Group Title: early stages of crystallization in alkali-silicate glasses
Title: The early stages of crystallization in alkali-silicate glasses
Full Citation
Permanent Link:
 Material Information
Title: The early stages of crystallization in alkali-silicate glasses
Alternate Title: Alkali-silicate glasses
Physical Description: xiv, 143 leaves. : illus. ; 28 cm.
Language: English
Creator: Kinser, Donald L., 1941-
Publication Date: 1968
Copyright Date: 1968
Subject: Crystallization   ( lcsh )
Glass   ( lcsh )
Metallurgical and Materials Engineering thesis Ph. D
Dissertations, Academic -- Metallurgical and Materials Engineering -- UF
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
Thesis: Thesis - University of Florida.
Bibliography: Bibliography: leaves 138-142.
Additional Physical Form: Also available on World Wide Web
General Note: Manuscript copy.
General Note: Vita.
 Record Information
Bibliographic ID: UF00097803
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 001133269
oclc - 20143525
notis - AFN0638


This item has the following downloads:

earlystagesofcry00kinsrich ( PDF )

Full Text






Dedicated to my wife,



The author would like to acknowledge the assistance

of his advisory committee, Drs. Rhines, DeHoff, Reid and

Carr, and especially his chairman Dr. Hench who contributed

a great deal in the form of discussions, advice and

encouragement throughout the course of this investigation.

The numerous discussions with Dr. Hren are also gratefully


Thanks are also due to numerous members of the

faculty, staff and students of the Metallurgy Department

for many stimulating discussions and a great deal of

experimental assistance. The aid of Mr. E. J. Jenkins

with electron microscopy techniques and Mr. D. A. Jenkins

with many experimental problems is gratefully acknowledged.

The author is also grateful for the financial

support of the National Science Foundation and the United

States Air Force without whose support this research

would not have been possible.







ABSTRACT ...........










INTRODUCTION ............


POLARIZATION .............



SUMMARY ..................

CONCLUSIONS ..............


BIBLIOGRAPHY .....................























: : : :


Table Page


LITHIA GLASSES .............................. 54

MOLE % LITHIA-SILICA GLASSES ................ 56

MOLE % SODA-SILICA GLASSES .................. 58

OF 30 MOLE % LITHIA GLASS AT 4800C .......... 71

POINTS ..................................... 107


Figure Page

1. Schematic melting procedure for 30
mole % lithia-silica glass ................ 20

2. Schematic diagram of x-ray source
and Guinier-DeWolff camera ................ 27

3. Photograph of sample holder employed
in electrical measurements ................ 32

4. Schematic diagram of audio frequency
measurement equipment ..................... 34

5. Schematic diagram of transformer
ratio arm bridge .............................. 35

6. Schematic diagram of radio frequency
measurement equipment ..................... 37

7. Schematic diagram of null detection
system ....................................... 38

8. Schematic diagram of DC measurement
equipment .................................... 40

9. DC polarization curve of 30 mole %
lithia-silica glass in as cast and heat
treated 5 hours at 5000C forms ............ 43

10. Conductivity versus log time curve for
33 mole % lithia-silica glass as cast ..... 45

11. Log DC conductivity versus reciprocal
temperature for 33 mole % lithia-
silica glass as cast ...................... 46

12. Schematic diagram of model used for
polarization calculation .................. 48

13. Replica electron micrograph of 30
mole % lithia-silica glass (30,000x)
as cast .................................... 59


Figure Page

14. Replica electron micrograph of 30
mole % lithia-silica glass (26,000x)
after 5 hours at 5000C ............ ....... 60

15. Replica electron micrograph of 30
mole % lithia-silica glass (31,000x)
after 10 hours at 500C ............... .... 61

16. Replica electron micrograph of 30
mole % lithia-silica glass (31,000x)
after 20 hours at 5000C ................... 62

17. Replica electron micrograph of 30
mole % lithia-silica glass (37,000x)
after 50 hours at 5000C ......... .......... 63

18. Electron diffraction pattern of as
cast 30 mole % lithia-silica glass
at room temperature ................. .. ... 66

19. Electron diffraction pattern of 30
mole % lithia-silica glass at 4800C
after 5 minutes at 4800C .................. 67

20. Electron diffraction pattern of 30
mole % lithia-silica glass at 4800C
after 14 minutes at 4800C ................. 68

21. Electron diffraction pattern of 30
mole % lithia-silica glass at 4800C
after 26 minutes at 4800C ................. 69

22. Electron diffraction pattern of 30
mole % lithia-silica glass at 4800C
after 2.5 hours at 480C .................. 70

23. Electron diffraction pattern of 30
mole % lithia-silica glass at 4800C
after 3.75 hours at 4800C ................. 70

24. Transmission electron micrograph 30
mole % lithia-silica glass at room
temperature as cast (41,000x) ............. 73



Figure Page

25. Transmission electron micrograph of
30 mole % lithia-silica glass at
4800C after 11 minutes at 4800C
(42,000x) ................................. 75

26. Transmission electron micrograph of
30 mole % lithia-silica glass at 4800C
after 35 minutes at 4800C ................ 76

27. Transmission electron micrograph of
30 mole % lithia-silica glass at 480C
after 2.3 hours at 4800C ................. 77

28. Transmission electron micrograph of
30 mole % lithia-silica glass at 4800C
after 4.8 hours at 4800C (23,000x) ....... 78

29. Transmission electron micrograph of
30 mole % lithia-silica glass at 5000C
after 1 hour at 5000C ........... ........ 80

30. Electron diffraction pattern of 30
mole % lithia-silica glass at 5000C
after 1 hour at 5000C ..... ............. 82

31. Transmission electron micrograph of
30 mole % lithia-silica glass at
room temperature after 5 hours at
5000C in bulk form (31,000x) 83

32. Tan 6 versus logio frequency for the
30 mole % lithia-silica glass as cast .... 85

33. Tan 6 versus logio frequency for the
30 mole % lithia-silica glass after
5 hours at 5000C .......................... 86

34. Tan 6 versus logo frequency for the
30 mole % lithia-silica glass after
10 hours at 500 C ...... ................... 87



Figure Page

35. Tan 6 versus logio frequency for the
30 mole % lithia-silica glass after
20 hours at 500 C ........................... 88

36. Tan 6 versus logo frequency for the
30 mole % lithia-silica glass after
50 hours at 5000C .......................... 89

37. Log of frequency maxima versus
reciprocal temperature for various
thermal treatments at 5000C ............... 90

38. Tan 6 versus logo frequency for
the 30 mole % lithia-silica glass
for various thermal treatments
measured at 800C ............................ 91

39. Tan 6 versus loglo frequency for
the 33 mole % lithia-silica glass
for various thermal treatments ............ 92

40. Tan 6 versus log10 frequency for
the 26.4 mole lithia-silica glass
for various thermal treatments ............ 94

41. Tan 6 versus logo frequency for
the 33 mole % soda-silica glass
for various thermal treatments ............ 95

42. Tan 6 versus logo frequency for
the 25 mole % soda-silica glass
for various thermal treatments ............ 96

43. Log DC conductivity versus reciprocal
temperature for the 30 mole % lithia-
silica glass .................................. 98

44. Log DC conductivity versus reciprocal
temperature for the 33 mole % lithia-
silica glass ................................ 100


Figure Page

45. Log DC conductivity versus reciprocal
temperature for the 26.4 mole %
lithia-silica glass ......................... 101

46. Log DC conductivity versus reciprocal
temperature for the 33 mole % soda-
silica glass ................................ 102

47. Log DC conductivity versus reciprocal
temperature for the 25 mole % soda-
silica glass ................................ 104

48. Thermal expansion curve for the 26.4
mole % lithia-silica glass as cast .......... 105

49. Phase diagram for the lithia-silica
system ..................................... 116

50. Free energy-composition diagram for
the lithia-silica system at 500C ........... 117

51. Summary of results for the 30 mole %
lithia-silica glasses ....................... 122

52. Thermal treatment time required to
develop loss peak versus composition
for lithia-silica glasses .................. 123

53. Phase diagram for the soda-silica
system ..................................... 125

54. Free energy-composition diagram for
the soda-silica system at 5500C ............. 126


Subscripts 1 and 2 refer to the parameters for the

matrix and dispersed phase respectively.

A Shape parameter defined by equation (1-11)

d Interplanar spacing

E Electric field

F Frequency

Q Quench rate

R Resistance

T Absolute temperature

t Sample thickness

t+, t Transferrence number of positive and negative
ion respectively

V Volume fraction of dispersed phase

x Distance parameter defined on Figure 12

tan 6 Tangent of the loss angle 6

e' Real part of complex dielectric constant

E"' Imaginary part of complex dielectric constant

E Static dielectric constant
Dielectric constant extrapolated to high

p Surface charge density

a DC conductivity



tan total

tan 6AC

Period of oscillation

Electric potential

Angular frequency (2nrf)

Tangent of the loss angle

AC component of the loss angle


Abstract of Dissertation Presented to the Graduate Council
in Partial Fulfillment of the Requirements for the
Degree of Doctor of Philosophy



Donald L. Kinser

December, 1968

Chairman: L. L. Hench
Major Department: Metallurgy and Materials Engineering

The initial stages of crystallization in glassy

systems are of importance because of their influence upon

the structure of the fully crystallized material. The

precise nature of the structural changes occurring in the

early stages of crystallization was the subject of this


This investigation has shown the appearance of a

metastable transition phase prior to the appearance of

the equilibrium precipitate in four of the five glasses

examined. The metastable phase enables the crystallization

behavior of these glasses to be explained as a process

involving the metastable precipitate as a nucleation site

for the equilibrium lithium disilicate phase.


The compositions examined include a 26.4, 30 and

33 mole % lithia-silica glass and a 25 and 33 mole % soda-

silica glass. In order to detect the structural changes

occurring during nucleation and crystal growth it was

necessary to apply new x-ray, transmission electron

microscopy, AC and DC electrical measurement techniques

to the glasses.




Crystallization of glasses has been the subject of

scientific interest since the early work of Reaumur (1739).

He crystallized a glass bottle by imbedding it in sand

and heating it to elevated temperatures for an extended

time and found that the resulting crystalline article

was extremely brittle and technically useless. In marked

contrast to this experiment the recent work of Stookey

(1962) led to the development of a class of crystallized

glass-ceramic materials (Pyrocerams) which are quite

strong (20,000 psi-50,000 psi) and as a result are

technically useful materials. The principal difference

between the two products which gives rise to the striking

contrast in mechanical properties is the extremely small

crystallite size of the order of 1 micron in the Pyroceram

materials as opposed to large crystals of the order of

1 millimeter (mm) for the Reaumur product. It is thus

evident that the understanding of the physical processes

which lead to an extremely fine grain size are of

considerable importance both from a technical and a

scientific point of view.

- 1 -

- 2 -

In the case of crystallization in a glass matrix

it has been generally accepted that the number of crystal

nuclei present in the glass determine the ultimate lower

limit to the grain size in the final crystalline product.

Extreme importance is thus necessarily attached to the

question of identity of the nuclei as well as the behavior

of these nuclei under varying conditions.

The objective of the present work is to determine

the nature and behavior of the structural changes occurring

during the nucleation of simple binary alkali silicate

glasses which in combination with other components form

the basis of the Pyroceram crystallized glass-ceramics.

In practice, the nucleation step in the preparation

of a glass-ceramic is carried out at a lower temperature

(nucleation temperature) than is the crystallization or

growth step. Experimental observation of the structural

changes on the size scale necessary to resolve nuclei in

the early stages of crystallization requires an experi-

mental technique capable of detecting and characterizing

small crystals in the glass matrix. This problem is in

some ways analagous to problems encountered in studies of

the precipitation stages in age hardening aluminum alloys.

One of the first experimental techniques employed in

aluminum alloys was DC conductivity studies which gave a

great deal of information about these systems. These

- 3 -

systems have subsequently been investigated by electron

microscopy, x-ray small angle scattering, and several

other techniques. For these reasons and others which

will become evident, electrical properties, electron

microscopy and x-ray diffraction have been used to study

the precipitation or nucleation stage in the crystalli-

zation of glasses. In order to understand and interpret

the results of the experiments described later it is

necessary to understand the conduction mechanisms oper-

ating in these glasses as well as the parameters which

affect them. The historical review which follows is an

attempt to place the present work-in proper perspective

with previous investigations.

DC Conduction

The DC behavior of glasses has been the subject

of study for over 200 years. Early workers in the area

include such eminent scientists as Franklin, Kohlrausch

(1847), Hopkinson (1876), the Curies and Maxwell (1891).

Other lesser known but important workers include

Warburg (1884), Tegetmeier (1890), Fousserau (1883) and


Warburg's early work established the presently

accepted fact that electrical conduction takes place by

the motion of the alkali metal ions in alkali-silicate

- 4 -

glasses. This was established by passing a direct current

through a test tube of "Thuringian"(CaO-MgO-Na20-K20-

A1203-Si02) glass which was filled with either mercury or

sodium amalgam which served as electrodes. Pure mercury

electrodes gave rise to polarization which caused the

current flowing to drop to less than one-thousandth of

its initial value in one hour. In the case of the sodium

amalgam electrodes the current was observed to be constant

over a protracted period of time; thus it was inferred

that the sodium from the amalgam was the charge carrier

in the glass.

Le Blanc and Kerschbaum (1910) repeated Warburg's

work and concluded, perhaps prematurely, that conduction

is due entirely to the motion of sodium ions present in

the glass. Kraus and Darby (1922) recognized that other

conduction mechanisms could also be operating, and confirmed

Faraday's law for charge transport in the glass by a

relatively simple mass transport experiment. The probable

error in the values of the transport number obtained is

too large to allow one to assert conclusively that the

conductivity was entirely ionic. More recently Kirby

(1950) has reported that the transference number of the

sodium ion is greater than 0.995. Russian workers have

also been active in this area as reported by Mazurin

(1965) but no values of the transference number are reported.

- 5 -

Hughes and Isard (1968) have recently made transference

measurements on ternary and higher order systems and

reported values very near unity.

From time to time various workers (Poole 1921,

Cohen 1957 and others) have reported work with many

glasses at high electric field strengths as well as work

in glasses containing transition metal oxides. They have

concluded that electronic conductivity is in fact

operative under either of the two conditions above. The

fact remains, and is almost universally accepted (Owen

1963, Morey 1954, and Mazurin 1965), that in alkali-

silicate glasses alkali ion motion is responsible for

electrical conduction except perhaps at high field


Thermal History Effects

Thermal history effects were first recognized by

Fousserau (1883) in the observation of a marked time

dependance of electrical conductivity during thermal

treatments near the annealing range. Extensions of that

work carried out by Fulda (1928), Mulligan, Ferguson and

Rebbeck (1925), Littleton and Morey (1933), Littleton

and Wetmore (1936), Foex (1944), Joyner and Bell (1953)

and others have shown that electrical conductivities in

glasses are extremely sensitive to thermal history. This

- 6 -

strong dependance of conductivity upon thermal history is

generally manifested as a decrease in the conductivity

with increasing thermal treatment. Concurrent with the

conductivity changes,densification is observed. This

observation gives one explanation of the conductivity

behavior. Densification of the glass structure leaves

less void space in the glass so that motion of the alkali

ions is hindered by steric constraints so that conductivity

is reduced during thermal treatments prior to the

appearance of the equilibrium crystalline phases.

Composition Effects

Compositional effects in binary alkali-silicate

glasses are generally well documented and not difficult to

rationalize. Consider first the case of increasing

alkali content in a binary glass. It is evident from

the conduction mechanism that an increased concentration

of the alkali or current carrying species should increase

the conductivity. This behavior has been observed in

the lithia, soda, potash and cesium glasses by Mazurin

(1965) and Owen (1963).

Examination of the molar equivalent composition

glasses in the series lithium, sodium, potassium and

cesium shows that the conductivity decreases in the series

in an inverse relation to the ionic radii. Once again

- 7

this is rather easily rationalized on the basis of steric


Electrode Effects

Ionic conduction gives rise to the phenomenon of

electrode polarization or time-dependent conductivity.

This has been recognized by relatively few investigators

since the early work of Warburg. Most investigations

are conducted under conditions which the investigators

apparently assume are non-polarizing.

Proctor and Sutton (1959, 1960) have examined

the polarization in an alkali-lead-silicate glass by

applying a DC potential and measuring the potential

distribution as a function of time. Their results agree

qualitatively with predicted behavior which assumes that

only one ion is mobile and it is blocked at the electrodes.

Conditions which should lead to the smallest

amount of polarization are (1) small field strength,

(2) short time of measurement or AC extrapolations and

(3) non-polarizing electrodes. The conditions above give

rise to serious problems.

The first solution, i.e., that of employing small

field strengths, introduces the problem of measuring

extremely small currents which in turn introduces obvious

experimental difficulties. Short time or AC measurements

- 8 -

require the definition of a time at which the electrode

charge build-up is negligible and thus introduces an

arbitrary variable. Non-polarizing electrodes are

difficult to prepare because of the requirement that the

activity of the charge carrying species in the glass be

equal to that of the corresponding species in the

electrode. This problem is soluble but it is not the

most difficult problem because the "electrode" with the

correct alkali ion activity must then be connected to

some electronic conductor to allow measurements to be

performed. Connection of the electrode to the measuring

circuit then introduces problems similar to the original

one. Electrode polarization in DC measurements has been

avoided in almost all the reported work by allowing only

very small currents to flow in the glass. This has given

rise to considerable experimental difficulty in measuring

the small currents. Some investigators have used AC

measurements of conductivities to circumvent the problem.

AC Behavior

AC electrical properties in the audio frequency

range have been the subject of extensive investigations

for only about 40 years. McDowell and Begeman (1929)

were the first to conduct comprehensive studies of AC

properties of glasses. They examined six glass

- 9 -

compositions including lead glasses, borosilicates and

lead borosilicates over a frequency range of 800 Hz to

1500 KHz and concluded that the dielectric losses could

be satisfactorily rationalized on the basis of the behavior

of individual "molecules and ions" rather than on the

basis of the heterogeneous theories of Maxwell (1892) and

Wagner (1914). Strutt (1931) examined five commercial

glasses including a borosilicate, a soda-lime-silicate

and a "heavy" lead glass and observed a strong correlation

between resistivity and dielectric loss which has been

noted by a great many subsequent workers. Strutt also

formulated an empirical equation (1-1) to describe the

behavior of the dielectric loss.

tan 6 = A exp aT (1-1)

The constants A and alpha in the equation depend upon the

glass composition as well as the measuring frequency.

Other workers, notably Stevels (1946, 1950) and Moore and

DeSilva (1952), have largely corroborated the validity

of Strutt's equation (1-1).

Robinson (1932) examined several glasses using

"non-polarizing electrodes" and concluded that the

observed behavior could be rationalized equally well by

either the Debye dipolar theory or the Maxwell-Wagners

heterogeneous dielectric theory.

- 10 -

In addition to the experimental work cited above,

a great deal of theoretical work was done in the period

since the observation of dielectric absorption by

Hopkinson (1876). Several books (Debye 1929, Smyth 1955

and Frohlich 1958) and innumerable papers treating the

ionic theory of dielectric loss in solids, liquids and

gasses have appeared so that discussion of these theories

is unnecessary. Most of the theories developed give

equations which are reducible in form to the equations

commonly called the Debye equations.

E- E
E =E + s (1-2)
00 1 + W2T2

S" (1-3)
1 + 2T2

t- ( E ]
tan = (1-4)
E + E 2 T2

Concurrently with the development of the ionic or

atomistic models, several workers commencing with Maxwell

(1891) have developed macroscopic or heterogeneous

theories to express the behavior of dielectric losses.

Maxwell's initial work developed the frequency dependance

of the dielectric loss for a stratified dielectric model.

This was extended by Wagner (1914) to a uniform dispersion

- 11 -

of spheres and subsequently by Sillars (1937) to spheroids.

A model for ellipsoids was developed by Fricke (1953).

The recent article by van Beek (1967) includes a summary

of equations characterizing the above systems and a great

many other types of dispersions. The equations below

from van Beek characterize a system of dispersed spheres

of conductivity 02 and dielectric constant E2 with volume

fraction V dispersed in a matrix of conductivity oa and

dielectric constant E1.

201 + 02 + 2Vv(o2 01)
S20 + 02 V (02 01) +

(2ai + 02) (62 E1) (2E1 + E2) (02 0i)
x (1-5)
|2ai + 02 V,(o2 u)12

2E1 + E2 + 2VVIE2 E)
E = E (1-6)
O 2E1 + E2 + V (E2 E1)

The relaxation time (T) of the system above is expressed


S= 2 + 2-V (E2 El)
T = O
201 + 02-V (02 o01

The analagous equations for dispersed spheroids

derived by Sillars are considerably more cumbersome than

the sphere model equations. They are given below.

- 12

oC + IA (1 vV + v (C2 1)

I1 + A

(1 V ) (02 Ci)

+ V 02

x 1 + Aa(2 2 o C a + A (E2-E1) (C2- 1

Ioi + Aa(1 VV) (2 i1 2

8 = -

T =



1 + Ia (1-V)+VI ([2-E8]

81 + A (1-VJ)(C2-E 1)

1 + Aa(1 V)(T' El)

a1 + A (1 V) ](2
a v


- 01]

For prolate spheroids (a > b)

A + n{
a ( 2 -) n12 3/2b
F bm

b 2

- li 1/2}


For oblate spheroids (a < b)

1 5
A = arc cos
a 1- (a)2 1 (a)213/2

For spheres (a = b)

A -
a 3

Because of the complexity of the above equations the

8 = 81



- 13 -

conditions of small volume fraction and oa >> 02 are

generally imposed in order to simplify the equations to

the form below.

T = o + A 2 (1-12)
A a2

S= e 1 + (1-13)

E2 E1

SV E iE + A (E2 1)

Most previous calculations have been made using the

simplified equations but the limitations imposed to derive

these equations must not be neglected.

It is apparent from the above discussion that two

methods of analyzing dielectric losses in materials are

the ionic, atomistic theory or the macroscopic, heterogene-

ous theory. In the past 10 years both types of analysis

have been applied. The work of Taylor (1957), Heroux

(1958), Isard (1962) and Prod'homme (1960) are examples

of the ionic or atomistic type analysis.

Taylor examined the AC properties as functions

of frequency and temperature in three soda-lime-silica

glasses, "commercial sheet glass," and a "Pyrex" type

- 14 -

borosilicate glass. He concluded that the dielectric

relaxation observed was a result of the motion of alkali

ions in the random glass structure. His analysis further

indicated that the distribution of relaxation times was

very similar for all the glasses examined. Prod'homme

essentially verified Taylor's results although there was

some disagreement as to the breadth of the distribution

of relaxation times. Barton (1965) has analyzed his

results in a manner similar to the workers above and has

been able to obtain qualitative results upon the depth

of the ion potential wells which indicate that the

potential well depths depend upon the "occupancy time."

To the knowledge of the author, none of the

workers who attribute dielectric losses to ion motion

have examined the effect of thermal treatments or thermal

history on either the dielectric losses or the distribution

of relaxation times.

The work of Isard (1962), Owen (1961) and Charles

(1963) is of importance because they analyzed their

results on the basis of a heterogeneous dielectric theory.

Owen concluded from analysis of the dielectric loss

behavior of CaO-B203-A1203 glasses that dielectric losses

occurred by the Maxwell-Wagner heterogeneous mechanism.

He further concluded that the borate rich phase was

distributed as a dispersed phase in an alumina rich matrix.

- 15 -

Observations were made on glasses thermally treated in the

annealing-transformation range but no interpretation was

given as to the morphological changes occurring.

Isard's work with a number of glasses has led to

the conclusion that the "classical theory of inhomogenity

can satisfactorily explain the main loss peak in glass."

However, he goes on to say that the high frequency (> 1MHz)

behavior requires an explanation based on the atomistic


Charles (1963) has examined the dielectric

behavior of a series of lithia-silica glasses with

different thermal histories and reached several conclusions

as to the effect of thermal treatments upon the morphology

of the glass. His results for different quenching

treatments were markedly different because of the

sensitivity of the metastable phase separation to

quenching rates. Analysis of his results gave infor-

mation on the morphological differences and the

connectivity of the various phases which was corrobo-

rated with replica electron microscopy.

Glass Structure

The current conception of the structure of glasses

stems largely from the x-ray diffraction studies of

Zachariasen (1932) and Warren (1937). Their work

- 16 -

indicated that in silicate glasses the average silicon-

oxygen distance is 1.62 A and the average number of

oxygens adjacent to a silicon atom is 4. These two

observations form the basis of a structural model of

tetrahedral SiO4 groups interlocking by sharing of

oxygen ions between adjacent tetrahedra. This model, or

models which closely resemble it, are generally accepted

as representative of most silicate glasses. A great

deal of literature exists in this area and is well

reviewed by Mackenzie (1960). Russian workers in the

glass structure area were first to propose a theory of a

microhetrogeneous glass structure. The early Russian

work exemplified by the work of Lebedev (1940), concludes

that the structure of glass is one of microcrystallites

with dimensions in the same size range as the interatomic

distances in crystals.

Glass structure on a larger scale (20 to 1,000 A)

is an area in which electron microscopy has led to a

detailed structure characterization. Electron

microscopy has shown evidence of liquid-liquid phase

separation in a variety of systems. The lithia-silica

and the soda-silica binaries are two systems which have

been shown to exhibit this type of behavior. The

pioneering work of Slayter (1952) and Prebus and

Michener (1954) has shown that silicate glasses contain

- 17

structural heterogenities in the size range of 20 to 200 A.

Subsequent investigations by Seward et al. (1967) and

Shaw and Uhlmann (1968) have shown that structural

features in the above size range are a general feature

of many glasses.

The work by Vogel and Byhan (1964) in lithia-

silica glasses has shown that most compositions in the

SiO2-Li20-2SiO2 phase field show structural heterogenities

whose existence and general behavior can be rationalized

by a metastable liquid-liquid separation. The existence

of the metastable miscibility gap can be inferred from

the "S" shaped liquidus on the phase diagram determined

by Kracek (1939).

Tran's (1965) work in soda-silica glasses has

shown a similar phase separation in glasses between 9

and 20 mole % soda. No evidence of separation was

observed in glasses of soda content greater than 19 mole %.

Subsequent work in both binary systems involving thermal

treatments in the annealing-transformation range has

shown the coarsening of the liquid-liquid separation

prior to the appearance of crystalline phases (Aver'yanov

and Koshits 1966).

The mechanism of liquid-liquid separation has

been investigated by various authors who have generally

agreed upon the nature of the separation. Depending upon

- 18 -

the composition of the glass, separation takes place

either by nucleation and growth or by spinodal decom-

position. Cahn (1968) has recently summarized the

thermodynamic arguments for spinodal decomposition as

well as discussed the various systems, both oxide and

metal, in which the mechanism has been reported. Cahn

and Charles (1965) have summarized the theory of phase

separation and applied their results to various glass

systems. Haller (1965), McCurrie and Douglas (1967) and

others have examined many of the systems discussed in

the above works and concluded that the observed structures

could be explained by a random nucleation and growth




Glass Preparation

The objective of this work was to examine the

effect of thermal treatment upon the electrical properties

of alkali-silicate glasses and to correlate those

properties with structural changes occurring in the

glasses. To accomplish this objective it is necessary

that the initial "state" of the glasses be well defined or

at least invariant insofar as the parameters used to

characterize the glasses could distinguish. With this in

mind an experiment to examine the possible glass prepa-

ration variables such as melting time, temperature of

melting, pouring temperature, alkali vaporization, quenching

rate, annealing time and annealing temperature was designed.

The above variables were examined by pouring a

series of 30 mole % Li2O-SiO2 glass samples as shown

schematically in Figure 1. The diagram depicts the

process of melting a glass for 24 hours at 13500C,

pouring several samples at two quenching rates followed

by annealing of these samples for periods of time from

1 to 12 hours. Next the melt temperature was increased

to 14500C and held for 24 hours and a similar set of

- 19 -

- 20 -

2 moad moia

b. 300'C

24 48 72
Time (hours)

Figure l.--Schematic melting procedure for 30 mole %
lithia-silica glass.

- 21-

samples was poured. Following this the melt temperature

was reduced to 13500C, held for 24 hours, and a similar

set of samples poured.

Samples were selected from the above melt and

electrical property measurements were carried out by

methods discussed later. Each of the possible types of

samples exhibited AC and DC electrical properties which

were identical within the limits of experimental error

with all others. From this result it is concluded that

none of the above-mentioned variables,over the range

examined, affects the initial electrical properties of the

30 mole % lithia-silica glass.

The possibility that the initial state of the

glass is different under the above-described conditions,

but is not detected by the techniques employed, was

examined by the following technique. Samples from each

class of specimens described above were thermally treated

and their electrical properties were re-examined. Once

again the electrical properties of the various types of

samples were identical even though the treatment had

changed the properties of the entire group. The property

changes of the group are discussed later. It is thus

concluded that variations of the variables set forth

above within the range examined do not affect the initial

structure of the glass.

- 22 -

X-ray fluorescence examination of the samples

melted for a total of 72 hours in a platinum crucible

showed no evidence of platinum.

In consideration of the above conclusions,

preparation of glasses for the remainder of this study

were made in the following manner.

1. Glass batches from materials of purity shown
in Table 1 and total weight of approximately
1 kilogram were weighed to an accuracy of
0.1 gram, giving an expected composition
accuracy of at least 1 part in 103.

2. The batch was then mixed in a jar mill without
balls for a minimum of 1 hour.

3. A 150 milliliter platinum crucible was then
filled from the batch and placed in an
electrically heated silicon carbide element
furnace at 13500C 50C or 14500C 50C,
depending on the glass being melted.

4. After approximately 15 minutes the crucible
was removed and refilled as necessary until
the crucible was full.

5. The crucible was then covered with a platinum
lid and the melt held at temperature for a
minimum of 24 hours.

6. After 24 hours the lid was removed from the
crucible and the crucible was replaced in the
furnace to allow it to come back to 13500C
prior to glass pouring.

7. The glass was then poured in a 17.5 mm diameter
steel mold. A tightly fitted plunger was
pressed into the molten glass, giving a sample
thickness between 3 mm and 8 mm and a diameter
of 17.5 mm.

- 23 -



Compound Weight %

Sodium Carbonate'
Na2C03 99.8
Chloride .0005
Nitrogen .0005
Phosphate .0005
Sulfate .001
Arsenic .0001
Calcium and Magnesium .005
Iron .0002
Potassium .001
Silica .005
Heavy Metals .0002

Lithium Carbonate2
Li2Cos 99.3
Na2Co3 0.2
Iron .0008
Sulfate 0.3
Chloride .0003
Calcium .0003
Phosphate .0001

Silicon Dioxide3
Si02 99.91
Iron .019
Alumina .08
Titania .009
Calcium and Magnesium Trace

IBaker Reagent.

2Foote Mineral Company.

3Pennsylvania Glass Sand Company.

- 24 -

8. The resulting glass button was quickly removed
(after approximately 30 seconds for cooling)
and placed in a furnace at 3000C 50C where
it was held for 1 hour and air cooled.

9. After each sample was poured, the remaining
glass was placed in the furnace and allowed
to come back to the melting temperature, while
the steel mold was chilled in tap water to
keep it at the same temperature for each

It was found that 1 hour at 3000C was the minimum

treatment which allowed the samples to be cooled to room

temperature without breakage. The residual stresses

remaining after this treatment were measured by standard

birefringence techniques and found to be 3,000 to 5,000 psi.


For electrical measurements, the faces of the

samples were ground parallel to within 0.05 mm with

silicon carbide metallographic paper and polished with

600 grit metallographic paper. This surface was then

cleaned with distilled water and quickly dried. In order

to vapor deposit the desired double guard ring gold

electrodes on the faces of the samples, paper masks were

affixed to the sample faces. These paper masks with an

inner diameter of 12.7 mm and an outer diameter of 15.9 mm

were cut from a gummed paper label with a modified

machinist's compass. The resulting masks were moistened

- 25 -

with distilled water and carefully affixed to the samples

so that they were concentric with the sample itself. The

samples were then placed in a specially designed holder in

a standard vacuum metallizer and the vacuum system was

pumped down to less than 1 micron. The samples were then

plated with gold from a tungsten basket placed about

100 mm from the sample. Total electrode thickness was

approximately 2 10-6 mm as determined from weighing

samples before and after plating. The samples were then

removed from the evaporator and placed in a furnace at

3000C 50C for 1 hour to increase the adherence of the

gold to the sample. Samples were removed from the

furnace, air cooled and the charred paper masks were

carefully removed, leaving the sample with gold electrodes

in a double guard ring configuration.


Samples for examination in the Guinier (1956) and

DeWolff (1947) x-ray camera were selected from the samples

poured for electrical measurements in order to insure that

their thermal history was identical to that for the

electrical samples. These samples were heat treated,

broken with a hammer and anvil, then ground in an alumina

mortar and pestle to pass a Tyler 200 mesh screen. The

ground glass-crystals were stored in a closed container

to prevent moisture pickup and resulting hydration.

- 26 -

The Nonius Guinier-DeWolff x-ray camera and copper

x-ray tube used for these experiments are shown schematically

in Figure 2. The Guinier-DeWolff camera, a vacuum path,

focusing quartz crystal monochromated powder camera, is

capable of examining 4 samples simultaneously. The samples

were held in 4 slots in a flat sample holder approximately

0.025 mm thick with Scotch #810 tape. The powder pattern

was recorded on Kodak Type NS double emulsion x-ray film,

developed 8 minutes at 230C in Kodak Type D-76 developer

and fixed 3 minutes at 230C in Kodak "Rapid Fix." It was

found that this technique gave the highest line intensity

with a tolerable fog level for a given exposure. Using

the above techniques, it was found that 0.1 weight %

lithium metasilicate crystal in a prepared lithium

disilicate glass standard could be detected. Detection

of crystals with this small weight fraction required a

50 hour exposure at 40 kilovolts and 20 milliamperes. The

resulting patterns were analyzed using a reader which

allowed the crystal lattice spacings to be read directly

from the film.

Thermal Expansion and Softening Points

Samples for thermal expansion measurements were

prepared by melting the glass in the same manner as for

the electrical samples. The dilatometric samples were

- 27 -



X-ray /
Source Sample

Figure 2.--Schematic diagram of x-ray source and
Guinier-DeWolff camera.

- 28 -

poured in a steel mold which tapered from a diameter of

12.7 mm to 15.9 mm in its five centimeter length. This

taper was necessary to allow rapid removal of the sample

from the mold to prevent breakage. Following pouring,

the samples were held in a furnace at 300C for 1 hour

to prevent breakage, then removed and air cooled. Samples

prepared in this manner were too highly strained to be

cut with a cutoff wheel so they were all heat treated

1/2 hour at their subsequent heat treatment temperature

to further remove strains. After this treatment the

samples were cut to slightly over 50.8 mm length with a

water cooled silicon carbide wheel and the ends were

polished to 50.80 mm .02 mm on 180 silicon carbide

metallographic paper.

The thermal expansions of the various thermally

treated glasses were measured in an Orton recording

dilatometer which is a quartz tube and push rod apparatus

with a linear variable differential transformer transducer

for measuring the expansion. Samples were separated from

the quartz tube and push rod with 0.025 mm platinum foil

to prevent reaction with the quartz. The apparatus was

set up to plot the thermal expansion curve as a continuous

function of the temperature.

- 29 -

Electron Microscopy

Replica Preparation

Samples to be examined by electron microscopy were

prepared in the same manner as the electrical samples and

heat treated in the bulk state. After thermal treatments,

the samples were fractured to expose a fresh surface from

the interior of the sample. The fracture surface thus

obtained was etched 1 minute in an aqueous 5 volume %

hydrofluoric acid solution. The samples were then placed

in an evaporator and a platinum preshadow was applied,

followed by a carbon film. The carbon film replica was

removed from the glass by immersing the sample in an

aqueous 2% hydrofluoric acid solution and allowing the

replica to float off the sample. This replica was placed

on a copper microscope grid and washed several times in

distilled water to remove the acid. Replicas were

examined in a Phillips EM 200 electron microscope using

standard techniques.

Transmission Preparation

Samples for examination by transmission electron

microscopy were poured and heat treated in the same manner

as the electrical samples. Samples were mechanically

thinned to approximately 0.2 mm prior to chemical thinning.

The mechanical thinning was accomplished by grinding and

- 30 -

polishing one face of the poured button, cementing the

polished face to a small flat piece of steel and grinding

to the final thickness. Samples of 3 mm diameter were

then cut from the thinned material while it was still

fixed on the steel block. The samples were cut by using

a hollow copper drill (3 mm ID) with a slurry of 400 grit

silicon carbide in a drill press. The resulting sample

blanks were removed from the steel block and the cement

was removed using ethylene dicloride in an ultrasonic


The chemical thinning was carried out by dimpling

the sample in the center, followed bya final thinning

operation. The dimpling operation was accomplished by

masking the outer edges of the samples with a lacquer

("Microstop") which did not allow the edges of the sample

to be attacked. The solution used for this operation was

made up of 10 parts hydrofluoric acid (48%), 5 parts

nitric acid and 14 parts acetic acid by volume. After the

dimpling operation, the masking material was removed in

acetone and the sample was carefully washed. At this

point the sample has a relatively thick edge with a thin

interior region which greatly facilitated handling. Final

thinning was accomplished by alternately dipping the

sample in hydrofluoric acid for short periods of time

(30 to 60 seconds) and examining the center portion for holes.

- 31 -

When the first visible hole developed, etching was stopped

and the sample was washed several times in distilled water.

The sample was then placed in an evaporator and a thin

film of carbon was evaporated on one surface to prevent

charging of the sample by the electron beam. The sample

was then placed in the heating stage holder (Phillips

PW 6560) with a large platinum aperture to facilitate

heat conduction.

The sample holder was then placed in the Phillips

EM-200 electron microscope with a rotating-tilting stage

and heated to the required observation temperature. The

remainder of the electron microscopy was then conducted

by standard techniques.

Electrical Measurements

Sample Chamber

DC and AC properties were measured over a range of

temperatures in a vacuum environment. The requirements

of electrical shielding in AC measurements and guarding in

DC measurements was of major importance in the sample

holder design. The measurements require that the leads

and contacts be made in a coaxial configuration. The

sample holder design shown in Figure 3 used in these

experiments utilizes a coaxial arrangement of leads as

far as possible. Where this is not possible, high quality

- 32 -



Figure 3.--Photograph of sample holder
employed in electrical measurements.

- 33 -

insulation with resistivity greater than 1 1014 ohm cm

at room temperature was used. Electrical contacts to the

samples and leads out of the sample chamber were platinum

to avoid oxidation and thermoelectric problems. The sample

chamber was vacuum sealed to allow all electrical

measurements to be conducted in a vacuum of less than

1 micron. The sample chamber and sample were heated by

means of an external nichrcme heating element in con-

junction with a Variac for temperature control. The

sample temperature, as measured with a chromel-alumel

thermocouple placed approximately 1 mm from the sample,

was constant over a series of electrical measurements to

10C of the set point

AC Measurements

Discussion of the electrical equipment is sub-

divided into the audio frequency (AF) equipment, the radio

frequency (RF) equipment and the null detection system.

The AF equipment, shown schematically in Figure 4, consisted

of a Wayne-Kerr B-221 transformer radio arm bridge in

conjunction with a Hewlett-Packard 651-A oscillator. A

schematic diagram of the bridge circuit is shown in

Figure 5 where ZS and Z are the standard and unknown

impedances respectively. The balance condition is

satisfied, as indicated by a null on the detector, when

- 34 -





Figure 4.--Schematic diagram of audio frequency
measurement equipment.

- 35 -


Figure 5.--Schematic diagram of transformer ratio
arm bridge.

- 36 -

equal currents flow in each half of the center tapped

transformer (T2). When this condition is satisfied, the

potential on the primary will be zero and the right hand

terminals of the unknown and standard will be at neutral

potential. The same voltage is applied to both unknown

and standard and for equal currents to flow in each half

of the primary of transformer T2; the real and imaginary

parts of the unknown impedance must be equal to those of

the standard. The instrument is designed to allow values

of the resistive and capacitative component to be read

directly from the instrument dials.

The RF measuring equipment, shown in Figure 6,

consisted of a Wayne Kerr B-601 bridge and the same

oscillator as was used with the AF equipment. The bridge

design is, in theory, similar to the AF bridge except that

the transformers and standards used are designed for use

in the RF range.

The null detection system, shown in Figure 7,

consisted of a General Radio 1232-A null detector used in

conjunction with a General Radio 1232-Pl RF mixer and a

Wayne Kerr 0-22-D beat frequency oscillator. The 1232-A

covers the frequency range 20 Hz to 100 KHz directly, so

that the AF and RF null signals up to 100 KHz were

detected directly. At frequencies greater than 100 KHz,

it was necessary to use a beat frequency technique to

- 37 -





Figure 6.--Schematic diagram of radio frequency
measurement equipment.

- 38 -




Figure 7.--Schematic diagram of null detection system.

- 39 -

reduce the signal frequency to the range of the 1232-A

detector. This was accomplished by mixing a local

oscillator signal of frequency 100 KHz greater than the

measurement frequency with the signal to be detected.

This gives a beat frequency signal proportional to the

original signal at a frequency of 100 KHz which can be

detected by the 1232-A detector.

DC Measurements

DC measurements were made with the equipment shown

schematically in Figure 8. The short time measurements

were conducted by displaying a signal proportional to the

current flowing in the sample on a Hewlett-Packard 140 A

oscilloscope with a 1420 A time base and 1402 A dual trace

amplifier and photographing the oscilloscope trace with

a Tektronix C-12 camera. This technique allowed the

sample conductivity to be measured in a time of less than

5 milliseconds in most cases. The minimum time depended

upon the range of the Keithly 416 high speed picoammeter

and no measurements were made in times less than the

response time of the picoammeter.

The DC potential for the measurements was furnished

by a Hewlett-Packard 6217 power supply which has a voltage

stability of less than 0.10% + 5 millivolts in 8 hours

with less than 200 microvolt AC ripple. The connections

- 40 -



r ----------------- I
HP-412-A I

Figure 8.--Schematic diagram of DC measurement

41 -

to the sample and picoammeter were made with guarded

leads as indicated in Figure 8. Sample currents greater

than 3 10-5 amps were measured on a Hewlett-Packard

412-A vacuum tube voltmeter and because of the large

time constant of this instrument, no measurements could

be made in times less than 1 second.



The objective of the polarization experiments was

to establish the behavior of the sample under a DC field

in order that the electrode polarization could be elimi-

nated as a cause of relaxation phenomena in the AC

measurements. This objective requires that the time

required to polarize the sample electrodes be greater

than the maximum period (minimum frequency) used in the

AC measurements.

Polarization results from a 30 mole % lithia glass

in the as cast state are presented in Figure 9. These

results are a combination of DC and AC conductivity

measurements. The "time" for the AC measurements was

taken as the reciprocal of the measurement frequency. This

definition of the time parameter results in a reasonably

good fit in the region of overlap of the two curves and is

thus considered to be a valid definition. The agreement

of the two curves is in fact remarkably good considering

the fact that in the overlapping region both measurement

techniques are approaching their time limits and the

accuracy is subject to larger errors in that time domain.

- 42 -

- 43 -

Time ISecondsl

Figure 9.--DC polarization curve of 30 mole % lithia-
silica glass in as cast and heat treated 5 hours at
5000C forms.

- 44 -

The general shape of the curves for the as cast

glass, as expected, shows increasing conductivity in the

short time (10-4 to 10-5 second) range, a flat intermediate

region (10-s to 10-1 second) and a region of electrode

polarization for times in the 10-1 to 10+3 second range.

The curve corresponding to the sample heat treated

5 hours at 5000C exhibits more structure than the as cast

glass. The short time and initial flat region is quite

similar to the as cast behavior up to about 2 10-3

second. At that point a dispersion appears and at longer

times the flat tries to reappear but is masked by the

appearance of electrode polarization. Discussion of the

dispersion in the heat treated sample will be presented

in the AC properties section for reasons that will be

evident later.

The appearance of the polarization problem is

illustrated by the series of conductivity-time plots in

Figure 10. It is evident that at low temperatures, when

the conductivity is low, the problem is unimportant but

at high temperatures the problem is very serious. The

effect of the polarization upon the measured DC conductivity

is shown in a more conventional form, the log conductivity

versus reciprocal temperature plot, for two arbitrarily

chosen times of measurement (approximately 1 second and

60 seconds) as shown in Figure 11. The problem of

- 45 -

TIME (Seconds)

Figure 10.--Conductivity versus log time
curve for 33 mole % lithia-silica glass
as cast.

- 46 -

lol T IC)

Figure ll.--Log DC conductivity versus
reciprocal temperature for 33 mole %
lithia-silica glass as cast.

- 47 -

polarization is less important in the low temperature

region but at higher temperatures the problem in its

most serious state causes an inversion in the slope of

the curve. It is thus evident that DC measurements made

without consideration of the problem of polarization are

virtually meaningless if they are measured over a

temperature-time range where the phenomeron shown above

is significant.

In order to develop a working criterion for

determining the effect of polarization on conductivity

measurements, a model consisting of a plane parallel

capacitor with mobile positive charges in the dielectric

and totally blocking electrodes was chosen. This model

is approximately the situation in an alkali silicate

glass capacitor. With a potential applied, this capacitor

(Figure 12) will have a surface charge equal to the

negative of the volume charge assumed to be uniformly

dispersed in the dielectric. It is assumed that total

polarization occurs when the back voltage due to the

surface charge is equal to the applied voltage and the

apparent conductivity is zero.

To allow comparison of the surface charge calcu-

lated from the model above with an experimentally derived

value, it was necessary to define electrode polarization

on the basis of the conductivity time curves. The criterion

- 48 -


Figure 12.--Schematic diagram of model used for
polarization calculation.

- 49 -

chosen was to assume that a conductivity drop of one full

order of magnitude represented total polarization of the

sample. The charge transported during the time required

for polarization to occur will be calculated from the

experimental results and compared with that predicted from

the model.

The electric field in the capacitor due to the

uniform surface charge on one electrode is given by

equation (3-1) from Page and Adams (1958).

E P (3-1)

The field due to the uniform negative charge of equal

magnitude but opposite sign to the surface charge is

given by equation (3-2) (Page and Adams 1958).

P x 2)
E = (3-2)

In order to obtain the potential between the two plates,

the two fields above are added and integrated over the

electrode spacing. As a result of symmetry, the uniform

volume charge does not contribute to the potential thus

the integration of (3-1) gives equation (3-3).

pt (3-3)
E: E:

- 50 -

Equation (3-3) gives the magnitude of the back potential

corresponding to a given polarization charge on or adjacent

to the cathodic electrode. It is thus possible to calcu-

late from equation (3-3) the magnitude of the surface

charge from the applied potential, sample thickness and

the dielectric constant obtained from AC measurements.

Substitution of the above values from the sample used to

obtain the polarization curves in Figure 10 yields a

surface charge of 9.80 10-4 coulombs/cm2.

In order to obtain an experimental value of the

surface charge, it is assumed that all the charge

transported is left on the electrodes. The charge

transported is the time integral of the current over the

time of polarization. Because the analytical form of the

current time behavior is not known, the integral above is

approximated by assuming that the current drops from its

initial value to zero in the time interval considered.

This assumption can be justified by reference to

Figure 10. The conductivity-time and current-time curves

have the same shape because of Ohm's law; hence Figure 10

can be looked upon as a current-time plot. The value of

the current transported during the polarization depicted

in the 1800C curve of Figure 10 calculated in the manner

described above yields a surface charge of 1.33 10-3


- 51 -

Comparing this value with the value calculated

from the model above, it is evident that the agreement is

good considering the approximations made. There are, at

least, two possible reasons for the differences. The

electrodes have been assumed to be totally blocking to

the charge carrier but it is possible that this is not

the case and part of the charge transported is not on the

surface of the dielectric. Another possible source of

the difference is the approximation of the shape of the

current-time curve. The approximation of the curve shape

in Figure 10 yields a value of the current which is too

high. These two factors are sufficient to explain the

observed difference and the experimental measurement.

The results above indicate that the problem of

polarization can be approached by two methods. The first

and most direct consists of measuring the polarization

behavior of each sample and analyzing the conductivities

directly from the polarization results. The second and

faster method is to calculate the conductivity required

to cause polarization in the measurement time used. This

would allow prediction of cases in which polarization

difficulties will arise. Clearly both techniques have some

inherent difficulties but either one or both methods should

be used to eliminate or reveal polarization problems in

DC measurements.

52 -

The DC measurements reported herein were made by

measuring representative polarization curves and taking

the conductivity values from the time independent portions

of those curves.




The objective of the x-ray studies was to detect

and identify the crystals appearing in the initial stages

of crystallization. The initial x-ray studies were

conducted to determine the lower limits of detection of

crystalline phases in a glassy matrix. Examination of a

series of lithium disilicate glasses mixed with 5.0, 1.0,

0.5 and 0.1 weight % crystalline lithium metasilicate

showed that the crystals in a 0.1 weight % standard could

be observed. A similar experiment with crystalline

lithium disilicate in lithium disilicate glass gave the

same lower limit of detection.

Results of the x-ray examination of the 30 mole %

lithia-silica glass following various heat treatments at

500C are presented in Table 2. The actual patterns are

not presented because the lines of interest are very weak

even after 50 hour x-ray exposures and photographic

reproduction is difficult. The phases present change with

thermal treatment at 5000C from the glassy material in

the as cast form, to a glassy material plus a phase

tentatively identified as crystalline lithium metasilicate

- 53 -

- 54




30 mole % Li2O 30 mole % Li2O
Standard Standard Heat Treated Heat Treated
Li20O2SiO2 Li20OSiO2 5 hrs at 5000C 50 hrs at 5000C

1 (A) Intensity d (A) Intensity d (A) Intensity d (A) Intensity








1 (M.S.)



1 (M.S.)




















LA L .1









- 55 -

with a weight fraction of approximately 0.1 weight %.

This 5 hour x-ray pattern shows seven lines, three of

which correspond to the stronger lithium metasilicate

crystal lines. The other four lines are weaker and, at

present, not conclusively identified. It appears that

the unidentified lines correspond to a transition phase

of lower symmetry than the orthorhombic metasilicate.

Indexing is difficult, if not impossible, because of the

small number of lines observed. Examination of the 10

and 20 hour samples show only a diffuse peak character-

istic of the glassy state with no evidence of crystalline

phases. The 50 hour sample exhibits the four strongest lines

corresponding to the equilibrium lithium disilicate crystal.

X-ray results for the 33 mole % lithia (lithium

disilicate) glass are shown in Table 3. The reaction

sequence observed in this glass is that of the glassy

material transforming to the equilibrium lithium

disilicate crystal with thermal treatment. The equilibrium

lithium disilicate precipitate is observed after 50 hours

at 5000C.

X-ray examination of the 26.4 mole % lithia glass

did not reveal the presence of any crystalline phases with

thermal treatments up to 50 hours at 5000C.

- 56 -



Standard Crystal 33 mole % Li20 33 mole % Li20
Li20-2SiO2 50 hrs at 5000C 100 hrs at 5000C

d (A) Intensity i d (A) Intensity d (A) Intensity

7.4 2 7.4 2

5.45 10 5.45 1 5.45 3

4.18 1

3.75 10 3.75 2 3.75 10

3.65 10 I 3.65 2 3.65 10

3.58 10 3.58 2 3.58 10

2.95 1

2.90 5 2.90 3

2.39 5 2.39 3

2.35 5 2.35 1 2.35 3

2.27 2

- 57 -

The results of x-ray examination of the 33 mole %

soda (sodium disilicate) glass are presented in Table 4.

These results show the appearance of the equilibrium

sodium disilicate crystal after 10 hours at 5500C.

Further treatments at this temperature only increase the

amount of crystalline sodium disilicate present in the


The x-ray results of the 25 mole % soda glass do

not show evidence of crystallization with treatments of

50 hours at 5500C.

Electron Microscopy

The objective of the electron microscopy was to

identify and determine the morphology of the phases

appearing during the initial stages of crystallization.

Replica Techniques

Replica electron micrographs of the 30 mole %

lithia glass are presented in Figures 13 to 17. The

micrograph of the as cast glass (Figure 13) exhibits a

droplike structure with separated regions in the size

range 0.1 to 0.5 micron.

It is known (Vogel and Byhan 1964) that silica

rich glasses are attacked by hydrofluoric acid much less

rapidly than lithia rich silicate glasses. From this

information and a knowledge of the shadowing direction,

it is evident that the drop regions in Figure 13 are richer

in silicon than the surrounding matrix.

- 58



Standard Crystal 33 mole % Na20 33 mole % Na20
Na20-2SiO2 50 hrs at 5000C 10 hrs at 550C

d (A) Intensity d (A) Intensity d (A) Intensity












- 59 -

Figure 13.--Replica electron micrograph of 30 mole %
lithia-silica glass (30,000x) as cast.

- 60 -

**~;P M

Figure 14.--Replica electron micrograph of 30 mole %
lithia-silica glass (26,000x) after 5 hours at

- 61 -

Figure 15.--Replica electron micrograph of 30 mole %
lithia-silica glass (31.,000x) after 10 hours at

- 62 -

Figure 16.--Replica electron micrograph of 30 mole %
lithia-silica glass (31,000x) after 20 hours at

- 63 -

Figure 17.--Replica electron micrograph of 30 mole %
lithia-silica glass (37,000x) after 50 hours at

- 64 -

The micrograph of the 30 mole % sample heat

treated 5 hours at 5000C is shown in Figure 14. This

micrograph exhibits the separated drop regions character-

istic of the as cast sample and further shows a secondary

drop-like separation in the matrix surrounding the primary

drops. This secondary separation in the matrix is the

separation of the lithium metasilicate in the matrix.

Figure 15 corresponds to the 30 mole % glass heat

treated at 5000C for 10 hours. This micrograph shows

some evidence of the' initial droplets but they are not as

pronounced as in the earlier thermal treatments. The

matrix exhibits a fine scale separated drop morphology

similar to that in the 5 hour treatment. Thus, the drop

regions are beginning to dissolve with thermal treatment.

The micrograph of the 30 mole % glass heat treated

20 hours at 5000C is shown in Figure 16. This micrograph

exhibits an overall uniformity of structure with only

traces of the primary drop separation.

The micrograph of a sample heat treated 50 hours

at 5000C is shown in Figure 17. This micrograph shows

remnants of the structure present at 5 hours but it is

generally more homogeneous than all the other glasses

examined in the above series. The raised regions are

the equilibrium lithium disilicate precipitate.

- 65 -

Transmission Electron Microscopy

Diffraction Patterns

The electron diffraction patterns taken during a

thermal treatment at 4800C are presented in Figures 18

to 23. The patterns initially show no crystallinity

followed by the appearance of lithium metasilicate and

lithium disilicate crystals. The interplanar spacings

obtained from the diffraction patterns and tabulated in

Table 5 indicate that during the early stages of

crystallization the crystalline lithium metasilicate

predominates, but in the latter stages the lithium

disilicate predominates.

The lattice spacings taken from the ASTM card file

(Smith 1967) are those corresponding to room temperature,

while those discussed above correspond to 4800C. The

results of Glaser (1967) with high temperature x-ray

diffraction on the lithium disilicate indicate that the

expected expansion in the {130} lattice spacing for example

is less than one percent. This is within the expected

experimental error in the electron diffraction measurements

so the comparison with room temperature lattice spacings

is a valid one.

Transmission Electron Micrographs

Figure 24 is a micrograph of the 30 mole % glass

before thermal treatment. This micrograph exhibits a

network structure which is undoubtedly a result of the

- 66 -

Figure 18.--Electron diffraction pattern of as cast
30 mole % lithia-silica glass at room temperature.

- 67 -

Figure 19.--Electron diffraction pattern of 30 mole %
lithia-silica glass at 4800C after 5 minutes at

- 68 -

Figure 20.--Electron diffraction pattern of 30 mole %
lithia-silica glass at 480C after 14 minutes at

- 69 -

Figure 21.--Electron diffraction pattern of 30 mole %
lithia-silica glass at 4800C after 26 minutes at

- 70 -

Figure 22.--Electron diffraction pattern of 30 mole %
lithia-silica glass at 4800C after 2.5 hours at 4800C.

F-_.ure 23.--Electron diffraction pattern of 30 mole %
lithia-silica glass at 480QC after 3.75 hours at 4800C.

- 71 -

U) U) U)

o o


I "i

Ir 0 --
0! 0^
/! 03i-

U C Cf~Cf

















< ^

r C r-C H co
o Ln co
co H r- n m H -

(N ir4 r4 -1 i)


ON O N ON C N 'D C Ln C) Co D N H
m om- C Cm N Nr in M H mr- M U in ID

in N N : mN N H H H Hi- +

> > > 4p

S--I N Ln D -i -I L
Cn o N <(D C) n'

N Ln ra')n rn m N N,4 N^

en n
(N C

(M r


r-i (N

+r c

Ln co
r~- r



O o



U~C) C) ) ) ) C

t t

0 Ci

CM )

H ii
cr f
xa 0 (

- 72 -

u~ i~ u~ ci u c~ i~ i I I
aaa~aaI I Il

C -

1 4
" y

;' (

on H on
-D 0 0
H 0O
*H *




4-1 Q

00 k.N

0 H O
CD OrI m

l m C


(2 C" H








- --

2S S22 S3 :S >

m m m m m U1 m m
Q a Q Q a z z

- 73 -



Figure 24.--Transmission electron micrograph 30 mole %
lithia-silica glass at room temperature as cast

- 74 -

carbon deposited on the glass surface. The carbon layer

partially masks the glass structure but the diffraction

pattern leaves no doubt that some of the structure present

is crystalline metasilicate. The carbon surface film is

discussed further in the following section. Figure 25 is

a micrograph taken at 480C after 11 minutes at 480C. It

shows the appearance of black spots which are apparently

crystals in a glassy matrix. The diffraction pattern at

this point in time shows that the crystals present are

primarily lithium metasilicate.

Bright and dark field micrographs taken after 35

minutes at 480C are presented in Figure 26. The dark

field micrograph was taken from the {170} diffraction ring

of the lithium disilicate so that the bright crystals in

the dark field image are lithium disilicate crystals.

Micrographs taken after 2.3 hours at 4800C are shown in

Figure 27. The dark field image corresponds to the {170}

diffraction ring in the lithium disilicate pattern hence

all the bright areas in the dark field micrograph are

lithium disilicate crystals. The disilicate crystal size

has increased in the time elapsed between Figure 26b and

Figure 27b, indicating the growth of the lithium disilicate

crystals in the glassy matrix. The last micrograph in

this series at 480C, shown in Figure 28, was taken after

4.8 hours. This micrograph shows the development of an

elongated morphology from the previously equiaxed morphology.

- 75 -

Figure 25.--Transmission electron micrograph of
30 mole % lithia-silica glass at 4800C after 11
minutes at 4800C (42,000x).

- 76 -


Figure 26.--Transmissicn electron micrograph of
30 mole % lithia-silica glass at 4800C after 35
minutes at 480C. (a) Bright field (b) Dark
field (46,000x).

,ye ... .', .

t, %1'?^ .- *o
.I. '-- .. *';4
+<.'^ <^*f.
*.; -%; A ?d M

- 77 -

Figure 27.--Transmission electron micrograph of
30 mole % lithia-silica glass at 4800C after 2.3
hours at 4800C. (a) Bright field (b) Dark field

- 78 -

Figure 28.--Transmission electron micrograph of
30 mole % lithia-silica glass at 4800C after 4.8
hours at 480C (23,000x).

- 79 -

The effect of the carbon film on the surface of

these glasses was investigated by evaporating a carbon

film similar to that used on the glasses on a thin mica

crystal and examining the evolution of the structure with

thermal treatment. The carbon film on mica exhibited an

initial structure similar to that shown in Figure 24.

Heating of the mica-carbon film to 480C and following

the microstructural changes revealed a lower volume

fraction of essentially the same black spots character-

istic of the glass micrographs in Figures 25 and 26a.

It is thus concluded that some of the black spots must

be carbon.

It was not possible to obtain a dark field image

from the weak and diffuse lithium metasilicate rings

during the sequence. This difficulty indicates that the

observed black spots may be crystalline carbon but no

carbon diffraction rings were observed. This leaves the

identity of the black spots open to question but it

appears that some of the black spots must be lithium

metasilicate. It is possible that the carbon masks the

lithium metasilicate crystals and hence they are not

observed except in the diffraction pattern.

Figure 29a and 29b are micrographs taken at 500C

after 1 hour at 5000C. The dark field image is taken

from the area circled on the diffraction pattern in

- 80 -


Figure 29.--Transmission electron micrograph of
30 mole % lithia-silica glass at 5000C after 1
hour at 5000C. (a) Bright field (b) Dark field
(31,000x) .

- 81 -

Figure 30. The important feature of these micrographs is

the extremely long crystals growing in the glass matrix.

The growth of the whiskers out of the glass matrix is

also an interesting feature.

A room temperature transmission micrograph of a

glass heat treated in the bulk form for 5 hours at 5000C

is shown in Figure 31. There is a pronounced similarity

between this micrograph and the replica micrograph of a

similar sample shown in Figure 14.

AC Results

The objective of the AC measurements was to monitor

the initial stages of crystallization and characterize the

structure and structural changes occurring during this


30 Mole % Lithia-Silica Glasses

The results of the AC measurements are presented

in the form of the AC loss angle (tan 6AC) as a function

of frequency and temperature. The tan 6AC values were

calculated from equation (4-1) following the method of

Charles (1963).

GTotal 1 1 1
tan 6 a- = tan 6AC + tan 6 +


- 82 -

Figure 30.--Electron diffraction pattern of 30 mole %
lithia-silica glass at 500C after 1 hour at 5000C.

- 83 -

Figure 31.--Transmission electron micrograph of
30 mole % lithia-silica glass at room temperature
after 5 hours at 500C in bulk form (31,000x)

- 84 -

The tangent of the loss angle is shown as a function of log

frequency for the various heat treatments at 5000C in

Figures 32-36. It can be noted from Figure 32 that the

as cast glass is free of loss peaks, but after a 5 1/2

hour heat treatment (Figure 33) at 5000C, large loss peaks

have appeared. The tolerance on the time of appearance for

the loss peaks was established by heat treating a sample in

1/2 hour increments and measuring the AC properties following

each heat treatment. Further heat treatments at the same

temperature for times up to 50 hours (Figures 34, 35 and

36) cause the magnitude of tan 6AC to decrease and the

peak location to shift to higher frequencies.

The temperature dependence of the frequency maxima

of the tan 6AC curves is shown in Figure 37. The frequency

maxima curves have the Arrhenius form typical of a thermally

activated process. The activation energy of the frequency

maxima-reciprocal temperature curves is 14.7 kilocalories

mole in all cases shown in Figure 37. The equivalence of

the loss process activation energy and the DC conductivity

activation energy indicates that the loss process is the

result of ionic motion. The loss behavior results are

summarized in Figure 38.

33 Mole Z Lithia-Silica Glasses

Results for the 33 mole % lithia glass are summarized

in Figure 39. The loss spectrum of this glass shows the

appearance of the loss peak between 2 and 5 hours heat

- 85 -



122 --o- -
02 -o--- I

2i i

24 I

s V

4 1
\l I.

_______ I 0" -.-.. .. .. ... -.-- ....-
2 1 4 6 e


Figure 32.--Tan 6 versus logo frequency for the
30 mole % lithia-silica glass as cast.

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - Version 2.9.7 - mvs