Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

I O U R N A L OF

ELSEVIER Journal of Non-Crystalline Solids 172-174 (1994) 1408-1412

'Window' effect in the analysis of frequency dependence


of ionic conductivity
H. Jain*, C.H. Hsieh
Department of Materials Science and Engineering, Lehigh University, 5 E. Packer Avenue, Bethlehem, PA 18015-3195, USA

Abstract

Usually the frequency dependence of ionic conductivity of solids is analyzed in the form of a power law or as the
deviation of imaginary part of electric modulus from ideal Debye peak. The power law formalism is shown to be sensitive
to the choice of frequency window employed in the analysis. The results demonstrate the inadequacies of either formalism
and a method is suggested of combining the two formalisms for parameterizing non-ideal electric response in the vicinity
of the electrical conductivity relaxation time.

1. Introduction unity. Here A' is another fitting parameter. They


concluded that the transition of s is actually
Various ionic conductors have similar dispersion a superposition of two distinct and separate pro-
of electrical conductivity, tr, as a function of fre- cesses with different exponents, namely, So and 1.0.
quency. Often this has been analyzed in the form of Here the last term becomes important only at rela-
a power law [1] tively low temperatures or high frequencies, i.e.,
Eq. (2) reduces to Eq. (1) at moderate T and (o.
t r = t r d c + A O 9 ~, 0 < S < 1. (1)
Alternatively, the conductivity dispersion is ana-
trdc is the dc conductivity, ~o = 2rcf is the angu- lyzed in the complex electric modulus M*
lar frequency and A and s are fitting parameters. ( = M' + iM") formalism [3],
The exponent s has been found to be material-
dependent [1]. Very recently Lim and co-workers [2] M*(~o) = e~l [1 - dtexp( - io)t)( - d~b/dt)],
have extended the range of temperature, T, and o
suggested (3a)
(7 ~--- O'd¢"Jl-A e f ° + A'(o 1"°, 0 < So < 1, (2) with the field decay given by Kohlrausch function
based on the observation that the exponent, s, in [4]
Eq. (1) has a transition with increasing temperature
q~(t) = exp - (t/r) ~, 0 < B < 1, (3b)
from s = 1 to a constant s = So which is less than
where e~ is the high frequency dielectric constant
* Corresponding author. Tel: + 1-6107584220. Telefax: + 1- and r is the electrical conductivity relaxation time.
610 758 4244. B is also a material constant [5], and has been

0022-3093/94/$07.00 © 1994 Elsevier Science B.V. All rights reserved


SSDI 0 0 2 2 - 3 0 9 3 ( 9 4 ) 0 0 0 1 5 - F
H. Jain, C.H. Hsieh / Journal of Non-Crystalline Solids 172 174 (1994) 1408-1412 1409

correlated to the activation energy of ionic conduc- 1 0 -a


N a20-0.4AI203-2,2SIO = glass
tion [5] and the average distance between conduct-
ing ions [6]. One may predict that /3 ~ 1 - s
because mathematically Eq. (3) would give
a oc to I # in the high frequency limit. However, A k
deviations are often observed in this formalism; for E
o
1 0 -r
example, the experimental M" data are higher than 0~
C /'/ ~i
the values given by Eq. (3) for frequencies much
larger than the M" peak frequency. It has been
suggested [7,8] that this discrepancy can be . ~G~
explained by considering the contribution of asym-
metric double well potential (ADWP) configura- 104
tions to the total conductivity. 1 01 10 = 10s 1¢ 0s
Experimentally one has a limited range fre- f (Hz)
quency window available for determining the elec-
trical parameters. If the fitting parameters in Fig. 1. The effect of frequency window width on s. The open
Eqs. (1)-(3) were to represent a true physical phe- circles represent experimental data.
nomenon, the choice of frequency window should
not determine s, fl or the transition of s. The pur- procedure. For analysis of lithium borate and
pose of this paper is to examine the question: does sodium aluminosilicate glasses above room tem-
the choice of frequency window affect s (or fl) or the perature, dc conductivity values are taken from
transition of s? complex impedance plots. For the data analysis of
the latter glass below the room temperature, trdc is
determined by extrapolating high temperature con-
2. Data analysis and discussion ductivity using the Arrhenius equation. Only those
a values are used which are not affected by the
We have analyzed the electrical conductivity electrode effects (as determined from complex im-
data of several solids: a sodium aluminosilicate pedance plots or the frequency dependence of
(Na20-0.4A1203-2.2SiO2) glass, a quartz crystal capacitance). Nearly the same values of the fitting
and a lithium triborate glass. The method of sample parameters are obtained when we use log(a)
preparation is described in Ref. [9] for sodium and log(f) as input data, and log(a)=
aluminosilicate glass, in Ref. [10] for quartz crystal log(ado + A*10 s log(:)) as the model expression to fit
and in Ref. [11] for lithium triborate glass. All the data without any weighting factors.
results reported here are based on data obtained Three frequency windows are chosen to examine
using precision capacitance bridge assembly (GR the effect of window width on exponent s of Eq. (1).
1616) at 10-105Hz. Details of conductivity The result is shown in Fig. 1, as an example, for ac
measurements are presented in Ref. [12]. conductivity data of sodium aluminosilicate glass
Conductivity data are analyzed according to the at 59.8°C. In this case the power law region with
power law expression, Eq. (1), using the non-linear s = 1 is at a much higher frequency than the experi-
regression procedure (NLIN) in SAS program [13]. mental range available so that Eq. (2) becomes
Experimental values of conductivity and corres- practically equivalent to Eq. (1). The important
ponding frequency are input as values of the vari- observation we find from the analysis is that, as the
ables a andf. The mathematical model defining the frequency window increases from the low frequency
relationship between a and f is: a = ad¢ + A * f s side encompassing higher and higher frequency re-
where A* and s are the fitting parameters and A* is gion, s decreases. Further, we note that s is most
related to A in Eq. (1) by A* = A 2 n ~. A weighting sensitive to the window width in the frequency
factor w = a-2 is used for each data set when evalu- region where conductivity just begins to deviate
ating the error sum of squares in the least squares from the dc plateau. Away from this 'bend' in log(a)
1410 H. ,lain, C.H. Hsieh / Journal of Non-Crystalline Solids 172-174 (1994) 1408-1412

1,2 a greater part of the experimentally available


1.1
N a20-0.4AI203-2.2SIO 2 glass window, which means that, effectively, the
frequency window moves to lower frequencies.
1.0 0 This shift should result in a higher s, as observed.
0.9
Since we have carefully chosen the a data that
is unaffected by electrode effects, this phenomenon
m 0.8 O
should not be caused by errors from electrode
H
0.7
o o
o
E~
effects. Apparently Nowick et al. found a plateau,
s = So, around room temperature because they
0.6
o 50 - 5k Hz confined the analysis to those frequencies for which
[] 500- 50k Hz
0.5 ff ~> 2adc [2b].
By shifting log(a) and log(to) axes with a scaling
0.4 i i i i i i i

0 50 100 150 200 250 300 350 400 factor determined by the dc conductivity and elec-
T (K) trical conductivity relaxation time, Kahnt [15] ob-
tained one 'master' curve of a(to) for several glasses.
Fig. 2. Variation of s with temperature, T, for two frequency So he concluded that all the materials have funda-
windows of the same width. mentally the same frequency dispersion. In order to
examine this suggestion, we plot a(to) and M"(to)
together for two different solids in Fig. 3. One is
vs. log(f) plot, s becomes less and less sensitive to sodium aluminosilicate glass which has high Na
the choice of frequency window. These observa- ion concentration ( ~ 28mo1% Na20) and the
tions imply that Eq. (1) does not describe the data other is a quartz crystal with total mobile ion
completely, particularly in the region of the bend. It concentration of several hundred ppm. For the
is noteworthy that this failure of Eq. (1) for describ- high mobile ion concentration glass which is typi-
ing the data is outside the uncertainty arising from cal of the glasses used by Kahnt, the M"(to) peak
errors in the measurements or the choice of dc frequency occurs where a(to) begins to increase rap-
conductivity. idly (Fig. 3(a)). For quartz crystal, the peak in
Next, following Lim and co-workers [2], ac con- M"(to) occurs, however, where a(to) is very slowly
ductivity data of sodium aluminosilicate glass over varying, i.e., much before the bend in the a(to) plot
the temperature range of 4.3-350K is analyzed (Fig. 3(b)). From a comparison of these two plots,
using power law to examine the transition of s. Two we suggest that the relative position of the bend in
frequency windows of the same width are chosen: a(to) with respect to the conductivity relaxation
50 Hz-5 kHz and 500 Hz-50 kHz. The result is time (as given by the frequency where M" is max-
shown in Fig. 2. Up to ~ 100 K we obtain s - 1.02 imum) could be a sensitive function of ion concen-
irrespective of the window position. This region is tration and possibly some other material property.
discussed by Lu et al. [14] in these proceedings. This effect is apparently obscured if comparison is
Above ~ 100 K, s shows a transition from s - 1.02 made among different solids with relatively high
to s < 1, but the transition is dependent on the mobile ion concentration. Fig. 3(b) shows that for
choice of frequency window. A lower frequency quartz there is a slowly varying afro) which contrib-
window reduces the transition temperature. Fur- utes to the non-Debye characteristics in the M"
ther, from ~ 300 to ,,~ 350 K, s increases again peak. Apparently, in high mobile ion concentration
rather than remaining constant (s = So) as reported solids investigated by us, the s < 1 and s = 1 re-
by Lim and co-workers [2] and shown by the gions are not so clearly resolved and we see a con-
broken line in Fig. 2. This increase can be under- tinuous variation from dc-like to s = 1 behavior.
stood as follows. We noted above from Fig. 1 that, Thus, in general, there is a material-dependent
when the frequency window width decreases by component in the a(to) plot. This component is
excluding higher frequency data, s increases. When most easily observed in the frequency range deter-
the temperature increases, the dc plateau covers mined by the conductivity relaxation time, or the
H. Jain, C.H. Hsieh ,' Journal o[ Non-Crystalline Solids 172- 174 (1994) 1408 1412 1411

1 0 "6 0.04 1 0 "e 0.08


(a) N a20-0.4AI203-2.2SIO 2 glass (b} Quartz
M" (223°C)
M" (79.80C) 298.8°C

I 0 "7 0.03 0.06

1 0 "e 2230C

U) b l0-, 0.02 ~.- 161.4°C 0.04

1 0 "10 120"C

1 0 "9 0.01 0,02

/
1 0 "1° 0 1 0 "1= ................. ' ....................... 0
101 102 103 104 10 s l0 s 101 1 02 1 03 1 04 1 0s 1 0~

f (Hz) f (Hz)

Fig. 3. Location of M" (to) peak with respect to conductivity dispersion in (a) glass and (b) high purity quartz crystal. Curves are to guide
the eye.

M" plot. For a discussion of various processes 0.05


lO'e,
observed from the frequency and temper- A: I~:0.53 Lithium Trlborate (#28)
B: s=o.6g
ature dependence of a, the reader is referred to C: s=0.75 ." . ' 0.04
1 0 .7 ~" ~ , C a formalism
Ref. [ 16]. ,~ 20 - 5k Hz . ' ~ /
¢ "~ \..j.
We emphasize that often we are forced to choose .'<o 0.03
a frequency window by the equipment. This is the 1 0 -e a
/ ' '--~
~..O ~
"~ ~B
a formallam
window of a(to). M"(og) formalism often selects 0.02

a smaller frequency window around ~ = 1. a(~o) A


1 0 -o
data are likely to be more accurate than M"(co) data 0.01

(when electrode effects are not dominant) because M" formalism - ~_


the error in M" includes error from conductivity as 1 0 "I° 0
101 102 10 s 104 105 10 e
well as capacitance measurements. So in order to
f (Hz)
determine the material response in the vicinity of
conductivity relaxation time, it appears logical to Fig. 4. Appropriate selection of frequency window
choose the window width from M"(to) plot and (20 Hz-5 kHz) for analyzing the data by the power law using
then use this window for analyzing a(~o) via Eq. (1). M"(to) formalism as the basis.
With this view, we have analyzed the conductivity
data of lithium triborate glass. The fl-value has
been determined earlier to be 0.53 [7] by using the
M"(~o) formalism. As shown in Fig. 4 and observed 20 Hz to 5 kHz, which is the range in which M"(~o)
by many others (for example, Ref. [3]), there is formalism shows a good fit to the experiment data.
a significant deviation of experimental M" data It gives s = 0.75. In either case, note that fl :~ l-s,
from the fitted curve at high frequencies. Two fre- which indicates that the often used approximation
quency windows are chosen to analyze the data in fl--- l-s should be used with caution. To confirm
a(~) formalism by using Eq. (1). One is from 20 Hz its validity, one must exclude contributions to
to 100 kHz, which is the same as the total window conductivity from sources like A D W P configura-
of datapoints. It gives s = 0.66. The other is from tions.
1412 11. ,lain, C.H. Hsieh / Journal of Non-Crystalline Solids 172-174 (1994) 1408-1412

3. Conclusions J.M. Tarascon and M. Armand (materials Research


Society, Pittsburgh, PA, 1993) p. 219.
From above analysis, we conclude the following. (b) A.S. Nowick, B.S. Lim and A.V. Vaysleyb, these Pro-
ceedings, p. 1243.
(1) The presence of window effect illustrates that [3] C.T. Moynihan, L.P. Boesch and N.L. Laberge, Phys.
conductivity power law (Eq. (1)) is only an approxi- Chem. Glasses 14 (1973) 122.
mate description of the actual frequency depend- [4] G. Williams, D.C. Watts, S.B. Dev and A.M. North, Trans.
ence of ionic conductivity. Faraday Soc. 67 (1971) 1323.
[5] K.L. Ngai, J.N. Mundy, H. Jain, O. Kanert and G. Balzer-
(2) The transition from s = 1 to s < 1 region
J611enbeck, Phys. Rev. B39 (1989) 6169.
with increasing temperature is sensitive to the [6] H.K. Patel and S.W. Martin, Phys. Rev. B45 (1992)
choice of frequency window. 10292.
(3) The frequency response in the vicinity of con- [7] K.L. Ngai, R.W. Rendell and H. Jain, Phys. Rev. B30
ductivity relaxation time is a material property, i.e., (1984) 2133.
it is not the same for all the ionic conductors. [8] K.L. Ngai and R.W. Rendell, in: Handbook of Conducting
Polymer, Vol. 2, ed. T.A. Skotheim (Dekker, New York,
(4) A more accurate description of material re- 1986) p. 967.
sponse in the vicinity of conductivity relaxation [9] C.H. Hsieh, H. Jain, A.C. Miller and E.I. Kamitsos, J.
time may be obtained by combining M"(o~) and Non-Cryst, Solids 168 (1994) 247.
it(to) formalisms. [10] H. Jain and A.S. Nowick, J. Appl. Phys. 53 (1982) 477.
[11] H.L. Downing, N.L. Peterson and H. Jain, J. Non-Cryst.
Solids 50 (1982) 203.
The authors wish to thank X. Lu for sharing his [12] H. Jain, in: Experimental Techniques of Glass Science, ed.
low temperature electrical data. This work is sup- C.J. Simmons and O.H. E1-Bayoumi (American Ceramic
ported by National Science Foundation. Society, Westerville, OH, 1993) p. 433.
[13] SAS/STAT User's Guide, ver. 6, 4th Ed., Vol. 2, 1989, SAS
Institute Inc., Cary, NC 27512, USA.
[14] X. Lu, H. Jain and O. Kanert, J. Non-Cryst. Solids, these
References Proceedings, p. 1436.
[15] H. Kahnt, Ber. Bunsenges. Phys. Chem. 95 (1991) 1021.
[1] A.K. Jonscher, Nature 267 (1977) 673. [16] H. Jain and O. Kanert, in: Defects in Insulating Materials,
[2] (a) B.S. Lim and A.S. Nowick, in: MRS Symp. Proc., ed. O. Kanert and J.M. Spaeth (World Scientific, Sin-
Vol. 293, Solid state Ionics I I k e d . G.-A. Nazri, gapore, 1993) p. 274.

You might also like