Design Criteria f o r Thermal Shock

November 1963
Table IV.

535

Constitution of NaP03-NaF Glasses with N a / P = 1.4, (6) = 5,to Na/P = 1.22, (n)* = 9 (Distribution of Total
Phosphorus in Weight Percent)
Na/P = 1.4
(El* = 5

-_-_-- -

Time of melting
(minutes)

N a / P = 1.33
(n)* = 6

Sodium

OMT*

Ortho$
Pyro
Tri
Tetra
Penta
Hexa

Septa
High fractions and
hypob

15

N a / P = 1.28
(?
=I
7 )*

__

_
p
i
7--

Sodium

N a / P = 1.22
(Z)* = 9

,---*_--_

Sodium

Sodium

glass?

OMT*

15

glass?

OMT*

15

glass?

OMT*

15

glass?

18.6
8.9
9.0
11.3
9.9
8.7
7.0

1.4
3.9
16.0
21.0
16.8
12.0
9.1

1.1
3.9
15.6
21.6
17.4
12.3
9.4

13.1
7.0
7.5
10.6
8.7
8.0
7.1

0.4
2.1
9.7
15.4
13.8
11.7
9.4

0.8
2.1
10.4
16.0
14.5
11.6
9.8

12.3
6.8
6.7
9.3
8.3
8.0
7.8

0.6
1.8
6.3
11.3
10.8
8.7
8.1

0.6
1.9
8.0
12.9
12.4
10.8

6.6
3.9
44
7.7
6.7
6.6
6.2

0.6
1.3
4.2
8.6
7.8
6.9
5.7

1.3
1.4
5.0
8.7
8.4
8.3
8.3

26.6

19.8

18.7

38.0

37.5

34.8

40.8

52.4

53.4

57.9

64.9

58.6

14.3
16.1

14.3
17.5

11.1
20.6

11.1
14.1

11.1
14.6

No. of molecules per 100 phosphorus atoms

Theoretical (loo/(%)*)

20

20

20

16.7

16.7

16.7

14.3

Experimental5
36.2
23.1
22.7
29.8
18.4
19.2
28.5
* OMT = zero melting time (see text).
1Sodium phosphate glass with ii = 4 (after Westman and Gartaganis,footnote 3(b)).
$ Orthophosphate monofluoro-orthophosphate.
5 Obtained by estimating the chain length distribution in high fractions and hypoly.

to the present data on the loss of fluorine at these temperatures

their melts should not contain any fluorine; indeed, some of
their experiments on the loss of fluorine from melts showed
that although 4 wt% of NaF was added to the composition,
only 0.01% fluorine remained in the glass. Hence, the
(c) is
structure breakdown observed by Williams et aZ.2(b)s

not due to the presence of fluorine but to the introduction of

additional Nal+ ions into the melt and the consequent increase in Na/P ratio.
Acknowledgment

The writers wish to acknowledge the many helpful discussions

with A. E. R. Westman and his continued interest.

Elastic Energy at Fracture and Surface Energy as

Design Criteria for Thermal Shock
by 0. P. H. HASSELMAN
Applied Research Branch, Research and Development Division, The Carborundum Company, Niagara Falls,

The physical properties which affect the propagation of cracks in specimens fractured by
thermal shock are discussed. The driving force
for crack propagation is provided by the elastic
energy stored at fracture. The mechanism of
energy dissipation which will tend to arrest the
propagating cracks is provided by the effective
surface energy required to produce the newly
formed crack surfaces. An expression is derived
applicable to a body of spherical shape for the
mean area traversed by cracks nucleated by
thermal shock. Three numerical examples are
given for materials with widely different physical
properties, and their fracture behavior is predicted. Good agreement with experiment was
obtained. Thermal shock damage resistance
parameters suitable for the relative comparison
of the degree of damage to be expected in
materials fractured by thermal shock are proposed. The criteria for a low degree of
damage are high values of Youngs modulus of
elasticity, Poissons ratio, and effective surface
energy and low values of strength. Recom-

New

York

mendations are made for the selection of materials for severe thermal shock, where the best
materials available are known to fail.

1. Introduction
current theoretical treatment of the material properties which influence the thermal shock behavior of
brittle ceramic materials generally is based on those
material properties which affect the nucleation of fracture.
The procedure followed for given specimen geometry and
heat-transfer conditions is to calculate the resulting thermal
stresses. The specimen is considered to have failed when the
HE

Received December 6, 1962; revised copy received June 12,

1963.

This work was based on an investigation sponsored by the

United States Air Force under Contract No. AF33(616)-6806.
A t the time this work was done, the writer was physicist,
Applied Research Branch, Research and Development Division,
The Carborundum Company. He is now a graduate student,
Department of Mineral Technology, University of California,
Berkeley, California.

.Xi

Journal of The American Ceramic Society-Hasselman

thern-a1 stresses exceed the tensile strength. Based on this

criterion of failure, expressions are derived for the maximum
thermal shock to which a body of given shape can be subjected
in terms of the material properties and heat-transfer parameters., *
Materials with good resistance to fracture by thermal shock
are characterized by high values of strength, thermal conductivity, and thermal diffusivity and low values of Youngs
modulus of elasticity, Poissons ratio, coefficient of thermal
expansion, and emissivity. For the verification of the developed theories a number of investigations have appeared
in the
I n the experiments accompanying the
theoretical approach the severity of thermal shock is adjusted
to the material under study.
In summary, thc approach to the theoretical treatment of
thermal shock resistance is based on the determination of the
physical properties and heat-transfer variables which govern
the nucleation of fracture only.
In industry, however, particularly in the evaluation of the
thermal shock behavior of firebrick, thermal shock tests are
employed which are based on a constant-temperature difference and heat-transfer mechanism. Different materials
therefore are subjected to a thermal shock of the same
severity. The relative thermal shock resistance of the materials studied is then evaluated by measuring the relative
eflect of thermal shock on the physical properties or physical
condition of the specimen. For instance, for a given number
of cycles the percentage loss of eight,^ percentage loss of
~ t r e n g t hor
, ~the change in permeability5 may be determined.
The number of cycles required to produce a given change in
strength5 (or complete failure) or to produce a crack of given
widthGmay also be determined.
These thermal shock tests do not indicate whether a given
test will nucleate fracture but indicate the degree of damage
which will result from fracture by thermal shock. The relative thermal shock resistance of the materials studied there-

( a ) W. R. Buessem, Die Temperaturwechselbestandigkeit

keramischer Massen (Resistance of Ceramic Bodies t o Temperature Fluctuations (Spalling)), Sprechsaal, 93 [6] 137-41
(1960)
( b ) W. D. Kingery, Factors Affecting Thermal Stress Resistance of Ceramic Materials, J . Am. Ceram. Soc., 38 [l]3-15
(1955).
(c) W. B. Crandall and J . Ging, Thermal Shock Analysis of
Spherical Shapes, ibid., pp. 44-54.
* Listing all the references for theoretical investigations
of thermal shock resistance is beyond the scope of this paper.
The references cited should be regarded as typical examples
only.
( a ) Bernard Schwartz, Thermal Stress Failure of Pure
Refractory Oxides, J . Am. Ceram. Soc., 35 [12] 325-33 (1952).
( b ) W. R. Buessem and E. A. Bush, Thermal Fracture of
Ceramic Materials Under Quasi-Static Thermal Stresses (Ring
Test), ibid., 38 111 27-32 (1955).
( 6 ) D. P. H. Hasselman, Thermal Shock by Radiation Heating,ibid.,46 [51 229-34 (1963).
( d ) D. P. H. Hasselman and W. B. Crandall, Thermal Shock
Analysis of Spherical Shapes, 11,ibid., [9] 434-37.
( a ) Standard Method for Basic Procedure in Panel Spalling
Test for Refractory Brick, ASTM Designation C 38-58, pp.
293-99 in 1958 Book of ASTM Standards, Part 5. American
Society for Testing Materials, Philadelphia, Pa.
( b ) Standard Method of Panel Spalling Test for High Duty
Fireclay Brick, ASTM Designation C 107-52, ibid., pp. 300301.
(c) Standard Method of Panel Spalling Test for Super Duty
Firelay Brick, ASTM Designation C 122-52, ibid., pp. 302-303.
( d ) Standard Method of Panel Spalling Test for Fireclay
Plastic Refractories, ASTM Designation C 180-52, ibid., pp.
304-306.
a ) C. W. Parmelee and A. E. R. Westman, Effect of
Thermal Shock on Transverse Strength of Fireclay Brick,
J . Am. Ceram. SOG.,11 [12]884-95(1928).
( b ) H. R. Goodrich, Spalling and Loss in Compressive
Strength of Fire Brick, ibid., 10 [lo] 784-94 (1927).

Vol. 46, No. 11

fore is not governed by those physical properties which affect

the nucleation of cracks but by those physical properties
which affect the propagation of cracks once nucleated.
It is the purpose of the writer t o discuss the principal material properties which affect the propagation of cracks under
conditions of thermal shock. From these material properties,
thermal shock damage resistance parameters are derived,
which are indicative of the relative resistance of materials
to damage after fracture by thermal shock has been nucleated.
The application of these parameters to the design of materials
for thermal shock and to the interpretation of thermal shock
tests is illustrated.

II. Theoretical Considerations

The energy criteria which affect the propagation of cracks
consist of (a) the elastic energy stored in the body being
fractured which provides the driving force for crack propagation and (b) the various mechanisms which dissipate energy
and which tend to slow down or to arrest the propagating
crack; these mechanisms can be grouped together under the
heading7 effective surface energy.
The criterion for crack nucleation or propagation is provided by Griffith,s who states that a crack will start propagating and will continue to propagate as long as the elastic
energy released from the stress field surrounding the crack is
as great or greater than the energy required to supply the
effective surface energy. A propagating crack which is
being slowed down can also derive energy from the stress
wave (the kinetic energy) accompanying the crack. An
exact calculation of the energy balance a t any time during the
propagation of a crack in a finite body fractured by thermal
shock is mathematically prohibitive, if not impossible. It
is reasonable to assume, however, that the distance (or area)
over which the crack will propagate is directly proportional
to the elastic energy stored a t the time of fracture, i.e., a t
the time the cracks were nucleated. The same conclusion
was also reached by Kingery.lcb) Since fracture by thermal
shock generally takes place in the absence of external body
forces, the energy absorbed during the process of crack propagation is supplied by the elastic energy stored in the body
only. In the following discussion, the absence of external
body forces is always assumed.
The material properties which govern the elastic energy
a t fracture and their relation can be established by considering a sphere subjected to thermal shock by heating.
At the time of maximum stress the temperature distribution
to a good approximation can be considered to be parabolic.
From this temperature distribution the thermal stresses can
be c a l ~ u l a t e d . ~Since a t fracture the maximum thermal stress
6R. Fourneau, P. Lapoujade, and L. Fredholm, fitude
Comparative, Experimentale et Statistique de Diverses Mkthodes
EuropCennes dEssais de Chocs Thermiques (Comparative,
Experimental, and Statistical Study of Various European
Methods of Measuring Resistance to Thermal Shock); pp.
207-20 in Transactions of the VIIth International Ceramic
Congress, London, 1960. The British Ceramic Society, Federation House. Stoke-on-Trent. 1961. 491 D W . : Ceram. Abstr..
1962,August, p. 204f.
Russian Standard for Chemical Stoneware GOST 473-53;
see also N. L. Pevzner, Improving Method of Determining
Heat Stability of Acid-Resistant Shapes, Steklo i Keram., 1 1
1101 18-19(1954): Ceram. Abstr.. 1955,Tune, D. 10712.
f J. J. Gilman, Cleavage, Dhctilitf, and-Tenacity on Crystals; pp. 193-224 in Proceedings of International Conference on
Atomic Mechanisms of Fracture, Swampscott, Mass., April
1959. Edited by B. L. Averbach, D. K. Felbeck, G. T. Hahn,
and D. A. Thomas. Technology Press of Massachusetts Institute of Technology, Cambridge, and John Wiley & Sons,
Inc., New York, 1959. 646 pp.; Ceram. Abstr., 1960, June, p.
151d.
8 A. A. Griffith, Phenomena of Rupture and Flow in Solids,
Phil. Trans. Roy. Soc. (London), A221 [4] 163-98 (1920); abstracted in J . Am. Ceram. Soc., 4 [6] 513 (1921).
9 B. A. Boley and J. H. Weiner, Theory of Thermal Stresses,
p. 302. John Wiley & Sons, Inc., New York, 1960. 586 pp.
A.

Design Criteria for Thermal Shock

November 1963

equals the tensile strength (failure by tension is assumed),

the thermal stresses can be expressed in terms of tensile
strength. Substituting the expressions obtained into the
elastic energy equation,lo the total elastic energy stored a t
fracture ( W ) can be calculated to be
W =

4?rb3St2(1- v )
7E

= radius of sphere.
Si = tensile strength.
Y = Poissons ratio.
E = Youngs modulus of elasticity.

Equation (1) indicates that the elastic energy stored at

fracture is proportional to the volume of the sphere and the
material parameter SL2(1 - v)/E. But for the factor (1 - Y),
the same conclusions can also be obtained by using the technique of dimensional analysis. Therefore, but for the
numerical constant and the factor (1 - v), equation (1) is
valid for any geometry and any type of thermal shock.
The effective surface energy (yeff)consists principally of
the sum of the following: (a) the thermodynamic free energy,
( b ) energy dissipated by anelastic deformation at the tip
of the crack, and (c) energy dissipated by plastic deformation
(dislocation motion) of a thin surface layer on the newly developed crack surfaces.
The actual magnitude of these various forms of energy
depends on the type of material, surface roughness, and temperature Experimentally observed values for these different forms of energy dissipation are of the order of lo2
to 5 X l o 3 ergs
5 X l o 3 to lo6 ergs cmm2, and lo6
to lo8 ergs
respectively. Experimentally i t is rather
difficult to obtain precise values for (yeff). For materials in
which stable cracks can be introduced, such as
and single crystals,I2which exhibit cleavage fracture, or materials which contain surface flaws of known size,11@)values
of (rerf)have been obtained by means of a static technique.
On the basis of crack velocities, (yeff)also has been determined
for glasses by a dynamic method.I3 For brittle, polycrystalline, refractory materials no satisfactory experimental
technique has been developed. However, for these materials
the value of (yeff)will be controlled principally by the thermodynamic surface energy, for which several methods of measurement are available. l 4
Another factor which may affect the propagation of cracks
through a body is the presence of elastic discontinuities in
the form of grain boundaries or hard and soft dispersions in
the body undergoing fracture. Elastic discontinuities will

Stcplien Timoslienko and J. N. Goodier, Theory of Elasticity,

2d ed., p. 148. McGraw-Hill Book Company, Inc., New York,
1951. 506pp.
l 1 ( a ) F. C. Roesler, Brittle Fractures near Equilibrium,
Proc. Phys. Soc. (London), 69 [442B] 981-92 (1956); Ceram.
Ahstr., 1957,April, p. 82d.

( 0 ) A. M. Bucchc and J . P. Berry, Mechanisms of Polymer

Failure ; pp. 265-80 in Proceedings of International Conference
on rltomic Mechanisms of Fracture, Swampscott, Mass., April
1959. Edited b y B. L. Averbach, D. K. Felbeck, G. T. Hahn,
and D. A. Thomas. Technology Press of Massachusetts Institute of Technology, Cambridge, and John Wiley & Sons,
Inc., New York, 1959. 646 pp.; Ceram. Abstr., 1960, June, p.
151d.
l 2 J . J. Gilxnan, Direct Measurements of Surface Energies
of Crystals, J . Appl. Phys., 31 [12] 2208-18 (1960).
I3E. B. Shand, Fracture Velocity and Fracture Energy of
Glass in Fatigue Range, J . Am. Ceram. Soc., 44 [1] 21-26
(1961).
l 4 ( a ) W. I). Kingery, Metal-Ceramic Interactions:
IV,
Absolute Measurement of Metal-Ceramic Interfacial Energy
and Iuterfacial Adsorption of Silicon from Iron-Silicon Alloys,
J . Am. Ceram. Soc., 37 [2] 42-45 (1954).
( b ) M. Humenik, Jr., and T. J. Whalen, Physicochemical
Aspects of Cermets; pp. 6-49 in Cermets. Edited by J. R.
Tinklepaugh and W. B. Crandall. Reinhold Publishing Corp.,
New York, 1960. 239 pp.; Ceram. Abstr., 1961,April, p. 105a.

537

disperse or scatteP any elastic wave (the kinetic energy)

accompanying the propagating crack. Energy is thereby
removed from the propagating crack, causing it to be arrested
more rapidly. The suggestions of Royston and BarrettlG
that soft or hard inclusions in a material will increase its
thermal shock resistance and the observations of Hasselman
and Shaffer? on the fracture behavior of elastically heterogeneous bodies subjected to thermal shock offer experimental
evidence for the foregoing phenomena.
An exact calculation where, and if, the cracks will be
arrested on the basis of the stress distribution in the body
during the instant of fracture is exceedingly complex. For
simplicity, it will be assumed that all the elastic energy
stored at fracture is transformed into effective surface energy
and that the body is stress free when the cracks are arrested.
When the cracks are arrested, some elastic energy will remain
in the body until thermal equilibrium has been established.
The foregoing assumption therefore will break down for
those bodies in which the cracks will propagate over a
relatively short distance only. The distance the cracks will
propagate also depends on the number of cracks nucleated.
By equating the total surface energy required for the
propagation of the cracks to the elastic energy stored a t
fracture (equation (l)),the mean area ( A ) over which ( N )
number of cracks will propagate in a sphere is given by*

If the calculated area ( A ) for a given material and sphere

size is greater than the cross-sectional area of the sphere
(@), the cracks will completely traverse the body and lead
to complete failure. If ( A ) is substantially larger than the
cross-sectional area of the sphere, a considerable quantity
of energy is still available. This excess energy will be transformed principally into acoustic energy and also into kinetic
energy of the fragments. Under these conditions the fracture process will be quite explosive. If the calculated area
( A )is smaller than the cross-sectional area of the sphere, the
cracks will be arrested and the sphere will remain intact.
Since, in a fractured body in which the cracks are arrested
some elastic energy remains, the actual area over which the
cracks will propagate will be less than the area calculated
by means of equation (2). Expressions similar to equation (2)
can be derived for other geometries and other types of thermal shock.
Since the relative degree of damage will be proportional to
the area over which the cracks will propagate, minimizing
the quantity ( A ) in equation (2) will result in maximum thermal shock resistance. This will require low values of strength
and high values of Poissons ratio, Youngs modulus of elasticity, and effective surface energy. Nucleating a large
number of cracks will also result in a lower degree of darnage.

15 (a) C. F. YinP and R. Truell. Scatterinp of a Plane Longitudinal Wave by a Spherical Obstacle in- an Isotropically
Elastic Solid, J . Appl. Phys., 27 [9] 1086-97 (1956).
( b ) N. G. Einspruch, E. T. Witterholt, and Rohn Truell,
Scattering of a Piane Transverse Wave by a Spherical Obstacle
in an Elastic Medium, ibid., 31 [5] 806-18 (1960).
l6 M. G. Royston and L. R. Barrett, Some Observations and
Reservations on Thermal Shock Theory, Trans. Brit Ceram.
Soc., 57 [lo] 678-85 (1958); Ceram. Abstr., 1960, November, p.
271j.
D. P. H. Hasselman and P. T. B. Shaffer, Factors Affecting
Thermal Shock Resistance of Polyphase Ceramic Bodies,
Wright Air Development Division, WADD-TR-60-749 (Part
11),Contract AF33(616)-6806(February 1962). 155pp.
* The factor 2 is introduced since the quantity (ref?)
refers to
the effective surface energy of one side of the crack only.

538

Journal of The American Ceramic Society-HasAelman

Two thermal shock damage resistance parameters

can now be introduced* given by

and

RIII/

The parameter R can be used to compare the relative degree of damage of materials with similar crack-propagation
properties, i e . , the same values of (rcff).The parameter
R l l l l can be applied to compare the degree of damage of
materials with widely different values of (refJ,
such as brittle
and ductile materials.

111. Discussion
First, some remarks are in order concerning the use and
interpretation of thermal shock tests. As discussed in
Section I, the thermal shock resistance of a material can be
determined by two distinctly different methods By the
first method the degree of difficulty of nucleating fracture by
thermal shock is determined By the second method the
degree of damage by thermal shock after fracture has been
nucleated is measured. In reporting the relative thermal
shock resistance of various materials, it must be stated on
which basis the materials were judged. Ratings such as
poor, good, or excellent are useful only when regarded
in the proper light. Comparing materials judged on the
basis of the nucleation oi fracture with materials judged
on the basis of the degree of damage may lead to the completely erroneous selection of materials for thermal shock
applications. One of the favorite techniques of evaluating
relative thermal shock resistance is the thermal cycling of
small (approximately l/z by 1/4 by 3 in.) modulus of rupture
bars between room temperature and some other fixed temperature near 14OO0C. The strength of the bars after thermal
cycling is then compared with the original strength before
thermal cycling. It is clear that this test is useful only for
evaluating the relative degree of damage by thermal shock.
As soou as one or all of the materials being tested do not fracture, the results of the thermal cycling test become rather
meaningless, since no relative measure is obtained of the
degree of difficulty of nucleating fracture The danger
exists that the same thermal shock rating will be given to a
material in which no fracture is nucleated and to a material
which fractures very easily but in which the cracks are difficult
to detect The results of a thermal cycling test will be useful
only when used in conjunction with another thermal shock
test which indicates the relative degree of difficulty of nueleating fracture.
Since the terms thermal shock resistance or spalling
resistance are used indiscriininately to describe both the
nucleation of fracture by thermal shock and the degree of
damage by thermal shock, it is proposed here to refer to the
resistance to nucleation by thermal shock as thermal shock
fracture resistance or thermal stress resistance and to the
resistance to damage by thermal shock as thermal shock
damage resistance
Probably the most significant result of the foregoing theoretical treatment of crack propagation in a body subjected
to thermal shock is the conclusion that for a low degree of
damage there must be low values of strength and high values
of Youngs modulus of elasticity and Poissons ratio. Since
in previous theoretical treatments of the physical properties
affecting thermal shock resistance only those properties which
affect the nucleation of fracture by thermal shock were considered, most research efforts to improve thermal shock resistance have been directed toward improving strength and finding materials with low Youngs modulus of elasticity. Since
the improvement in strength which can be obtained for a
given material often is only marginal, for severe cases of thermal shock this approach frequently has led to more and more
disastrous results. In many cases it simply does not pay to
overcome the problem of failure by thermal shock by improving strength.

Vol. 46, No. 11

Actually, by deliberately weakening the body it is assured

that fracture occurs at a relatively low stress level. Rather
than fracture occurring catastrophically, the nucleated cracks
then will propagate only over a relatively short distance and
will become arrested. In this manner, the body retains its
shape and will still be useful for many applications. Deliberately lowering strength to improve thermal shock resistance appears to have been the approach taken by Aldred.l*
Another method of lowering strength is to deliberately introduce slits or cracks19 in the surface of the body. These
act as stress concentrators and will lower strength. In this
manner the number of cracks nucleated also may be increased.
An increase in the degree of damage with an increase in
strength has been observed by Heindl and PendergastZ0and
by Morgan.21
Since a body with high strength, for a given value of
Youngs modulus of elasticity, Poissons ratio, and effective
surface energy, will shatter completely, its strength after
thermal shock can be considered to be zero. Also, a body
with near zero strength, although undergoing only a minimal
amount of damage,21will still have a low value of strength
after thermal shock, simply because of its initial low value of
strength before thermal shock. As a consequence, there must
be some intermediate value of strength and a resulting degree of damage such that the strength after thermal shock is
a maximum. Therefore, although high values of the thermal
shock damage resistance parameters Rand R are desirable, it is clear that these parameters should not be maximized by letting the strength ( S I )approach zero. The optiwhich will
mum value of strength before thermal shock (Sf*)
result in maximum strength after thermal shock can be derived approximately as follows: As indicated by the results of Morgan21 the plot of strength after thermal shock
versus strength before thermal shock takes approximately
the form of a parabola, concave downward and symmetric
around a line parallel to the strength-after-thermal shock axis.
The optimum value of strength (St*)is then approximately
one half the value of strength where the cracks just traverse
the body completely, i.e., in the case of the sphere where the
quantity ( A ) in equation (2) equals the cross-sectional area
of the sphere. Setting ( A ) in equation (2) equal to nb2, the
optimum value of strength becomes
(4)

In the derivation for St* only strength is Considered as a

variable. In practice, however, it will be difficult, if not impossible, to change strength without affecting any one or
all the other variables. The foregoing quantitative approach
for the determination of (S,*) should be considered as exploratory only. It is recommended that research on the development of materials with good thermal shock damage resistance
be directed toward those materials with maximum strength
after thermal shock.

* For definitions of the parameters R, R, and R see footnote

Ub).
It should be acknowledged here that Kingery (footnote l ( b ) )
also suggested the use of thermal shock damage resistance

parameters. Kingery concluded, however, that these parameters

were identical to the thermal stress resistance parameters R,
R, and R. The results of the present work do not support this
conclusion.
18 F. H. Aldred. Ceramic Materials, U. S. Pat. 3,025,175,
March 13, 1962; Cevam. Abstr., 1962, August, p. 192c.
19 G. Ficker, Ceramic Structural Unit Resistant to High
Temperatures and to Variations in Temperature, Brit. Pat.
735,068, November 8,1955.
*o R. A. Heindl and W. L. Pendergast, Results of Laboratory
Tests of High Duty and Super Duty Fire-Clay Plastic Refractories, Am. Ceram. Soc. Bull., 36 [l]6-13 (1957).
21W. R. Morgan, Thermal Shock Effect on Transverse
Strength of Clay Bodies, J . Am. Ceram. Soc., 14 [12] 913-23
( 1931).

Design Criteria for Thermal Shock

November 1963

Graphite, owing to its exceptionally low value of Youngs

modulus of elasticity, takes a special position with regard to
thermal shock resistance. From the point of view of thermal shock fracture resistance, graphite can be considered to
be excellent, not only because of its low value of Youngs
modulus of elasticity but also because of its high thermal
conductivity. However, also because of its low value of
Youngs modulus, the elastic energy stored at fracture is
exceptionally high and exceeds that for many high-densityhigh-strength ceramics From the point of view of degree
of damage, graphite is rather susceptible to failure. Efforts
t o develop graphite materials for severe thermal shock conditions, such as electrodes, generally have resulted in bodies
of lower density, larger grain size, and relatively low strength 22
Much confusion appears to exist in the literature with regard to the elfect of porosity on thermal shock resistance.
I t is clear that the effect of porosity can be investigated from
the points of view of both nucleation of fracture and degree
of damage. The effect of porosity on strength and on Youngs
modulus of elasticity is generally such as to reduce the strain
a t fracture (ratio of S t / E ) . Thermal conductivity is also
lowered substantially The overall effect of porosity is to
lower thermal shock fracture resistance, as demonstrated
by Coble and K i n g e r ~lor
~ ~sintered alumina From the
point of view of crack propagation, however, the effect of
porosity on strength and on Youngs modulus of elasticity
is to reduce substantially the elastic energy stored at fracture Porosity, therefore, tends to reduce the degree of
damage in agreement with the observations of Parmelee and
W e ~ t n i a n , ~B
a~r t ~ c hKato
, ~ ~ and Okuda,2Gand Richardson 26
The results of Kato and OkudaZ5are of particular interest
since these investigators determined the effect of porosity on
the thermal shock fracture resistance as well as on the thermal
shock damage resistance of the same porcelain bodies Their
results showed that porosity lowered the thermal shock fracture resistance but increased the thermal shock damage resistance, in agreement with the foregoing postulates Pores
are often cited as acting a5 crack arrestors, and in this manner
are thought to increase thermal shock damage resistance
The principal d e c t of porosity, however, is to decrease substantially the elastic energy stored a t fracture
The effect of metal fiber reinforcements on the thermal
shock resistance of refractory oxides as investigated by Baskin, Arenberg, and I-1andwerkz7and by TinklepaughZ8can be
interpreted in terms of high values of effective qurface energy
The favorable properties of these fiber-reinforced ceramics
are based principally on the frictional force required to pull the
metal fiber5 from the ceramic matrix 29 The work done in
pulling the fibers from the matrix can be regarded as a form
of effective surface energy Assuming that the strength
of the composite is equal to the maximum frictional force on
the fibers pcr unit area and assuming that the fibers are perpendicular to the plane of the propagation crack, the effective surface energy can be expressed by
2LfE =

%sti

1 = average distance fibers must be pulled from matrix

fiber length)

(I

1/4

For the molybdenum-fiber-reinforced alumina investigated

by TinklepaughZ8with S t = 20,000 psi and fiber length of
in., the effective surface energy equals 313 in.-lb per in.2
This is an extremely high value
or 1.08 X lo8 ergs
compared with those usually observed for brittle materials.
Incorporating high-strength fibers in a ceramic matrix therefore can raise the effective surface energy substantially.
Since the effective surface energy (rerf)
has been considered
of significance only in the theoretical treatment of fracture
nucleation or propagation, it may not be readily appreciated
that this property can be considered of engineering value in
thermal shock damage resistance. To illustrate its practical
significance by means of equation (2), some sample calcula-

539

tions were carried out for the area traversed by cracks in

spheres 2 in. in diameter. Three distinctly different materials
were selected: ( a ) a high-strength brittle porcelain, (b) a
low-strength brittle fireclay body, and (c) a high-strength
ceramic reinforced by molybdenum fibers. The effective
surface energies of the first two materials are governed principally by the thermodynamic surface free energy, whereas
the effective surface energy of the third material is governed
principally by the frictional forces between the molybdenum
fibers and the alumina matrix, as discussed in the foregoing.
Listed in Table I are the values of the physical properties
of the three materials. Some of the values were obtained
from the literature; others, out of necessity, were estimated.
The number of cracks formed was assumed to be equal to
five, based on the experimental observations of Crandall and
Gingl(c) and Hasselman and Shaffer.? Included in Table
I is the prediction of the type of lailure to be expected by
comparing the calculated area of the crack with the crosssectional area of the sphere (approximately 20 cinz). Obviously, the porcelain sphere has fractured quite explosively,
whereas the spheres of the fireclay and the fiber-reinforced
alumina have retained their shape and can still be used for
many applications. For the comparison of the predicted and
observed types of failure of the fireclay body and the fiberreinforced alumina it was assumed that the experimental results that were obtained on rectangular bodies could be extended to bodies of spherical shape. For the fireclay body
and the fiber-reinforced alumina to have maximum strength
after thermal shock, the optimum values of strength before
were calculated to be 580 and 407,500
thermal shock (S,*)
psi, respectively. Morganz1 found an optimum value of
cross-bending strength for fireclay bodies of approximately
1600 psi. For these types of materials, the actual tensile
strength is usually approximately one half the cross-bending
strength. The value OF (St*) as observed by Morgan21
then becomes approximately 800 psi, in good agreement with
the calculated value. The exceptionally high value of
(St*) for the fiber-reinforced alumina suggests that if composites could be manufactured approaching this value of
strength, a material would be obtained with high thermal

( a ) E. A. Neel, A. A. Kellar, and J . J. Zeitsch, Research

and Development on Advanced Graphite Materials: Vol.
VII, High Density Recrystallized Graphite by Hot Forming,
Wright Air Development Division, WADD-TR-61-72, Contract
AF33(616)-6915 (Scptember 1961). 70 pp.
( b ) C. A. Klein, Pyrolytic Graphite, Intern. Sci. Technol.,
1 (8160-68 (1962).
2 3 R ,I,. Coble and W. D. Kingery, Effect of Porosity on
Thermal Stress Fracture, J . Am. Ceram. SOC.,38 [l] 33-37
(1955).
24 Otto Bartsch, Ueber die Schlagbiegefestigkeit von ke-

ramischen Masseii und Glasern und ihre Beziehung zur Temperatmwechselbestandigkeit von Schamottemassen (Impact
Bending Strength of Ceramic Bodies and Glasses and Its Relationshin to the Resistance to Thermal Shock of Fireclav Bodies).
Her. Deut Keram. Ges., 18 Ill] 465-89 (1937); Cevam. Abstr:,
17 [ll]354 (1938).
26 Shuzo Kato and Hiroshi Okuda, Relations Among Porosity,
Thermal Shock Resistivity, and Some Physical Constants of
Porcelain. N a ~ o v a Komo Giiutsu Shikensho Hokoku.. 8 .151.
37-43 (1659); d & m . Ab%., 1959, November, p. 287a.
zGH. M. Richardson, Choice of Refractories for Linings of
Blast Furnaces; pp. 13540 in Transactions of the VIIth
International Ceramic Congress, London, 1960. The British
Ceramic Society, Federation House, Stoke-on-Trent, 1961.
491 pp.; Ceram. Abstr., 1962, August, p. 204s.
27Y.Baskin, C. A. Arenberg, and J . H. Handwerk, Thoria
Reinforced by Metal Fibers, Am. Ceram. SOC.Bull., 38 [7]
34548 (1959).
28 J. R. Tinklepaugh, Metal Reinforcement and Cladding
of Cermets and Ceramics; pp. 170-80 in Cermets. Edited by
J. R. Tinklepaugh and W. B. Crandall. Reinhold Publishing
Corp., New York, 1960. 239 pp.; Ceram. Abstr., 1961, April,
p. 105a.
29 J. R. Tinklepaugh; private communication.

Journal of The American Ceramic Society-Hasselman

540
Table I.

Vol. 46, No. 11

Calculation of Area Traversed by Propagating Cracks ( A ) and Prediction of Type of Failure for Spheres 2 In. in
Diameter Fractured by Thermal Shock

27rStZ(1- v)b3
7NEr0.tt
.. . ~
Condition of failure: Calculated value of A greater than cross-sectional area of sphere (about 20 cmz)
General equation: A

Material properties

Tensile strength (SO

dyne cm-2 (psi)

Youngs modulus ( E )
dyne cnir2 (psi)
Poissons rate ( u )
Effective surface energy
( Y r f i ) (ergs cm-*)
Calculated area traversed by
single crack ( A ) (cnia)*

Predicted type of failurc

Observed type of failure
Correctly predicted

~~~

Fireclay body P

Porcelaint

1.38 X loQ
(20,000)
2.07 X 1012
(30 X lo6)
0.25
2

10q

1.05 x 103

Complete failure
Complete failure
Yes

3.45 x 103
( 5001

3.45 x loll$
(5 x 106)
0,253
2

1.38 x 109
(20,000)
3.45 x 1012
(50 X 106)

0.25$
108

1033

4.0

Alumina reinforced by molybdenum

fibers11

Cracks arrested
Cracks arrested
Yes

1.22 x 10-2
Cracks arrested
Cracks arrested
Yes

* Xuinbcr of cracks nuclcatcd (N)

was assumed to be equal to five.

t Footnote l(c).

$ Assumed values.

Footllotc 21.

I Footnote 28 and R. S.Truesdale, J. J. Swica, and J. R. Tinklepaugh, Metal Fiber Reinforced Ceramics, Wright Air Development
Center, WADC-TR-58-452, Contract AF33 (616)-5298 (December 1958). 36 pp.

shock fracture resistancc as well as high thermal shock damage resistance.

Since the probability that a crack will completely traverse
a body will depend on the ratio of the total elastic energy
stored at racture to the cross-sectional area of the body,
it is clear that the larger the body undergoing fracture by
thermal shock, the greater becomes the degree of damage.
An increase in thermal shock damage resistance therefore can
be attained simply by decreasing the body size.
Since it has been shown that the effective surface energy
can play a significant role in the fracture behavior of materials
subjected to thermal shock, it is suggested that greater emphasis be placed on the quantitative determination of the
mcchnnisrns by which energy is dissipated during crack propagation
In this manner, more accurate values for (rcff)
are obtained with the result that the fracture behavior of brittle materials subjected to thermal shock can be predicted
with greater accuracy Ways also may be developed by
which the effective surface energy can be increased Incorporating into thc brittle ceramic matrix a continuous or dispersed relatively ductile phase will increase the value of
(-y.rf) and thereby can lead to an increase in thermal shock
damage resistance. The same result is obtained by increasing
the degree of anelasticity of the material. Of course, an
increase in plasticity or anelasticity also will lead t o an increase in thermal shock fracture resistancz.
The foregoing results and conclusions should be regarded
as valid only for the first cycle of thermal shock. It is entirely possible that crack propagation on subsequent thermal
cycles is again governed by those physical properties which
affect the nucleation of fracture. Also, during the time of
thermal cycling, the chemical properties of the material at
the tip of the crack may change, owing to the adsorption OF
gases on the surface This will affect the value of (reff)
and could lead to further crack propagation. During the
cycling treatment, small particles of material may become
dislodged, enter the crack, and gradually widen the cracks
by means of a wedge action. The presence of the cracks
also may affect the temperature distribution considerably.
Rased on the theories advanced in this paper, two directions can now be taken in the selection of materials for thermal shock applications. The criterion of lack of nucleation
of fracture by thermal shock dictates the selection of materials
with low coefficients of thermal expansion, Youngs modulus of

elasticity, Poissons ratio, and eniissivity and high values of

strength and thermal conductivity. Since the avoidance of
fracture by thermal shock is always desired, this is the direction which should always be pursued first. For those cases of
severe thermal shock where even in the best materials available fracture will be nucleated, the selection of materials
should be based on a low degree of damage, which is dictated
by low values of strength and high values of Youngs modulus
of elasticity, Poissons ratio, and effective surface energy.
The approach of a low degree of damage should always be
accompanied, however, by a search for, or development of,
those materials which will resist the nucleation of fracture by
thermal shock.

IV.

Summary and Conclusions

The principal factors that affect the propagation and arresting of cracks were discussed. The driving force for crack
propagation is provided by the elastic energy stored at fracture. The mechanism which tends to arrest the propagating crack is provided by the effective surface energy that is
required to produce the newly formed crack surfaces.
An expression was derived for the mean area over which
cracks nucleated in a body of spherical shape will propagate.
This expression permits the prediction of whether failure by
thermal shock will be catastrophic or whether the cracks will
be arrested. Sample calculations were carried out for three
materials of widely different properties and their fracture
behavior was predicted. Good agreement with experiment
was obtained.
Thermal shock damage resistance parameters applicable to
a comparison of the relative damage to be expected in materials fractured by thermal shock were derived. The criteria For a low degree of damage are high values of Youngs
modulus of elasticity, Poissons ratio, and effective surface
energy and low values of strength. Examples were cited
from the literature and interpreted in terms of these criteria.
For cases of very severe thermal shock for which no materials
exist, in which fracture is not nucleated, it was recommended
that the selection of materials be based on a low degree of
damage.
Acknowledgment

The writer is indebted to his colleagues in the Physics Department of The Carborundum Company for many helpful discussions and for review of the manuscript.