Analytical solutions of the heat diffusion equation for 3

omega method geometry

Jean Yves Duquesne, D. Fournier, Christian Frétigny

To cite this version:

Jean Yves Duquesne, D. Fournier, Christian Frétigny. Analytical solutions of the heat diffusion
equation for 3 omega method geometry. Journal of Applied Physics, 2010, 108 (8), pp.086104.
�10.1063/1.3486441�. �hal-00544156�

HAL Id: hal-00544156

https://hal.science/hal-00544156v1
Submitted on 7 Dec 2010

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
Analytical solutions of the heat diffusion equation for 3ω method
geometry

J.-Y. Duquesne1 , D. Fournier1 , and C. Frétigny2

1
Institut des NanoSciences de Paris, UMR 7588, CNRS,
UPMC, 140, rue de Lourmel, F-75015 Paris - France
2
laboratoire de Physico-Chimie des Polymères et des Milieux Dispersés,
UMR 7615, UPMC, CNRS, ESPCI,
10 rue Vauquelin, F-75005 Paris, France

Abstract
“3ω” experiments aim at measuring thermal conductivities and diffusivities. Data analysis relies
on integral expressions of the temperature. In this paper, we derive new explicit analytical formu-
lations of the solution of the heat diffusion equation, using Bessel, Struve and Meijer-G functions,
in the 3ω geometry for bulk solids. These functions are available in major computational tools.
Therefore numerical integrations can be avoided in data analysis. Moreover, these expressions
enable rigorous derivations of the asymptotic behaviors. We also underline that the diffusivity
can be extracted from the phase data without any calibration while the conductivity measurement
requires a careful one.

1
 The so-called 3ω method is a well-known method for the measurement of thermal conduc-
tivity in bulk and thin film solids [1] [2] [3] [4] [5]. A finite-width metallic strip is designed
on the surface of the sample and acts both as a heater and a temperature sensor. It is fed
with an AC-current. The heating power has an AC component which generates a thermal
pseudo-wave in the sample, at double frequency. The temperature of the strip and, con-
sequently, its resistance are then modulated at double frequency. An harmonic voltage at
triple frequency results in and is used to measure both amplitude and phase of the pseudo-
thermal wave, at the surface. For that purpose, a careful calibration of the setup is required.
From these values, recorded as a function of frequency, both thermal conductivity and dif-
fusivity can in principle be derived. The theoretical basis for the analysis of the data relies
on an integral expression of the temperature. Approximate expressions have been derived in
the low frequency [2] and high frequency regimes [6]. Series expansions and semianalytical
solutions of the temperature have been put forward recently in the case of a finite thickness
strip on a finite solid [7] [8]. In the present paper, we consider a semi-infinite solid and
we derive analytical expressions for the temperature distribution at the surface and for the
mean temperature of the metallic strip (neglecting the strip thickness and thermal boundary
resistance). This formulation uses analytical functions which are all convenient for compu-
tation. This greatly simplifies the fitting of experimental data and extraction of thermal
parameters since numerical integration is avoided. Moreover, a rigorous derivation of the
asymptotic behaviors can be performed and constants which are usually derived numerically
are expressed as a function of fundamental mathematical constants.
The theoretical analysis of the “3ω” method stems from the infinitely narrow line source
of heat, located on the planar surface of a semi-infinite medium [2, 9]. In that case, the
temperature reads:
P
T0 (r) =
K0 (qr) (1)
πΛ
p
where q = iω/D. Λ and D are the thermal conductivity and diffusivity of the solid,
respectively. P is the magnitude of the power supplied per unit length, at angular frequency
ω (twice the current angular frequency). r is the distance from the line. K0 is the zeroth-
order modified Bessel functions of second kind.
T0 (r) can be considered as a Green function for the problems of surface heating invariant
along one axis of the surface. For the 3ω method, let us consider such a source, homogeneous
in an infinite strip parallel to the y-axis and located at the surface (−b < x < b) (see figure

2
 FIG. 1: (Color online) Heat source geometry (red strip) and coordinate axis.

1). The temperature on the surface reads:

+b
P
Z
T (x) = K0 (q|x − ξ|) dξ (2)
2bπΛ −b

This expression is usually written as a convolution product and is calculated thanks to

Fourier transforms [2]. This leads to an integral expression of the temperature at the surface:
Z +∞
P sin(bλ)
T (x) = p e+iλx dλ (3)
2bπΛ −∞ λ λ2 + iω/D
3ω experiments are sensitive to the mean temperature hT i of the strip. hT i can be estimated
by averaging the previous expression across the line [2]. We get:
Z +∞
P sin2 y 1
hT i = p dy (4)
πΛ 0 y 2
y + iΩ
2

where Ω is the reduced circular frequency:

b2 ω
Ω= (5)
D
Indeed, the integral expression (3) can be replaced with an analytical expression. Equa-
tion (2) can be written using Bessel and Struve functions, thanks to the following relation
derived from recurrence relations on Bessel functions and from an integral tabulated in [10]:
πz
Z
K0 (z)dz = [K0 (z)L−1 (z) + K1 (z)L0 (z)] (6)
2
Kn is the nth-order modified Bessel functions of second kind. Ln is the nth-order modified
Struve function. We define the Ξ function as:

Ξ(z) = K0 (z)L−1 (z) + K1 (z)L0 (z) (7)

We get:
P h p 
T (x) = (1 − X) Ξ iΩ(1 − X)2 (8)
4Λ p i
+ (1 + X) Ξ iΩ(1 + X)2

3
FIG. 2: (Color online) Temperature of the strip: Magnitude of (4Λ/P )T (x) versus the reduced
coordinate x/b, for different reduced circular frequencies Ω. Equation (8).

FIG. 3: Temperature of the strip: Phase of T (x) versus the reduced coordinate x/b, for different
reduced circular frequencies Ω. Equation (8).

where X is the reduced coordinate.

x
X= (9)
b
At the center of the strip (x = 0):
P √ 
T (0) = Ξ iΩ (10)
2Λ

Figures 2 and 3 display the magnitude and phase of T (x), computed with equation (8)
[11]. Indeed, we have plotted the reduced magnitude (4Λ/P )T (x) and the phase of T (x)
versus the reduced coordinate X, for different values of the reduced circular frequency Ω.
The integral expression (4) of the mean temperature can also be replaced with an ana-
lytical expression. Starting from equation (2), the average temperature of the line can be

4
 6

5
In-phase

Strip mean temperature

4

0
Out-of-phase
-1
0.001 0.01 0.1 1 10 100 1000
Reduced circular frequency Ω

FIG. 4: (Color online) Strip mean temperature versus reduced circular frequency Ω. In-phase and
out-of-phase components of (4Λ/P ) < T >. Solid line: full expression. Equation (14). Dashed line:
low frequency approximation. Equation (15).

written as:
2b 2b
P
Z Z
hT i = 2 K0 (q|u − v|) dudv (11)
4b πΛ 0 0

The integrand is invariant under u and v interchange. By symmetry, the integral is twice
the integral evaluated on a rectangular triangle:
Z 2b Z u
P
hT i = 2 du K0 (q(u − v)) dudv (12)
2b πΛ 0 0

After variables substitutions (t = u/2b and z = (u − v)/u), we get:

2P 1 1
Z Z
hT i = tK0 (rzt)dtdz (13)
πΛ 0 0
√
where r = 2qb = 2 iΩ. This double integral is related to a Meijer G-function [12]. We get:

iP 
1, 32

hT i = − G22 iΩ (14)
4πΛΩ 24 1,1, 12 ,0

Solid lines in figures 4 and 5 display the in-phase component, out-of-phase component,
amplitude and phase of the reduced mean temperature (4Λ/P )hT i, versus reduced circular
frequency, computed with equation (14) [11].
Phase frequency dependance can be of interest since this measurement does not require
any absolute calibration of the setup but provides an easy way to measure the sample
diffusivity. Indeed, as it is clear from the temperature expressions above, phase of the
temperature is a universal function of the only reduced frequency Ω. Therefore, the sample

5
 6 0

Strip mean temperature - Amplitude

Strip mean temperature - Phase ( ° )

5
-10

4
-20
3
Phase
-30
Amplitude
2

-40
1

0 -50
0.001 0.01 0.1 1 10 100 1000
Reduced circular frequency Ω

FIG. 5: (Color online) Strip mean temperature versus reduced circular frequency Ω. Solid line:
phase of < T >. Dashed line: amplitude of (4Λ/P ) < T >. Equation (14).

diffusivity can be extracted with a simple logarithmic shift of the phase curve in figure 5,
along the abscissae, to fit the experimental data.
In both low and high frequency regimes, we can derive useful approximate expressions for
T (x) and hT i, using asymtotic expansions of the the Bessel and Struve functions. Asymptotic
analysis of equation (13) gives:
 
P 1 3 π
hT i ≃ − ln Ω + − γ − i (15)
πΛ 2 2 4

for ω → 0 (i.e ω << D/b2 ), and

P
hT i ≃ √ (1 − i) (16)
2Λ 2Ω
for ω → +∞ (i.e. ω >> D/b2 ).
γ ≃ 0.5772 is the Euler constant. Equation (15) is well known and is extensively used
to measure thermal conductivities. However, we notice the real constant term (3/2 − γ ≃
0.9228) which is here derived explicitly for the first time. To the best of our knowledge, its
approximate value was derived previously numerically.
p
In the high frequency limit, we may notice that the thermal diffusion length 2D/ω
becomes much smaller that the heater width. Then, the above result is equivalent to the
surface temperature of a uniformly heated semi-infinite medium [9]. However, in this fre-
quency region, thermal behavior can be affected by the thermal load of the heater [7].
We have revisited the theoretical backgrounds of the so-called “3ω” method, in the case of
a semi-infinite medium with a finite width but vanishing thickness heater. We have derived
analytical expressions of the surface temperature and of the mean temperature of the strip

6
line. The main benefits are: (1) simplification of experimental data analysis since numerical
integration is avoided; the involved Bessel, Struve and Meijer-G functions are implemented
in major calculation software and can be easily computed (2) rigorous derivation of the
asymptotic regimes at both low and high frequencies.

[1] D. G. Cahill and R. O. Pohl. Phys. Rev. B, 35:4067, 1987.

[2] D. G. Cahill. Rev.of Sci.Instr., 61:802, 1990.
[3] T. Yamane, N. Nagai, S. Katayama, and M. Todoki. J. Appl. Phys., 91:9772, 2002.
[4] C. E. Raudzis, F. Schatz, and D. Wharam. J. Appl. Phys., 93:6050, 2003.
[5] J.-Y. Duquesne. Phys. Rev. B, 79:153304, 2009.
[6] I. K. Moon, Y. H. Jeong, and S. I. Kwun. Rev.of Sci.Instr., 67:29, 1996.
[7] H. Wang and M. Sen. Int. J. of Heat and Mass Transfer, 52:2102, 2009.
[8] S. P. Gurrum, W. P. King, and Y. K. Joshi. J. Appl. Phys., 103:113517, 2008.
[9] H. S. Carslaw and J. C. Jaeger. Conduction of Heat in Solids. Oxford University Press,
Oxford, 1959.
[10] M. Abramowitz and I. A. Stegun. Handbook of Mathematical Functions with Formulas,
Graphs, and Mathematical Tables, 9th ed. (Dover, New York, 1964)
[11] Computation was performed with mpmath, Python library for arbitrary-precision floating-
point arithmetic (http://code.google.com/p/mpmath/). Results were also reproduced using
commercial computational tools Mathematica (Wolfram Research Inc.) and Maple (Waterloo
Maple Inc.).
[12] From the Mellin inversion formula expressing K0 (rzt) we have computed directly the double
integral and after simplification we obtain an integral form which corresponds to the Meijer
 
22 1, 32
definition formula (9.301) of G24 x 1,1, 1 ,0 in [13].
2

[13] I. S. Gradshteyn and I. M. Ryzhik. in Table of Integrals, Series, and Products, 7th ed., edited
by A. Jeffrey and D. Zwillinger (Elsevier, Academic, Amsterdam, Boston, Heidelberg, London,
New York, Oxford, Paris, San Diego, San Francisco, Singapore, Sydney, Tokyo, 2007).