Journal of ELECTRONIC MATERIALS, Vol. 39, No.

9, 2010
DOI: 10.1007/s11664-010-1265-6
Ó 2010 TMS

Thermal Conductivity Measurements on Challenging Samples

by the 3 Omega Method

A. JACQUOT,1,2 F. VOLLMER,1 B. BAYER,1 M. JAEGLE,1 D.G. EBLING,1

and H. BÖTTNER1

1.—Department of Thermoelectric Systems, Fraunhofer-Institut für Physikalische Messtechnik IPM,

Heidenhofstraße 8, 79110 Freiburg, Germany. 2.—e-mail: alexandre.jacquot@ipm.fraunhofer.de

The 3 omega method is the best established method for measuring the thermal
properties of thin films (>100 nm) and nanowires. Theoretically, the method
could be applied to many more types of samples, leading to new knowledge,
but to date little effort has been made to extend its applicability. An enabling
set of technologies has been tested at the Fraunhofer-IPM. The technologies
developed encompass a new design of microheater, the measurement of bulk
samples with a prefabricated microheater on adhesive tapes and polymer
sheets, the measurement of tiny bulk samples glued to the underside of
polymer sheets, a fully automated experimental setup, and a new numerical
tool adapted to the new type of heater. The new design of microheater and
software were validated using float glass as a reference material. A micro-
heater on adhesive tape was used to measure accurately the thermal prop-
erties of sintered thermoelectric materials. The thermal conductivity of a very
small melt-spun nanocomposite sample glued to the underside of a Kaptonä
sheet was measured. The potential of the new design of microheater to mea-
sure very thin (nm) films is discussed.

Key words: 3 Omega, thermal properties measurement, thermal

conductivity, heat capacity, SPS materials, bismuth telluride,
bismuth antimony telluride

INTRODUCTION almost impossible to isolate the microheater prop-

erly from the sample. Practicable ways of making
Since Cahill pioneered the 3 omega method for
the insulating layer are by sol–gel techniques or by
measuring bulk insulating material about 20 years
using postbaked photoresists, but these films do not
ago,1 it may be pointed out that very few bulk
reliably prevent leakage. If the insulating film is
thermoelectric materials have been measured using
made thicker, a lower frequency (1 Hz) has to be
it. The major field of application for the 3 omega
used so that the thermal wave penetrates deep into
method is rather the measurement of relatively thin
the thermoelectric material, making the contribu-
films or superlattices2–4 and nanowires,5 where its
tion of the insulating film to the temperature rise
accuracy is undisputed. Still, the 3 omega method is
smaller. In fact, nobody uses low frequencies, not
very appealing for measurements in a cryostat,
because it is experimentally impossible, but because
because the method uses between two and up to four
in this case, the thermal wavelength may be com-
electrical contacts to measure the thermal proper-
parable or larger than the microheater length itself.
ties, a record in simplicity that remains unbeaten to
In this case, the assumption that under the heater
this day. If the 3 omega method is not the work-
the heat transfer is two dimensional is no longer
horse for the measurement of thermal conductivity
valid. Recently, a microheater approximately in the
of bulk thermoelectric materials, it is because it is
form of a spiral was presented, but the numerical
tools failed to account for all the features of the
(Received July 10, 2009; accepted May 3, 2010; phase and amplitude plotted as a function of fre-
published online May 25, 2010) quency.6 In this article the heat transfer below the

1621
1622 Jacquot, Vollmer, Bayer, Jaegle, Ebling, and Böttner

Fig. 1. Exploded assembly drawing of the spiral-like microheater on its substrate, design of the microheater, and cross-sectional view of the finite
elements (meshing) used to calculate the heat transfer beneath the microheater. The four pads are the connections for current leads (I+,I) and
voltage leads (V+,V). The hatched area of the exploded view is the plane of the cross-sectional view of the meshing.

spiral-like heater is simulated by the heat transfer Seebeck coefficient is also possible with this mea-
below concentric rings instead of modeling the surement system.12 For room-temperature mea-
spiral-like heater as a disk. Since a ring has an axis surements, the Fraunhofer-IPM uses a sample
of symmetry and since the asymmetry of the spiral- holder with Peltier elements for temperature regu-
like heater is negligible, this symmetry can be used lation. In this case the temperature does not change
to calculate the three-dimensional (3D) heat trans- by more than 1 mK and stabilizes very quickly.*
fer using a two-dimensional (2D) mesh (Fig. 1), and Measurements are made in vacuum. This is impor-
the numerical simulation of Jacquot et al.7 can be tant when the measurement is done on membranes
easily extended. The method will be designated the or at high temperatures. The numerical tool that
concentric 3 omega method. The concentric method comes with the measurement system is an improved
is tested on float glass and then used to measure the version of the Windows-based software developed by
thermal conductivity of melt-spun spark plasma Jacquot et al.7 The material properties used in the
sintered (SPS) (Bi,Sb)2Te3. Since there is a demand numerical simulation are listed in Table I.
for measurement systems and methods for mea-
suring small samples, the Fraunhofer-IPM CONCENTRIC METHOD
extended the on-membrane method introduced by
Validation
Jacquot et al.8 by using an alternating current (AC)
instead of a direct current (DC) in the microheater. A bolometer with the shape shown in Fig. 1 was
Preliminary results of measurements on Kaptonä fabricated by photolithography on a sample made of
sheet, Polyamide films, and melt-spun (Bi,Sb)2Te3-2 float glass. The diameter of the spiral-like heater,
mol% PbTe nanocomposite are presented.9,10 the strip width, and strip separation were 920 lm,
20 lm, and 40 lm, respectively. The temperature
amplitude (a) and phase (b) were measured with a
EXPERIMENTAL PROCEDURES
power per unit area of 7521 W m2 and plotted as a
The measurement system used for this work function of the inverse square root of frequency
is a further development of a previously presented (Fig. 2). The agreement between experiment and
system.11 Validation was carried out by measur- the simulation done with the meshing shown in
ing the thermal conductivity of quartz glass Fig. 1 is very good over the whole frequency range
(1.31 W m1 K1) from Schott Lithotec (Lithosilä). for both amplitude and phase. Hence, it can be
The system can perform highly automated mea-
surements. The instrumentation is used together *The temperature controller is a Thorlab TED350. Depending on
with a cryostat from Cryovac, thus enabling mea- the temperature sensor used, temperature stability is better than
surements between 70 K and 700 K. The sample 1 mK over 24 h, and measurement resolution of 0.001°C and
holder for the cryostat was developed at the temperature accuracy of ±0.1°C are achievable (see technical
Fraunhofer-IPM. Measurement of the cross-plane data of the TED350 available online).
Thermal Conductivity Measurements on Challenging Samples by the 3 Omega Method 1623

Table I. Thermophysical properties at room temperature of materials used in the numerical simulations
Thickness Density Thermal Conductivity Heat Capacity
(lm) (kg m23) (W m21 K21) (J kg21 K21)

Float glass 1000 2450a 1.07b 790a

Kaptonc 25 1090 0.28 1420
Siliconc 500 2330 150 711
Adhesive film 15 830d 0.13b 1690d
Meltspun-SPS-(Bi,Sb)2Te3 1500 7730d 1.00b 160d
Polyamidee 10 1140 0.226 1500
(Bi,Sb)2Te3 films 1.1 7730d 2.06b 160d
Nanocomposite (Bi,Sb)2Te3-2 mol% PbTe 40 7730d 1.80b 160d
a
The density and heat capacity of float glass are assumed to be not far from the properties of LithosilTM quartz glass; bOur work (best
curve fitting); cData from Ref. 7; cData taken from www.MatWeb.com and assumed to be the same as the properties of the silicone-based
fluid Durathermä; dData taken from www.MatWeb.com and assumed to be the same as those of Bi2Te3; eData from Ref. 6.

measurement results done on liquids if the proper-

ties of the material labeled 1 and 2 are set to those
of the liquid. The measurement is simplified
because it is possible to measure not only the heat
capacity but also the thermal conductivity with the
same spiral-like heater. It was noticed in Ref. 6 that,
when the thermal wave penetration depth is much
smaller than the diameter of the spiral-like heater
and smaller than the distance between two rings,
the integral form of the Cahill formula gives a rea-
sonable value for the measured amplitude. This
means that, if the strip width of the spiral-like
heater is much smaller than the distance between
two rings, a frequency range exists where the line-
arized (Eq. 1) form of the Cahill formula may also
work quite well. In this case, the slope of the real
Fig. 2. Amplitude (j) and phase (d) measured at room temperature
with a spiral-like microheater on float glass. The solid lines are cal- part of the temperature amplitude plotted as a
culated with the data of Table I. function of the logarithm of the frequency may be
used to extract the thermal conductivity. If the
thermal wave is much larger than the distance
between two rings but much smaller than the
concluded that it is correct to simulate the heat
diameter of the spiral-like heater, the product of
transfer of a microheater in the form shown in
heat capacity and thermal conductivity can be
Fig. 1 with concentric rings. This is a remarkable
extracted by plotting the amplitude as a function of
result since the spiral-like heater is not perfectly
the inverse square of frequency.13 To date, the state
symmetric whereas concentric rings are. Neverthe-
of the art was to measure the heat capacity using a
less, taking into account eight rings and taking the
wide heater and the thermal conductivity using a
average diameter of the spiral-like heater seems to
much narrower heater,14 the equations used to
be sufficient. The new microheater design and
extract the heat capacity and the thermal properties
numerical tool may also simplify and greatly
from the measurement results being only valid
improve the accuracy of the measurement of ther-
in two extreme configurations and over a very
mal properties (heat capacity and thermal conduc-
restricted frequency range almost impossible to
tivity) of solids and liquids.13,14 The measurement of
determine experimentally.
a liquid by the 3 omega method as proposed by
Birge and Nagel13 will be more accurate with a
Melt-Spun SPS (Bi,Sb)2Te3
spiral-like heater because of the decoupling between
the microheater width, necessarily large to measure A bolometer in the shape shown in Fig. 1 was
the heat capacity accurately, and the electrical fabricated by photolithography on a sample made of
resistance of the microheater, which should be kept Kapton adhesive tape. The microheater had the
to a reasonable level (larger than 10 Ohm). Since it same geometry as the one used for the measure-
is possible in the simulation to put the microheater ment done on float glass. The adhesive tape was
between the material labeled 2 and 3 in Fig. 1, the stuck to a silicon wafer, and the temperature
simulation can be used for the interpretation of amplitude (Fig. 3a, curve S) and phase (Fig. 3b,
1624 Jacquot, Vollmer, Bayer, Jaegle, Ebling, and Böttner

Sample

(1)
(3)

(2)

(4)

Fig. 4. (1) A supporting Kapton sheet with two integrated micro-

heaters (metal strips with four electrical contacts) prefabricated.
(2) The sample of which the thermal conductivity has to be measured
was glued with a polyamide film onto the underside of the Kapton
sheet just underneath one microheater. (3) The assembly (mem-
brane + sample) was then fixed with silver paint onto an aluminum
plate (frame) having two rectangular openings so that the metal
strips pass exactly across the middle of these openings. (4) The
comparison between the measurements made with both micro-
heaters enables the determination of the thermal conductivity of the
sample.

with only one sample, for example, by using two

spiral-like heaters of different diameter and/or by
making the measurement and interpreting the
results over a broad frequency range. Since the
diameter of the spiral-like heater may be much
Fig. 3. (a) Amplitude and (b) phase measured at room temperature smaller than the straight strip currently used
with an adhesive tape on which a spiral-like microheater was (length between 2 mm and 4 mm), much smaller
microfabricated, when the adhesive tape is on a silicon wafer (S) or samples can be measured.
SPS material (T).

ON-MEMBRANE METHOD
curve S) were measured at a power per unit area of Preliminarily results are presented for the first
3451 W m2 to obtain a comparison between the measurements of thermal conductivity of a flake
experiment and simulation of the properties of the (1.5 mm 9 2 mm 9 0.04 mm) of (Bi,Sb)2Te3-
adhesive film (glue underneath Kapton tape), since 2 mol%PbTe made by melt-spinning.9,10 A micro-
the thermal properties of Kapton and silicon are heater in the form of a metallic strip 5 mm long and of
known (Table I). Then, the adhesive tape was stuck width 10 lm has been microfabricated by photoli-
to the sintered sample, and the temperature thography on a Kapton sheet 50 lm thick (Fig. 4).
amplitude (Fig. 3a, curve T) and phase (Fig. 3b, The nanocomposite was glued with a polyamide
curve T) were measured again. We emphasize that spray to the underside of the Kapton sheet. The sheet
just crude polishing of the sample is required. It is was glued with silver paint onto a frame made of
not necessary to microfabricate the microheater on aluminum with an opening of 4 mm 9 2 mm. The
the sample surface. When the frequency is lower distance between the voltage (1 mm) wires connected
than 0.4 Hz, the difference between curves S and T to the microheater was chosen to be significantly
is large, especially for temperature amplitude. The smaller than the length of the flake. The real part of
additional temperature rise of curve T compared the amplitude and the phase shift were measured at a
with curve S is a function of the thermal properties heating power per unit length of 2 W m1 and are
of the sintered material. Using the 3 omega method, shown in Fig. 5a and b. A Kapton sheet, a Kapton
a thermal conductivity of 1 W m1 K1 has been sheet with a 10-lm- to 20-lm-thick polyamide film,
measured, which is in very good agreement with a and a Kapton sheet with a 1.1-lm-thick (Bi,Sb)2Te3
result obtained by the laser-flash method by film sputtered on the underside have also been mea-
Netzsch (1.03 W m1 K1). The 3 omega method sured (Fig. 5). The material underneath reduces the
may offer the additional advantage of being able to temperature rise at low frequency (4 Hz). At higher
measure the anisotropy of the thermal conductivity frequency, the thermal wavelength is smaller than
Thermal Conductivity Measurements on Challenging Samples by the 3 Omega Method 1625

Fig. 5. (a) Amplitude and (b) phase measured at room temperature and calculated (c and d) with the data of Table I in the case of the
microheater on a membrane. The materials underneath the membrane are listed in the insert to (a).

the thickness of the Kapton sheet, so the thermal Below 0.1 Hz the thermal wavelength is compa-
wave is not affected by the material underneath. The rable to the size of the opening. There is heat flow
1.1-lm sputtered film cannot be measured accurately along the length of the membrane that is not taken
with the experimental configuration used here into account in the numerical simulation. Heat loss
because of the large difference in thickness between by radiation could also decrease the temperature
the Kapton sheet (50 lm) and the sputtered layer amplitude, since the amplitude measured at low
(1.1 lm). This conclusion is also valid for the poly- frequency is rather high. Obviously, better experi-
amide film. Since the thickness of the flake of nano- mental conditions can be found for this pioneering
composite is comparable to the thickness of the experiment.
Kapton sheet, it decreases the temperature ampli-
tude significantly enough to be measured with good
DISCUSSION OF THE MEASUREMENT
accuracy. Numerical simulations have been per-
OF NANOMETER-THICK FILMS
formed to obtain greater insight into these experi-
mental results (Fig. 5c and d). There is good The effect of the width of a microheater in the
semiquantitative agreement between experiment form of a strip on the measurement accuracy of thin
and simulation for frequencies down to 0.1 Hz. The films can be illustrated by the effect of a change in
thermal conductivity obtained (1.8 W m1 K1) is microheater width on the real part of the complex
just slightly lower than that measured on SPS temperature rise ðDTÞ~ arising from the substrate (1)
material of the same composition.10 and from the film (2)1,11,15:
1626 Jacquot, Vollmer, Bayer, Jaegle, Ebling, and Böttner

 
Pl 1 1 k the form of sintered materials or tiny flakes. Pho-
~
ReðDTÞ ¼  lnð4pf Þ þ ln þ 0:92 ; (1)
pk 2 2 qcp b2 tolithography is no longer needed by the end-user,
and there is no longer any need for electrical insu-
lation directly on measured materials even if the
materials are electrically conducting. This should
Pl  e increase interest in the 3 omega method and enable
ReðDT~F Þ ¼ ; (2)
2b  kF measurement of thermal properties from low to
moderate temperatures and at lower cost. Micro-
where Pl is the heating power per unit length, k is heaters in the form of rings and the numerical tool
the thermal conductivity of the substrate, f is the presented herein may enable measurement of
frequency of the current in the microheater, 2b is extremely thin films or very small samples in the
the strip width, q and cp are the density and heat very near future.
capacity of the substrate, e is the film thickness, and
kF is the cross-plane thermal conductivity of the
ACKNOWLEDGEMENTS
film.
The temperature rises in the substrate and film This work was supported by the German Federal
are functions of the strip width and increase Ministry of Education and Research (BMBF) within
when the strip width decreases. Nevertheless, the the Joint Project VEKTRA and by the Fraunhofer-
increase of temperature in the substrate is a loga- Gesellschaft, Grant Number: WISA 815 020.
rithmic function of the width, whereas the temper-
ature rise in the film is a much stronger (1/2b) REFERENCES
function of the width. As a rule of thumb, the best 1. D.G. Cahill, Rev. Sci. Instrum. 61, 802 (1990).
accuracy is obtained with a narrow strip as long as 2. S.M. Lee and D.G. Cahill, J. Appl. Phys. 81, 2590 (1997).
it is possible to make a narrow strip. By photoli- 3. T. Borca-Tasciuc, R. Kumar, and G. Chen, Rev. Sci. Instrum.
thography, it is possible to make strips with well- 72, 2139 (2001).
4. V. Giraud, J. Cluzel, V. Sousa, A. Jacquot, A. Dauscher,
defined widths equal to or larger than 10 lm. Very B. Lenoir, H. Scherrer, and S. Romer, J. Appl. Phys. 98,
accurate (5%) measurements can be done in this 013520 (2005).
case when the film is no thinner than 100 nm. By 5. B. Feng, W. Ma, Z. Li, and X. Zhang, Rev. Sci. Instrum. 80,
e-beam lithography it is possible to make the strip 064901 (2009).
100 times narrower. Truly nanometer-thick films 6. M. Jaegle and A. Jacquot, Sixth European Conference on
Thermoelectrics, July 2–4, Paris (2008), pp. 2–11.
should be measurable with very good accuracy. If 7. A. Jacquot, B. Lenoir, A. Dauscher, M. Stölzer, and
this has not been done up to now, it is because it was J. Meusel, J. Appl. Phys. 91, 4733 (2002).
not possible to make the microheater long enough for 8. A. Jacquot, G. Chen, H. Scherrer, A. Dauscher, and
the heat transfer to be two dimensional. In addition, B. Lenoir, Sens. Actuators A Phys. 117, 203 (2005).
9. D.G. Ebling, A. Jacquot, M. Jägle, H. Böttner, U. Kühn, and
the electrical resistance of the microheater would L. Kirste, Rapid Res. Lett. 1, 238 (2007).
have been much too high. A microheater in the form 10. D.G. Ebling, M. Jaegle, M. Bartel, A. Jacquot, and
of a ring can be easily made by e-beam lithography, H. Böttner, J. Electron. Mater. doi:10.1007/s11664-009-0825-0
and the measurement results can be easily inter- (2009).
preted with the numerical tool presented in this 11. A. Jacquot, J. Konig, and H. Bottner, 25th International
Conference on Thermoelectrics, Wien, 6–10 August (2006),
article. The resistance of the microheater may be p. 184.
easily tuned by varying the diameter of the ring. 12. B. Yang, J.L. Liu, K.L. Wang, and G. Chen, Appl. Phys. Lett.
80, 1758 (2002).
13. N.O. Birge and S.R. Nagel, Rev. Sci. Instrum. 58, 1465
CONCLUSIONS (1987).
14. I.K. Moon, Y.H. Jeong, and S.I. Kwun, Rev. Sci. Instrum. 67,
It has been demonstrated experimentally that the 29 (1996).
3 omega method can be used to measure the ther- 15. D.G. Cahill, M. Katiyar, and J.R. Abelson, Phys. Rev. B 50,
mal properties of bulk thermoelectric materials in 6077 (1994).