Strengthening by Ion Exchange
by MARTIN E. NORDBERG, ELLEN 1. MOCHEL, HARMON M. GARFINKEL, and JOSEPH S. OLCOTT
Research and Development Laboratories, Corning Glass Works, Corning, New York
This paper is a brief review of the physical and
chemical methods of strengthening glass, in par-
ticular that due to the large for small ion exchange
resulting from treatment in molten salt at tem-
peratures below the annealing range for t h e
glass. Abrasion of treated glass is shown to re-
duce markedly the strength of experimental
alkali-lime-silica glasses as well as commercial
glasses of this and other types. Treated alkali-
alumina-silica a n d alkali-zirconia-silica glasses,
however, are less affected by abrasion. Strength
after abrasion increases with alumina or zirconia
content, reaching 117,000 psi for cane of a 35%
alumina glass. The explanation for the alumi-
num or zirconium effect may involve a n unusual
ionic environment caused by their presence in the
lattice.
1. Introduction
LASS has a conlbination of desirable properties which
G constitute a unique asset for many modern-day needs
when compared with other available materials such
as metals, plastics, and most crystalline conglomerates.
Among these might be listed transparency, hardness, good
durability, low cost, ease of forming, and nondeformability.
But glass also has an undesirable property which has se-
verely limited its use throughout the ages, and this is its
vulnerability to breakage, or lack of strength.
During the last seventy years or so a great deal of study
has gone into the understanding and improvement of strength.
From studies of interatomic forces, backed up by experimental
strengths obtained on fibers drawn under ideal conditions,
it is well known that the intrinsic strength of glass is about
2,000,000 psi. However, surface flaws ranging from the
submicroscopic Griffithl flaws to the major abrasions result-
ing from rough treatment of the surface during use markedly
reduce this strength.
II. Physically Strengthened Glass
Strengthened glass has been known since the seventeenth
century, when Prince Rupert drops were first studied. Here,
rapid quenching sets up a system of high surface compression
with a balancing high interior tension, giving high strength
as long as the compression layer is not pierced, Similar
stresses occurred, but usually to a lesser degree, in most
manufactured glass not cooled slowly enough. The resultant
strength was not useful, however, because there was not a
uniform compressive stress over the entire surface. Tensile
stresses a t the surface often resulted in spontaneous break-
age. An actual improvement in usable strength was obtained
by substantially eliminating these tensile stresses, i.e., by
annealing the glass.
The first real purposeful strengthening was carried out by
Otto Schott,2 using the overlay process of a low-expansion
surface glass on a high-expansion core glass. After cooling,
the greater contraction of the core compared with the sur-
face layer put the surface in compression and the core in
tension. This technique helped to define the requirement
for the useful strengthening of glass, namely, an outer com-
pression layer of sufficient thickness that it could withstand
abrasion, with the glass core in tension.
215
The most widely known and used strengthening process
today is chill tempering, the commercial use of which dates
back to 1930. Strength increases up to 30,000 psi can readily
be obtained. There are drawbacks to tempering, however,
which limit its application. Articles may be shaped so that
it is impossible to obtain sufficiently uniform stress distribu-
tion. For example, articles which have sharp reentrant angles
do not temper satisfactorily. Depending on the expansion
of the glass and the desired strengthening, there is a minimum
thickness required for tempering to be able to set up the neces-
sary stress distribution. For some applications, one of the
most serious limitations to tempering is an insufficient final
strength, even when other conditions are met.
111. Chemically Strengthened Glass
In contrast to this physical method of strengthening, there
are other methods of strengthening whereby the chemical
composition of the surface of the glass differs from that of the
interior In one method, a composite article is formed, the
outer glass or glass-ceramic having a lower thermal coefficient
of expansion than the interior. The composite may come
from using two different glasses, as with the Otto Schott
overlay process, or from chemically changing the compo-
sition or structure of the outer layer by high-temperature ion
exchange and/or by cry~tallization.~Thus the composite
cooled from above its annealing temperature will, at room
temperature, have the surface in compression and the interior
in tension, with a resultant increase in strength.
Another method is strengthening by large for small ion
exchange at a relatively low temperature, as described
recently by Kistler4 and by Acloque and T o ~ h o n . ~
This is the
Presented at the Fall Meeting of the Glass Division, The Ameri-
can Ceramic Society, Bedford, Pa., October 12, 1963. Received
November 4, 1963; revised copy received January 7, 1964.
The writers are, respectively, senior research associate, Chem-
istry Research Department, research chemist, Chemistry Re-
search Department, manager, Physical Chemistry Research
Department, and director of chemical research, Research and
Development Laboratories, Technical Staffs Division, Corning
Glass Works.
1 A. A. Grifith. “Phenomena of RuDture and Flow in Solids.”
Phil. Trans. Roy.’Soc. London, A221, i63-98 (1920); abstracted
in J . A m . Ceram. Soc., 4 [6] 513 (1921).
2 E. Berger, “Unbreakable and Malleable Glass,” Nuturwzssen-
schuften, 2 5 , 79 (1924).
( a ) J. S. Olcott and S. D. Stookey, “Strengthening by Surface
Crvstallization” : DD. 400403 in Advances in Glass Technolom-
Technical Papers ;(Sixth International Congress on Glass, W-ish-
ington, D. C., July 1962. Compiled by The American Ceramic
Society and sponsored by the International Commission on Glass.
Plenum Press, New York, 1962. 661 pp.; Ceram. Abstr., 1962,
~~
September, p. 227i.
( b ) H. M. Garfinkel. D. L. Rotherniel. and S. D. Stookev. _.
“Strengthening by Ion Exchange,” ibid., p i . 404-11.
(c) H. P. Hood and S. D. Stookey, “Method of Making a Glass
Article of High Mechanical Strength and Article Made Thereby,”
U. S. Pat. 2,779,136, January 29, 1957; Ceram. Abstr., 1957,
June, p. 1328.
4 S. S. Kistler, “Stresses in Glass Produced by Nonuniform
Exchange of Monovalent Ions,” J . Am. Ceram. Soc., 45 [2] 59-68
(1962).
6 Paul Acloque and J. Tochon, “Measurement of Mechanical
Resistance of Glass After Reinforcement”; pp. 687-704 in Sym-
posium sur la resistance mecanique du verre et les moyens de
l’ameliorer (Colloquium on Mechanical Strength of Glass and
Ways of Improving It), September 25-29, 1961, Florence, Italy.
(Footnote 5 continued on p . 216)
 Journal OJ l’ha American Ceramic Society- Novdberg et al.
BEFORE ION EXCHANGE
I AFTER ION EXCHANGE
Fig. 1 . Crowding from low-temperature exchange of KC for Na+
ions.
method discussed in this paper. When the ion cxchange is
carried out below the so-called transformation range, the
“crowding” or stress produced by the larger ion taking the
place of the smaller ion is not relieved and permanent con-
pression remains. This is shown in an oversimplified fashion
in Fig. I . ’The ion exchange may cause an increase in the
expansion cocfficient of the surface glass which reduces the
compression on cooling to room but this effect
is relatively small.
IV. Abrasion
For any strengthening process to be of practical iinpor-
lance, it is necessary that the resulting strength be retained
during service of the article. Classes used for almost every
application, e.g., windows or containers, are subjected to some
degree of surface abrasion which lowers the strength.
In this laboratory, a scratch type of service abrasion is
reproduced by rubbing the surface of the glass in a direction
normal to that of the breakage with 150-grit abrasive paper.*
The depth of the resultant scratches has been estimated to
be in the range 20 to 3Op. The thickness of the compression
layer, thereforc, should be in excess of this amount in order
to retain high strcngth.
An impact type of abrasion is produced by tumbling glass
cane specimens in a ball mill containing SO-grit silicon car-
bide. * This procedure has been arbitrarily standardized.
Ten specimens are rolled together with 200 cm3 of 30-grit
S i c for 15 minutes in a half-gallon ball-mill jar turning at 90
to 100 rptn. Microscopic examination of a tumble-abraded
specimen shows it to have rather shallow craters, 1 0 to 20p
deep. Correlation of depth of the compression layer with
modulus of rupture values indicates, however, that the com-
pression layer must be more than 8Op deep to retain modulus
of rupture values above 50,000 psi after abrasion. This
indicates that there are cracks extending from the bottom of
the crater down into the compression layer another 40 to Bop.
With either scratch or tumble abrasion, strength values are
obtained that arc usually lower than those €or unabraded
Published by Union Scientifique Continentale du Verre, Char-
leroi, Belgium, 1962. 1044 pp.; Ceram. Abstr., 1963, January,
p. 31;.
6 Alfred Sendt, “Ion Exchange and Diffusion Processes i n
Glass” (in Gerrxan); pp. 307-32 in Advances iri Glass Technol-
ogy-Technical Papers of Sixth International Congress on Glass,
Washington, L). C., July 1962. Compiled by The American
Ceramic Society and sponsored by the International Commission
on Glass. Plenum Press, New York, 1962. 661 pp.; Ceram.
Absti., 1962, September, p. 2272..
* Manufactured by thc Electro-Minerals Division, The Car-
borunduin Company, Niagdra Palls, New York.
Vol. 37, No. 5
specimens but with less spread between the individual values
than for unabraded specimens. What is more important is
that these strength values more nearly reflect the glass
strength during later use where abrasion is almost certain to
occur.
V. Experimental Procedures
Unabraded strength and its reduction due to abrasion
depend on both the strengthening treatment and the coni-
position of the glass being strengthened. The strengthening
treatment may affect both the depth of the compression
layer and the compression a t the surface, which may in turn
affect the lrardness or abrasion resistance of the surface.
The glasses were treated in molten alkali nitrates held to
within 5 “C of the desired temperature in Nichrome-element
furnaces. The salts were contained in I- or 3-liter type 308
stainless steel beakers. Specimens 4 in. long were cut from
drawn cane, nominally 0.25 in. in diameter, and placed in
stainless steel racks which supported them separately and
vertically and touched the glass specimens only a t the ends.
After the desired treatment time, the rack holding the rods
was removed from the hot salt bath, cooled in air, and washed
in warm tap water. Washing was carried out within 1 hour
after reinoval from the salt bath. After abrasion by tech-
niques described in the foregoing the modulus of rupture was
determined by breaking on an 8.9-cm span with a Tinius
Olsen Electromatic testing machine, using a double central
knife-edge with 1.8-cm spacing.
With few exceptions, five rods of a given composition were
used in each experiment. The total number of specimens
varied up to 28 in the 1-liter salt bath and up to 50 in the ?-
liter bath.
Baths were reused but never to the point where the con-
tamination level of the ion corning from the glass exceeded 1%
calculated as the nitrate. The data points given are simple
averages of five or more individual values and, where shown,
the deviations are standard deviations from the mean.
VI. Glass Composition
Important as the actual treatment may be, the present
work shows that the effect of glass composition on abraded
strength can be far greater. The effect of composition is
shown in Table I, which lists a group of commercial glasses
and two experimental glasses, all treated for l ( i hours in
potassium nitrate a t 400“C. Unabraded strengths are also
shown. It can be seen that although conventional soda-
lime glasses may show high unabraded strengths, the high
strength is not retained after abrasion.
It is obvious that there is some minimum level of soda
required to obtain sufficient ion exchange for strengthening.
Hut the data in this table indicate that other elements also
play an important part. The borosilicate glass used in
Pyrex-brand laboratory ware is not strengthened by this
method. This probably is not due entirely to its low alkali
content (47, NazO), since the experimental borosilicate glass
with much higher alliali also is not strengthened. In par-
ticular, it can be seen that both boron and lead depress the
amount of ion exchange and the strengthening.
Of great interest is the high strength retained after abra-
sion in the experimental soda-alumina-silica glass. By com-
parison with the soda-lime-silica bulb glass it can be seen that
this is not due primarily to an increase in the amount of re-
action, but suggests an effect of the alumina in the glass net-
work.
The strengths of simple soda-alumina-silica glasses are
compared with those of simple soda-lime-silica glasses in
Fig. 2. All were treated for JB hours in potassium nitrate at
40OOC. The numbers beside each point give the weight of
potassium exchanging for sodium ion per square centimeter
of surface. Note that the quantity of reaction decreases
markedly with increasing CaO content, whereas it changes
but little, if a t all, with increasing A1203 content. The re-
M a y 1964 Strengthening by Ion Exchange 217
Table I. Strengthening by Ion Exchange ( 1 4 Hours, 400°C)

Approximate composition
Production glasses
7
Experimental glasses
_____h--.p.--p

Soda-alumina-
-
(wtC/o) nuit) Sheet Borosilicate 1,ead Borosilicate silica

73 72.5 81 56 75 54.5
16.5 15 4 4 15 16
8 2
1 19
10 2
2
MgO
I'bO 30
TiOa 4 5
K exchanged (mg/cmZ)
+ 05 0 15 0.04 0.2 0 16 0 5
Modulus of rupture X 10 - 3 (psi)
N o abrasion 69 j z 28 49 f 21 29 & 8 12 & 3 34 f 13 87 =I=
2
150 grit 17 =t5 8 f 0 3 10 f 1 6 f l 16 f 8 68 f 5

I
0

-0 80-
w
3

o 20% Na20
I I I I X 10%Na20
5 10 15
A1203 or CaO (wt%)
Fig. 2. Strength of Alp03 VS. C O O glasses. Each contains 16% I I I I I I I 1
NoyO. Treatment: 16 hours in K N 0 3 a t 400"C, 150-grit obra- 5 10 15 20 25 30 35
sion. Numbers indicate mg per cm2 of KaO. Alumina (wt'/o)
Fig. 3. Strength of alumina glasses. Treatment: 16 hours a t
380°C, 150-grit abrasion. Numbers indicate mg per cm2 of KzO.

duced amount of reaction a t higher lime concentrations, in

this case, may be the explanation for the reduced strengths.
In other words, there may be insufficient compression
depth to resist the 150-grit abrasion at 10 and 15% lime.
Glasscs containing 5% of the modifying oxides have about
equal strengths in spite of the Fact that the quantity of
reaction varies by almost a factor of two. At this low level
of modifying oxide the CaO apparently has an effect a t
least equal to that of AlpOaon the network structure. The
abraded strength decreases sharply, however, as CaO is
either increased or reduced.
The effcct of higher alumina content in simple soda-
I I I
alumina-silica glasses is shown further in Fig. 3. The most 5 10 15 20
important information shown in this figure is again the marked Zirconia (wt%)
effect of alumina content on the 150-grit abraded strength. Fig. 4. Strength of zirconia glasses. Treatment: 16 hours at
The quantity of reaction, noted beside each point, is essen- 380°C, 150-grit abrasion.
tially constant up to 25% A1203,and again cannot account
for the great increase in strength with increasing alumina.
I t is also interesting to note that lowering the soda in
the glass from 20 to 10yoreduces the abraded strength some- If these soda-alumina-silica or soda-zirconia-silica glasses
what but does not change the direct correlation between are treated in the same way as in the foregoing and then
strength and alumina content. tumble-abraded, very little of this strength will be retained
The writers found a correlation between strength and zir- because the depth of the compression layer is not sufficient
conia content in soda-zirconia-silica glasses similar to that to withstand this degree of abrasion. If, however, lithia-
bctwcen strength and alumina content, as shown in Fig. 4. alumina-silica glasses are treated for 4 hours at 40OoC in a
In this system, the effect of increased soda is more pro- sodium nitrate bath and then tumble-abraded, the strengths
nounced than in the alumina glasses. The increase in soda shown in Fig. 5 are obtained. In this case, the mole percent
from 1 5 to 20% gives nearly a 50yo increase in strength a t rather than the weight percent of LizOwas kept constant and,
?()yo zirconia levels. This compares with the 15% strength therefore, the weight percent of LipO varied slightly, from
increase in ZOyo alumina glasses when the soda increases 9.3 to 8.2, as A1203was increased from 14 to 35%. The error
from 10 to 20%. produced by this variation of LizO is small. The rate of
218 Journal of The American Ceramic Society-Nordberg
look
B4ot
I I 1 I I 1 I I
5 10 15 20 25 30 35
Alumina (wly,)
Fig. 5. Strength of LizO-Alp03-SiO~ glasses (9% Lb.0).
ment: 4 hours a t 4OO0C, tumble abrasion.
Fig. 6. Stress profile after exchange o f Na+ far Li+. The left
side i s the view through air and the right side through gloss. (Cour-
tesy of J. S. McCartney, Technical Services Research, Corning
Gloss Works.)
sodium for lithium ion exchange is so much faster than the
potassium for sodium exchange, hecause of the small ion
sizes involved, that there is no problem in obtaining suf-
ficient compression-layer depth in a reasonable time and the
treated glasses can withstand tumble abrasion. Once suf-
ficient depth is obtained, the governing factor for strength
is shown to bc largely the alumina content of the glasses.
An illustration of the stress profile in a plate of glass that
was strengthened by exchanging sodium for lithium ions in
the surface layers is given in Fig. 6, an actual photograph
taken from the work of J. S. McCartney in this laboratory.
The vertical black line shows the position of the air-glass
interface. A glass plate was treated for 4 hours in sodium
nitrate at 40OOC. A section was then cut from the plate in
such a way that a well-collimated beam of polarized light could
be passed through it in a direction parallel to the original
treated surface. The beam was then passed through a Babi-
net compensator and a polarizing analyzer.
The displacement of the Rabinet fringes upward near the
surface of the specimen indicates a compressive stress, the
downward displacement a tensile stress. The plane of zero
stress, or neutral plane, is usually 0.010 to 0.020 in. from the
surface.
Accurate determination of the depth from the birefringence
pattern proved to be difficult because of index variations
and consequent beam distortion. This was minimized by
using a specimen thickness of only 0.100 in. Even then the
et al. Vol. 47, No. 5
70 -
60-
- 5.0-
40 10 PO 30
(MIN)
Treat- Fig. 7. Rate o f ion exchange in glasses with 2570 Ale03 and
16% NazO.
575’ 550’ 525” SOP0 4500
I I
1.2 1.3 I .4
(-+-*K)~ lo3
Fig. 8. Arrhenius plot showing the dependence of rate of K t
uptake on temperature.
method is applicable only for the ion pair (Na+ for Lif)
exchange, in which case the actual ion substrate causes a
decrease in index to compensate at least partly for the
increase caused by the compressive stress.
VII. Treatment Variables
Returning to the soda system, the solution to the problem
of obtaining high tumble-abraded strengths from a high-
alumina glass is to form a compression layer deep enough to
withstand this kind of abrasion. To do this, the extent of
the treatment can be increased either by treating for longer
periods of time or by treating at higher temperatures.
Previous workers7have shown that alkali ion exchange in a
glass is a diffusion-controlled reaction; i.e., the rate of the
reaction follows Fick’s diffusion law, as seen in Fig. 7. This
means that it is necessary to treat a glass four times longer
to double the amount of reaction.
Activation energy can be obtained from the usual plot of
log of the amount of reaction divided by the square root of
time, Q / z / i ; versus the reciprocal temperature, as shown in
Fig. 8. This energy is 28.9 kcal per mole.
With the particular glass of Fig. 7, as in those of Fig. 2,
16 hours of treatment does not give a sufficient amount of
reaction, and therefore there is not the depth required for
tumble-abrasion resistance. Doubling this treatment does
7 ( a ) Bela von Lengyel, “Die Diffusion von Kalium in Glass”
(Diffusion of Potassium in Glass), 2.Physik. Chem., A167 [3]
295-311 (1933); Ceram. Abslr., 13 [Q]233 (1934).
( 6 ) Gunther Schulze,“Formation of Poorly Conducting Layers
During Electrolysis of Glasses,” Ann. Physik., 37,435-71 (1912).
May 1964 Space Charge and Electrode Polarization in Glass, II
loot
I5MIN IHR , 4HR I
I 16HR
10 20 30
JTlME I M I N )
Fig. 9. Effect of temperature on strength of a glass containing
2 5 % A1203 and 16% NasO. Treatment: 150-grit abrasion.
give abrasion resistance but requires 64 hours of treatment.
A better way would be to treat the glass at higher tempera-
tures, as shown in Fig. 9. In 15 minutes there is sufficient
depth €or high strength, e.g., above 20,000 psi, after tumble
abrasion, even at 575°C. At 1 hour, good strengths are ob-
Space Charge and Electrode Polarization in Glass, II
by PAUL
Philco Research Laboratories, Philco Corporation, a Subsidiary of Ford Motor Company, Newpart Beach, California
This paper is Part IT of a two-part series review-
ing the present understanding of space charge and
electrode polarization in glass and the inter-
relations of these concepts with the usually meas-
ured electrical properties of glass. Emphasis is
placed on a theory describing the distribution of
static space charge in material having only a
single mobile charge carrier. Charge dissocia-
tion and recombination are considered and the
carriers are assumed to move under diffusion and
conduction. The charge distributions predicted
by the theory are discussed and are compared
with the results of experiments which were
undertaken specifically to verify essential fea-
tures of the predictions. The theory is confirmed
qualitatively by the observed charge distribution
shapes and asymmetries. Quantitative confirma-
tion is also shown in computed values of mobility,
diffusion constant, and contact potential ratios.
This theory has here been expanded to include a
linearized ac treatment, and the phenomena pre-
dicted by this treatment fit the known data semi-
quantitatively. A large low-frequency relaxation
mechanism is predicted. Previous theory and
experiment were briefly reviewed in Part I. In
Part I1 the predicted and observed space charges
are conceptually related to various phenomena
such as conduction, electrode polarization, di-
electric absorption, and dielectric loss. Needed
improvements in the present theory and a num-
ber of ideas for more experiments are outlined in
219
tained at 550°C, near the strain point of the glass, but a t
575 "C strengths are lower, indicating considerable stress
release. After 4 hours, stress release occurs a t 550"C, so
that higher strengths are obtained at lower temperatures
even with thinner compression layers. Thus, i t can be seen
that the time and temperature of treatment can be adjusted
with respect to the abrasion resistance and strength required.
This is in keeping with the discussion by Acloque and Tochon.6
VIII. Summary
A review and brief summary of what is known about this
low-temperature ion-exchange method of increasing the
strength of a glass follows.
Figure 1 represents the crowding which occurs when a
larger alkali ion is exchanged for a smaller alkali ion. The
temperature at which this reaction takes place is too low for
much relaxation of the stresses created, so that the surface
layer remains in compression and high strength results.
Experimental data have shown that alumina in an alkali
glass greatly increases the abraded strengths which can be
obtained by low-temperature ion exchange. Although there
is some increase in amount of ion-exchange reaction owing to
the presence of alumina in the glass, it is not sufficient to ex-
plain the great strength increase due to the alumina.
M. SUTTON
order to stimulate efforts to acquire a more de-
tailed understanding of these phenomena.
1. Space Charge and Electrode Effects in Glass
(I) Approach
As described in Part I, a detailed quantitative theory
encompassing conduction and dielectric absorption does not
exist for silicate glass. It is not clear exactly how ion motion
affects them both. Does space charge develop in all cases
and, if it does, how does it affect the current flow? Is ion
motion alone responsible for current flow or are electron
drifts involved? How does alternating current differ from
direct current? What is the source of the large reverse
potentials? Is the steady state ever attained? Do current
flows or space charge affect dielectric breakdown or dielectric
loss?
It is apparent from these questions that a great deal needs to
be done to understand even the simplest phenomena satis-
Presented in part at the Sixty-Fourth Annual Meeting, The
American Ceramic Society, New York, N. Y., April 30, 1962
(Symposium on Electrical Phenomena in Ceramics, No. 4-1s-62).
Received April 15, 1963; revised copy received October 28, 1963.
The writer is manager, Applied Physics Department, Physics
Laboratory, Philco Research Laboratories, Philco Corporation.
For Part I (introduction and review) see J . Am. Ceram. Soc.,
47 [4]188-94 (1964).