OPEN ACCESS
The purpose of this paper is to investigate a steadystate method enabling the measurement of the thermal conductivity of an electric conducting wire. Instead of varying the pulsation of a periodic current as done in the 3ω method, the proposed method is based on steadystate measurements realized for different lengths of the same wire. The model developed takes into account an external heat transfer coefficient (conduction + radiation) often neglected in previous works. A sensitivity analysis shows the possibility of estimating simultaneously the wire thermal conductivity and the external heat transfer coefficient. The experimental results obtained with Chromel and Copper wires are in good agreement with references values (less than 5% deviation).
wire, thermal conductivity, heat transfer coefficient, steady state, estimation method
The measurement of thermal properties of wires (or fibers) is essential when there are used in temperature sensors (as thermocouples for example), superconducting materials [1] or microelectronic devices.
Figure 1. Schema of the method used for thermal conductivity measurement of an electric conducting wire
Most of the methods used for the estimation of the thermal conductivity of electric conducting wires are based on the typical experimental device represented in Figure 1: The two extremities of a wire of length L are kept at a constant temperature T_{a} and they are connected to an electrical generator. An electric current l gets through the wire, producing by Joule effect.
The mean temperature rise in the wire is deduced from the variation of its electrical resistance R measured by a four probe method. It is assumed that the wire electrical resistance varies linearly with its mean temperature as follows:
$R={{R}_{0}}[1+\alpha (\overline{T}{{T}_{a}})]$ (1)
where $\overline{T}$ is the wire mean temperature defined as:
$\bar{T}\text{=}\frac{1}{L}\int _{0}^{L}T(x)dx$ (2)
α is the temperature coefficient of the electrical resistance and R_{0} is the electrical resistance of the wire at the temperature T_{a}. The heat produced inside the wire is transferred to its environment by two ways:
The methods based on the previously described device may be classified in two types:
The proposed method intends to conserve the simplicity of the stationary method but without being forced to neglect the external heat transfer. This is made possible by realizing several steadystate measurements for different lengths of the same wire. The idea has already been used by Volkein and Kessler [10] for thin films thermal conductivity measurements but the applicability was not fully demonstrated. This paper will to demonstrate that thermal conductivity of wires may be estimated with a good precision using the proposed method.
The remainder of this paper is organized as follows: Section 2 describes the experimental device and the measurement method. Section 3 presents the modeling of the wire mean temperature rise θ by heat transfer analysis, both with a sensitivity study of this temperature rise θ to the length L of the wire, to its thermal conductivity λ and to the heat transfer coefficient h. Section 4 presents the experimental results obtained with Copper and Chromel wires, confirming the feasibility of thermal conductivity estimation by the proposed method.
The experimental device is represented in Figure 2.
Figure 2. Schema of the experimental device
Both extremities of the wire that must be characterized are sandwiched and pressed between two brass blocks: their high thermal conductivity ensures a uniform temperature in the blocks and their high thermal capacity ensures a constant temperature (equal to the environment temperature T_{a} ) during the experiment. The distance L between the two brass blocks is measured by a caliper with a 0.02 mm precision. The wire extremities are connected to a Tektronik PWS2185 Power Supply that produces (and measures) a constant intensity current I_{0 }through the wire. The electric potential difference between the two brass blocks (adding no electric resistance since the electrical conductivity of brass is high) is measured by a Tektronik DMM4040 Multimeter. To reduce the convective part in the external heat transfer, the device is set in a vacuum chamber where the pressure may be lowered to 3´10^{2} mbar.
The wire means temperature rise $\bar{θ}=\bar{T}T_a$ is related to the stationary value $U_L$ of the electrical potential difference by Eq. (1). Thus, the measurement of $U_L$ and of $I_0$ enable to estimate the wire mean temperature rise by:
$\bar{θ}=\bar{T}T_a=\frac{1}{α}(\frac{U_L}{R_0I_0}1)$ (3)
where:
α=$\frac{k}{R_{20°C}}$ (4)
The expression of $\bar{T}$ as a function of λ, h and L will now be developed and it will be shown that the measurement of U_{L} for several values of the wire length L enables the estimation of the thermal conductivity λ of the wire.
The following hypotheses have been assessed:
Figure 3. Schema of the thermal balance on an infinitesimal part dx of the wire
For modeling the temperature T(x) of the wire, where x is the distance from the middle of the wire, a thermal balance has been realized on the wire part between x and x+dx as represented on figure 3, it may be written:
${{\phi }_{con{{d}_{x}}}}+{{\phi }_{gen}}={{\phi }_{con{{d}_{x}}}}+{{\phi }_{conv}}$ (5)
${{\phi }_{con{{d}_{x}}}}\text{=}\lambda \text{S(}\frac{dT}{dx}\text{)}$ (6)
${{\phi }_{gen}}={{\phi }_{0}}Sdx$ (7)
${{\phi }_{con{{d}_{x}}}}\text{=}\lambda \text{S(}\frac{dT}{dx}{{\text{)}}_{x+dx}}$ (8)
${{\phi }_{conv}}=hPdx[T(x){{T}_{a}}]$ (9)
and: P is the wire perimeter: P = πD (D is the wire diameter)
S is the wire cross section: $S\text{=}\frac{\pi {{D}^{2}}}{4}$
ϕ_{0}(x) is the heat flux per volume unit produced by Joule effect in the wire, given by:
${{\phi }_{0}}(x)=\frac{16{{\rho }_{0}}I_{0}^{2}}{{{\pi }^{2}}{{D}^{4}}}\{1+\alpha [T(x){{T}_{a}}]\}(W{{m}^{3}})$ (10)
where ρ_{0} is the electrical resistivity of the wire (Ω m).
Setting θ(x)=T(x)T_{a}, this system of equations leads to:
$\frac{{{d}^{2}}\theta }{d{{x}^{2}}}\frac{hP}{\lambda S}\theta +\frac{{{\phi }_{0}}}{\lambda }=0$ 0≤x≤L/2 (11)
The boundary conditions are:
$\frac{d\theta }{dx}=0$ (null flux for symmetry reason) at x=0 (12)
θ = 0 at x=L/2 (13)
Eq. (11) can also be written:
$\frac{{{d}^{2}}\theta }{d{{x}^{2}}}(\frac{4h}{\lambda D}\frac{16{{\rho }_{0}}\alpha I_{0}^{2}}{\lambda {{\pi }^{2}}{{D}^{4}}})\theta +\frac{16{{\rho }_{0}}I_{0}^{2}}{\lambda {{\pi }^{2}}{{D}^{4}}}=0$ (14)
If: $\frac{4h}{\lambda D}\frac{16{{\rho }_{0}}\alpha I_{0}^{2}}{\lambda {{\pi }^{2}}{{D}^{4}}}>0$, let us consider:
$K=\frac{16{{\rho }_{0}}I_{0}^{2}}{\lambda {{\pi }^{2}}{{D}^{4}}}$ (15)
And: $\beta _{1}^{2}=\frac{1}{\lambda }(\frac{4h}{D}\frac{16\alpha {{\rho }_{0}}I_{0}^{2}}{{{\pi }^{2}}{{D}^{4}}})$ (16)
Taking into account the boundary conditions Eq. (12) and Eq. (13) the resolution of Eq. (14) leads to:
$\theta (x)=\frac{k}{\beta _{1}^{2}}[1\frac{\cosh ({{\beta }_{1}}x)}{\cosh ({{\beta }_{1}}L/2)}$ (17)
The mean temperature rise can be calculated by:
$\bar{\theta}(x)=\frac{2}{L}\int_0^{L/2}\theta (x)dx$ (18)
And finally:
$\overline{\theta }=\overline{T}{{T}_{a}}=\frac{K}{\beta _{1}^{2}}[1\frac{2\tanh ({{\beta }_{1}}L/2)}{{{\beta }_{1}}L}]$ (19)
If $\frac{4h}{\lambda D}\frac{16{{\rho }_{0}}\alpha I_{0}^{2}}{\lambda {{\pi }^{2}}{{D}^{4}}}<0\\let us consider: \,\,\,\,\beta _{2}^{2}=\frac{1}{\lambda }(\frac{16{{\rho }_{0}}I_{0}^{2}}{{{\pi }^{2}}{{D}^{4}}}\frac{4h}{\lambda D})$ (20)
Taking into account the boundary conditions Eq. (12) and Eq. (13) the resolution of Eq. (14) leads to:
$\theta (x)=\frac{k}{\beta _{2}^{2}}[\frac{\cos ({{\beta }_{2}}x)}{\cos ({{\beta }_{2}}L/2)}1]$ (21)
Eq. (18) and Eq. (21) lead to:
$\bar{\theta }=\bar{T}{{T}_{a}}=\frac{K}{\beta _{2}^{2}}[1\frac{2\tan ({{\beta }_{2}}L/2)}{{{\beta }_{2}}L}1]$ (22)
With the hypotheses that h=0 and that the heat flux ϕ_{0} dissipated in the wire by Joule effect is constant and equal to:
${{\phi }_{0}}=\frac{4UI_{0}^{2}}{\pi {{D}^{2}}L}$ (23)
Eq. (14) becomes:
$\frac{{{d}^{2}}\theta }{d{{x}^{2}}}+\frac{{{\phi }_{0}}}{\lambda }=0$ (24)
Let us consider: ${{K}_{1}}=\frac{{{\phi }_{0}}}{\lambda }$ (25)
The resolution of Eq. (24) taking into account the boundary conditions Eq. (12) and Eq. (13) lead to:
$\theta (x)={{K}_{1}}(\frac{{{L}^{2}}}{4}\frac{{{x}^{2}}}{2})$ (26)
And finally: $\bar{\theta }=\bar{T}{{T}_{a}}=\frac{1}{3}\frac{U{{I}_{0}}L}{\lambda \pi {{D}^{2}}}$ (27)
Then the thermal conductivity λ of the wire can be deduced by:
$\lambda =\frac{1}{3}\frac{U{{I}_{0}}L}{\pi {{D}^{2}}\overline{\theta }}$ (28)
Several authors such as Moon [2] estimate the thermal conductivity l by using Eq. (28). It will be further shown than in some cases using Eq. (28) may lead to an estimation error up to several percents so that it is always better to use Eq. (19) or Eq. (22).
The reduced sensitivities of the temperature $\overline{T}(L)$ to the parameters λ and h are defined as $\lambda \frac{\partial \overline{T}}{\partial \lambda }(L)$ and $h\frac{\partial \overline{T}}{\partial \lambda }(L)$. If they are not proportional on a wire length interval [L_{1}; L_{2}] then it will be possible to estimate separately λ and h from several mean temperature measurements realized for different wire lengths in this interval [11].
The values of $\lambda \frac{\partial \overline{T}}{\partial \lambda }(L)$ and $h\frac{\partial \overline{T}}{\partial \lambda }(L)$have been calculated numerically by Eq. (19) or Eq. (22), for example:
$\lambda \frac{\partial \overline{T}}{\partial \lambda }(L)=\lambda \frac{\overline{T}(1.001\lambda ,h,L)}{0.001\lambda }$ (29)
The three following practical cases (corresponding to further experiments) have been analyzed:
The following thermal properties have been considered in the simulations:
The calculations have been done with an external heat transfer coefficient h=2 W m^{2} K^{1} corresponding to an experiment under vacuum where convection may be neglected and with a wire emissivity around 0.3.
Figures 4 to 6 represent for each material the following parameters as a function of the wire length L:
Figures 4 to 6 show that the precision of Eq. (28) (in which the external heat transfer is neglected) decreases when the wire length or its diameter increases. One can also see on figure 6 that for a Copper wire with a length lower than 2 cm the sensitivity of the mean temperature to the thermal conductivity λ is at least thirty times greater than its sensitivity to the external heat transfer coefficient h. It is thus possible to fix h to an approximate value and then to deduce l solving Eq. (19) or Eq. (22) from a single temperature measurement realized with a wire length lower than 2 cm.
Concerning the thinner Chromel wire, the length should be lower than 1 cm to obtain a ratio of the sensitivities to λ and to h to greater than 30. The uncertainty measurement on a wire with a length lower than 1 cm may become significant so that in this case it is better to realize several measurements with different wire lengths L_{i} to estimate separately the thermal conductivity λ and the external heat transfer h. This separate estimation is possible because the ratio of the sensitivities represented in Figure 4 and 5 is not constant when the wire length varies. The estimation will be realized by minimizing the sum of the quadratic difference defined as:
$S\text{=}\sum\limits_{i=1}^{n}{{{[{{\overline{T}}_{\exp }}({{L}_{i}}){{\overline{T}}_{\bmod }}({{L}_{i}})]}^{2}}}$ (30)
where n is the number of wire lengths L_{i} tested, using the LevenbergMarquart algorithm [12].
Figure 4. Mean temperature rise and reduced sensitivities of $\overline{T}$ for a Chromel wire with a diameter D=0.254 , (I_{0}=0.06 A)
Figure 5. Mean temperature rise and reduced sensitivities of $\overline{T}$ for a Chromel wire with a diameter D=0.127 mm (I_{0}=0.015)
Figure 6. Mean temperature rise and reduced sensitivities of $\overline{T}$ for a Copper wire with a diameter D=0.0254 mm (I_{0}=0.02 A)
First, the temperature coefficients for the electrical resistance of each wire were measured. The experimental results are: α_{Copper}=3.88×10^{3} K^{1} and α_{Chromel}=3.26×10^{3} K^{1}. The first value is very close to the value α_{Copper}=3.93×10^{3} K^{1} given by Fallou [13].
Eq. (15) shows that K is proportional to D^{4}. Thus, the relative standard deviation of K would be four times the standard deviation of D if Eq. (15) was used to calculate K. Since the measurement of the electrical resistance of a thin wire is much more accurate than the measurement of its diameter, the resistance per length unit $\frac{R}{L}$ of each wire has been measured and then K has been calculated by:
$K\text{=}\frac{16(1+\alpha \theta )I_{0}^{2}}{\lambda \pi {{D}^{2}}}\frac{R}{L}$ (31)
Moreover, when using Eq. (31) instead of Eq. (15), it is no longer necessary to know the value of the electrical resistivity ρ_{0} of the wire to calculate K.
Table 1 presents the measured values for each of the three wires.
The mean temperature $\overline{T}$ of a Copper wire with a length L=2.1 cm and a diameter D=0.025 mm has been measured using the experimental device described in Figure 1, under an air pressure of 3×10^{2} mbar. The measurement has been repeated three times.
Table 1. Experimental values of the linear resistance of the wires

Chromel 
Chromel 
Copper 
D[mm] 
0.127 
0.254 
0.025 
L [m] 
1.044 
1.070 
0.200 
R [Ω] 
55.81 
15.33 
6.987 
$\frac{R}{L}$ [Ω m^{1}] 
53.46 
14.33 
34.94 
The measured values of $\overline{T}$ and the values of the thermal conductivity λ calculated by Eq. (19) or Eq. (22) on one hand and by Eq. (28) on the other hand are presented in Table 2. For the estimation of using Eq. (19) or Eq. (22), two limit values h_{min} and h_{max} of the external heat transfer coefficient h have been considered, enabling the estimation of two extreme values λ_{min} and λ_{max} of λ.
Table 2. Experimental values of $\overline{T}T_a$; estimated and reference value of the thermal conductivity λ of Copper
D [m] 
0.025 mm 

$\overline{T}T_a$ [K] 
16.65 K 

$s\overline{T}$ [K] 
0.12 K 

λ [W m^{1} K^{1}] 
Eq. (19) (h=6 W m^{2} K^{1}) 
393.0 
Eq. (19) (h=2 W m^{2} K^{1}) 
379.6 

Eq. (28) 
370.9 

Reference [14] 
386 
The low pressure value decreases the convective heat transfer and the external heat transfer is essentially done by radiation. The radiation heat transfer coefficient of a surface at temperature T is lower than 4σT^{3}, where σ is the StefanBoltzmann constant (σ=5.67×10^{8} W m^{2} K^{4}), corresponding to an approximate value of 6 W m^{2} K^{1 }for a black body at a temperature of 25°C. Thus, the extreme values h_{min}=2 W m^{2} K^{1} and h_{max}=6 W m^{2} K^{1} have been considered.
Compared to a reference value of 386 W m^{1} K^{1 }[14], the relative deviation is less than 2 % that is quite acceptable. It may be noted that using the approximate Eq. (28) would lead to a greater deviation around 4%.
For the Chromel wires the mean temperatures of wires with lengths between 15 mm and 85 mm have been measured under a constant pressure of 6.10^{2} mbar. An algorithm minimizing the sum of the quadratic errors between the experimental values and the values calculated with Eq. (19) enabled the simultaneous estimation of the thermal conductivity λ and of the external heat transfer coefficient h for each wire. The experimental points and the theoretical curves are presented in Figure 7. The values of λ and h obtained using Eq. (19) are presented in Table 3.
Furthermore, the specific heat c of the Chromel has been measured with a Setaram mdsc3 differential scanning calorimeter and its density ρ has been estimated by weighing a sample and measuring its dimensions. The thermal diffusivity a has been estimated by the flash method [15] applied on a plane sample. The thermal properties obtained at 20°C are:
ρc=3.83×10^{6} J m^{3} K^{1}, a=4.54×10^{6} m^{2} s^{1}
leading to λ=17.4 W m^{1} K^{1}.
The estimated values of the Chromel thermal conductivity obtained by our method are very close to the values obtained by coupling the results of the calorimetric and flash methods, with a deviation less than 4% that is quite acceptable.
Figure 7. Experimental (□) and modeled ( — ) temperature for Chromel wires: a) D=0.127 mm and b) =0.254 mm
Table 3. Estimated values of the wire thermal conductivitiy λ and of the external heat transfer coefficient h for two Chromel wires
Chromel Wire 
λ 
h 

[W m^{1} K^{1}] 
[W m^{2} K^{1}] 

Experiment 
D=0.127 mm 
18.0 
5.8 
D=0.254 mm 
18.2 
4.5 

Reference 
 
17.4 
 
The steadystate method proposed in this paper enables the measurement of the thermal conductivity l of an electrical conducting wire or fiber. The expression of the wire mean temperature has been established as a function of the wire thermal conductivity l and of the external heat transfer coefficient h. This expression has been used to realize a sensitivity analysis that demonstrated the possibility of the simultaneous estimation of l and h from several steadystate measurements carried on different wire lengths. The estimation of l is realized without making the assumption that the external heat transfer is null. The theoretical model and the estimation method have been validated by an experimental study carried out with a Chromel wire and with a Copper wire. The estimated values of the thermal conductivities are in good agreement with references values (less than 5 % deviation).
The extension of the method to thinner and shorter wires or fibers
D 
wire diameter, m 

h 
external heat transfer coefficient, W.m^{2}.K^{1} 

L 
wire length, m 

n 
number of wire lengths tested 

P 
wire perimeter, m 

R 
electric resistance, W 

R_{0} 
electric resistance at initial temperature, W 

S 
wire section area, m^{2} 

s$\bar{T}$ 
standard deviation of $\bar{T}$, °C 

T 
wire temperature, °C 

T_{a} 
ambient temperature, °C 

$\bar{T}$ 
wire mean temperature, °C 

x 
distance from the middle of the wire, m 

Greek symbols 

a 
temperature coefficient of the wire electrical resistance, K^{1} 

ϕ_{0} 
heat flux per volume unit, W.m^{3} 

λ 
thermal conductivity, W.m^{1}.K^{1} 

ρ 
density, kg.m^{3} 

ρ_{0} 
wire electrical resistivity, W.m 

θ 
wire temperature rise, K 

$\bar{θ}$ 
wire mean temperature rise, K 

ω 
alternative current pulsation, rd.s^{1} 
[1] Bonura M, Senatore C. (2013). Thermal conductivity of industrial Nb3Sn wires fabricated by various techniques. IEEE Transactions on Applied Superconductivity 23(3): 6000404. https://doi.org/10.1109/TASC.2012.2234203
[2] Moon J, Weaver K, Feng B, Chae HG, Kumar S, Baek JB, Peterson GP. (2012). Thermal conductivity measurement of individual poly(ether ketone)/carbon nanotube fibers using a steadystate dc thermal bridge method. Review of Scientific Instruments 83: 016103. https://doi.org/10.1063/1.3676650
[3] Bougeois O, Fournier T, Chaussy J. (2007). Measurement of the thermal conductance of silicon nanowires at low temperature. Journal of Applied Physics 101: 016104. https://doi.org/10.1063/1.2400093
[4] Bhatta RP, Mohr RK, Brandys M, Pegg IL, Dutta B. (2010). High temperature thermal conductivity of platinium microwire by 3ω method. Review of Scientific Instruments 81: 114904. https://doi.org/10.1063/1.3496048
[5] Lu L, Yi W, Zhang DL. (2001). 3ω method for specific heat and thermal conductivity measurements. Review of Scientific Instruments 72: 2996. https://doi.org/10.1063/1.1378340
[6] Zhang X, Fujiwara S, Fujii M. (2000). Shorthotwire method for the measurement of the thermal conductivity of a fine fibre. High TemperaturesHigh Pressures 32: 493500. https://doi.org/10.1068/htwu374
[7] Choi TY, Poulikakos D, Tharian J, Sennhauser U. (2005). Measurement of thermal conductivity of individual multiwalled carbon nanotubes by the 3ω method. Applied Physics Letters 87: 013108. https://doi.org/10.1063/1.1957118
[8] Dames C, Chen G. (2000). 1ω, 2ω, 3ω methods for measurements of thermal properties. Review of Scientific Instruments 76: 124902. https://doi.org/10.1063/1.2130718
[9] Liang J, Saha MC, Altan MC. (2013). Measurement of thermal conductivity of carbon fibers using wirebased 3 omega method. 28th Technical Conference of the AmericanSocietyforComposites, State College, Pennsylvania University, USA, pp. 593599.
[10] Volklein F, Kessler E. (1984). A method for the measurement of thermal conductivity, thermal diffusivity, and other transport coefficients of thin films. Physica Status Solidi 81(2): 585596. https://doi.org/10.1002/pssa.2210810222
[11] Beck JV, Arnold KJ. (1977). Matrix Analysis for linear parameter estimation in parameter estimation in engineering and science. John Wiley and Sons, New York. https://doi.org/10.1002/aic.690240233
[12] Marquart DW. (1963). An algorithm for leastsquares estimation of nonlinear inequalities. J. Appl. Math. 11: 431441.
[13] Fallou M. (1998). Résistivité des métaux. Techniques de l’Ingénieur, Paris
[14] Holman JP. (1990). Heat transfer. 7th edition. McGrawHill, New York