Journal of Non-CrystaUine Solids 105 (1988) 307-312 307

North-Holland, Amsterdam

ANALYSIS OF IONIC CONDUCTIVITY IN ALKALI AND MIXED-ALKALI

ALUMINOSILICATE GLASSES

G. De MARCHI and P. MAZZOLDI

Dipartimento di Fisica dell'Universitd, and Centro Interuniversitario Struttura della Materia, Via Marzolo 8, 35131 Padova, Italy

A. MIOTELLO
Dipartimento di Fisica dell'Universit~ and Centro Interuniversitario Struttura della Materia, 38050 Povo (Trento), Italy

Received 13 April 1988

Revised manuscript received 5 July 1988

The ionic conductivity in single-alkali aluminosilicate glasses has been analyzed on the basis of a previously published
empirical relation derived for binary alkali silicate glasses. Moreover ionic conductivity in mixed-alkali aluminosilicate glasses
has been analyzed on the basis of a previously published phenomenological model derived from the thermodynamics of
irreversible processes. Some features of the microscopic structure of aluminosilicate glasses and their connection to transport
phenomena have been emphasized. Some aspects of the mixed-alkali effect for ionic conductivity have been revised for low
alkali concentration silicate glasses.

I. Introduction sodium concentration and independent of AI/Na

Many different structural models [1-6] have
been proposed to explain some characteristic
changes of the physical properties, in particular
the electric conductivity increase, which soda
alumino-silicate glasses exhibit at a molar ratio
A1203/Na 2° = 1. This paper is devoted to analyze
ionic transport in alumino-silicate glasses by using
some phenomenological and empirical equations
recently proposed by the present authors [7-9] for
ionic transport both in single and mixed-alkali
silicate glasses. The main results of this paper will
be discussed paying attention to the models whose
basic concept is that aluminum can be either
four-fold and/or six-fold coordinated with oxygen
depending on the glass composition. Because of
the four-fold coordination state of aluminum the
introduction of each A13+ should cause one non-
bridging oxygen to disappear. XPS photoemission
spectroscopy studies [6], performed on soda
alumino-silicate glasses, indicate that the non-
bridging oxygen content is directly proportional to
0022-3093/88/$03.50 © Elsevier Science Publishers B.V.
(North-Holland Physics Publishing Division)
ratio. There are however some other XPS experi-
mental studies performed in aluminum silicate
glasses (see e.g. Smets and Lommen [10]) showing
that the incorporation of aluminum oxide in
sodium sificate glass results in the disappearance
of non-bridging oxygen atoms. However, in the
context of the present paper, by analyzing some
features of ionic conductivity in alumino-silicate
glasses, in particular in the case of mixed-alkali
glasses, it will appear that the XPS results of Lam
et al. [6] are quite plausible. During incorporation
of aluminum oxide in sodium silicate glasses the
charge neutrality condition is achieved by means
of a uniform transfer of electronic charge density
from the sodium ions to the alumino-silicate net-
work. This more uniform charge density distribu-
tion in the glass network modifies the potential
barrier for sodium movements so that it is
smoother relative to that in soda silicate glasses.
The ionic conductivity then increases as A1203
concentration increases. Moreover X-ray fluores-
cence analysis of alumina-containing glasses indi-
308 G. De Marchi et al. / Ionic conductivity in aluminosilicate glasses
cates that aluminium goes into six-fold coordina-
tion when the A1/Na ratio exceeds 1, a result
which supports the proposed structural model [3].
More recently, in order to explain the ionic
conductivity of binary alkali silicate glasses, Maz-
zoldi and Miotello [7,8], introduced empirically an
exponential scahng behaviour in the conductivity
trend:
o = o0 e x p [ - E ( A - c ) / k a T ] ,
where c is the molar fraction of alkali oxide, A is
a parameter connected to the delocalization de-
gree of the alkali valence electron and E(eV/molar
fraction) is an effective activation energy. Obvi-
ously eq. (1) is meaningful, in empirically de-
scribing ionic conductivity, only for ( A - c ) > 0.
Taking into account the previous results reported
by Lam et al. [6] and Day et al. [3] on alumina-
containing glasses, we first try to apply eq. (1) (see
accordingly discussion in ref. [7]) to the ionic
conductivity of these glasses by utilizing the data
from the literature [11] concerning conductivity of
glasses containing alkah ions of a single species.
We note that some effect of composition on the
activation energy for conduction in an alkali
alumino-silicate was shown also by Isard [12]. As
previously stated, we expect the effective energy E
parameter to decrease according to the more uni-
form charge density distribution when aluminum
is four-fold coordinated with oxygen in the glass
network. Such an effect should be more evident at
an A1/Na ratio equal to one. Because of the
six-fold coordination for aluminum, beyond
A I / N a = 1, a conductivity behaviour similar to
that observed for three-component glasses [13],
was expected and confirmed by the very few avail-
able data from the literature. Such an agreement
between our proposed model, based on the non-
bridging oxygen presence and the experimental
conductivity data, suggests that the mixed-alkali
effect also takes place in alumino-silicate glasses,
at least as far as ionic conductivity is concerned.
Such a hypothesis has been confirmed by the
analysis of the conductivity data on mixed alkali
glasses reported by Hayward [14] for (Na, K)
AISi308-SiO 2 systems and by Lapp and Shelby
[15] for (Li, Na) mixed alumino-silicate glasses.
2. Experimental results
2.1. Ionic conductivity in soda alumino-silicate
glasses
The experimental data, collected by different
authors [1] on soda alumino-silicate glasses, have
been found to fit generally conductivity equation
(1) very well. The data reported in fig. 1 refer to
(1) different A1/Na molar ratios. In particular the
lower curve is obtained by using the experimental
data up to an A1/Na ratio of about 0.4, where the
presence of aluminum does not seem to affect
electrical conductivity. The upper curve refers to
the experimental conductivity data of samples with
an A1/Na ratio equal to one. The two different
slopes are due to the different effective energy E
parameter values. The value of E calculated for
the lower curve is the same (0.68 + 0.04 eV/molar
fraction) as the one obtained for soda silicate
glasses [7,8,13]. The slope of the upper curve cor-
responds to an E value equal to 0.49 eV/molar
fraction within a 10% error. For the A parameter
we maintained the same value as the one obtained
for soda sihcate glasses [7]. The variation of the E
-2
i -6
-8 , ,
10 20 30
4 -1
10 (A-c)/T ( K )
Fig. 1. Experimental conductivities [5,11] versus alkali con-
centration function (A - c ) / T (see text), c is the alkali oxide
molar fraction. The lower curve is obtained by using the
experimental data up to an A I / N a ratio of about 0.6, whereas
the upper curve is up to A 1 / N a = 1. The data refer to a N a 2 0
molar concentration in the range 6-20% and temperatures
100-600 o C.
 G. De Marchi et al. / Ionic conductivity in aluminosilicate glasses 309

parameter is observed only for an aluminum con- alkali effect. Lapp and Shelby [15] still observed
tent such that the A1/Na ratio is higher than this effect in lithium soda alumino-silicate glasses
about 0.4. Such results suggest that most probably at a comparable alkali oxide and alumina con-
at low aluminum content the Na interacts centration of about 20 mol%.
with O:-Si4+(O1/2)4- groups rather than with A C conductivity studies performed by Hay-
A13+(O1/2)4- ones, both forming the network ward [14] on five (Na, K)A1Si3Os_SiO 2 series in
glass structure. The E value decreases as the A13 ÷- the 350-650°C temperature range, showed an
(O,/2) ~- number increases, reaching a minimum apparent disappearance of mixed-alkali effect at a
whenA1/Na = 1. We want to underline that it total alkali oxide concentration lower than 10
seems more correct, from the point or view of the mol% and the existence of a "pseudo" mixed
physical mechanism involved in ionic conductiv- alkali effect at an alkali oxide molar concentration
ity, to regard the sodium migration as connected in the 10-13% range. The large deviation from an
to a Na interaction with the molar forming net- additivity behaviour in some of the physical prop-
work groups rather than with the non-bridging erties, particularly in the electrical conductivity,
oxygens. This fact may also explain the non-bridg- when one alkali oxide is replaced by another in
ing oxygen presence at A1/Na = 1, as reported by the glasses, is commonly named the mixed alkali
Lam et al. [6]. From this point of view, the charge effect. Nevertheless it does not mean that the
density distribution in the network plays the de- mixed alkali effect occurs only when the electrical
terrninant role in the ionic transport mechanism in conductivity evidences a deep minimum as func-
the glass. tion of c i / ( c i + cj) where ci indicates the molar
fraction of the i-alkali oxide species involved in
2.2. Mixed alkali effect in aluminosilicate glasses the phenomena. A previously reported phenom-
enological model derived from the thermody-
Now we want to analyze the ionic transport in namics of irreversible processes of conductivity
mixed alkali alumino-silicate glasses in order to has been shown to account for many different
best clarify the mechanism governing the mixed mixed-alkali ionic conductivities in silicate and

. GLASS 1 - GLASS 2 GLASS 3 GLASS 4 _ GLASS 5

'x \\o \\~' x\ "o xx

O~ \, 'o
\x\ \\\ \\x,\ ~xx \xx x~x~
\ \ i
I~ \ \\ 0
~ -4 ,
,
,
\
\~\\ \\\x \\\\~_ ~ ~\
\\
\\
x
"" \ x \

\o ~\ o,
\ \\ \x /
\ \\ ~e /
\\ • •
\\e ~-- \\e \\
o \ \
\\ \ • T=673K 1
\
\\ a T=773 K I
-6 ~T=873~
I I I I I I I I I I I I I I I I I I I I I n n t I
0 0.2 0,4 0.6 0.8 1. 0 0.2 0.4 0.6 0,8 1. 0 0.2 0.4 0.6 0.8 1. 0 0.2 0.4 0,6 0.8 1. 0 0.2 0.4 0.6 0.8 I.
K/(K+ Na)
Fig. 2. Electrical conductivity of mixed-alkali glasses at different temperatures. The symbols are the experimental data [14] and the
dashed lines the theoretical results (see table 1).
310 G. De Marchi et al. / -Ionic conductivity in aluminosilicate glasses

germanate glasses, in a wide temperature range

[8,9,16]. The conductivity in a system where two o
-6 /
different species, i and j, are present in a dilute GLASS 6
/
concentration, may be expressed as /

G ~-- (~ii -}" Gjj "q- 2o, j, (2) /

¢
where Oil and ojj represent the contribution of the o \\ I
r -7
E
\
\
/
non-interacting i, j species to the total conductiv- J=
_o.
o \ /I
\ /
ity and G~j accounts for the i-j interaction. Notice \\
ca \\
that the interaction 2oij is required according to o
]D
the general thermodynamical principles (entropy \\ //
production increase for irreversible processes). -8 b /
\ /
Generally such a contribution is disregarded in \ //
\\ /
conductivity equations for mixed-alkali glasses or
\"- J°/ T = 373 K
written as ogj + oj~ thus neglecting Onsager's rela-
tions. A non-linear scaling behavior of conductiv-
! I
ity in o, and ojj has been proposed through the 0 0.5 1.
empirical relations [9] N a / ( N a + Li)

Gii = o~ exp[ - E / ( 2 - ni)/kaT], Fig. 3. Electrical conductivity of a (Li,Na) aluminosilicate

glass. The circles are the experimentaldata [15] and the dashed
o,j: (3) line the theoreticalresults (see table 1).

O,j=G~j exp[-Eij/knT],
where ni,j=ci,j/(ci,j+ cj,i) and ci,j is the i, j
alkali oxide molar fraction. (A notation change,
n;.j instead of ci,s with respect to refs. [8,9] has
been introduced in order to avoid possible confu-
sion with the term c of eq. (1).) Notice that when
n~ (or nj) = 1, i.e.. when alkali of single species are
present the term exp[ - E~,j(2 - ni,j)/knT ] relaxes
to exp[-Ei,j/kBT] where Ei,j= E ( A - f i , j ) with
E, A, c defined in eq. (1). E and A assume the
values pertinent to glasses containing alkali of a
single species.
The replacement of /-alkali oxide by j in a
series of glasses causes the possible cationic sites
for i to reduce, determining an exponential scaling
dependence similar to the one introduced for bi-
nary alkali silicate glasses (eq. (1)) in the conduc-
tivity. The formation probability of unstable i-j
alkali complexes is connected to the o~j interac-
tion term.
We want to emphasize that, if the two scaling
conductivity laws introduced by eqs. (1, 3) are
similar in their mathematical expression, from a
microscopic point of view very essentially differ.
In fact, in the first case an alkali concentration
reduction causes the non-bridging oxygen number
and consequently the available cationic sites to
decrease. In the case of mixed-alkali systems the
replacement of one alkali species by the other
does not affect the total number of non-bridging
oxygens (the total alkali concentration being con-
stant), but modifies the available occupancy sites.
In our conductivity equations, these facts are
related to the Ei(Ej) term which, in mixed alkali
systems, is rescaled by the factor 1/(c i + cj) with
respect to E. This is quite an important fact since
in our scheme for the mixed-alkali effect with
respect to ionic conductivity, the phenomenon is
not connected to the presence of a deep conductiv-
ity minimum, as is usually considered, but to the
change in the energetic features of the transport
processes. The structure of the minimum in the
conductivity function depends on the oij term
connected to the i-j interaction mechanism which
occurs only for (c i + cj) > 0.1 (alkali oxide molar
fraction)
In figs. 2(a-e) and 3 we show, along with the
experimental data [14,15], the theoretical o curves
obtained through eqs. (1-3) by means of a best-fit
procedure. The chemical compositions and the
calculated values of the relevant parameters char-
 G. De Marchi et aL / Ionic conductivity in aluminosilicate glasses 311

Table 1
Glass experimental composition and relevant parameters pertinent to eqs. (2,3).

Glass Alumino- Total alkali Tot. o6 E6 o~ e~ og E;J Ref.

silicate glass oxideconeentr, alkali/A1 (ohm-cm)- 1 (eV) (ohm-cm)-I (eV) (ohm-cm)-1 (eV)
(tool%) ratio
1 Na20-K20 6.7 1 0.8 +0.2 0.4 0.8 +0.2 0.78 - - [11]
2 Na20-K20 8.14 1 1 5:0.2 0.4 1.0 0.76 - - [11]
3 NaEO-K20 10.0 1 1.2 5:0.2 0.4 1.5 0.75 1.2 0.806 [11]
4 Na20-K2 O 11.1 1 1.1 5:0.1 0.4 1.2 0.73 19.0 0.93 [11]
5 Na20-K20 12.5 1 1.25 5:0.25 0.4 1.75 5:0.25 0.68 14.0 0.86 [11]
6 Na20-Li20 20.0 1 1.5 0.45 2.0 0.5 oij = 2)<10 - 9 (ohm-cm) -1 [12]

acterizing the glass series are reported in table 1.

-9
In fig. 4 we report the Arrhenius plot of (2aq)
parameters for glasses 3 - 5 (see table 1).
vo It has to be pointed out that the o u parameter
value is zero in the glasses 1 and 2 where the total
alkali oxide concentration is lower than 10 mol%.
12
This fact is not surprising because, at very low
total alkali concentrations an i-j interaction prob-
ability and consequently an i-j "unstable" com-
I I I I
plex formation are negligible. In fact, as we previ-
ously outlined [7,8,13], this low alkali concentra-
tion regime is characterized by a localized state of
'~ 8 GLASS5 the alkali atom valence electron, corresponding to
r an abrupt conductivity decrease.
o-9

c 10 3. Conclusions
11 A previously reported phenomenological model
(including some empirical relations) both for
-12 single-alkali and mixed-alkali silicate glasses once
I [ [ I
-8
again appears to be capable of accounting for the
ionic conductivity of alumino-silicate glasses. Such
~- 9 ~ GLASS4 a remarkable result permits a unified view of ionic
conductivity behaviour in silicate glasses at differ-
=-10 ent compositions and gives an interpretation of
e~
--=-11 the possible microscopic mechanism governing the
transport processes. Moreover it has been proved
-12 that the mixed-alkali effect, in the ionic conductiv-
ity, is not connected to the appearance of a deep
-13 "~
I I I I
minimum in conductivity behaviour but instead to
1.1 1.2 1.3 1.4 1.5 the changes in the energetic features of transport
lrr x lo 3 (K "1) when one alkali species is gradually substituted by
Fig. 4. Arrhenius plot of (2oij) parameter concerning the another.
results reported in fig. 2(c, d, e).
312 G. De Marchi et al. / Ionic conductivity in aluminosilicate glasses

References [9] A. Miotello and P. Mazzoldi, J. Non-Cryst. Solids 94

[1] D.E. Day and G.E. Rindone, J. Am. Ceram. Soc. 45
(1962) 489.
[2] D.E. Day and G.E. Rindone, J. Am. Ceram. Soc. 45
(1962) 496.
[3] D.E. Day and G.E. Rindone, J. Am. Ceram. Soc. 45
(1962) 579.
[4] J.E. Shelby, J. Appl. Phys. 49 (1979) 5885.
[5l V.K. Hunold and R. Bruckner, Glastechn. Bet. 53 (1980)
149.
[6] D.J. Lain, A.P. Paulikas and B.W. Veal, J. Non-Cryst.
Solids 42 (1980) 41.
[7] P. Mazzoldi and A. Miotello, J. Non-Cryst Solids 94
(1987) 175.
[8] P. Mazzoldi and A. Miotello, J. Non-Cryst. Solids 95&96
(1987) 897.
(1987) 181.
[10] B.M.J. Smets and T.P.A. Lommen, Phys. Chem. Glasses
22 (1981) 158.
[11] N.P. Bansal and R.H. Doremus, Handbook of Glass
Properties (Academic Press, New York, 1986).
[12] J.O. Isard, J. Soc. Glass Technol. 43 (1959) 113T.
[13] G. De Marchi, P. Mazzoldi and A. Miotello, J. Non-Cryst.
Solids, to be published.
[14] P,J. Hayward, Phys. Chem. Glasses. 17 (1976) 54.
[15] J.C. Lapp and J.F. Shelby, J. Non-Cryst. Solids 95&96
(1987) 889.
[16] G. De Marchi, P. Mazzoldi and A. Miotello, J. Non-Cryst.
Solids 104 (1988) 211.