Compact high-temperature cell for Brillouin scattering measurements
Stanislav V. Sinogeikin,
a)
Jennifer M. Jackson,
b)
and Bridget O’Neill
Department of Geology, University of Illinois at Urbana–Champaign, Urbana, Illinois 61801
James W. Palko
Department of Materials Science and Engineering, University of Illinois at Urbana–Champaign, Urbana,
Illinois 61801
Jay D. Bass
Department of Geology, University of Illinois at Urbana–Champaign, Urbana, Illinois 61801
共Received 20 July 1999; accepted for publication 27 September 1999兲
A compact ceramic high-temperature cell for Brillouin spectroscopy was designed and tested. The
cell can be mounted onto a three- or four-circle goniometer and allows collection of the full set of
elastic constants of minerals to temperatures in excess of 1500 K from samples with dimensions of
100⫻ 100⫻ 20
m or smaller. As a test of the instrument the single-crystal elastic constants of MgO
were measured to 1510共10兲 K, and are found to be in excellent agreement with earlier independent
results. The high-temperature cell should be useful for other types of spectroscopic measurements,
and is especially useful in situations where spectral properties vary with the scattering geometry.
© 2000 American Institute of Physics. 关S0034-6748共00兲02201-2兴
I. INTRODUCTION
A knowledge of elastic and other fundamental properties
of materials at high temperature is important from both the
technological and scientific points of view. High-temperature
elasticity data provide fundamental information on the inter-
atomic forces in solids. They are important in geophysics for
constructing reliable mineralogical models of the Earth’s in-
terior by comparing laboratory elasticity data on minerals
with seismologically determined properties. For geophysical
purposes and for materials that may be used in high-
temperature applications, it is obviously advantageous to
characterize materials at elevated temperatures.
Brillouin scattering provides a means of studying the
high-temperature elasticity of materials. Being a purely opti-
cal method, it does not require any mechanical contact be-
tween the sample and measuring equipment, and can be per-
formed on very small samples with dimensions comparable
to the size of the focused beam. In addition, Brillouin spec-
troscopy can be used to study other properties at high tem-
perature such as the refractive index,
1,2
photoelastic
properties,
3
hypersonic attenuation,
4
and the mechanisms of
phase transitions.
2,4–9
Despite the considerable possibilities described above,
there have been relatively few high-temperature Brillouin
spectroscopy experiments reported in the literature. Most
have typically used bulky furnaces with limited optical ac-
cess and restricted possibilities for data collection, and the
crystal samples typically exceeded several millimeters in
each dimension. Although this does not present a problem
for work with common materials, some materials of interest
cannot be synthesized with crystal sizes more than a few
hundreds of microns. Other limitations may include 共1兲 dif-
ficulty in the precise orientation of samples, 共2兲 very limited
angular access to the sample, which restricts the number of
scattering directions and elastic constants that can be ob-
tained, especially for symmetries lower than cubic, and 共3兲
difficulty in measuring the refractive index as a function of
temperature independently.
8,10
To overcome these problems we designed a ceramic
high-temperature cell for single-crystal Brillouin spectros-
copy to temperatures in excess of 1000 °C. The cell is com-
pact 共about 5 cm in the maximum dimension兲, and easily fits
onto any standard three- or four-circle goniometer, allowing
a number of additional applications such as x-ray scattering,
etc. Our Brillouin cell utilizes a symmetric scattering
geometry
11
which yields velocity measurements independent
of the refractive index. This is important because the refrac-
tive index can change appreciably with temperature. Because
the cell mounts onto a goniometer, one obtains very good
control of the scattering geometry, and any phonon direction
within a plane of the platelet can be easily sampled by
changing the
-angle setting of the goniometer. As a result,
the complete set of elastic constants as a function of tempera-
ture for crystals with symmetry higher than monoclinic can
be obtained from a single sample if the sample plane inter-
sects all three crystallographic axes.
12
Finally, the cell allows
us to work with crystals approximately 100⫻100⫻ 20
mor
smaller.
13
It should be straightforward to modify the cell for
experiments in a controlled atmosphere.
II. CELL DESCRIPTION
The geometrical design of the high-temperature Bril-
louin cell is similar to that of a large-opening Merill–Bassett
diamond anvil cell
14
共Fig. 1兲. However the cell described
here is designed to generate high temperature only, not pres-
a兲
Author to whom correspondence should be addressed; electronic mail:
sinogeik@uiuc.edu
b兲
Current address: Dept. of Civil Engineering and Geological Sciences, Uni-
versity of Notre Dame, Notre Dame, IN 46556.
REVIEW OF SCIENTIFIC INSTRUMENTS VOLUME 71, NUMBER 1 JANUARY 2000
2010034-6748/2000/71(1)/201/6/$17.00 © 2000 American Institute of Physics
Downloaded 09 Jan 2007 to 131.215.240.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp
sure. The cell consists of five major components: the cell
base, window supports, windows, heater, and thermocouples
共Fig. 2兲, each of which is described below.
The main body of the cell is constructed from machin-
able alumina silicate ceramic, which is inexpensive and is
easily obtained from a variety of manufacturers and distribu-
tors 共e.g., McMaster–Carr Supply Company兲. This type of
ceramic is rated to withstand temperatures of up to 1370 K,
and has proved reliable for sample temperatures well in ex-
cess of 1500 K. For higher temperatures, a different machin-
able ceramic can be readily substituted 共e.g., RESCOR 960,
Cotronics兲. Although RESCOR 960 ceramic is more expen-
sive, it has better thermal shock and mechanical properties at
higher temperatures than alumina silicate ceramic.
The advantage of using a machinable ceramic is that it is
soft and easily shaped with conventional tools. Since these
types of ceramics are usually quite brittle, care must be taken
not to chip sharp edges during machining. After the parts are
machined, they are heat treated according to the manufactur-
er’s instructions to obtain their maximum strength and ther-
mal stability. Note that during heat treatment the alumina
silicate ceramic can expand slightly 共1.8% in our case兲,
which must be taken into account when fitting with metal
parts such as the goniometer socket. A number of machin-
able ceramics 共e.g., RESCOR 960兲 do not require heat treat-
ment and can be used for a more precise match of cell com-
ponents.
A. Base
The base of the cell 共Fig. 3兲 consists of two main parts:
a goniometer socket and a ceramic base. The goniometer
socket was machined from brass. Alternatively, commercial
sockets/height adapters can be obtained from any x-ray
equipment supplier.
The upper part of the cell base, to which the main body
of the cell is mounted, was made from machinable alumina
silicate ceramic. The spacer for mounting window supports
in the upper part of the base is machined with a thickness
which is slightly greater than the thickness of the two win-
dows to allow for the thickness of the sample and the ther-
mocouples. We used 1/8 in. 共3.175 mm兲 thick windows and
a thickness for the spacer bar of 6.5 mm to ensure a 0.15 mm
gap between the windows.
Since the cell is mounted on a goniometer, special care
should be taken to prevent the goniometer from overheating.
The cell can be water cooled to minimize the temperature at
the base. Due to the low thermal conductivity of the ceramic
and holes for water cooling 共which decrease the effective
area for heat transfer兲, the base of the cell remained rela-
tively cool 共60–70°C兲 without water cooling even at the
highest temperature achieved 共1510 K兲.
A strip of 1 mm thick stainless steel was placed between
the lower and upper parts of the base and bent appropriately
so it could provide support for electrical connectors 共Fig. 2兲.
A piece of insulating ceramic fiber strip was placed in be-
tween the connector support and the ceramic base. The go-
niometer socket and ceramic base were connected with two
screws.
B. Window supports
Top and side views of the window support plates are
shown in Fig. 4. The window supports were machined from
1.5 in. diam rods of machinable alumina silicate ceramic.
The cone has an overall opening of 100°, which allows its
use in a symmetric scattering geometry with up to a 90°–95°
scattering angle. Since no pressure is applied to the cell, the
FIG. 2. View of the entire high-temperature cell. 1—Goniometer socket;
2—insulating ceramic plate; 3—metal support for the thermocouple and
heater connectors; 4—ceramic base; 5—threaded holes for water cooling;
6—heater connector; 7—ceramic window support; 8—screws; 9—fused
silica window; 10—ceramic guide pins; 11—ceramic spacer; 12—
thermocouple connectors.
FIG. 1. Schematic diagram of Brillouin scattering in the high-temperature
Brillouin cell. TC—thermocouples; HW—heating wires.
FIG. 3. Base of the high-temperature cell. 1—Goniometer socket;
2—ceramic plate; 3—metal support for the heater and thermocouple con-
nectors; 4—insulating ceramic fiber strip; 5—ceramic base; 6—threaded
holes for optional water cooling; 7—screw well for connecting the ceramic
base with the metal base; 8—spacer for mounting window supports;
9—holes for ceramic guide pins.
202 Rev. Sci. Instrum., Vol. 71, No. 1, January 2000 Sinogeikin
et al.
Downloaded 09 Jan 2007 to 131.215.240.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp
diameter of the central hole in the supporting plate can be
nearly as large as that of the windows.
The parallellness of the window supports is controlled
by proper dimensions of ceramic spacers 共one bar of the base
and one tube on top sitting on guide pins兲. The distance
between window supports can be increased by putting addi-
tional washers on top of the ceramic spacers.
The supporting plates have five through holes: three for
ceramic guide pins and two for screws. The guide pins 共stan-
dard ceramic thermocouple insulators兲 ensure proper posi-
tioning of the windows and connect the window supports
with the base. In one of the two support plates the screw
holes are threaded.
Before heat treatment we made grooves 1–1.5 mm deep
which are used to recess and secure the thermocouple and
heating wires 共Fig. 4兲.
On the outer side of the window supports we cut 1–2
mm deep steps to allow gluing of additional silica-glass win-
dows, which eliminate contact of the main windows with air
and reduce the temperature gradient inside the cell. The
trade-off is that the outside windows introduce additional
astigmatism and reflect an appreciable part of the incident
and scattered beams, diminishing the quality of the Brillouin
spectra.
C. Windows
Brillouin spectra are geometry sensitive, and the scatter-
ing angle must be carefully controlled in any experiment. To
ensure that no part of the incident or scattered beams is cut
off, the ratio of window height to outside window diameter
should not exceed a critical value which is defined by the
refractive index of the window material used, the scattering
geometry, and the numerical aperture of the focusing and
collecting lenses. The maximum height to width ratio for the
windows is easily calculated through Snell’s law. For ex-
ample, in our experiments we used windows made of fused
silica. At a wavelength of ⫽ 5000 Å, the room temperature
refractive index of the fused silica is about 1.46 and does not
change appreciably with temperature. For a window thick-
ness of 1/8 in. 共3.175 mm兲 the diameter of the outside sur-
face of the window should not be smaller than 4.8 mm in
order to use an 80° symmetric scattering geometry with f
⫽ 3.3 lenses and achievea1mmworking area inside the
cell. Failure to meet these conditions will result in vignetting
共shadowing part of the incident and/or scattered cones of
light兲 and systematically lower the observed Brillouin shifts.
A discussion of vignetting effects on Brillouin spectra is be-
yond the scope of this article; we refer the reader to the
discussion by Oliver et al.
15
who considered an optical ge-
ometry identical to ours in the context of Brillouin scattering
with a diamond anvil cell. We note only that the dimensions
of our high-temperature cell are easily altered to suit almost
any optical configuration without significantly changing the
performance of the cell.
The windows were machined from 1/8 in. 共3.175 mm兲
thick disks of fused silica with polished parallel faces. For
windows 6 mm in diameter the disks were cut into 6
⫻ 6 mm square or octagonal blocks. These blocks were glued
with superglue to 6 mm diam silica-glass rods 共with the pol-
ished surfaces of the blocks perpendicular to the axis of the
rod兲. These rods were secured in a drill press, and the blocks
were ground to the desired conical shape with a conventional
diamond file.
The fused silica windows proved adequate to tempera-
tures of up to 1500 K, but can lose optical quality or soften at
higher temperatures. The silica may also react with certain
materials at high temperature. Alternative materials used for
windows were sapphire and cubic zirconia. The advantage of
these two materials are higher temperature stability, lower
reactivity, and higher refractive indexes. This reduces the
minimum window diameter-to-height ratio, which will result
in a more uniform temperature distribution, and allows work
with larger sample areas. The disadvantages are higher
strength 共which makes the machining of windows more dif-
ficult兲, high acoustic velocities and velocity anisotropy
共which could complicate the Brillouin spectra of very thin
samples兲, and higher astigmatism due to the higher refractive
indexes.
Sapphire windows were also machined from 1/8 in. thick
polished sapphire disks. The windows, shaped as truncated
octagonal pyramids, were cut using a fine diamond saw. Cu-
bic zirconia windows were made from round diamond-cut
jewelry obtained from a commercial gemstone dealer. A flat
inner facet was hand ground and polished parallel to the
table of the gemstone using a metal jig and a series of silicon
carbide abrasive films 共30, 5, 1
m grit size兲. Excess mate-
rial on the sides was ground off with silicon carbide sandpa-
per. Alternatively, the windows from either of the above ma-
terials can be easily machined using a lapidary faceting
machine.
Note that only two parallel facets of the windows should
have near-optical quality. The sides of the windows should
be coarse ground, which makes it easier to glue the windows
to window supports and to glue thermocouples to the win-
dows.
The windows were attached to the window supports with
FIG. 4. Top view and cross section of the ceramic window support plate
with the attached window. 1—Window; 2—ceramic window support plate;
3—holes for guide pins; 4—screw hole; 5—grooves for thermocouple
wires; 6—grooves for heater wires; 7—high-temperature cement;
8—optional outer window.
203Rev. Sci. Instrum., Vol. 71, No. 1, January 2000 Brillouin scattering
Downloaded 09 Jan 2007 to 131.215.240.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp
either Cotronics 903 HPRT alumina adhesive or Cotronics
906 magnesia adhesive. We found that, since the alumina
adhesive contains much finer powder than the magnesia ad-
hesive, it is easier to work with, especially where very fine
work is required. Nevertheless the magnesia adhesive shows
less reactivity and better stability at high temperature.
D. Heater
The core of the heater was machined from a 1/2 in. diam
rod of alumina silicate ceramic. The rod was threaded inside
and outside with a 1/4 and 1/2 in.⫻40 tpi tap and die, respec-
tively, cut into 5 mm thick disks, and heat treated. A high
ratio of the height to the inner diameter of the heater is very
important for creating stable temperature conditions and
maintaining small temperature gradients.
It was found that the double-coil design of the heater
共the heating wire wound inside and outside of the heater
core兲 produces more stable temperatures and smaller tem-
perature gradients in the sample chamber. It also reduces the
temperature of the heater wires needed to maintain a given
temperature in the sample chamber, which allows higher ex-
perimental temperatures to be obtained. It was found benefi-
cial to use different heating wires for the internal and exter-
nal coils with higher power density for the internal coil. We
used 0.01 in. Pt–30Rh wire for the internal coil and 0.005 in.
Pt or Pt–30Rh for the external coil. Outside the heater, the
two coils were connected together in parallel and lead to the
high-temperature ceramic connector outside the cell.
The heater was powered with a digital dc power supply
rated to a maximum of 18 V and 10 A. The temperature was
controlled to 0.5° by controlling the output voltage. At any
temperature the voltage did not exceed 14.6 V with a corre-
sponding current of 6.6 A. At the highest temperature 共1510
K兲 the output power was 96.6 W.
To further increase the effectiveness of the heater, we
packed all empty space in between the window supports with
insulating ceramic fiber tape. This additional insulation de-
creases both power consumption and temperature gradients
inside the cell.
E. Temperature control
The temperature was measured directly by two thermo-
couples fixed with high-temperature cement to the window.
The cement also served to insulate the thermocouple wires
from heating and conducting heat to the thermocouple 共TC兲
junctions. The thermocouples were placed as close to the
sample and as far away from the heater as possible.
We used type K 共Chromel–Alumel兲 thermocouples
made of 0.005 in. diam wire. At temperatures above 1000 °C,
the thermocouple wires oxidize and reliability can be lost
after long durations at high temperatures 共although covering
them with an appropriate cement inhibits the oxidation兲.
Therefore, at temperatures in excess of 1000 °C, Pt-based
thermocouples 共types B, S, R兲 are prefered, even though the
sensitivity and reliability of type K thermocouples at lower
temperatures are superior.
The temperature in the cell is stable over time. Even at
the highest temperatures, after thermal equilibration the tem-
perature fluctuations did not exceed 2° over several hours of
operation. Since the cell is made from materials with very
low thermal conductivity, the thermal gradient inside the cell
共between thermocouples兲 is small. The thermal gradient is
not perfectly symmetric. Preliminary measurements in the
direction of the maximum thermal gradient show that this
gradient does not exceed 10°/mm at temperatures below
1000 K and 20°/mm at temperatures around 1500 K. The
thermal gradient is very sensitive to the choice of materials
for the cell and the configuration of the heater and windows.
III. ELASTICITY OF MgO TO 1500 K
MgO is an ideal substance for testing our new furnace. It
is cubic in structure, and therefore its elasticity is character-
ized by only three independent elastic constants. The high-
temperature single-crystal elasticity of MgO has been mea-
sured by a number of different experimental techniques up to
its melting point. The experimental measurements include
ultrasonic 共rectangular parallelepiped resonance兲 studies
16,17
and Brillouin measurements.
18
In addition, MgO is a very
good Brillouin scatterer, large flawless single crystals can be
easily obtained, and perfect cleavage along 兵100其 planes
makes the orientation of samples easy. MgO is nonreactive,
and the sample does not deteriorate in air even at tempera-
tures exceeding 1500 K.
The single-crystal elasticity measurements were per-
formed on two samples of MgO by Brillouin scattering in
two separate runs up to a temperature of 1510共10兲 K.
The slab of a synthetic single crystal of MgO was pol-
ished on both sides along the 共100兲 direction to a thickness of
⬇100
m and cleaved into plates of ⬃1⫻ 2 mm. These
plates were glued to the fused silica window with Cotronics
high-temperature alumina or magnesia adhesive. Because of
strong capillary forces some of the adhesive flowed under the
sample and therefore lifted it from the surface of the window
by 20–30
m 共Fig. 5兲. Subsequent optical goniometry indi-
cated that the angle between the face normals of the window
and sample faces did not exceed 1°.
In both runs we utilized 80° symmetric platelet scatter-
ing geometry with an Ar-ion laser (⫽ 5145 Å) light source
FIG. 5. Sample mounting on a window. The thermocouples shown are nor-
mally attached to an opposite window.
204 Rev. Sci. Instrum., Vol. 71, No. 1, January 2000 Sinogeikin
et al.
Downloaded 09 Jan 2007 to 131.215.240.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp
and a six-pass tandem Fabry-Pe
´
rot interferometer. Further
description of the Brillouin system is given elsewhere.
19
It is
worth noting that the scattered light is filtered with a combi-
nation of a dispersing prism and a pinhole. This filters out all
light except for the green region. Thus, thermal radiation
does not show as a high background in Brillouin spectra, a
typical example of which is shown in Fig. 6.
In the first run, alumina adhesive was used to attach the
sample and thermocouples to the cell window. Data were
obtained at temperatures of 295, 523, 773, and 1023 K
共shown by circles in Figs. 7 and 8兲. Above 1050 K, one of
the thermocouples broke. In addition, the optical quality of
the MgO sample started to deteriorate, probably due to vapor
reaction with the alumina adhesive, and the run was stopped.
We found that the Chromel thermocouple wires had reacted
with the alumina adhesive, and one of them was completely
destroyed during the run. The alumina adhesive also reacted
with the silica windows at high temperature, although the
reaction was restricted to the contact zone only.
In the next run, we used magnesia adhesive to glue the
thermocouples and sample to the window, and no reaction
was observed to the maximum temperature of 1510 K. The
data were collected at temperature of 300, 473, 673, 873,
1073, 1173, 1273, 1373, and 1510 K 共shown by squares in
Figs. 7 and 8兲. After the data collection at 1273 and 1510 K
the temperature was returned to 1073 K to check if there was
any difference in velocity measurements made on increasing
and decreasing temperatures. No systematic difference was
observed, and both elastic moduli and aggregate acoustic ve-
locities were in mutual agreement.
At temperatures of 295共1兲, 1073共5兲, and 1510共10兲 K the
data were collected in more than 10 crystallographic direc-
tions 共Fig. 9兲 which allowed us to solve for both orientation
and single-crystal elastic moduli.
12
Our results show that the
calculated phonon directions at the three different tempera-
FIG. 6. Brillouin spectra of MgO in the 关011兴 direction. Solid lines—Room
temperature spectra; dashed lines—spectra at 1510共10兲 K. The intensities of
the peaks are scaled to fit on the same plot. At 1510 K, the actual intensity
of the Brillouin peaks was a factor of 5 higher than at room temperature.
Note the low background, even at 1510 K; this indicates that thermal emis-
sion was effectively filtered out.
FIG. 7. Single-crystal elastic constants of MgO as a function of temperature.
Open symbols 共circles and squares兲 represent two different runs from this
study. Closed symbols show ultrasonic data of Sumino et al. 共Ref. 16兲 and
of Isaak et al. 共Ref. 17兲 for comparison. The size of the symbols is bigger
than the experimental uncertainty.
FIG. 8. Aggregate acoustic velocities in MgO as a function of temperature.
The symbols have the same meaning as those in Fig. 7.
FIG. 9. Acoustic velocities in MgO as a function of crystallographic direc-
tion 共
angle兲 at room temperature 共closed symbols兲 and at 1510 K 共open
symbols兲. Acoustic velocities calculated from the best-fit elastic moduli at
room temperature 共solid line兲 and at 1510 K 共dashed line兲. The 关010兴, 关011兴,
and 关001兴 crystallographic directions are marked.
205Rev. Sci. Instrum., Vol. 71, No. 1, January 2000 Brillouin scattering
Downloaded 09 Jan 2007 to 131.215.240.9. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp
