the
special cases of double extraction have been treated in present lead oxide vidicon, that the current versus
the literature. The most common example is that of the voltage curve for strongly absorbed (blue) light is either
back-biased p-n junction in which one obtains a linear leading into saturation or a space-charge-limited
saturated (or nearly saturated) current of volume curve leading into saturation.6 ,7 In the linear case, those
generated carriers already at a voltage of a few kT/e.2 carriers that escape the thin region in which they are
The back biased long p-i-n structure has also been generated are drawn to the opposite electrode. The
treated but only for the case of a semiconducting i fraction of generated carriers that escape is proportional
section with equal electron and hole lifetimes.s
to the applied voltage.
The problem treated here can also be thought of as a
Red light (~6300 A) on the other hand is reasonably
back biased, long p-i-n structure but one for which the uniformly absorbed throughout the volume of the lead
i section is an insulator and for which the lifetimes of oxide vidicon layer. The current versus voltage curves
electrons and holes are generally widely different. The reported by Inslee have, in general, four parts: a lowcontacts are assumed to be blocking in the sense that voltage linear section, a transition region, a more or less
the currents they inject are small compared with the extended section in which the current increases as the
volume generated currents. While the contacts inject square root of the applied voltage and finally a saturated
negligible currents, they can readily accept carners current at higher voltages. s It was the attempt to
understand this behavior that led to the present analysis.
coming out of the volume. 4
A particularly significant example is the current
It should be pointed out that there is a variety of
versus voltage curve for weakly (uniformly) absorbed situations for which the present analysis should be valid
light in the lead oxide vidicon (Plumbicon)" Here one or, at least, a good guide to interpreting data. In general
2823
I. INTRODUCTION

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 210.212.192.132 On: Thu, 05
May 2016 09:08:25

2824

A.

M.

GOODMAN AND

if one applies a voltage across any thin homogeneous

insulator with blocking contacts, the current generated
by any penetrating radiation (light, electrons, or x
rays) should show evidence of the square-root dependence on voltage.
B. Qualitative Description of Photoconductivity in
Weakly Absorbed Light
We consider the physical configuration shown in Fig.
1 (a). The insulator (photoconductor) is in the form of
a slab or layer of thickness 1 in the x direction and
infinite in the y and z directions so that the problem is
one dimensional. There is a transparent conducting
contact at each face of the layer. The contacts are noninjecting for both electrons and holes; i.e., carriers can
only be extracted from or flow out of the photoconductor
into the contacts. A voltage V is applied between the
contacts resulting in an electric field E in the photoconductor. Initially-before the flow of current-the
field is uniform; E= V /1.
The incident light is turned on at t=O; it is weakly
absorbed, resulting in uniform generation of holeelectron pairs throughout the volume of the photoconductor. The mean carrier drift lengths are Wn = jJ.nETn for
electrons and Wp=jJ.pETp for holes where the jJ.'s are the
carrier drift mobilities and the T'S are the "lifetimes"
before deep trapping. If both Wn and Wp are much larger
than 1, all of the photogenerated carriers are extracted
and no net charge is stored in the photoconductor layer.
If, however, either Wn or Wp or both are smaller than 1,
trapped charge will accumulate in the layer altering the
electric field. The electric field will continue to change
until a (steady state) condition is reached in which there
is no further net trapping of charge in any region of the
layer and, consequently, equal currents of electrons
and holes are extracted at the contacts. In this case,
electrons and holes are trapped at the same rate at any
point in the layer and the T'S become recombination
lifetimes.
Let us consider a specific case in which jJ.pTp> jJ.nTn and
charge is trapped in the layer. As shown schematically
in Fig. 1 (b) the net trapped charge is negative and the
electron energy-band curvature is concave downward.
Thus, the electric field is increasing in the region
adjacent to the positive contact enhancing the extraction of electrons and decreasing their probability of
being trapped. Conversely the electric field is decreasing
in the region adjacent to the negative contact diminishing the extraction of holes and increasing their probability of being trapped. This process continues until a
steady-state condition is obtained.
In the steady state, the photoconducting layer can
be described in terms of the dominant electronic
processes which occur in the 3 regions shown schematically in Fig. 1 (c). In regions 1 and 3, most of the
generated carriers escape without recombining; the
electrons from region 1 are extracted at the positive

A.

ROSE

contact, the holes from region 3 are extracted at the

negative contact, and the holes from region 1 and the
electrons from region 3 drift into region 2. In region 2,
the recombination rate is approximately equal to the
generation rate; it follows that n= gTn and p= gTp where
nand p are the electron and hole densities, respectively,
and g is the volume generation rate. The electron and
hole current densities J nand J p in the steady-state

J;'

>-

l!)

Q:

co~<;)

LI.I

LI.I

\)C~\O~ ~

~~Q

eV

Q:
~

U
LI.I

..J
LI.I

(a)

>-

l!)

Q:

LI.I

LI.I

Z
0
Q:
~

U
LI.I

..J
LI.I

x
(b)

>-

l!)

Q:

LI.I

LI.I

Z
0

Q:
~

LI.I
..J
LI.I

L----+----~------~~--+x

>-

iii

LI.I

LI.I
Q:
Q:

::l

_ _- L _ _ _ _~ _ _ _ _ _ _~4---+x

(d)

FIG. 1. Energy-band diagrams and current density distribution for a photoconducting layer with voltage V applied between
two noninjecting contacts: Energy-band diagrams (a) in the
dark (before current flow), (b) in uniformly absorbed light with
net trapping of negative charge, (c) in uniforml>: abs~rb~d li~ht
with steady-state current flow; (d) current densIty dlstnbutIon
of holes and electrons corresponding to (c) .

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 210.212.192.132 On: Thu, 05
May 2016 09:08:25

EXTRACTION OF PAIRS FROM INSULATORS

condition are shown qualitatively as a function of

position in Fig. 1 (d). The steady-state solution requires,
of course, that the sum of I n and J p be independent of
position, i.e., the divergence of the total current density
must be zero.
In reality, of course, the dominant processes referred
to above are not the only ones that occur. Some recombination does occur in regions 1 and 3, and generation and recombination are not precisely balanced
throughout region 2.

2825

>-

'"'"
UJ
Z
UJ

...'"
u

UJ

..J

UJ

L---+---+-------~----+x

~1----~--1_------+7r_~

e
E

E;, i= 1,2,3
g
I
j

jt
J
Jo

J,,;, i= 1,2,3
Jpi, i= 1,2,3
1
li' i= 1,2,3
n

p
v

V
Vo
Vi, i= 1,2,3
Wn
Wp

x
(J

II. LIST OF SYMBOLS

u;

drift length ratio (JJ.n'Tn/ JJ.p'Tp)

magnitude of the electronic charge (C)
Electric field (V/cm)
electric field in region i
rate of generation of hole-electron pairs
(events/cm3 sec)
current (A)
normalized current density (dimensionless)
normalized peak transient current density
current density (A/cm2)
saturation value of current density
electron current density in region i
hole current density in region i
layer thickness (cm)
thickness of region i
electron density (cm-a)
hole density (cm-a)
normalized voltage (dimensionless)
applied voltage (V)
saturation value of applied voltage
voltage drop across region i
drift length for electrons
drift length for holes
position in the layer relative to positive
contact (cm)
ratio of free to trapped carriers
electron mobility (cm2/V sec)
hole mobility (cm2/V sec)
electron lifetime (sec)
hole lifetime (sec)

...z
w
'"'"u=>

III. AN APPROXIMATE THEORETICAL MODEL

A. Formulation, Approximations, and Solution

We consider here an approximate model of the

steady-state behavior described qualitatively in Part B
of Sec. I and show that it leads to an analytic expression
for the current density J as a function of the applied
voltage V and the physical parameters describing the
photoconducting layer. The approximate model is
shown in Figs. 2(a) and (b) ; comparison with Figs. 1 (c)
and (d) shows the nature of the approximations. The
dominant processes referred to previously in the qualitative description are here considered to be the only

w
c

FIG. 2. Energy-band diagram and current density distribution for the theoretical model with steady-state current flow: (a)
energy-band diagram, (b) current density distribution of holes
and electrons.

processes; recombination is equal to generation at all

points in region 2 and does not occur at all in regions 1
and 3. The electric field is assumed to be constant in
each region. The effect of a nonconstant field is considered in the appendix and shown to be relatively
unimportan t.
In region 1, the electric field El and the extent of the
region, O<X<ll, are related by the requirement that
the carriers with the smallest JJ.'T product (assumed here
to be electrons) must pass through without recomb ining. 9 Thus, in this model,
h=JJ.n'TnEl.

( 1)

Since El can be written as Vl/ll, we have

II = (JJ.n'TnV l )1/2,

(2)

where VIis the voltage across h. Since we have neglected

recombination in region 1 and the contacts are noninjecting, the electron current density J nl and the hole
current density J pl can be written as
(3)
and
Jpl=egx

( 4)

in the region 0 <x <ll, where g is the volume generation

rate of electron-hole pairs, e is the magnitude of the
electronic charge, and J n2 is the electron current in
region 2.
Similar reasoning shows that

(5)
(6)

J n3 = eg(l-x) ,

(7)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 210.212.192.132 On: Thu, 05
May 2016 09:08:25

2826

A.

M.

GOODMAN AND

A. ROSE

The voltages across the 3 regions are

V1=E1l1=lNIJ.nTn,
V2=E212=E2[l- (lr+la) ]=blr[l-lr(l+b) J//JonTn

(18)
(19)

and
(20)

We can now set the sum of Vl, V 2, and Va equal to V, the

total voltage, and solve the resulting equation for II as a
function of V. The result is

1 _ -bl+[b212+4(1-b) V/JonTnJl/2
12(I-b)
FIG.

3. Normalized current density versus normalized voltage

for various values of the drift length ratio b.

and

(21)

The current flowing through the layer J is the total

current generated in regions 1 and 3 where no recombination occurs:

(8)

(22)

in region 3, [(ll+~) <.x<..l= (l1+~+13)], where Vs=

Eala is the voltage across the region, Ea is the electric
field in region 3, and J p2 is the hole current density in

Equations (21) and (22) can be combined to giveJ(V);

region 2.
In region 2, the generation and recombination rates
are equal;
(9)
The electron and hole current densities in region 2
may then be written as

J=[gel][I+b]
X {-b+[h2+4(I-h) VIJ.nTn/12]1/21/[2(I-b)].

It is convenient to put Eq. (23) in a normalized form by

introducing the dimensionless current and voltage
variables j and v defined in the following way:

(11)

where & is the electric field in region 2. From Eqs. (10)

and (11), we see that

11 =P.pTpE2

v=V/Vo,

(25)

Jo=gel

(26)

is the saturated current density obtained when regions

1 and 3 occupy the entire layer and

Vo= (1+b2)l2//JonTn(1+b)2

(27)

is the applied voltage necessary to make 11+13=1. The

result is

(l+b){ -b+[b2+4(1-b) (l+b 2 )v/(1+b)2]1/2}

J=
[2(I-b)]
.

(13)

and
From Eqs.

(24)

where

(12)
the drift.length ratio. The electron and hole current
must be matched at the boundaries between adjacent
regions as shown in Fig. 2 (b). By equating Eqs. (4)
and (11) at x=ll and Eqs. (7) and (10) at X=t1+~'
we obtain

j=J/Jo
and

(10)
and

(23)

(28)
B. Discussion

ta=/JonTnE2.

(14)

(12), (13), and (14),

la/II = b

(15)

and

la+lt= (1+b)ll'

(16)

By combining Eqs. (1) and (13), we obtain

&/E1=/JonTn//JopTp=b.

(17a)

Similarly, by comparing Eqs. (5) and (14), we see that

(17b)

It was previously assumed that b=IJ.nTn/P.pTp<1.

This assumption was employed in Eqs. (1) and (5)
where the thicknesses of regions 1 and 3 were determined
by the smaller /JoT product, viz., /Jon1'n. If, however,
/JopTp</JonTn then P.pTp would be used in Eqs. (1) and (5).
The results would be the same as those already obtained in Eqs. (21)-(28) except that wherever b
appears, it would be replaced by (3=/JopTp//JonTn=b-1 and
wherever IJ.nTn appears it would be replaced by /JopTp.
The trapped charge in tha t case would be positive and
the energy-band curvature would be concave upward
rather than concave downward as shown in Fig. l(b)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 210.212.192.132 On: Thu, 05
May 2016 09:08:25

2827

EXTRACTION OF PAIRS FROM INSULATORS

and (c). If j.LpTp= j.LnTn there is no trapped charge and no

curvature of the energy bands; this special case will be
considered shortly.
A plot ofj versus v is shown in Fig. 3 for various values
of the parameter b. The model predicts the existence of
4 distinct regimes of the dependence of j on v.

1 ,

..f

~ 0.1

(1) Low voltage or linear regime

[vb2(1+b)2j4(1-b) (1+b2 ) J.

-THEORY
RCA-H587. Vo -45 VOLTS
a AMPEREX-29504. Vo '45 VOLTS
RCA-CC3. Vo -104 VOLTS

In this case dj/dv=constant andj is linear with v.

(2) Intermediate voltage or transition regime
[v,""b2 (1+b)2j4(1-b) (1+b2)].
.01

Here djj dv is decreasing with increasing voltage and the

variation of j with v becomes increasingly sublinear.
(3) High voltage or square-root regime

In this range, dj/ dv decreases with increasing voltage as

and j increases as Vl{2.
(4) Very high voltage or saturation regime [v5>lJ.
Equation (28) does not apply in this case. Physical
reasoning must be used. All of the generated carriers are
extracted from the layer; there is no recombination. The
current is independent of voltage; j= 1.

v-112

It is of interest to examine some "special" or limiting

cases of the solution.

(a) Equal hole and electron drift lengths

(j.LnTn

= J.lpTp,

i.e., b=l).

The net trapped charge is zero and the electric field is

uniform throughout the layer. Only the linear and
saturation regimes of j (v) are possible.

FIG.

.02

.05

0.2
V -VIVo

0.1

0.5

5. Comparison of theory with experimental data.

and 2. Thej(v) dependence is dominated by the square

root and saturation regimes.
In order to compare experimental data with the
theory, it is desirable to reduce the data to a normalized
form and plot it on top of Fig. 3. This procedure allows
a check of the functional form of j (v), the evaluation of
Vo, and permits an approximate evaluation of b. This
will be done in Sec. IV. Alternatively, if one picks a
particular value of j, say, ja, then for each value of b
there exists a corresponding value of v, viz., Va. By
inverting Eq. (28), we obtain
Va=

{[2(1-b)j",+b(1+b) J2
-[b(1+b)J2}j[4(1-b) (1+b2 )].

(29)

(b) Electron drift length much less than hole drift

length (1)>~0).

In Fig. 4 we have plotted b against v", for 7 values of JIlt

(0.005,0.01,0.02,0.05,0.1,0.2,0.5). One Can then pick
a value ofj" from the experimental data.and evaluate b
directly from Fig. 4. This information along with Eq.
(27) allows the evaluation of J.lnTn;

The layer is effectively composed of only 2 regions-l

(30)

and j.LpTp is then, of course, ILnTnjb.

Further, as may be verified from Eq. (28), the
normalized voltage at which the extrapolated linear
regime intersects the saturated current level (j= 1) is a
direct (approximate) measure of b. FromEq. (28) it can
be seen that the location of the linear regime must
correspond to j at vb2j4.

b
0.5

0.2

IV. COMPARISON WITH EXPERIMENT

1153_L:;5~2...L.J..J-lll.ll.:-+-,~1J..lU=--!:-'-'-~JlL.,,--O-,LLLlllJ.L,-.LLJ.--'-'-'.u.u
10

10'3 2
'10'2
va-VIVo

10'1

FIG. 4. Drift length ratio b as a function of Va for 7 values of j",.

See text for explanation.

In order to compare the theory with experiment, the

photocurrent versus voltage characteristics of 3 leadoxide photoconductive pickup tubes were fitted to the
appropriate portion of Fig. 3; the result is shown in
Fig. S. The e~,:perimental data were provided by Inslee
who will describe the details of the measurements in a

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 210.212.192.132 On: Thu, 05
May 2016 09:08:25

2828

A. M.

GOODMAN AND A. ROSE

future report. s The agreement is generally good except

for two regions where the departure of the experimental data points from the theoretical curves is
apparent. (1) The data points show a gradual saturation
of current with voltage. The abrupt saturation at
V = Vo predicted by the theory is a result of the simplifying assumption that no recombination at all occurs in
regions 1 and 3. In fact, of course, this is not true as was
previously discussed in Sec. III and the gradual saturation is what one should expect. (2) At values of vl,
the experimental points for the two RCA tubes fall very
close to the b=0.1 curve; however, the points for the
Amperex tube tend to fall off at low v and do not lie along
any line of constant b. This might indicate that the }.IT
values are not constant throughout the lead oxide layer
in the Amperex tube.
At normal light levels (corresponding to photocurrent densities up to 10-7 }.IA/cm2), Inslee has found
that the observed photocurrent is linear with light
intensity.s
V. SPACE-CHARGE-LIMITED CURRENT EFFECTS

In region 1 of Fig. 2 (a), the density of holes increases

from left to right starting from near zero at the left
(x=O). The density of electrons in this region on the
other hand, increases from right to left starting from
near zero on the right (x = 11). This disparity in density
means that over most of region 1, the electrons are not
neutralized by an equal density of holes and vice versa.
In brief, charge neutrality does not hold. The resultant
space charge mayor may not have a significant effect
on the currents depending on the relative magnitudes
of the actual current and the maximum available
current. The latter is the space-charge-limited current
that can be drawn across region 1 at the voltage appearing across this region.
For example, if the space-charge-limited current is
large compared with the actual current, one can neglect
the effects of space charge as has been done in Sec. III
of this paper. On the other hand, if the space-chargelimited current is less than the total photogenerated
current in region 1, the observed current will be the
smaller of the two, namely the space-charge-limited
current. For this case, the problem must be re-examined.
The following is an approximate argument for the
current-voltage dependence in the regime of spacecharge-limited currents. To anticipate the result we will
obtain the same half-power dependence of current on
voltage as was obtained in the absence of space-charge
limitation, but for distinctly different physical reasons. To
emphasize this last point, we note, first, how the halfpower dependence arose in the absence of space-charge
limitation, i.e., in the }.IT limited case. For ease of
argument we consider here the case b1 in the halfpower regime. Under these conditions most of the
voltage is across region 1 and the current generated in
region 1 is substantially the total current. The length of

region 1 is [as in Eq. (1)]:

11= E 1}.1nTn = (Vt/it)}.InTn

(31)

or
(32)
Since we have taken VI to be substantially equal to
the applied voltage and the observed current to be
substantially that which is photogenerated in 11, we
read Eq. (32) immediately to say that the photocurrent
is proportional to the square root of the applied voltage.
In the case of space-charge-limited currents, the
primary relation to be satisfied is:

where () is the ratio of free to trapped carriers and }.I the

free-carrier mobility.1O It is also true (for VI = applied
voltage) that
]=egl1.
(34)
Combination of Eqs. (33) and (34) yields
II a:

(V11/2/ gl/4) ,

(35)

so that, again, the photocurrent J is proportional to the

square root of the voltage. The physical reasons for the
half-power dependencies of Eqs. (32) and (35) are
clearly quite different. A further point to be noticed is
that the photocurrent, from Eqs. (34) and (35) is
proportional to the 3/4 power of g, the generation rate or
light intensity. In the space-charge-free case, the
photocurrent was proportional to the light intensity
since II was independent of g.
From Eq. (35) we see that for a given generation rate,
it varies as the 1/2 power of VI. For some sufficiently
large value of applied voltage, say V o', It will become
equal to the sample length I and saturation of the
photocurrent will occur; for larger values of applied
voltage, the photocurrent will be constant. From Eq.
(35) it can be seen that the saturating value of voltage
Vo' will then vary as gl/2.
In general, one would expect that the space-chargelimited regime would set in at lower current densities
for higher bandgap materials where trapping plays a
larger role.
Also, for a given bandgap, as in PbO, one can expect
that for increasing light intensities, the photocurrents
will move from the space-charge-free to the spacecharge-limited regime. A test of this shift would be the
3/4 power dependence on light intensity.
A measure of the space-charge constraint is that the
space-charge-limited current across a 10-}.I thickness of a
material with a bandgap of 2 V, at an applied voltage
of 30 V is in the order of 10-16 A/cm2 if one assumes that
the Fermi level is pinned at the middle of the gap by a
high density (> 1015/cm3 ) of traps at that energy level.
The fact that currents far in excess of this value,
namely, currents of the order of lQ-6 A/cm2 are observed means that the Fermi level is about 0.6 V below

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 210.212.192.132 On: Thu, 05
May 2016 09:08:25

2829

EXTRACTION OF PAIRS FROM INSULATORS

the band edge and that the trap density between this
energy and the middle of the bandgap is less than
1015/cm3. For a more detailed explanation see Ref. 10.
VI. TRANSIENT EFFECTS
The model we have presented in Sec. III describes the
steady-state behavior of the photoconducting layer in
weakly absorbed light. However, it also allows some
insight into the nature of the transient behavior of the
photocurrent when the light is turned on or off.
Two transient phenomena which are often-but not
always-observed (in connection with the operation of
vidicons in weakly absorbed light) are "red fade" and
"after image." Red fade is a term used to describe a
relatively large initial photoresponse when the light is
turned on, followed by a decay in photoresponse to some
lower steady-state value. After image is a term used to
describe a decaying vestigial current through the
photoconductor after the light is turned off. The names
red fade and after image arose from the visible effect
viewed on a color kinescope when the associated transient current phenomena occur in a vidicon used as the
red-channel photoconductive pickup tube.
In order to understand the red fade we return to the
situation discussed earlier in part B of Sec. I. Initially,
with some voltage V < Vo applied and before the light is
turned on, the field in the photoconductor is uniform;
E= V It. The incident light is turned on at (=0 producing uniform generation of electron-hole pairs
throughout the volume of the photoconductor. The
carriers drift in the electric field: some are extracted at
the contacts and the remainder are trapped. If, as
previously assumed, P-pTp>P-nTn, then the extracted hole
current is greater than the extracted electron current
and there is net trapping of negative charge. This
trapped charge alters the electric-field distribution in
such a way as to enhance the extraction of electrons and
to diminish the extraction of holes until the extraction
currents are equal and the steady state is reached. The
diminution of the hole current is more pronounced than
the enhancement of the electron current so the net
effect is one of an "overshoot" in photocurrent before
the steady state is reached. The decay of the photocurrent from its peak value to the steady-state value is
the red fade. The normalized peak transient current
density jt can be calculated from the formula

jt= [(b+ l)bv- (b 2+ 1) V2+V2 (b 2trl/"+e-b1v ) J/b2 (36)

obtained by generalizing an expressionll for the electron
transient current to account for both electrons and holes.
Both the steady-state current density j and the peak
transient current density jt are plotted versus v in Fig. 6
for the case b=O.1. The crossover ofj andjt is a result of
the simplifying assumptions used in deriving j. In fact
of course, j saturates gradually as shown previously in
Fig. 5, and no crossover is expected. This is indicated
qualitatively by the dashed part of the curve. When v is

1~~~~~~Uill~~~~~~~~
5 1<)2

5 1<j"1

=VIVo
FIG. 6. Peak transient and steady-state normalized photocurrent
densities versus normalized voltage for b=O.l
V

very small (vb) , both the photocurrent and the overshoot are small. Most of the holes and electrons are
trapped (and subsequently recombine) but the net
trapped charge is small. As v gets larger more net
positive charge is extracted initially and the net trapped
negative charge in the layer becomes larger in the steadystate condition; there is a correspondingly greater
change in the energy-band curvature in the steady state.
As the applied voltage becomes still larger (b<v<l),
all of the holes and more of the electrons are extracted
initially resulting in less trapped charge. When the
applied voltage is equal to or greater than the saturation value (v>l), all (or nearly all) of the holes and
electrons are extracted; the steady-state trapped charge
is very small, the electric field remains essentially uniform, and j,;:::::,jt. Thus the initial transient (red fade)
can be significantly decreased or eliminated by keeping
v> 1.
After image is directly related to the red fade; it is
due to current flow through the photoconductor (after
the light is turned off) resulting from the thermal
release of carriers previously trapped during the
establishment of steady-state current flow. Thus we see
that the after image will be small when the initial
transient is small and large when the initial transient is
large.
As a practical matter we observe that the optimum
operation of a lead oxide vidicon in red light will occur
when the applied voltage is at or above the saturation
value; in this case, the sensitivity is maximized and the
red fade and after image are minimized.
In the foregoing discussion, it was assumed that the
time for readjustment of the electric field pattern within
the layer is long relative to the time necessary for the
establishment of steady-state carrier densities. For the
case of low applied voltage and/or high light intensity,
the inverse is likely to be true; the red fade will then tend
to be suppressed. Moreover, when the applied voltage is .
very low and region 1 is correspondingly small, those
carriers that are generated in region 2 but do not
recombine there provide a small but significant additional component of current. The actual steady-state

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 210.212.192.132 On: Thu, 05
May 2016 09:08:25

2830

A.

M.

GOODMAN AND

A.

ROSE

current will then be slightly larger than the value computed from our approximate solution [Eq. (28) J and
will more closely approach the value of jt. Thus the red
fade will be further diminished at low v. At v< 10-4, still
other arguments can be given to show that the photocurrent overshoot should be vanishingly small.

into the layer from the positive contact. Equation (A4)

may then be used to evaluate (-1).

ACKNOWLEDGMENT

By inserting Eq. (AS) into Eq. (A3) and simplifying,

we obtain

We are indebted to J. A. Inslee for many valuable

discussions and for permission to use his unpublished
experimental data.
APPENDIX: THE EFFECT OF A NONCONSTANT
ELECTRIC FIELD
In Sec. III the thickness of region 1 was set equal to
the drift length of the electrons in region 1:
(AI)
where, for simplicity, EI was assumed to be constant
and equal to VI/II. In order to evaluate the propriety of
this assumption, we now consider the effect of a non constant electric field E on the drift length Wn of a mobile
electron with a lifetime Tn. The time dt required for the
electron to travel a distance Idx I is I dx//-I"E I it follows
that

T,,=

-dx

Wn

w"

/-I"E

/-In

1 - - == -

(E-l),

(A2)

where (E-l) denotes the average value of the reciprocal

electric field over the drift length W n Therefore,
(A3)
which reduces to the simpler expression /-InTnE only in
the case of a constant electric field.
If the electric field EI in region 1 is not constant but
instead varies linearly in the range O<x<it, due to a
constant trapped charge density, then
EI = Celt-x) /It](2VIllt) ,

(A4)

where VI is the voltage across 11 and x is the distance

(E-l)=Wn- 1

jWo (~)(~)
dx
2V II-x
n

(E-l) = (lN2VIwn ) In[lI/(lI-Wn ) ] .

In[ll/ (ll-Wn ) J= 2 (/-I"Tn V1/II 2) = 2.

(AS)

(A6)

From Eq. (A6) we may obtain the ratio w..III; the result
IS

(A7)

The actual distribution of trapped charge is neither a

sheet of charge at X=/I (as would be necessary for a
constant field in region 1) nor a constant density in
region 1 (as would be necessary for a field varying
linearly with distance in region 1). The actual distribution probably lies somewhere between these extremes.
The actual value of W will in any case be not very far
from the value It computed for the constant-field
approximation.
M. A. Lampert, Rept. Prog. Phys. 27, 329 (1964).
W. Shockley, Bell System Tech. J. 28, 435 (1949).
3 A. Herlet and E. Spenke, Z. Angew. Phys. 7, 149 (1955).
4 J. J. O'Dwyer [J. App!. Phys. 39, 4356 (1968) ] has also
treated a double extraction problem with contact~ of this type.
However, in that case, the generation of electron-hole pairs is
due to collision ionization in the body of the insulator; it is therefore spatially nonuniform and electric field dependent in contradistinction to the spatially uniform, field-independent generation
in the problem to be considered here.
6 L. Heijne, Philips Res. Repts. Suppl. 4 (1961).
6 P. K. Weimer and A. D. Cope, RCA Rev. 12, 314 (1951).
7 F. J. Du Chatenier, Philips Res. Repts. 23, 142 (1968).
8 J. A. Inslee (private communication).
9 Statistically, about e- l of the carriers generated in region 1
(whose extent is one "electron drift length") will recombine.
However, as the magnitude of the electron drift length changes
with applied voltage, the ratio of surviving carriers to generated
carriers will remain the same. Thus, the recombination in region 1
will not affect the form of the current-voltage curve. A similar
argument holds for region 3.
10 A. Rose, Concepts in Photoconductivity and Allied Problems
(Wiley, New York, 1963), Chap. 4.
11 N. F. Mott and R. W. Gurney, Electronic Processes in Ionic
Crystals (Dover, New York, 1964), p. 122.
1

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 210.212.192.132 On: Thu, 05
May 2016 09:08:25