Abstract

The transient plane source (TPS) is a well-known method for measuring the thermal conductivity, thermal diffusivity, and specific heat capacity of solids, liquids, pastes, and powders. The method assumes that the heat source delivers a constant power throughout the measurement. However, in real-world measurements, the effective power is reduced due to the temperature coefficient of resistance (TCR) and the heat capacity of the sensor. In this paper, we present first-order corrections to the average temperature of the sensor, accounting for the power reduction caused by the TCR and the heat capacity of the TPS sensor. To verify the corrections, TPS measurements were performed on Pyroceram 9606 at multiple initial powers. The measured properties without the corrections showed a significant trend with the power used. The first-order corrections significantly decreased the trend of the measured thermal conductivity with power, validating the improvements the corrections make to the TPS method.

1 Introduction

The transient plane source (TPS) is a well-known method for measuring the thermal conductivity, thermal diffusivity, and specific heat capacity of solids, liquids, pastes, and powders. The method makes use of a resistive element acting as a heat source and a temperature sensor. This resistive element, also known as the TPS sensor, is shaped into a bifilar spiral and is enclosed in an electrically insulating sheet. At the start of the measurement, the TPS sensor is placed in between two large, identical, homogeneous samples in thermal equilibrium. The TPS sensor undergoes Joule heating upon the application of electrical current. At the same time, the TPS sensor records its own temperature. The thermophysical properties of the samples can be obtained from the temperature transient recorded by the TPS sensor with Gustafsson's mathematical model [1,2].

It was pointed out by Gustavsson and Gustafsson [3] that the power delivered by the TPS is not constant due to factors such as power loss in the electrical leads, the temperature coefficient of resistance (TCR) of the TPS sensor, the heat capacity of the sensor, etc. In this paper, we present a semi-analytical expression for the average temperatures of the TPS sensor with a first-order correction to the delivered power. The first-order correction includes the effects of the sensor heat capacity and the TCR of the sensor. To verify the correction, TPS measurements were made on Pyroceram 9606. Pyroceram 9606 was selected because it can be obtained from the market easily, is made with high quality, and has a measured thermal conductivity value reported in literature [4]. The measured properties of Pyroceram 9606 using Gustafsson's model and the corrected model were then compared over different initial powers.

2 Theoretical Framework

Suppose the TPS sensor is made up of m concentric, equally spaced rings, with outermost radius a, placed at the origin. To simplify the derivation, we assume that m>10 so that the TPS can be treated as a disk source instead of a concentric ring source [1]. Suppose that this sensor is placed in an infinite, homogeneous medium of thermal conductivity Λ and thermal diffusivity κ. Since the source is made up of concentric rings, the system is expected to have azimuthal symmetry. Thus, the temperature distribution T=T(r,z,t) follows the heat equation [5]:
(1)
where Q(r,z,t) is the source term. The system is initially kept at a constant temperature T0, which can be represented by the initial condition T(r,z,t=0)=T0. It is convenient to make the substitution ΔT(r,z,t)=T(r,z,t)T0, which represents the change in temperature from time t=0 on Eq. (1), giving
(2)
where ΔT(r,z,t=0)=0. We apply the disk approximation on the source term by writing the equation
(3)
where P0 is the constant power delivered by the sensor, δ(z) is the delta function, and u(t) is the Heaviside step function. From Eqs. (2) and (3), the average temperature of the sensor
(4)
can be obtained, where t0 is the required time offset parameter for a disk source approximation. The explicit solution for ΔT(r,z,t) is given by Carslaw and Jaeger [5], and ΔTave(t) is provided in Gustafsson's work as [1]
(5)

where I0 is the modified Bessel function of the first kind of order zero.

2.1 Power Variations Due to the Sensor Temperature Coefficient of Resistance and Sensor Heat Capacity.3

To correct for the power variations on the average temperature rise ΔTave(t), the constant power P0 in Eq. (3) must be replaced with a power function P(t), defined as
(6)
where Pel(t) is the contribution from the electrical circuit and PSHC(t) is the contribution from the sensor heat capacity. According to literature, the power contribution from the sensor heat capacity can be approximated as [2,3]
(7)

where P0 is the initial power delivered to the TPS sensor and C is the sensor heat capacity in Joules per Kelvin. The minus sign indicates that a nonzero heat capacity reduces the power delivered to the TPS sensor to heat up the samples.

On the other hand, to calculate for Pel(t), we refer to the diagram in Fig. 1. The TPS sensor, with electrical resistance R(t), is connected in series with a reference resistor with constant resistance RREF with two lead resistances to the sensor, each RL/2. The voltage V across the circuit is
(8)
Fig. 1
Diagram for a constant voltage circuit
Fig. 1
Diagram for a constant voltage circuit
Close modal
which is kept constant throughout the measurement. Thus, the power Pel(t) delivered to the TPS sensor is given by
(9)
Let R(t)=R0+ΔR(t), where R0 is the resistance of the TPS sensor at time t=0, and ΔR(t) is its change in resistance from R0 at time t. Equation (9) then becomes
(10)
For a TPS measurement, the condition RREF+RL+R0ΔR(t) holds true. Setting x=RREF+RL+R0 and ε=ΔR(t), the denominator of Eq. (10) can be rewritten as (x+ε)2=x2(1+ε/x)2. Since ε/x is small, the denominator can be expanded with a binomial series expansion
(11)
Substituting Eq. (11) to Eq. (10) and replacing x and ε with their definitions gives
(12)
Since terms of order (ΔR)2 are negligible, Eq. (12) simplifies to
(13)
Equation (13) can be written as a sum of a constant term and a time-varying term, i.e.
(14)
where the constant term P0 is the initial power delivered to the TPS sensor, and ΔPel(t) is the power variation due to the change in sensor resistance over time. If RREF+RL=R0, the power variationΔPel(t)=0, consistent with the result in Ref. [3]. However, in most circuits, particularly commercial systems, which have fixed reference resistors and predesigned sensors, the condition RREF+RL=R0 is rarely met. Thus, the power variation in Eq. (14) must be used to get the average temperatures of the TPS sensor. Since the change in resistance ΔR(t) can be expressed in terms of the average temperature rise of the sensor according to the equation
(15)
then, the power variation in Eqs. (13) and (15) can be written as
(16)
while the initial power P0 from Eq. (13) is given by
(17)
Thus, Eq. (16) can be expressed in terms of P0
(18)
Substitution of Eqs. (7) and (15)(18) to Eq. (6) gives the overall power variation in the TPS sensor
(19)
where β=α(RREF+RLR0)/(RREF+RL+R0).

2.2 Obtaining Sensor Temperatures From the Power Variation P(t).

Applying the correction in Eq. (19) to Eqs. (1) and (2) gives
(20)
where P(t) is defined in Eq. (19). Taking the Laplace and Hankel transform of Eq. (20) gives
(21)
where σ is the Hankel variable, s is the Laplace variable, and q=s/κ. The solution to Eq. (21) at the sensor location, z=0, is given by
(22)
Taking the inverse Hankel transform of Eq. (22) and averaging over r using Eq. (4) gives
(23)
where Δθ¯ave(s) is the Laplace transform of the average TPS sensor temperatures without any corrections
(24)

where I1(x) is the modified Bessel function of the first order, and L1(x) is the modified Struve function of the first order.

The average temperature rise of the sensor ΔTave(t) can be obtained from Eq. (23) by taking its numerical inverse Laplace transform. In this paper, de-Hoog's algorithm is used, which treats the inverse Laplace transform as a Fourier transform. The convergence of the Fourier transform is then accelerated by a nonlinear double acceleration with a Padé approximation [6,7].

2.3 Measurement Procedure.

During a measurement, the TPS records the timestamps ti with the corresponding measured temperatures Ti. Thus, a list of pairs (ti,Ti) is obtained, where i200 [2]. The measurement also records the reference resistance RREF, the initial resistance R0, the sensor TCR α, and the heat capacity of the sensor C. To get the thermal conductivity, λ, and thermal diffusivity, κ, the sum-of-squares, S, is minimized
(25)

where β and C are fixed parameters determined before the start of the measurement, t0 is the time offset parameter introduced in Sec. 1, and Tf is the temperature offset added in the model to account for contact resistance [1]. To minimize Eq. (25), algorithms such as the Levenberg-Marquardt algorithm can be used to find the best-fit parameters t̂0, T̂f, Λ̂, and κ̂ [8].

3 Experimental Validation

To verify the new model, TPS measurements were performed on Pyroceram 9606 with Thermtest Instruments TPS 3317 sensor, a heat source made up of 16 concentric rings and a radius of 6.4 mm (Fig. 2). Pyroceram 9606 is used because it has a known thermal conductivity of 3.913 Wm−1K−1 at room temperature (23 °C) [4]. The measured TPS sensor temperatures were collected using Thermtest Instruments Measurement Platform-1. Measurements on Pyroceram 9606 were performed at multiple initial powers, from 200 to 1200 mW, with increments of 100 mW, all performed at room temperature (23 °C). Each measurement was 10 s long. The sample dimensions were large enough that the probing depth of the measurement at 10 s does not exceed the sample boundaries [1]. The thermal properties were then obtained from the sensor temperatures using Gustafsson's model [1] and the improved model in Eq. (24). The disk approximation used in Eq. (24) is valid since the TPS sensor has greater than 10 rings [1].

Fig. 2
(a) The Thermtest TPS 3317 sensor and (b) the Thermtest Instruments Measurement Platform-1
Fig. 2
(a) The Thermtest TPS 3317 sensor and (b) the Thermtest Instruments Measurement Platform-1
Close modal

4 Results

The plot of thermal conductivity (and thermal diffusivity) plotted against power is displayed in Fig. 3. The cross marks represent measurements using Gustafsson's model, while the gray circles represent measurements using the model in Eq. (23). In the thermal conductivity plot, the dashed line represents the reference value reported in Ref. [4]. Notice that there is little to no difference in the measured thermal properties with Gustafsson's model and Eq. (23) at the lowest power (200 mW). However, as the initial power is increased, the measured thermal properties exhibit a trend with respect to the power applied. To better quantify this trend, the percent deviation in measured thermal conductivity against the reference value of 3.913 Wm−1K−1 [4] is plotted in Fig. 4. For measurements performed at 200 mW, the percent deviation is low (below 0.5%), but as the initial power increases, the measured thermal conductivity becomes larger relative to the reference value. Without any power corrections, the percent deviation in thermal conductivity at 1200 mW is 3.15%, compared to 2.3% when a power correction is applied. The change in conductivity with power was measured to be 0.18 Wm−1K−1 per watt (4.7% per watt) without this correction, and 0.095 Wm−1K−1 per watt (2.4% per watt) with this correction, nearly a 50% improvement.

Fig. 3
(a) Plot of measured thermal conductivity λ and (b) plot of thermal diffusivity κ against the initial power of the TPS P0
Fig. 3
(a) Plot of measured thermal conductivity λ and (b) plot of thermal diffusivity κ against the initial power of the TPS P0
Close modal
Fig. 4
Percent deviation of measured thermal conductivity with power
Fig. 4
Percent deviation of measured thermal conductivity with power
Close modal

The power correction is a modest correction on the measured thermal conductivity. However, it does not improve the thermal diffusivity measurements, as seen in Fig. 3(b). At the lowest power (200 mW), the thermal diffusivity readings are similar. However, as the power is increased, the measured thermal diffusivity increases with Gustafsson's model, while it decreases with the power correction model. The measured thermal diffusivity at 1200 mW is 1.4% higher relative to the 200 mW measurement using Gustafsson's model, while the measured thermal diffusivity using the power correction model decreased by −1.8% relative to the 200 mW measurement.

5 Conclusion

A more accurate description of the TPS sensor is presented in this work, accounting for the change in power due to the TCR and the finite heat capacity of the sensor. The corrections were verified with TPS measurements on Pyroceram 9606. Using the power correction model made the thermal conductivity readings less sensitive to the initial power used relative to Gustafsson's model, where a constant power is assumed.

Funding Data

  • National Research Council of Canada through their Industrial Research Assistance Program (IRAP), with project code NRC-IRAP (No. 928449; Funder ID: 10.13039/501100000046).

Conflict of Interest

David Landry and Renzo Flores are employed by Thermtest Inc. and as such have beneficial interest in the work.

Data Availability Statement

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.

Nomenclature

a =

TPS sensor radius (m)

C =

sensor heat capacity (J/m2K)

I(t) =

electric current passing through the TPS sensor (A)

I1(x) =

modified Bessel function of order 1

J1(x) =

Bessel function of order 1

L1(x) =

modified Struve function of order 1

m =

number of concentric rings on the TPS sensor

P0 =

initial power delivered by the TPS sensor (W)

P(t) =

power delivered by the TPS sensor (W)

q =

s/κ (m−1)

Q(r, z, t) =

source term

r =

radial coordinate (m)

RL =

total electrical resistance of the leads to the sensor (Ω)

RREF =

electrical resistance of the reference resistor (Ω)

R0 =

initial electrical resistance of the TPS sensor (Ω)

R(t) =

electrical resistance of the TPS sensor (Ω)

s =

Laplace variable (s−1)

t =

time (s)

ti =

ith timestamp reading (s)

t̂0 =

fitted time offset parameter (s)

T0 =

initial temperature of the system (K)

Tf̂ =

fitted temperature offset parameter (K)

Ti =

ith temperature reading (K)

T(r, z, t) =

temperature of the system at coordinates (r, z) at time t (K)

u(t) =

Heaviside step function

U =

voltage across the bridge (V)

Greek Symbols
α =

temperature coefficient of resistance (K−1)

β =

α (RREFR0)/(RREF + R0) (K−1)

δ(z) =

delta function (m−1)

ΔP(t) =

P(t) – P0 (W)

ΔR(t) =

R(t) – R0 (Ω)

Δθ¯ave =

Laplace transform of the TPS temperatures for a constant power P0 (K s)

ΔT =

T(r,z,t) – T0 (K)

ΔTave =

average temperature of the heat source (K)

ΔT¯ =

Laplace transform of ΔT(r,z,t) (K s)

κ =

thermal diffusivity (m2/s)

κ̂ =

fitted thermal diffusivity (m2/s)

Λ =

thermal conductivity (Wm−1K−1)

Λ̂ =

fitted thermal conductivity (Wm−1K−1)

σ =

Hankel variable (m−1)

Footnotes

3

The treatment here is similar to Gustavsson and Gustafsson [3], but with RS=RREF.

References

1.
Gustafsson
,
S. E.
,
1991
, “
Transient Plane Source Techniques for Thermal Conductivity and Thermal Diffusivity Measurements of Solid Materials
,”
Rev. Sci. Instrum.
,
62
(
3
), pp.
797
804
.10.1063/1.1142087
2.
ISO 22007-2
,
2015
, “
Plastics — Determination of Thermal Conductivity and Thermal Diffusivity — Part 2: Transient Plane Heat Source (Hot Disc) Method
,” ISO, Geneva, Switerzland, accessed Jan. 15, 2025, https://www.iso.org/standard/81836.html
3.
Gustavsson
,
M. K.
, and
Gustafsson
,
S. E.
,
2003
, “
On Power Variation in Self-Heated Thermal Sensors
,”
Proceedings of the 27th International Thermal Conductivity Conference and the 15th International Thermal Expansion Symposium
, Knoxville, TN, Oct. 26–29, pp.
338
345
.https://books.google.se/books?hl=en&lr=&id=ZXOU8HI8JZIC&oi=fnd&pg=PA338&dq=%22On+Power+Variation+in+Self-Heated+Thermal+Sensors%22&ots=AlAg-ULWOA&sig=Ndp1TFoS8XfqlI-1Y47HWimkKkw&redir_esc=y#v=onepage&q=%22On%20Power%20Variation%20in%20Self-Heated%20Thermal%20Sensors%22&f=false
4.
Antoniadis
,
K. N.
,
Tyrou
,
A.
,
Assael
,
M. J.
,
Li
,
X.
,
Wu
,
J.
, and
Ebert
,
H.-P.
,
2020
, “
Reference Correlations for the Thermal Conductivity of Solid BK7, PMMA, Pyrex 7740, Pyroceram 9606 and SS304
,”
Int. J. Thermophys.
,
41
(
7
), pp.
1
35
.10.1007/s10765-020-02678-9
5.
Carslaw
,
H.
, and
Jaeger
,
J.
,
1959
,
Conduction of Heat in Solids
,
Clarendon Press
,
Oxford, UK
.
6.
Kuhlman
,
K. L.
,
2013
, “
Review of Inverse Laplace Transform Algorithms for Laplace-Space Numerical Approaches
,”
Numer. Algorithms
,
63
(
2
), pp.
339
355
.10.1007/s11075-012-9625-3
7.
de Hoog
,
F. R.
,
Knight
,
J. H.
, and
Stokes
,
A. N.
,
1982
, “
An Improved Method for Numerical Inversion of Laplace Transforms
,”
SIAM J. Sci. Stat. Comput.
,
3
(
3
), pp.
357
366
.10.1137/0903022
8.
Nocedal
,
J.
, and
Wright
,
S.
,
2006
,
Numerical Optimization
,
Springer Science & Business Media
, New York.