Ultra-high performance wearable thermoelectric coolers with less materials

Thermoelectric coolers are attracting significant attention for replacing age-old cooling and refrigeration devices. Localized cooling by wearable thermoelectric coolers will decrease the usage of traditional systems, thereby reducing global warming and providing savings on energy costs. Since human skin as well as ambient air is a poor conductor of heat, wearable thermoelectric coolers operate under huge thermally resistive environment. The external thermal resistances greatly influence thermoelectric material behavior, device design, and device performance, which presents a fundamental challenge in achieving high efficiency for on-body applications. Here, we examine the combined effect of heat source/sink thermal resistances and thermoelectric material properties on thermoelectric cooler performance. Efficient thermoelectric coolers demonstrated here can cool the human skin up to 8.2 °C below the ambient temperature (170% higher cooling than commercial modules). Cost-benefit analysis shows that cooling over material volume for our optimized thermoelectric cooler is 500% higher than that of the commercial modules.

This work investigated the combined effect of the thermal resistance and TE material properties on TEC performance. The manuscript showed that the optimized design achieved human skin cooling to 8.5oC below the ambient temperature, with a better cooling over material volume.
(2) Page 4, it is stated that "the effect of external thermal resistances on TEC performances has not been well-examined". There are quite amount of researches focusing on the impact of thermal resistances on TEC performances. The authors should give a complete review on this topic and explain why it hasn't been well-examined. In addition, the Introduction part should be more concentrated on the points that would lead to this work's innovations instead of talking too much about well-known things. When referring other works, the manuscript should directly touch the most critical and relevant points, not just list the references there.
(3) Page 6, "Despite these discoveries, the performance of TE devices during real deployments has remained poor [58], indicating a strong influence of operating environment on the device performance." The work [58] referred here mainly talks about high demand of breakthrough in TE material development. Apparently it doesn't indicate the strong influence of operating environment on the device performance. (4) Page 13, when designing on-body applications and personalized cooling TEC, what criteria should be the best to judge the quality of the TEC: temperature cooled, cooling heat flux, COP, or defined cooling/material volume and cooling/input power (Fig. S6)? Previous work [31-32] found human body emits thermal energy at the rate of ~25 mW/cm2 into the ambient environment. Meanwhile, [59] presented the findings on thermal comfort of human body under personal thermal management with a conclusion that different body parts may require different temperatures. Is the greater temperature drop naturally producing better thermal comfort? Therefore, a well justified logic foundation is required to support the conclusions made from Fig. 6. Also, an explanation might be needed for the reason why switching from cooling/area (W/cm2) of Fig. 2 and Fig. 3(a,b) to cold-side temperature (oC) of Fig. 3(c,d), Fig. 4, and Fig. 6. (5) Instead of directly applying the designed TECs to human body test, experiments under a controllable environment (e.g., fixed temperature or heat flux conditions) should be completed in order to more accurately examine the TEC performance. The heat transfer rate of the fin and heat flux from skin should be measured in the experiments, or at least the estimates should be verified.
This paper focuses on the optimum design of Thermoelectric Coolers (TEC) operating under huge thermally resistive environment for localized wearable cooling applications. Nevertheless, the theoretical modeling and experiments conducted in this study were under the atmosphere temperature of 22 °C, which is not a typical hot condition requiring personal cooling. To demonstrate the cooling effect of TEC and potential energy saving in building HVAC systems through expansion of set-points as articulated in the introduction of this research, modeling and experiments should be carried out at an ambient temperature of above 26 °C (4 °F more than the normal set-points to achieve 20% saving in HVAC energy for cooling). Also, it is better to study the cooling effect of TEC at different ambient temperatures and show the influence of environment on its performance.
Besides, to maintain thermal comfort in hot environment, TEC should remove a certain amount of heat from human body. Based on the FOA of ARPAE's DELTA program, an additional 23W of heat removal is required if the up-bound of neutral set-points is increased by 4 °F. It is suggested to show heat removal the proposed TEC, given its covering area and weight. Furthermore, the large temperature reduction in localized skin area covered by the TEC can cause severe localized cold discomfort. It is advised to evaluate the local cold discomfort. Moreover, without an effective heat rejection in a hotter environment than the ambient temperature (22 °C) investigated in this study, it is advised to show whether the cooling effect is stable.

Reviewer #1 (Remarks to the Author):
This manuscript presented a study of TECs for the use in cooling of human body, through both theoretical modelling and experimental measurement. Is sounds an interesting and valuable works. However, a few items may need further strengthen and emphasis: (1) mathematical equations that support the operation of the TECs cooling should be defined and For convenience of reviewer, the modeling section in the revised manuscript is presented below.

Modeling
Fig. 10(a) schematically illustrates a TEC module placed between a heat source and a heat sink.
A TEC consists of several p-and n-type legs connected electrically in series and thermally in parallel. When electric current is passed through a TEC, holes in p-type and electrons in n-type legs move in the direction of the current, which carry thermal energy thereby creating cold and hot sides of the TEC module. TE effect is a complex phenomenon as it is a combination of four different effects: Seebeck effect, Peltier effect, Thomson effect, and Joule heating 60 (please refer Supplementary information for more details). An accurate TEC model, therefore, requires multi-physics coupled expressions. Building upon this fundamental understanding, a simplified onedimensional energy equilibrium model is presented here, which will be used for explaining the key observations. The complex three-dimensional model, which is used to obtain the numerical results in this paper, is presented in the next section.

Simplified one-dimensional energy equilibrium model
The thermal resistances due to heat source, heat sink, and TEC constitute a thermal circuit as shown in Fig. 10(b). In the simplified one-dimensional model, it is assumed that the TEC is operating under steady state condition, material properties are temperature-independent (ignoring Thomson effect), and the heat losses to the environment are negligible. Performing the energy balance on the cold and hot sides of TEC, it can be shown that, 61 where Q is the heat flow from the heat source (termed as cooling capacity of the TEC), Q is the heat flow to the heat sink, Q is heat flow through TEC because of the internal temperature difference, Q is Peltier effect generated cooling on the cold-side, Q is Peltier effect generated heating on the hot-side, and is the contribution due to Joule heating. These terms can be mathematically expressed as 62 : where T and T are the core temperatures of the heat source and sink, R and R are the total thermal resistances of the heat source and sink, T and T are the temperatures of the hot-and cold-sides of the TEC, R and R are the internal thermal and electrical resistances of the TEC, n is the total number of pairs of p-and n-type legs, and α is the net Seebeck coefficient given as α = α − α , where α and α are the Seebeck coefficients of pand n-type TE legs, respectively. Lastly, as shown in Fig. 10(c), I is the current in the electrical circuit because of a current source. Using equations (4)-(11), the cooling capacity (Q ) of a TEC can be expressed as: V shown in Fig. 10(c) is the Seebeck voltage, which is given as 63 : The input electrical power P by the current source is given by 64 : The coefficient of performance (COP) of TEC is measured using relation: It should be noted that cooling capacity (Q ) and COP of TEC are a function of the internal temperature difference ∆T = T − T , which is usually not known. When R and R are negligible, it can be assumed that T ≈ T and T ≈ T . However, for the applications where heat source and/or sink resistances are considerably high, cooling capacity and COP are greatly affected by these external resistances.
The cooling capacity of TEC constitutes of three components (Equation (12)): Peltier cooling (Q ), Joule heating (Q ), and the internal heat flow (Q ). In order to increase cooling capacity, Q should be increased, while Q and Q should be lowered. Increase in Seebeck coefficient (α) increases Q (Equation (9)), thus it has a positive effect on cooling capacity. Joule heat Q is a function of TEC internal electrical resistance (R ), which is given as 65 : where is the number of leg pairs, is the leg height, is the leg cross-sectional area, σ and σ are electrical conductivity of p-and n-type legs. Increase in electrical conductivity (σ) decreases R and thus reduces Q (Equation (11)). Hence, it also has a positive effect on cooling capacity. Likewise, the internal thermal resistances of a TE module is given as 65 : where κ and κ are thermal conductivity of p-and n-type legs. Lowering the value of material thermal conductivity (κ) increases R , which reduces Q (Equation 8) and thus enhances cooling capacity. (c) The electrical-circuit of TEC depicting the Seebeck voltage and internal electrical resistance.

Three-dimensional coupled multi-physics model
The simplified one-dimensional model presented in the previous section is a decoupled model where thermal and electrical relations are solved separately. One-dimensional model explains the physics and provides fairly accurate results (error up to 10%) 61,66 when thermal gradient is small, material properties are temperature-independent, and contact resistances are small 67 . However, for a more robust analysis, a three-dimensional model is needed to account for temperature dependent material properties and various types of losses. In order to obtain numerical results provided in this paper, the complex three-dimensional equations of thermoelectricity in steady state are used, which are given as 68 : Current density vector and heat flux vector for TEC model in three dimensions, are given as 68 : = αT − κ∇T.
where V is the electrostatic potential, T is the absolute temperature, and κ, σ, and α denote the temperature dependent thermal conductivity, electrical conductivity, and Seebeck coefficient of the TE materials. Equations (18)-(21) are coupled and thus need to be solved numerically.
(3) comparison between the modelling and experimental results should be made intensively;   Under optimal electric current, the cooling over TE material volume is highest for the ultra-low

The information below has been added in the revised Supplementary information.
It can be noted that due to high thermal resistivity of the human skin and the ambient air along with the size constraint of the heat sink for on-body applications, wearable TECs operate in extremely high thermally resistive environment. This is evident from the heat transfer coefficient of human skin and heat sink, which is less than 100 W/m 2 -K. leg aspect ratio of 1.0, the low FF TEC has fill factor of 12% and leg aspect ratio of 1.0, and the ultra-low FF TEC has fill factor of 5.2% and leg aspect ratio of 1.6. The internal electrical resistance measured using four probe method was found to be 228 mΩ, 75 mΩ, and 185 mΩ for high FF, low FF, and ultra-low FF TEC modules, respectively.

Following information has been added in the revised introduction.
Prior studies have reported a very strong influence of external thermal resistance on the performance of TECs 33-35 and suggested that the optimal TEC should have internal thermal resistance in the range of 40-70% of the total thermal resistance of the system 36 . However, most of these studies were focused on electronic cooling using TECs. For on-body applications, the research in the field of thermoelectric, so far, has been focused towards thermal energy harvesting from body heat 37-43 . The thermoelectric modules with fill factor less than 20% have been recommended for on-body applications, which brings challenges in terms of mechanical applications under different ambient conditions and an optimal design for wearable TECs is obtained. The ultra-low fill fraction module proposed in this paper exhibits 170% higher cooling, while utilizing five times less TE materials than the commercial TEC.

Reviewer #2 (Remarks to the Author):
This work investigated the combined effect of the thermal resistance and TE material properties on TEC performance. The manuscript showed that the optimized design achieved human skin cooling to 8.5°C below the ambient temperature, with a better cooling over material volume. (

A portion of the revised introduction is given below.
Prior studies have reported a very strong influence of external thermal resistance on the performance of TECs 33-35 and suggested that the optimal TEC should have internal thermal resistance in the range of 40-70% of the total thermal resistance of the system 36 . However, most of these studies were focused on electronic cooling using TECs. For on-body applications, the research in the field of thermoelectric, so far, has been focused towards thermal energy harvesting from body heat 37-43 . The thermoelectric modules with fill factor less than 20% have been recommended for on-body applications, which brings challenges in terms of mechanical

The portion of the revised introduction is given below.
Prior studies have reported a very strong influence of external thermal resistance on the performance of TECs 33-35 and suggested that the optimal TEC should have internal thermal resistance in the range of 40-70% of the total thermal resistance of the system 36 . However, most of these studies were focused on electronic cooling using TECs. For on-body applications, the research in the field of thermoelectric, so far, has been focused towards thermal energy harvesting from body heat 37-43 . The thermoelectric modules with fill factor less than 20% have been recommended for on-body applications, which brings challenges in terms of mechanical Therefore, a well justified logic foundation is required to support the conclusions made from Fig.  6. Also, an explanation might be needed for the reason why switching from cooling/area (W/cm2) of Fig. 2 and Fig. 3(a,b) to cold-side temperature (oC) of Fig. 3(c,d), Fig. 4 We have added following discussion in the revised manuscript.
It is important to note that higher cooling by the low and ultra-low FF TECs does not naturally result in a better thermal comfort. In fact, a large temperature drop on human skin can be quite uncomfortable; therefore, for practical purposes, users need to be provided with a temperature controller to control the cooling based on thermal preferences. None-the-less, since the low and ultra-low FF TECs proposed in this study have capability to generate a larger temperature drop than the high FF TEC, they are expected to be more effective in hot-weather conditions. Fig. 8 (a)-(c) depict the cold-side temperature of the different TECs deployed on human body under different ambient temperatures. It can be noted that the initial skin temperature is higher at higher ambient temperature, indicating the fact that the human body temperature increases with increase in ambient temperature. The average initial skin temperature was noted to be 30.6°C under ambient temperature of 22°C, 32.1°C under ambient temperature of 26°C, and 34.6°C under ambient temperature of 32°C. In Fig. 8(a), when ambient temperature is 22°C, the high FF TEC generates minimum cold-side temperature of 25.2°C, which increases to 27.7°C under ambient temperature of 26°C, and 31.7°C under ambient temperature of 32°C. In Fig. 8( interesting to note that the ultra-low fill factor TEC generates least dynamic and static cold-side temperature and thus maximum cooling followed by the low fill factor TEC and the high fill factor TEC.

Following information has been added in the revised manuscript.
In order to mitigate the effect of human factors on experimental results, the experiments on different TEC modules were also performed under controlled environment and the outcomes were compared with the results obtained on human body. The controlled environment consisted of a heat source at fixed temperature (34°C) and a thermal resistor of known thermal conductivity (0.94 W/m-K) and dimensions: 16 mm × 18 mm × 10 mm placed between heat source and TEC to mimic the thermal resistance of the human skin. Fig. 5 Following information has been added in the revised manuscript.  increases with increase in electric current, but it decreases with increases in ambient temperature.
The peak cooling flux lies in the range of (a) 85-110 mW/cm 2 for High FF TEC, (b) 120-140 mW/cm 2 for low FF TEC, and (c) 115-130 mW/cm 2 for ultra-low FF TEC.

Following information has been added in the revised manuscript.
It is important to note that higher cooling by the low and ultra-low FF TECs does not naturally result in a better thermal comfort. In fact, a large temperature drop on human skin can be quite uncomfortable; therefore, for practical purposes, users need to be provided with a temperature controller to control the cooling based on thermal preferences. None-the-less, since the low and ultra-low FF TECs proposed in this study have capability to generate a larger temperature drop than the high FF TEC, they are expected to be more effective in hot-weather conditions. Fig. 8 (a)-(c) depict the cold-side temperature of the different TECs deployed on human body under different ambient temperatures. It can be noted that the initial skin temperature is higher at higher ambient temperature, indicating the fact that the human body temperature increases with increase in ambient temperature. The average initial skin temperature was noted to be 30.6°C under ambient temperature of 22°C, 32.1°C under ambient temperature of 26°C, and 34.6°C under ambient temperature of 32°C. In Fig. 8(a), when ambient temperature is 22°C, the high FF TEC generates minimum cold-side temperature of 25.2°C, which increases to 27.7°C under ambient temperature of 26°C, and 31.7°C under ambient temperature of 32°C. In Fig. 8( Fig. 8(c), the ultra-low FF TEC module generates minimum cold-side temperature of 22.3°C under ambient temperature of 22°C, 24.9°C under ambient temperature of 26°C, and 28.7°C under ambient temperature of 32°C. It is important to note that when ambient temperature is 32°C, the high FF TEC is unable to cool the skin temperature to 30.5°C, which is typically the skin temperature in normal ambient condition of 22°C. This implies that in hot climate the high FF TEC is ineffective to cool the human body. The low FF and the ultra-low FF TECs, on the other hand, can be observed to cool the human skin below 29°C, when ambient temperature is 32°C, emphasizing the fact that these modules are quite effective even in extreme climate. Fig. 8(a)-(c). Cold-side temperature for different TECs deployed on human body under different ambient temperatures. Increase in ambient temperature negatively impacts the cooling performance of the TECs. In a hot climate (ambient temperature ~32°C), the high FF TEC is unable to cool the skin temperature to its normal temperature (~30.5°C), whereas the low and ultra-low FF TECs are effective even in extreme climate. Circles illustrate the experimental data whereas solid lines show the modeling results.
(2) Besides, to maintain thermal comfort in hot environment, TEC should remove a certain amount of heat from human body. Based on the FOA of ARPAE's DELTA program, an additional 23W of heat removal is required if the up-bound of neutral set-points is increased by 4 °F. It is suggested to show heat removal the proposed TEC, given its covering area and weight.
Authors' reply: Thanks again for the suggestions. Figure 8(   (3) Furthermore, the large temperature reduction in localized skin area covered by the TEC can cause severe localized cold discomfort. It is advised to evaluate the local cold discomfort.

Following information has been added in the revised manuscript.
Authors' reply: We agree that the high cooling does not naturally produce better thermal comfort. In fact, a very large temperature drop on human skin could be quite uncomfortable.
However, as stated in FOA of ARPA-E DELTA program, one of the key advantages of using a localized cooling thermal management system is that it enables users to adjust the cooling based on their thermal comfort. For practical purposes, therefore, the users need to be provided with a temperature controller to control the temperature drop as per the desired comfort level.

We have added following explanation in the revised manuscript.
It is important to note that higher cooling by the low and ultra-low FF TECs does not naturally result in a better thermal comfort. In fact, a large temperature drop on human skin can be quite uncomfortable; therefore, for practical purposes, users need to be provided with a temperature controller to control the cooling based on thermal preferences. None-the-less, since the low and ultra-low FF TECs proposed in this study have capability to generate a larger temperature drop than the high FF TEC, they are expected to be more effective in hot-weather conditions. Fig. 8 (

Following information has been added in the revised manuscript.
In order to mitigate the effect of human factors on experimental results, the experiments on different TEC modules were also performed under controlled environment and the outcomes were compared with the results obtained on human body. The controlled environment consisted of a heat source at fixed temperature (34°C) and a thermal resistor of known thermal conductivity (0.94 W/m-K) and dimensions: 16 mm × 18 mm × 10 mm placed between heat source and TEC to mimic the thermal resistance of the human skin. Fig. 5 The authors have addressed most of the reviewer comments given in last round of review. Few minor revisions are necessary before publication: (1) Manuscript page 3, "The wearable TECs, therefore, experience a very low heat load and an extremely high thermally resistive environment." Some quantitative numbers or even a summary table for comparison of the extremely high thermally resistive environment will be helpful.
(2) Manuscript page 4, " The thermoelectric modules with fill factor less than 20% have been recommended for on-body applications, which brings challenges in terms of mechanical stability." Will the TE modules fabricated in this work face the similar problem because of low and ultra-low FF factors?
(3) Manuscript page 4, "The ultra-low fill fraction module proposed in this paper exhibits 170% higher cooling,..." Higher cooling power or temperature drop? Than commercial TEC? (4) Manuscript page 11, reference(s) is needed to support the claim "For body cooling, however, cold-side temperature drop (ΔT) can be considered a more appropriate criterion to judge the effectiveness of TECs as temperature change can be easily perceived by the human body." Reviewer #3 (Remarks to the Author): This paper provided an comprehensive study on the optimum design of Thermoelectric Coolers for wearable cooling applications. The author has tried to address my concerns about cooling power and local thermal comfort of such a device, which are crucial to the potential of such a concept.
Based on their experimental work and theoretical modelling under the condition that there is close contact between the device's cold side and human skin, the maximum cooling power of different designs ranges from 85~140 mW/cm2. To achieve 23W cooling required for expanding the indoor temperature by 2 oC, the required cooling area would be between 160-270 cm2. Although this is just about 1~2% of the body surface area, but it is substantial, considering one can only use the TC to cover the possibly exposed lower arm and leg areas where one can withstand relatively lower temperature. Furthermore the weight of the TC would also be a major concern in practical applications. a 2 in2 (12 cm2) TC assembly may weigh 100g. 160~270 cm2 would weigh 1.3-2.3 kg. High weight, large impermeable and inflexible area covering the body surface and local cold discomfort may prohibit its practical application. If one uses just one of such TC at the wrist, it only provide 1~2W cooling. Furthermore, I wish to know the heat sink temperature at the peak cooling power under different environmental temperature. For safety reasons, the maximum heat sink temperature should be less than 42 oC. This can be an issue when it is used in hot environment.

Reviewer #1 (Remarks to the Author):
The authors have made necessary amendments and I am satisfied with the current presence of the paper, and thus, suggest to accept it for publication.
Author's reply: Thanks, and we much appreciated your comments and suggestions.

Reviewer #2 (Remarks to the Author):
The authors have addressed most of the reviewer comments given in last round of review. Few minor revisions are necessary before publication: (1) Manuscript page 3, "The wearable TECs, therefore, experience a very low heat load and an extremely high thermally resistive environment." Some quantitative numbers or even a summary Boiling and condensation 2500-100,000 Human skin 20-100 (2) Manuscript page 4, " The thermoelectric modules with fill factor less than 20% have been recommended for on-body applications, which brings challenges in terms of mechanical stability." Will the TE modules fabricated in this work face the similar problem because of low and ultra-low FF factors?
Author's reply: The fabrication of low fill factor thermoelectric module certainly requires changes in manufacturing steps. However, with advancements in automated drop-n-reflow technology, we think that these changes can be accommodated. In this study, the optimum TEC modules were fabricated using dies that guide the placement of legs and provide support structure. This results in better alignment and mechanical properties of modules.
(3) Manuscript page 4, "The ultra-low fill fraction module proposed in this paper exhibits 170% higher cooling,..." Higher cooling power or temperature drop? Than commercial TEC?
Author's reply: The ultra-low fill fraction module proposed in this paper exhibits 170% higher temperature cooling than the commercial TEC. We have modified the sentence in the paper. different designs ranges from 85~140 mW/cm 2 . To achieve 23 W cooling required for expanding the indoor temperature by 2 o C, the required cooling area would be between 160-270 cm 2 .
Although this is just about 1~2% of the body surface area, but it is substantial, considering one can only use the TC to cover the possibly exposed lower arm and leg areas where one can withstand relatively lower temperature.
Author's reply: (2) Furthermore the weight of the TC would also be a major concern in practical applications. a 2 in2 (12 cm 2 ) TC assembly may weigh 100g. 160~270 cm2 would weigh 1.3-2.3 kg. High weight, large impermeable and inflexible area covering the body surface and local cold discomfort may prohibit its practical application. If one uses just one of such TC at the wrist, it only provide 1~2W cooling.
Author's reply: We have added this information in the revised Supplementary information. (3) Furthermore, I wish to know the heat sink temperature at the peak cooling power under different environmental temperature. For safety reasons, the maximum heat sink temperature should be less than 42 o C. This can be an issue when it is used in hot environment.
Author's reply: The heat sink temperature at the optimal current for maximum cooling for our TECs were noted to be 45-55°C, which is bit higher than the threshold value suggested by the reviewer. However, it should be noted that the hot-side temperature of the TEC can be substantially reduced by modifying the heat sink, as shown in Figure R1. In this study, we have used an off-the-shelf heat sink purchased from McMaster-Carr [4]. As shown on Figure R1 (a), the maximum hot-side temperature in this case is ~ 47°C. Temperature contours illustrated in Figure R1(b)-(d) shows that the hot-side temperature decreases by increasing the number of fins and adjusting the fin dimensions. It is also interesting to note that lower hot-side temperature and thus lower temperature difference between hot-and cold-sides of TEC results in higher cooling due to reduced reverse thermal current.