You are on page 1of 4

Solid-State Electronics, 1975, Vol. 18, pp. 1039-1042. Pergamon Press.

Printed in Great Britain

TRANSIENT CHARGING CURRENTS IN INSULATORS

H. J. WINTLEt
School of Electronic Engineering Science, University College of North Wales, Bangor, Gwynedd, Wales

(Received 6 January 1975)

Abstract--It is shown that charging currents in good insulators can be used to distinguish between different empirical
carrier injection laws, if most carriers are trapped according to the Walden model. Schottky injection and power law
injection are considered in detail. The current falls as t-", 0 < n < l. For power law injection the isochronal current
density ./ ~ Ek, where E is the applied field, and it is shown that the apparent power k - I for 0.6 < n < 0.9, whatever
the true power law exponent p. This accords with the experimentally observed superposition theorem. Some
limitations of the power law injection process proposed by Levine are discussed.

INTRODUCTION dependence alone is not adequate to distinguish between


The analysis of steady state current flow in insulators is the various empirical injection laws since it in fact gives
ambiguous[i] since a given set of experimental data can only the index n which relates to trapping, One must also
be made to fit a number of the available models equally study the isochronal field dependence which is implicit in
well, despite the widely different algebraic form of the the factor ]~to".
corresponding theoretical expressions, Consequently, it is This study of the field dependence is a potentially
also desirable to obtain data for the time dependent useful method for distinguishing between injection
absorption currents which flow following step voltage mechanisms, and it has already been employed[10] in an
excitation. When there is ready injection and little or no analysis of the current-time curves of poly(ethylene
trapping, this transient current flow will be space charge terephthalate) (PET)[12]. There it was possible to
limited (TSCLC) and this case has been extensively demonstrate that the assumed Schottky injection mechan-
studied [2, 3]. Somewhat similar results follow for material ism was not consistent with the data, and that power law
in which carrier exit from the sample is blocked [4, 5]. We injection provided a better fit.
are interested in the present paper in the case for which The purpose of this paper is to point out that while
there is both severe trapping and an injection limited some care has to be exercised in employing this analysis if
current[6, 7], and we concentrate upon systems for which the trapping is severe, i.e. if n ~ 1, it is normally possible
the injection is field dependent. This appears to cover a to distinguish unambiguously between the two cases.
number of commercially important systems running from Some discussion of the validity of the available models is
selenium (linear field dependence)[8] to plastic insulators presented.
(perhaps Schottky injection).
A rather general description of the absorption current THEORY
has been given by' Walden[9]. It is assumed that the It has already been shown (eqn 11 of[10]) that a
injected charge is heavily trapped, causing a reduction in pathological result follows from a power law injection if
field at the injecting electrode. This irr turn reduces the n = 1: for times t >>to, the transient current is a function
particle current density, and the current falls with time. It of time only, independent of the applied field. In practice,
has been shown that the displacement current cannot be this does not occur (see[13, 14] and references in[10]).
neglected at early times[10], but that subsequently the Consequently, in the normal case either there is incom-
total current density follows the relation plete trapping, or the injection Jaw is a more abrupt
function of field than a simple power law, or the Walden
j 0: .it [to/(to+ t)]" t >>to (1) model is too simplified.
Incomplete trapping could be identified by the time
where j~ is the initial particle current density, to is a dependence which would be of the form t-", n < l, at long
characteristic time defined in[10] and n < 1 is an index times. The corresponding isochrone for power law
which approaches unity as the trapping becomes increas- injection is
ingly effective. We have written only the particle current
term in this equation, and we emphasise that it applies not ] ~ E k t >>to (2)
only to the case of injection currents increasing exponen-
tially with the field, but also to Schottky law and power k = p2(1 - n)/(p - n)
law injection. It is thus of much wider application than has
been suggested by one author[l 1]. It follows that the time where p is the index appearing in the injection law, and k
is the apparent power law obtained at constant time.
tPermanent address: Department of Physics, Queen's Univer- For the moment we set aside the question of the
sity, Kingston, Ontario, Canada. adequacy of the simplifications, and ask what happens if

1039
SSE VoI. 18, No. 12--A
1040 H.J. WINTLE

there is sharp field dependence of the injection, but still and now it should be possible to separate the two cases
with complete trapping in the bulk of the specimen. We given satisfactory experimental data.
choose the Schottky barrier as an appropriate example.
From the previous results[10] we have APPLICATION OF THE THEORY
One example of the identification of the empirica
j ~ e x p {flsE~/2(l - n ) / k T } x E n n t " t >> to (3)
injection law has already been mentioned[10]. It wa~
shown that the data of Taylor and Lewis[12] on PE]
where /3, is the Schottky coefficient. It follows that the
corresponded to a power law injection with k = 1.79
current isochrones should exhibit a Schottky type of
n = 0.79 and p = 4.7. A number of similar cases haw
behaviour during the transient period. Again a pathologi-
been listed in [10], along with some examples for which
cal case arises when n = 1, when a simple E m behaviour
power law is appropriate.
is forecast. The question arises as to whether the two
We now consider some recent data for low densit~
cases can be distinguished with any certainty when n
approaches unity. We show in Fig. 1 the behaviour of
eqn (3) for various values of n, using Schottky plot
coordinates, while in Fig. 2(a) we show the behaviour of
eqn (2) for various values of k. This figure is com-
plemented by Fig. 2(b) which shows how p, n and k are
related. The isochronal log j vs E ~2 curves are identical
for n = 1 (Schottky law) and k = 0.5 (power law). In the
latter case, the decay rate measured by n is less than unity
over most of the range of p shown in Fig. 2(b). If we
assume that n can be found to within -+ 0.05, then we see
that the effect of the two injection laws could be
distinguished unless p > 10 or p < 1.1. If the experimental
value of n is 0.9, then the log j vs E ~t2 has a much less
marked curvature in the Schottky case than any of the ~2
power law plots in the lower field region 8
4-
0
(E": < 3 MVmm 1~2,or E < 105 V cm -~) g~

0.~ C
I I
0 5 10 15
E 1/2 ( M V l m f / 2

18

10 P

)[ 1 I J
0 5 10 15 0 012 014 (~6 016 110
E 1/2(MV/m) 1/2
rq

Fig. 1. Plot of isochronal currents in insulators under Schottky Fig. :2. (a) Plot of isochronal currents in insulators under powe
injection law conditions, eqn 3, in the time-dependent region. law injection (eqn 2) in the tin]e-dependent region. Values n
Schottky plot coordinates are used, and a dielectric constant of 2.3 apparent power exponent k are marked on the graphs. Sam,
is assumed. Values of the trapping index n are given on the coordinates as in Fig. 1. (b) Plot of apparent power exponent k a
graphs. a function of true exponent p and trapping index n.
Transient chargingcurrents in insulators 1041

polyethylene. R6hl and Fischer[13] have presented from the insulator, and is therefore under forward bias.
results of ](E) at various temperatures for currents This situation is shown in Fig. 3.
measured up to ] 0 4 s e c after voltage application. They 2. In a metal-insulator-metal structure, one junction
showed that a power law fitted their isochronal results, will be forward biased and the other reverse biased. What
while a Schottky plot also gave an approximate fit. The distinguishes an insulator from a semiconductor is that the
slope of their Schottky plots was about 50% greater than supply of carriers is negligible in the insulator, so the
/3s. The slopes in Fig. 1 are all less than /3,, so that particle current extracted from the insulator at the
Schottky injection and trapping is unable to account for forward biased junction is very small, and will barely
their data. Close inspection of their Schottky plots reveals exceed the reverse flow from the metal. At the other
curvature of the form shown in Fig. 2(a), and the authors contact, however, the rate of carrier injection from the
in fact showed that their results fitted a power law quite metal will be enhanced by the proposed surface-state
well. Their decay rates did not show a constant value of n mechanism. This is therefore the junction of (relatively)
at all temperatures, but the change of slope with time was easy carrier flow, and the conventional label of this as the
relatively slow. Their results, and the value of p deduced reverse-biased junction conveys a false impression.The
from them, are given in Table 1. Where n is constant, their difference between the net particle currents at the two
results correspond to a square law injection. junctions is accounted for by trapping somewhere in the
Another set of data has been obtained by Das Gupta et bulk until the final steady state is reached. This is also
al. [15, 16]. This work again refers to readings taken in the shown schematically in Fig. 3.
time dependent region, albeit the rate of change is slow, 3. There must be a depletion region adjacent to the
and the Walden model can be tested. Once more a power surface, which implies that there is a considerable
law fit is appropriate, with k = 2.7, n - 0 . 5 and conse- inhomogeneity between surface and bulk. The material
quently p - 5 . would otherwise behave as a conventional semiconduc-
POWERLAW CHARGING tor. It is conceivable that this inhomogeneity is not
We have shown that power law injection seems to essential and that the appropriate space charge is itself
occur into a number of high resistivity materials. If such controlled by the surface states[19, 20], with or without
power law injection is indeed a widespread feature of significant interpenetration of the two clouds[21]. The
metal-insulator contacts, then Fig. 2(b) shows how the simple space charge model[22-24] is inappropriate since
apparent linearity or near-linearity[17] comes to be in this case the interface is transparent to carriers. Clearly
obeyed. For any reasonable amount of trapping, say the details require further elucidation but they are not
0.6 < n < 0.9, then the apparent power law k is approxi- germane to the rest of this paper.
mately unity for a wide range of values of p. Thus the 4. Whatever the details of free and trapped charge
superposition theorem will appear to be obeyed, at least in adjacent to the contacts, the image forces are omitted in
limited form, for the charging currents. The discharge Levine's model. This must arise because the screening
currents will be discussed later. length is shorter than the distance from the interface to
A possible explanation for power law injection has been the peak of the Schottky injection potential, and again this
given by Levine [18]. He showed that a metal- implies that the region adjacent to the interface must be
semiconductor junction held in reverse bias coud be relatively conducting.
controlled by the filling of surface states. If the surface 5. On applying a d.c. voltage step, there will be an
states are exponentially distributed in energy, then the c o n v e n t iooaf reversQ
power law is obtained. We have p =- Eo/kT, where Eo is an
energy characterising the trap distribution. The more
uniformly distributed the traps are in energy, the larger is conventional
E(,. The same model can apply to insulators with the
following provisos.
1. For the sake of definiteness we consider an insulator
in which such charge motion as does take place is ~ ' ~ - - n - t y p e carriers
dominated by electrons. Then we define a forward biased
junction in the conventional way employed for metal-
semiconductor Schottky barriers: an interface at which
inJection I d r i f t drift <injection
the insulator is negative with respect to the metal
electrode will tend to eject a greater number of carriers
net net
Table I. Injection into polyethylene, data from R6hl and
Fischer[13}
Fig. 3. Sketch of energy levels in a metal-insulator-metalsystem.
Tempez,atupe, °C 88 76 55 uO 23 The dielectric is assumed to carry current by means of electrons,
k (obsevved) !. 5 l. 8 2.0 2.7 3.1 and the junctions are labelledwith the conventionalnames used in
semiconductor technology. The particle current densities imping-
n (observe~) O. 35 0,23 0,12 0.164 0.41 '% ing on the two planes of contact are illustrated schematically
p (f~om Eq.{2)) l.g 2.1 2.1 3. u 4.8
below the junctions, and it is seen that under injection conditions,
the "reverse" junction carries the greater net current, The
difference between the two net current flows is stored in the
'% n not constant with time insulator.
1042 H. J. W1NTLE

initial period during which displacement current will form mechanism which follows a power law with respect to the
a significant fraction of the total current. There may also field at the source electrode can account for a number of
be appreciable extraction of carriers from the region experimental observations on polymeric insulators and
adjacent to the drain electrode. Thereafter, the current that the t " decay law follows quite readily from this
flow will be controlled by the (conventional) reverse empirical law. While the true index p could vary widely,
biased junction, leading to Levine's power law injection the apparent poweer law law exponent k lies near unity as
and thus to the time dependence described by Fig. 2. long as most of the charge is trapped. Such values of k do
While it is conceptually simple, this model due to occur in typical experiments.
Levine raises a good many problems of detail which must
be satisfactorily explained. It does suggest, however, that Acknowledgements--The author wishes to thank Dr. J. Hirsch
the band-tailing and consequent localised levels which are and Dr. D. K. Das Gupta for a number of helpful discussions,
believed to exist in the bulk of amorphous solids [25] may Prof. T. J. Lewis for his hospitality while the author was on leave
at Bangor, and the National Research Council of Canada for their
extend to the surface, leading quite naturally to the financial support.
postulated exponential energy distribution of the surface
states. REFERENCES
1. J. J. O'Dwyer, The Theory of Electrical Conduction and
DISCUSSION Breakdown in Solid Dielectrics. Oxford University Press,
It is must be asked at this stage whether the Walden London (1973).
2. A. Many and G. Rakavy, Phys. Rev. 126, 1980 (1962).
model is really adequate to account for the absorption 3. Many references are given in H. J. Wintle, J. Appl. Phys. 43,
current phenomena in insulators. The two main simplify- 2927 (1972).
ing assumptions which enable an analytic solution to be 4. A. von Hippel, E. P. Gross, J. G. Jelatis and M. Geller, Phys.
obtained for the j(E,t) characteristic are that the trapping Rev. 91,568 (1953).
5. R. B. Comizzoli and F. K. Manasse, J. Appl. Phys. 39, 4868
index n and the trapping depth ~ remain constant. Each (1968).
of these assumptions is liable to be violated to some 6. J. Mort, F. W. Schmidlin and A. I. Lakatos, J. Appl. Phys. 42,
extent in real samples [26]. It is not clear how the index n 5761 (1971).
might vary with time, so no progress has yet been made 7. H. Scher, D. Pal and J. Mort, J. Appl. Phys. 44, 2908 (1973).
with an analysis of this situation. On the other hand, it has 8. H. Seki, Phys. Rev. B2, 4877 (1970).
9. R. H. Walden, J. Appl. Phys. 43, 1178 (1972),
been shown that with a steadily increasing depth 6, there 10. H. J. Wintle, J, Non-Cryst. Solids 15, 471 (1974).
is little change in the case of Schottky injection [27], even 11. A. V. Ferris-Prabhu, Solid-State Electron. 16, 1086 (1973).
though as time goes on the peak of the injection barrier 12. D. M. Taylor and T. J. Lewis, J. Phys. D. 4, 1346 (1971).
will almost certainly move through the position of the 13. P. R6hl and P. Fischer, Koll. Z Z. Polymere 251,947 (1973).
14. R. J. Munick, J. Appl. Phys. 27, 1114 (1956).
previously trapped space charge, To date, therefore, the
15. D. K. Das Gupta, M. E. Tindell, R. D. Wingrove and K.
Walden model would seem suitable for describing more Joyner, Elektrostatische Aufladung, pp. 73-86. Dechema,
complex situations than the model strictly covers. Frankfurt (1974).
We now turn to the converse question of discharge 16. D. K. Das Gupta, K. Joyner and R. S. Brockley, private
currents. Frequently, superposition is found to hold[17] communication.
17. M. E. Baird, Revs. Mod. Phys. 40, 219 (1968).
and both dipolar and tunelling[28] systems lead naturally 18. J. D. Levine, Solid-State Electron. 10, 1083 (1974).
to this result. The case of discharge from spatially 19. J. D. Levine and P. Mark, Phys. Rev. 144, 751 (1966).
distributed traps when the specimen is shorted has also 20. E. H. Nicollian and A. Goetzberger, Bell Syst. Tech. J. 46,
been studied theoretically [29], and the t-" law was found 1055 (1967).
21. J. G. Simmons, J. Phys. Chem. Solids 32, 1987 (1971).
to hold. However, the dependence on the previously 22. N. F. Mott and R. W. Gurney, Electronic Processes in Ionic
applied field was not established. In addition, this Crystals pp. 168-172. Oxford University Press, London
treatment differs markedly from another more generalised (1940).
analysis[30] in the assessment of the current density 23. A. Rose, Concepts in Photoconductivity and Allied Problems
pp. 134--138. Interscience, New York (1963).
through a plane of zero field. Until these uncertainties
24. C. G. Garton, J. Phys. D. 7, 1814 (1974).
have been resolved, it is not possible to say whether the 25. W. E. Spear, Adv. in Phys. 23, 523 (1974).
absorption and resorption currents display a similar time 26. J. Hirsch, private communication.
dependence because they arise from similar causes, or if 27. H. J. Wintle, submitted for publication.
this apparent obeyance of the principle of superposition 28. See H. J. Wintle, J. Appl. Phys. 44, 2514 (1973); and references
therein.
arises accidentally from quite different charging and 29. A. H. M. Shousha and L. Young, Thin Solid Films 8, 383
discharging mechanisms. (1971).
In summary, we have shown that a carrier injection 30. B. Gross and M. M. Perlman, J. Appl. Phys. 43, 853 (1972).

You might also like