Sie sind auf Seite 1von 10

Mechanism of mechanical failure of sapphire at high temperature

Daniel C. Harris,t Frederick Schrnid,* J. J. Mechoisky, Jr and Y. L. Tsai


tChemjstj.y Division, Research Department
Naval Air Warfare Center, Weapons Division, China Lake CA 93555
Crysta1 Systems, 27 Congress Street, Salem MA 01970
Department of Materials Science and Engineering, University of Florida, Gainesville FL 32611-2066

ABSTRACT
The strength of sapphire decreases more rapidly with increasing temperature than does the strength of
polycrystalline alumina and many other ceramics. Twinning on the rhombohedral plane (1 1 02) at elevated
temperature induced by compression along the crystallographic c-axis [000 1 1 appears to initiate failure and
accounts for the decreased strength. The tensile strength of sapphire along the a- [1 1 OI or c-axes is constant
to within -30% between 20 and 800C. Compressive strength along the a-axis is also constant to within
20%. However, compressive strength along the c-axis falls by >95% (from 2000 MPa to less than 100
MPa) between 20 and 800C.

Keywords: sapphire, mechanical strength, fracture, mechanical failure, twinning, infrared window,

midwave infrared material

1. INTRODUCTION
Sapphire (single crystal A1203) is the premier midwave (3-5 j.im) infrared window material for applications
requiring high mechanical and thermal durability.15 Although it is more resistant to thermal shock than other
oxide midwave window materials such as spinel, yttria and aluminum oxynitride6, sapphire is stressed to its

limit in some environments. Unlike its polycrystalline competitors, sapphire suffers a marked loss of
mechanical strength at temperatures above 20C,713 and this loss of strength limits the resistance to thermal
shock. This paper describes the mechanism by which sapphire loses strength at elevated temperature.
Fig. I displays crystallographic and mineralogical nomenclature for sapphire. The 3-fold symmetry axis (the
optical axis) is designated as the c-axis. The plane perpendicular to c is called the basal plane. Three a- and

three rn-axes are interspersed at 30 angles in the plane perpendicular to c. The a- and rn-planes are
perpendicular to the a- and rn-axes. Six r-planes (rhombohedral planes) are inclined at 57.6 to the c-axis.

2. FLEXURE STRENGTH VARIES WITH TEMPERATURE AND ORIENTATION


Flexure specimens from previous work13 are shown in Fig. 2 and the test geometry is illustrated in Fig. 3.
Specimens were made from Hemlite grade sapphire grown by the heat exchanger method (HEM)14 at
Crystal Systems. Bars and disks were polished to a nominal scratch/dig specification of 60/40 using random
circular motion, and the edges of bars were polished to a 0.08-0.13 mm radius. Mechanical testing was
conducted at the University of Dayton Research Institute with an Instron testing machine and silicon carbide
fixtures. Results of 4-point flexure tests with bars and ring-on-ring equibiaxial flexure tests with disks are
summarized in Tables 1 and 2 and Fig. 4. In Table 1, the average strength and standard deviation are given
for each type of specimen at each temperature. The number in the box beneath the strength in Table 1 is the
percent of specimens that failed at the outer support points, as shown in Fig. 3. Specimens failing at these

points usually gave just two fragments with lengths of approximately 42.5 and 2.5 mm. Sometimes,
secondary fracture of the 42. 5-mm fragment occurred.

16 / SPIE Vol. 2286


Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/08/2013 Terms of Use: http://spiedl.org/terms

O-8194-1610-X/94/$6.OO

[00011
(a)

[2113]

(m)

[r]

[n]

Fig. 1.
Two views of

[1010]

[1102]

sapphire crystal
showing mineralogical and
Miller index

[1
[mJ

notation. The
c-axis is a 3fold axis of
rotation.

(a)

2mmI

38 mm

.
Disk 2
(900 Disk)

Disk 1
(00 Disk)

45 mm

Bar 4

Bar 3
(m -Axis bar)

Bar 5
(a -Axis bar)

(C -Axis bar)

Fig. 2. Sapphire flexure specimens. In a previous publication,13 the diagrams associated with bars 4 and 5
were mistakenly interchanged. The previous paper is corrected by interchanging the labels in Figs. 3 and 4.
10mm

Fig. 3. Geometry of 4-point


flexure test. Load is applied
to the inner rollers, while the
outer rollers provide support.

-a
2.5
mm Many high-temperature
failures observed here

Span of tensile surface where


failure is normally expected

SPIE Vol. 2286 / 17


Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/08/2013 Terms of Use: http://spiedl.org/terms

Table 1. Four-point flexure strength (MPa) of sapphire bars in Fig. 2*


rn-Axis bars
Temperature

I % Failing

(C)

c-Axis bars

a-Axis bars

Strength standard deviation Strength standard deviation Strength standard deviation


at outer supportj

1% Failing at outer support

I%

Failing at outer supporti

2Q

70333%

105534%

500

95623%

78435%

60428%

II

1000

69%I
520 12%

[1
27%
I1
24437%
I1

80141%

169

490

58%I

58715%

1400

I7I

19%

I9I

47713%
167%I

*40 bars of each type were tested at each temperature using the geometry in Fig. 3 with a crosshead speed
of 0.508 mm/mm.

Table 2. Ring-on-ring equibiaxial flexure strength (MPa) of sapphire disks in Fig. 2*


Temperature
(C)

0 Disk

90 Disk

Strength standard deviation

Strength standard deviation

60730%
36948%

1053

20D

14211%

800

*20 disks of each type were tested at each temperature using a test fixture with a load radius of 7.94 mm
and a support radius of 15.88 mm and a crosshead speed of 0.508 mm/mm.

1200

1000

0.

800

600

0)

Fig. 4. Summary of flexure


tests from Tables 1 and 2.

C')

400
200

500

1000

1500

Temperature (C)

18 / SPIE Vol. 2286


Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/08/2013 Terms of Use: http://spiedl.org/terms

Key observations from Tables 1 and 2 include the following:


1.

Table 1 shows that sapphire mechanical properties in the m and a directions are essentially the same, and

differ from properties in the c-direction.


2. The strength of c-axis flexure bars fell by 84% (from 1055 to 169 MPa) between 200 and 1000C. Flexure
strength of a- and rn-axis bars declined less than 38%.
3 . The flexure strength of OO disks in Table 2 fell by 87% between 200 and 800C, while the strength of 900
disks declined only 39%.
4 . Flexure bars are expected to fail on the tensile surface between the two inner loads. However, the majority
of a- and rn-axis flexure bars tested at 500C failed at the outer support point (Fig. 3). In one case (a-axis
bars, 1000C), 95% of the failures occurred at the outer support.

3. FLEXURE SPECIMENS ARE TWINNED


Virgin sapphire samples used in this work were excellent-quality single crystals. After flexure testing,
twinning was apparent in samples broken at 800C when specimens were viewed between crossed
polarizers. Microscopic examination also revealed twinning in samples tested at 500C. Fig. 5 shows
twinning and crushing at the support point of an rn-axis bar tested at 1000C. This particular specimen failed
at the other support point not shown in the photograph, and broke into one long piece and one short piece.
Fig. 6 shows two twin planes in a 0 disk tested at 800C. Both terminate at the fracture line running along the
diameter of the disk. One of the twins crosses a radial fracture line, so the twin probably formed prior to the
radial fracture.

Support point

Fig. 5. Twinning at outer support point of in-axis


flexure specimen 3-85 tested at 1000C. The bar
broke at the other support point not shown here.
Fig. 2 shows the crystal axes of the bar and Fig. 3
shows the geometry of the test. This photograph
was taken with the sample sandwiched between
crossed polarizers, using bottomside illumination.

Fig. 6. Twinning observed in 0 disk 1-29 used

for equibiaxial flexure test at 800C. This

photograph was taken with the disk sandwiched

between crossed polarizers, using bottomside


illumination. Crystal axes in the disk are shown in
Fig. 2.

SPIE Vol. 2286 / 19


Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/08/2013 Terms of Use: http://spiedl.org/terms

Fig. 7. Schematic representation of twinning.


Planes of atoms in the twinned section do not
have the same orientation as atoms in adjacent
sections. Previous observations of twinning

Twin

induced by flexure of sapphire at elevated

temperature showed that the twinned regions had


thicknesses in the range 1 to 50 im.15

Twinning, which should be distinguished from slipping, is a reorientation of the planes of atoms in one slice

of a crystal relative to the alignment in neighboring regions (Fig. 7). Twins appear very sharp in crosspolarized light because of the shift in orientation. In contrast, slipping does not give rise to regions with
different orientations, so slip boundaries are not as distinct as twins in cross-polarized light. The predominant
twin plane was identified as the rhombohedral plane in several cases where the angle of the twin was measured

under a microscope. (It would be difficult to distinguish r planes from n planes in Fig. 1 based on our
measurements, but the r plane has been identified as the prominent twin plane in previous studies. 121516)
Measurement of load vs. time gives strong evidence that extensive twinning occurred in most high-temperature

flexure tests prior to fracture. Fig. 8 shows the force exerted by the sample on the test fixture as a function of
time. The trace at the left, which was typical of all tests at 20C, is characteristic of most brittle materials: the
load increased almost linearly until the specimen broke. At all test temperatures 500C, a variety of features
such as the ripples at the right side of Fig. 8 were observed. We interpret each ripple as a twinning event
leading to temporary stress relief, as described by Scott and Orr.2

Bar 3-43
20C

Failure

Bar 4-83
500C

point

Twinning events

40 lb

Fig. 8. Graphs showing


load vs. time in 4-point
Failure
point

flexure tests at 20C (left)


and 500C (right).
Ripples in the high

temperature test are

interpreted as twinning
events.

20 / SPIE Vol. 2286


Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/08/2013 Terms of Use: http://spiedl.org/terms

4. FAILURE MODES IN 4-POINT BENDING


All bars tested at 20C shattered into numerous small shards, which is typical of a strong, brittle material that

stores a great deal of energy before failure. Fracture origins were located in four instances. All were

associated with flaws in the tensile region on the bottom surface between the load points in Fig. 3.

The m- and a-axis bars in Fig. 2 had similar failure modes at elevated temperature (500C), and these are
shown in Fig. 9. Microscopic examination showed crushing and twinning at the outer support points on the
bottom surface of the bar in Fig. 3. Specimens that broke into more than two pieces did not show twinning on
the fracture surfaces that were not in contact with the outer support. We interpret these observations to mean
that twinning caused by compression at the outer support points was the primary cause of failure. Secondary
fractures that sometimes occurred after the primary event did not involve pre-existing twins.

The typical failure mode shown for c-axis bars in Fig. 9 is qualitatively different from the failure mode of mand a-axis bars. c-Axis bars at elevated temperature (500C) fail on the tensile surface beneath the load
points not at the support points. Twins were observed on every fracture surface that was examined.
rn-Axis

bar C

c-Axis bar f a

IC

crushing

Fig. 9. Left: Typical twin pattern observed in m- and a-axis bars tested at 1 000C. Failure originated at the

outer support point for many specimens. Right: Typical twin pattern observed in c-axis bars tested at
500C. Failure originated beneath the inner load point on the tensile surface.

5. A HYPOTHESIS
Why do the a- and rn-axis bars in Table 1 behave qualitatively differently from c-axis bars? The a- and rn-axis
bars do not lose strength as much as the c-axis bars, but the majority of a- and rn-axis bars fracture at the outer
support point instead of fracturing in the high-tensile-stress region. Why do the 0 disks in Table 2 (in which
the c-axis is perpendicular to the plane of the disk) lose more strength than the 900 disks? This contradicts the
behavior of the bend bars, in which most strength is lost when c is in the plane of the bar.

A common thread that appears to explain these observations is the hypothesis that the weakness of
sapphire at elevated temperature is due to c-axis compression. Such behavior is unusual, since
material strength is normally limited by response to tension, not compression.

Bend bar observations are explained as follows: Fracture of a- and rn-axis bars at 500C is initiated by
compression in the c direction by the outer support rollers. Compression where the roller meets the specimen
in Fig. 3 is not normally a source of failure, because most materials are stronger in compression than in

tension. Compression does not cause failure at the inner load points in Fig. 10 because of the biaxial
compressive stress at these points. The material under the inner load points is in compression from two
directions, whereas the material above the outer support points is in compression in one direction and tension
SPIE Vol. 2286 / 21
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/08/2013 Terms of Use: http://spiedl.org/terms

in the other direction. Failure occurs when (1) compression initiates twinning and (2) twins
interact with other twins or with pre-existing flaws in a tensile stress field.

Fig. 10. Compression in the vertical direction


causes failure at the outer support points where

Compression
interacts with

Compression

tension:

compression:

= failure

compression interacts with tension. Compression


in the vertical direction at the inner load points

interacts with compression in the horizontal


direction, so it does not cause failure.

interacts with

= no failure

In c-axis bend bars, compression in the c direction (parallel to the axis of the bar) on the top surface of the bar
in Fig. 3 initiates twinning at elevated temperature. When these twins propagate to the lower (tensile surface),
twin-twin or twin-flaw interactions lead to failure on the tensile surface.
To explain why disks appear to behave in the opposite manner from bars, we suggest that c-axis compression
initiates failure where the support ring contacts the disk. Fractography that might confirm this explanation has
not yet been carried out.

6. TESTING THE HYPOTHESIS


We postulate that sapphire is weak when compressed on the c-axis at elevated temperature, because such
compression initiates twinning. Confirmation of this hypothesis was obtained by measuring the strength of
sapphire in compression and tension along the a and c directions, using specimens shown in Fig. 11.
Table 3 shows that the tensile strength of sapphire is similar on the a- and c-axes, and the tensile strength does
not change much between 200 and 800C. The compressive strength of sapphire is essentially equal on the aand c-axes at 20C. The compressive strength on the a-axis decreases by only 21% between 20 and 800C.

However, the compressive strength drops by more than 95% for c-axis compression between
20 and 800C. Representative c-axis compression samples are shown in Fig. 12.

Fig. 11. Compression cylinder and tensile


specimen drawn to correct relative scale. The

Compression specimen
3.18 mm diameter
6.35 mm length

______________
______________

..+-

__________

cylindrical surface of the compression


specimen and the narrow central gauge

__________

to an optical finish. Both types of samples

section of the tensile specimen were polished

were prepared with the c or a crystal axes

Tensile specimen

oriented along the cylindrical axis.

6.35 mm gauge diameter


165 mm length

22 ISPIE Vol. 2286


Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/08/2013 Terms of Use: http://spiedl.org/terms

Table 3. Tensile and compressive strength (MPa) of sapphire specimens shown in Fig. 11 .
Temperature

Orientation of
axis of sample

Tensile strength range

a
a
c
c

302 17
314 48
432 53

20
800
20
8000

(2 samples tested)

292 2

Compressive strength
standard deviation

2002 293 (5 samples)


1581 92 (5 samples)
1972 457 (5 samples)
41 19 (7 samples)t

*Tensile loading rates were 310 lbs/s (1.4 kN/s) and compressive loading rates were 1200 lbs/s (5.3
kN/s). These rates were intended to give strain rates similar to those of flexure tests.

tAn eighth sample was anomalous, with a strength of 618 MPa. The average for all 8 samples was 113
201 MPa.

Fig. 12. c-Axis compression cylinders before testing


(left), after testing at 20C (center) or after testing at
8(X)C (right,).

Microscopic examination of tensile specimens after fracture showed no evidence of twinning. The fracture
origins were at the polished surface of the gauge section. Broken compression specimens tested at 800C (aand c-axis samples) exhibited extensive twinning. However, two intersecting twin systems were observed on
the loaded end of four c-axis samples while only one set of parallel twins was seen at the end of the one a-axis

sample that was investigated (Fig. 13). Fracture of the c-axis specimens may have originated at the

intersection of twins.

Fig. 13. Left: Loading surface of c-axis compression cylinder tested at 8(X)C showing intersecting systems
of twins. Right: Loading surface of a-axis cylinder tested at 800C showing only one set of parallel twins. It
is postulated that fracture originates at the intersection of twins. Tensile specimens showed no twinning.
SPIE Vol. 2286/23
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/08/2013 Terms of Use: http://spiedl.org/terms

Twinning in sapphire under compression at elevated temperature has been seen before, Scott and pl2
observed twinning during c-axis compression at elevated temperature and stated that "when twins on different

systems intersected, extensive fracture invariably initiated at the intersection." Heuer15 noted rhombohedral
twinning in sapphire flexure specimens and suggested that twin-subgrain and twin-twin interactions "probably
nucleated fracture." While we saw such effects at temperatures 500C, Wiederhorn, Hockey and Roberts1
observed that "plastic deformation by dislocation motion or twin formation and growth plays no role in the
fracture process at temperatures below 400C."

7. FUTURE WORK
In earlier experiments by Hurley9, a-axis compression specimens of sapphire suffered little loss of strength up

to 1300C, while c-axis strength fell drastically above 800C. Fig. 14 compares our results to those of
Hurley, carried out with the same size specimens. The dotted lines show that sapphire in our study has a
lower compressive strength at 800C.

2000

a.

1500

Fig. 14. Comparison of compression


strength of sapphire as a function of

C) 1000

temperature in the a and c directions to


data presented by Hurley.9

C)
CO

Cl)

C)

I..

a.

C)

200

400

600

800

1000

1200

Temperature (C)

What accounts for the different behavior? Hurley's sapphire was grown by pulling from a melt, using the
"EFG" method.17 The sapphire used in our study was grown by the heat exchanger method (HEM)14 and
contains low levels of subgrains and other defects. Radford and Pratt18 made a systematic study of the effect
of impurities on the mechanical properties of sapphire in the temperature range 1200 to 1650C using
compression tests. That study showed that the yield stress of sapphire was significantly increased by addition
of impurities. Chromium doping increases the hardness of sapphire and increases its yield strain in high
temperature plastic deformation.19 Impurities may therefore be important for increasing the strength of
sapphire at elevated temperature.

Hurley's EFO sapphire was drawn to the required 3 mm diameter and the surface was left as-grown. Our
HEM specimens were machined from a boule and polished which probably creates more surface flaws than
would be found in an as-grown cylinder. Hurley's tests were conducted under argon and ours were in air.
Radford and Pratt18 showed that the high temperature yield strength of impurity-doped sapphire increased
when tested in argon rather than in air.

24 / SPIE Vol. 2286


Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/08/2013 Terms of Use: http://spiedl.org/terms

Different surface quality and different levels of crystalline perfection could account for the differences in
response at elevated temperature in Fig. 14. It is possible that crystalline defects might increase the activation
energy for twinning and thereby increase the c-axis compression strength of sapphire at elevated temperature.
A deliberate investigation of factors controlling the activation energy for rhombohedral twinning in sapphire
could lead to a significant increase in the performance of this material.

8. SUMMARY
Sapphire loses strength at temperatures above 500C because compression on the crystallographic c-axis
causes twinning on rhombohedral planes. The intersection of different twin systems, or the interaction of
twins with pre-existing flaws, appears to initiate mechanical failure.

9. ACKNOWLEDGEMENTS
We are indebted to G. A. Graves, D. E. McCullum and S. M. Goodrich at the University of Dayton Research
Institute for their dedicated efforts to achieve accurate test results.

10. REFERENCES
1.

C. P. Khattak and F. Schmid, "Production of near-net-shape sapphire domes using the heat exchanger method," Proc.

SPIE 1992,
2.

1760, 41-47.

J. w. Locher, H. E. Bates, W. C. Severn, B. G. Pazol and A. C. DeFranzo, '80-mm EFG sapphire dome blanks yield

high-quality low-cost single-crystal domes," Proc. SPIE 1992, 1 760, 48-54.


3. B. G. Pazol, R. DeVito, P. J. Giguere and P. S. Kiefner, "Development of sapphire windows for use in high-quality JR
imaging systems," Proc. SPIE 1992, 1 760, 55-65.
4. S. Biderman, A. Horowitz, Y. Einav, 0. Ben-Amar, D. Gazit, A. Stern and M. Weiss, 'Production of sapphire domes by
the growth of near-net-shape crystals," Proc. SPIE 1991, 1535, 27-34.
5. D. C. Harris, Infrared Window and Dome Materials, SPIE Press, Bellingham, Washington, 1992.
6. J. S. Lin and L. B. Weckesser, "Thermal shock capabilities of infrared dome materials," Johns Hopkins APL Technical

Digest 1992, 13, 379-385.


7. E. A. Jackman and J. P. Roberts, "The strength of single crystal and polycrystalline corundum," Phil Mag. 1955, 46,
809-8 11.

8. J. B. Wachtman, Jr. and L. H. Maxwell, "Strength of synthetic single crystal sapphire and ruby as a function of
temperature and orientation," J. Am. Cer. Soc. 1959, 42, 432-433.
9. G. F. Hurley, "Mechanical behavior of melt-grown sapphire at elevated temperature," Applied Polymer Symposium No.
21 , John Wiley & Sons, New York, 1973, pp 121-130.
10. S. M. Wiederhorn, B. J. Hockey and D. E. Roberts, "Effect of temperature on the fracture of sapphire," Phil Mag. 1973,
28, 783-796.
1 1. R. L. Gentilman, E. A. Maguire, H. S. Starrett, T. M. Hartnett and H. P. Kirchner, " Strength and transmittance of
sapphire and strengthened sapphire," J. Am. Cer. Soc. 1981, 64, Cl 16-C 117.
12. W. D. Scott and K. K. Orr, "Rhombohedral twinning in alumina," J. Am. Cer. Soc. 1983, 66, 27-32.
13. J. W. Fischer, W. R. Compton, N. A. Jaeger and D. C. Harris, "Strength of sapphire as a function of temperature and
crystal orientation," Proc. SPIE 1990, 1326, 11-22.
14. F. Schmid and D. Viechnicki, "Growth of sapphire disks from the melt by a gradient furnace technique," J. Am. Cer.
Soc. 1970, 53, 528-529.; F. Schmid, C P. Khattak and D. M. Felt, "Producing large sapphire for optical applications," Am.
Cer. Soc. Bull. 1994, 73, 39-44.
15. A. H. Heuer, "Deformation twinning in corundum," Phil. Mag. 1966, 13, 379-393.
16. K. P D. Lagerlof, A. H. Heuer, J. Castaing, J. P. Rivire and T. E. Mitchell, "Slip and twinning in sapphire (a-A1203),
J. Am. Cer. Soc. 1994, 77, 385-397.
17. H. E. LaBelle, "EFG the invention and application to sapphire growth," J. Crystal Growth, 1980,50, 8-17.
18. K. C. Radford and P. L. Pratt, "The mechanical properties of impurity-doped alumina single crystals," Proc. Brit. Ceram.
Soc., 1970, 15, 185-202.
19. M. V. Klassen-Neklyudova, V. G. Govorkov, A. A. Urusovskaya, N. N. Voinova and E. P. Kozlovskaya, "Plastic
deformation of corundum single crystals," Phys. Stat. Sol. 1970,39, 679-688.

SPIE Vol. 2286 /25


Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/08/2013 Terms of Use: http://spiedl.org/terms

Das könnte Ihnen auch gefallen