1. Introduction
As the size of electronic devices decreases with further miniaturization techniques, there has been the problem of a large heat generation in small areas. Thermoelectric coolers are an alternative to efficiently dissipate heat generated in such conditions. The knowledge and control of irreversible phenomena that result in an increase of the temperature is, in this point of view, crucial. In many situations, the irreversibilities in the device are directly related to its optimum performance. Occasionally, optimal performance is obtained when the entropy generation is a low minimum under the operating constrictions. In this work, we study the thermal performance of a two-stage cooler with a dual purpose: (1) to theoretically model the transport of heat in the device and qualitatively compare theoretical predictions with experimental measurements of the temperature difference achieved between the hot and the cold faces in different operating conditions; and (2) to determine the relationship between global entropy production in the cooler and thermal performance measured by the thermal figure of merit [
1]. The operating conditions are defined by the currents flowing in each of the stages of the cooler. The experimental device is similar to those reported in the literature [
2,
3,
4,
5,
6,
7], which are used for measuring thermal properties of thermoelectric materials or devices which are similar to the one studied here. Typically, the theoretical models used to describe transport phenomena in thermal devices are based on global energy balances at the ends and interfaces of the devices. Here, we use a local steady-state model that allows us to describe the spatial distribution of the relevant physical properties of the problem. Models similar to the one described here can be found in [
3,
4,
8,
9]. Our experimental results fit well with the theoretical predictions and the analysis on the relationship between entropy production and thermal performance reveals that, for each operating regime characterized by a specific thermal figure of merit, there are two entropy production regimes. The operating conditions of the device should be fitted in the lowest entropy production regime. This manuscript is divided in sections as follows: in
Section 2, the two-stage cooler and the experimental procedure to measure the difference of temperature between the cold and the hot ends of the device are described as well as the operating conditions. In
Section 3, a theoretical model used to study the heat transport and calculate the stationary temperature on the cold side, while the hot temperature remains constant, can be found. In
Section 4, the thermoelectric parameters as well as the flowing electric currents in each of the stages are prescribed, and our theoretical results, comparison with measurements and results on the relation entropy production-thermal figure of merit are exposed and discussed. Finally, the concluding remarks are contained in
Section 5.
2. Experimental Procedure
The experimental set-up consists of two Peltier modules connected thermally in series and electrically independent (see
Figure 1). The thermoelectric modules are denoted by
M. The modules are made of Bismuth Telluride alloys. The side length and width is
mm and
mm, respectively. The module array is mounted on a flat metal plate with measures 15 cm × 8 cm. This hot plate (or hot wall) is kept at a constant temperature by means of a warm bath. Hot water is continuously forced to circulate by a microprocessor dosing pump (Milton Roy, Houston, TX, USA) in a circuit containing a rectangular frame where the hot plate is located. The pump is denoted by
B. Each direct electric current
I is obtained from a regulated power supply
V (Model 1696, B&K Precision Corporation, Yorba Linda, CA, USA) . The range of electric currents is between 0 to 1.517 A. In steady state, the temperature was measured in the geometrical center of the cold wall (top of module one
) by means of a thermocouple (K type) connected to a True RMS Multimeter (EX470, Extech Instruments, Nashua, NH, USA). The injected currents through the modules generate a heat flow from the cold to hot wall, thus the temperature in the cold wall is diminished. In order to avoid thermal decoupling, thermal paste (SILITEK Corporation, Taiwan) was placed between the two modules and the thermocouple-cold wall joint. The error in the temperature data was obtained by adjusting a normal distribution to the data at same points. The results of the experimental measurements are presented in
Section 4.
4. Results and Discussion
Before presenting and discussing the theoretical results, we depart from the experimental measurements for a single and a coupled system.
Figure 3a shows the temperature difference between the hot and cold sides
as a function of the injected electric current for a single thermoelectric device. In this figure, we can appreciate a
parabolic behaviour with a maximum
25 K, whereas
Figure 3b presents
elliptic isovalues of
as function of the two electric currents, when the system is composed on two thermoelectric devices. For this array of devices, the maximum of temperature difference is
34 K, that is, nine degrees of extra cooling are obtained when using two devices instead of a single one. For both cases, the single and two thermoelectric system, the electric current is normalized with
A, which is the optimal current for a single device.
As in previous studies concerning thermoelectric devices [
1,
10,
11], our theoretical results come from considering doped Silicon as working material, whose properties have been published before in [
12]:
Wm
·K
,
·m
,
VK
and
m
·s
, where
a is the thermal diffusivity; as the hot side temperature, we take
K. We must note that these values do not correspond to our experimental case. However, since our model is a one-dimensional approximation of the heat transport, the objective is a qualitative comparison, and as it is shown in
Figure 4, the aim is reached.
Figure 4a shows the temperature difference
as a function of the electric current for a single thermoelectric device (Equation (6)). Comparing with its experimental homologous,
Figure 3a, the
parabolic behaviour with a maximum is found.
Figure 4b contains the dependence of
when the system is composed on two thermoelectric devices (Equation (17)), presenting
elliptic isovalues. Theoretical results (
Figure 4) derived from a simple model agree qualitatively with the experimental ones (
Figure 3) showing a maximum
for a given set of electric currents. A quantitative comparison is far from been accomplished, since our model is a one-dimensional representation of a three-dimensional problem. However, if we calculate the percentage of the maximum extra cooling obtained when using two devices instead of a single one, that is
, it is the same for both the experimental and theoretical results. The maximum extra cooling that could be reached is 36%.
The temperature profiles along the
x-axis for different sets of electric density currents are shown in
Figure 5. The thermoelectric number one is contained in the domain (−1, 0), whereas the second one is in (0, 1). The cold (
) and hot (
) side temperatures are located at
and
, respectively. As seen in Equation (13), every segment of the profiles is a parabolic curve. The blue line has been previously obtained in the study of an optimal currents for a single thermoelectric [
1]. In this case, the cooling obtained is 3.1 K. The black line denotes the profile for the optimal set of electric currents, obtaining a maximum cooling of 4.2 K, whereas the red line shows the case when the cooling is zero. For this case, the electric currents are high enough for the Joule effect dominate over the Peltier effect. In
Figure 5b, it can be seen the theoretical profile of the local entropy production for different values of the electric current densities in the thermoelectric materials. As in
Figure 5a, the black line represents the optimal case, the red one the zero cooling case, and the blue one another case with non-vanishing cooling. In the zero cooling, the entropy production is at the maximum at the cold and hot sides and is at a minimum at the coupling boundary. This minimum is explained by the fact that the temperature has a maximum and the temperature gradient vanishes at the same position. It is interesting that the minimum displaces toward the right (
) when the current densities approach the optimal values. The discontinuity
in the entropy production for the optimal case (black line) is due to the fact that the electric density current in each segment is distinct.
The total entropy generation
in the devices can be seen in
Figure 6. The dependence of the entropy for a single device shows a
quadratic behaviour by increasing the electric current (see
Figure 6a). For the two-device system, the entropy presents
circular isolines increasing their magnitude as the electric currents are larger.
Finally,
Figure 7 presents the thermal figure of merit
as a function of the total entropy generation
. See the
Appendix for a deduction of the thermal figure of merit
. We must note that both quantities depend on the electric density current applied to the system. For a single device (see
Figure 7a), the figure of merit increases quickly as the total entropy generation does too, it reaches a maximum and then decreases slowly. We can note that aside from the maximum value, for each figure of merit value, there are two entropy production values. This is an interesting result since it provides evidence that the device can work in a smaller entropy production regime with the same thermal figure of merit. The same behaviour is found for the coupled devices (see
Figure 7b). For
, the maximum of thermal figure of merit is larger than the maximum for a single device. This demonstrates the desirability of using two coupled thermoelectric materials rather than one. If
is larger than 0.4, the maximum of the thermal figure of merit decreases drastically.