Solar thermoelectric generators (STEGs) often require long thermoelectric (TE) legs and efficient cooling at the cold side to increase the temperature difference across TE legs and, thus, the power output. We investigate the effects of direct side-wall air cooling of TE legs on the power output of STEGs fabricated with high aspect-ratio as well as V-shaped p-type and n-type TE couples without additional heat sinks. Wire-type metallic TE materials are welded together to create V-shape TE leg arrays without additional electrodes and attached to a ceramic plate with a solar absorber on top to complete the STEG. The power generation performance of the STEG is investigated with varying wind speed under concentrated solar irradiation. Finite element simulation is performed to further analyze the heat transfer and thermoelectric performance. We find that although sidewall air cooling helps to keep the cold-side temperature cooler in both natural and forced convection regimes, it can also lower the hot-side temperature to reduce the net temperature difference and, thus, the power output and efficiency. Partial thermal insulation of TE couples can further enhance the power output under forced air convection by keeping the hot side temperature higher. The developed STEG achieves a maximum power density of 230 μW/cm2 and a system efficiency of 0.023% under 10 suns with natural convection. The low efficiency was mainly due to the low ZT of the metallic TE materials used and the unoptimized leg length. Our simulation shows that the system efficiency can be improved to 2.8% with state-of-the-art Bi2Te3 alloys at an optimal leg length.

Solar thermoelectric generators (STEGs) have gained significant interest in the field of renewable energy as a promising alternative to traditional photovoltaic technologies for converting solar energy into electricity.1 In recent years, research and development efforts have been focused on enhancing the efficiency and reliability of STEGs to make them a competitive and viable solution for sustainable power generation.2–4 In a STEG, solar irradiation is first converted to heat by a solar absorber, and then the heat is converted to electricity with a thermoelectric module based on the Seebeck effect.5,6 In a STEG, solar absorbing paints are typically used as the absorber layer on a ceramic, electrically insulating substrate. The overall efficiency of a STEG is limited by TE conversion.

The efficiency of an ideal TEG is given by
μ T E = μ Carnot 1 + Z T ¯ 1 1 + Z T ¯ + T C T H ,
(1)
where T H and T C are, respectively, the hot-side and cold-side temperatures of the TEG, μ Carnot = T H T C T H is the Carnot efficiency, and ZT is the unitless thermoelectric figure of merit defined as Z T = S 2 σ κ T, where S, σ, and κ are, respectively, the Seebeck coefficient, electrical conductivity, and thermal conductivity of the material used. T is the absolute temperature. Z T ¯ is the average figure of merit over the temperature range from T C to T H. Efficiency is fundamentally limited by the Carnot efficiency, as the rest of the right-hand side of Eq. (1) is always smaller than 1. To achieve high efficiency and, thus, a large power output from a TEG for a given solar input, a large temperature difference ( Δ T = T H T C ) and a large Z T ¯ for the temperature range are desired.

Concentrated solar irradiation is typically used to increase the heat flux into the TEG and a heat sink is attached at the cold side to increase the temperature difference across the TE legs. Telkes's work7 in 1954 is one of the early studies about STEGs constructed with Bi2Te3 alloys and ZnSb. The efficiency reached 3.35% under 50 times the optical concentration. Later, Goldsmid et al.8 carried out a similar study with flat panels and optically concentrated configurations and showed a system efficiency of 0.6% and a generator efficiency of 1.4%. The low efficiency was because the energy absorbed by the collector was reduced and three-quarters of this energy was lost directly to the surroundings. Rush9 investigated a flat panel STEG configuration specifically for space applications with a maximum efficiency of 0.7% in 1964. In 1982, a concentrated STEG prototype was developed by Dent and Cobbel10 for high solar concentration using a solar furnace and PbTe material, allowing a larger temperature difference. However, later, there were significant doubts raised on these experiments due to the discrepancy between the reported STEG efficiency and the predicted TEG efficiency. Omer11 came up with a conceptual experiment using a commercial Bi2Te3 module and an infrared heat pump, which produced an efficiency of 0.9%. Later, a cost-competitive experimental study with a larger solar concentration proposed by Amatya and Ram12 reported a system efficiency of 3%, which is closer to that of Telkes.7 There were a few papers published on improving efficiency by enhancing various aspects of designing STEGs. A detailed study was presented by Kraemer et al.13 using a large flat panel for solar absorption and concentration with a single pair of TE legs. Later, they demonstrated a peak efficiency of 7.4% with segmented TE legs and improved thermal design resulting in reduced heat losses.14 

Baranowski et al.1 developed a model that accurately predicted heat fluxes and heat transport in TE materials for common STEG designs and showed that a system efficiency of 13%–15% is achievable with a hot side temperature of 1000 °C. Similarly, Gang Chen15 investigated the theoretical efficiency of STEGs with thermal concentration effects and showed that a 12% efficiency is possible with a 10-fold optical concentration and a 200-fold thermal concentration. Li et al.16 improved system integration and reduced energy loss in their prototype STEG, leading to increased total efficiency and the ability to predict system performance for different TE materials. Additionally, Sundarraj et al.17 emphasized the significance of improving TE material properties with nanostructures to attain higher efficiency in their review and, subsequently, developed a laboratory-scale HSTEG system18 that achieved a maximum electrical power output of 4.7 W and an electrical efficiency of 1.2%.

Several modeling studies have incorporated different types of heat losses in solar thermoelectric generators (STEGs). Telkes7 has investigated the impact of air conduction losses on the performance of STEG cells using an effective heat transfer coefficient with radiation heat losses. Xiao et al.19 demonstrated that a well-designed thermal system that effectively leveraged the unique properties of each thermoelectric material, utilizing a three-stage thermoelectric module, can achieve a total solar energy conversion efficiency of up to 10.52%. Subsequent to the previous study, Song et al.20 developed a mathematical model for STEGs that accounts for various heat losses, resulting in a peak electrical efficiency of 5.2%. Additionally, Luo et al.21 proposed a multi-physics numerical model for the entire TEG system, which highlighted the potential for further development of comprehensive and accurate models for predicting TEG performance.

Despite recent progress in thermoelectric generators (TEGs), the influence of air convection at the sidewall of TE legs remains an understudied area,18 with inconclusive results reported in literature. To address this gap, we investigated the impact of sidewall air cooling on the power output and the efficiency of TEGs using a model system made of long wire-type, V-shaped TE couples. The high-aspect ratio TE couples provide a high heat transfer area from sidewalls to surrounding air convection, which enables sidewall air cooling as a significant variable for the performance of STEGs. The proposed STEG design does not require extra heat sinks at the cold side as it utilizes sidewall air cooling. The direct junction of TE legs without additional electrodes to create V-shape TE couples is useful to eliminate the extra resistance added by electrodes and their contacts. The use of metallic TE materials without electrodes and heat sinks would provide a significant reduction in materials and system costs compared to state-of-the-art TEGs. We measured the voltage and power output performance of the model TEG under different wind speeds through the TE leg arrays and conducted a detailed heat transfer analysis using finite element simulation. This study aims to provide critical insights into complex heat transfer mechanisms that occur in TEGs, especially the role of air convection at the sidewall of TE legs, to optimize the TEG design for improved energy conversion efficiency.

To study the effect of sidewall convection on TE legs, we employed long, high aspect-ratio, wire-type metallic TE materials as the TE legs in our STEG to utilize the high surface area of the surroundings. State-of-the-art TE materials such as Bi2Te3 alloys are fragile and easily fracture due to their layered crystal structure, so they cannot be easily manufactured into long wires.22 We chose type-E thermocouple materials, chromel and constantan, as TE materials because they can be easily made into a wire form with high aspect-ratio and have one of the highest Seebeck coefficients among metals with excellent high-temperature stability.23 The n-type and p-type metallic wires of diameter ∼0.8 mm and length 5 cm were welded together to make direct junctions between them without additional electrodes involved and to make V-shaped TE leg arrays as shown in Fig. 1. The V-shaped TE arrays were bonded to a hexagonal boron nitride (hBN) substrate with a high-temperature epoxy. The hBN substrate offers both high thermal conductivity and electrical insulation. The other side (top side) of the substrate is painted with a high-absorptance solar paint (Pyromark High-Temperature Paint) with a thin (∼0.9 mm) layer. Figure 1(a) shows the picture of the fabricated STEG with V-shaped TE couples being tested under solar irradiation. Total 36 couples of TE legs were bonded on a 3 × 3 cm2 hBN substrate.

FIG. 1.

(a) Picture of the fabricated V-shape TE couple STEG, (b) V-shape array geometry used in the STEG, and (c) schematic of a concentrating solar experiment setup generator without additional heat sinks.

FIG. 1.

(a) Picture of the fabricated V-shape TE couple STEG, (b) V-shape array geometry used in the STEG, and (c) schematic of a concentrating solar experiment setup generator without additional heat sinks.

Close modal

Figure 1(c) shows the schematic of our concentrating solar TE experiment setup. A Fresnel lens was used to concentrate solar irradiation from a solar simulator (Oriel Sol 3A 94063A solar simulator, 6 × 6 collimated beam, 69 922 power supply, 1000 W Xe lamp) onto the top plate of the STEG. The concentration ratio is fixed to 10 in this study. At the backside of the substrate, the V-shape TE legs are exposed to air for air convection cooling. No additional heat sink was used in the system. A small fan was used to create straight wind blowing through the TE legs with varying wind speed from 0 to 4 m/s. The absorber side (top side) was shielded from the wind with a plastic wall to minimize additional heat loss by fan air from the hot side. A thermocouple was attached at the top surface to measure the hot-side temperature. Using a current source, voltage–current and power–current plots were obtained to characterize the power generation performance of the STEG.

ANSYS Thermal Electric simulation tool was used for the simulation to analyze and optimize the device performance.21 The suggested V-type TE design was modeled in the ANSYS Platform, and the respective thermal and electrical boundary conditions were given accordingly as per the experimental setup. To find the best model configuration, there are two parameters that need to be optimized for the device—the length of the TE legs and the number of pairs attached within the device area. These two variables and the convective heat transfer coefficient at the sidewall of the TE legs were varied independently and the hot-side and cold-side temperatures and the open circuit voltage were obtained from the simulation and optimized for the highest power output. The conditions for the simulation and experiment were chosen by considering the concentration ratio heat flux, temperature-dependent material properties, and some other parameters used in our solar experiment. The parameters used in simulation and solar experiments are summarized in Table I. When the ambient energy flux over the aperture receiver is uniform, the geometric and optical concentration ratios are equal, i.e., C geo = C opt.24 Hence, the concentration ratio can be calculated as the ratio of the concentrator area to the beam size at the receiver or the absorber. Since the average flux over the aperture is set by the solar simulator, the flux over the absorber area was calculated with the effective lens area (81.07 cm2) and the beam diameter at the absorber plane (3.21 cm), which resulted in the solar concentration C = 10.0 suns, which is used as a simulation input. Total heat input is given by
Q rad , in = G C τ α A total ,
(2)
where G is the solar irradiation intensity per unit area for one sun, τ is the transmittance through the concentrator lens, α is the average absorptance of the solar absorber, and A total is the area of the solar absorber. Thermal radiation emissivity at the absorber plate was taken into consideration for the radiation output from the plate. The calculated heat input at the top plate is 6.919 W. The solar experimental conditions used for our study are summarized in Table I.
TABLE I.

Experimental conditions used for solar TE generation.

ParameterValue
Effective lens area 81.1 cm2 
Solar beam diameter on the absorber plate 3.21 cm 
Absorber plate area, Atotal 3 × 3 cm2 
Solar concentration, C 10 suns 
Solar irradiation for one sun, G 1 kW/m2 
Solar transmittance of Fresnel lens, τ 0.95 
Solar absorptance of the absorber, α 0.9 
Thermal radiation Emissivity of the top absorber plate, ε 0.8 
ParameterValue
Effective lens area 81.1 cm2 
Solar beam diameter on the absorber plate 3.21 cm 
Absorber plate area, Atotal 3 × 3 cm2 
Solar concentration, C 10 suns 
Solar irradiation for one sun, G 1 kW/m2 
Solar transmittance of Fresnel lens, τ 0.95 
Solar absorptance of the absorber, α 0.9 
Thermal radiation Emissivity of the top absorber plate, ε 0.8 

At 22 °C ambient room temperature and a calm air situation, the radiation emissivity of the two TE materials is around 0.1, and the heat transfer coefficient of natural convection is set to be 8 W/m2K. The chosen E-type thermocouple materials, an excellent thermocouple with excellent properties, both elements being resistant to corrosion and capable of operating at temperatures up to 2000 °F, and their high thermoelectric power are of advantage. The material composition of chromel is 90% Ni and 10% Cr with few amounts of Si, Fe, and Mn. Constantan is 55% Cu and 45% Ni with small percentages of Mn and Fe. We used the temperature-dependent properties of chromel and constantan—thermal conductivity, electrical conductivity, and Seebeck coefficient in the simulation.23,25 Note that for chromel, a constant Seebeck coefficient of 21 μV/K has been used due to the unavailability of temperature-dependent data for our chromel. The material properties of chromel and constantan are displayed in Fig. 2.

FIG. 2.

(a) Resistivity of chromel and constantan used in our experiments as a function of temperature; (b) Seebeck coefficient of constantan as a function of temperature.

FIG. 2.

(a) Resistivity of chromel and constantan used in our experiments as a function of temperature; (b) Seebeck coefficient of constantan as a function of temperature.

Close modal

The two main parts in the solar experiment of the device are to check the device’s performance on a normal operating condition of a clear and windless day with only solar irradiation and natural convection and performance on a windy day with forced air convection. The STEG device was placed at the appropriate distance from the Fresnel lens to achieve the target solar concentration. Top surface temperature, current, and voltage values were recorded as data by a LabVIEW program.

Four different wind speeds were tested: no wind (0 m/s), slow wind (2 m/s), medium wind (3 m/s), and fast wind (4 m/s). To measure the open circuit voltage and short circuit current of the thermoelectric generator device, an external source meter was used as an input current source. This current source mimics a variable load resistance to obtain the load voltage and current simultaneously. The power output is obtained from P = V × I as a function of I. A thermocouple was used to measure the top surface temperature, all saved in the LabVIEW program in real time.

Figure 3 shows the V–I and P–I curves for our STEG device under 10 sun concentration and no wind. The x- and y-axes intercepts show the short-circuit current ( I s c ) and the open circuit voltage ( V o c ), respectively. From the VI characteristic curve in Fig. 3, we find that V o c = 183.6 mV and I s c = 44.8 mA since the open-circuit voltage is determined by the Seebeck effect in the multiple TE pairs in series as
V o c = N pair S pair Δ T ,
(3)
where N pair (=36) is the number of TE leg pairs; S pair is the average total Seebeck coefficient of a pair of p-type and n-type TE legs over the temperature range, i.e., S pair = S p S n; and Δ T is the temperature difference across each TE leg, assuming Δ T across a p-type leg is approximately the same as that across an n-type leg. With S pair 61 μ V / K, Δ T is found to be 83.6 K. The internal resistance of the STEG is obtained as the slope of the linear VI curve, which is found to be 4.195 Ω. From the PI curve, the maximum power output of 2.07 mW or the maximum power density of 230 μW cm−2 is obtained at I = 22 mA, i.e., under the electric load matching condition. (Figure 3)
FIG. 3.

Voltage and power density of the developed STEG under no wind and 10 sun concentration as a function of current.

FIG. 3.

Voltage and power density of the developed STEG under no wind and 10 sun concentration as a function of current.

Close modal

Figure 4 shows the results of varying wind speed tests for the STEG. Open-circuit voltage, top surface temperature, and the estimated temperature difference across the TE legs from Eq. (2) are shown as the wind speed is varying in real time in the plot. When the STEG is in no wind situation, the hot side temperature reaches 191 °C at the steady state, and the temperature difference is found to be ∼83 K. When the wind speed is increased to 2 m/s, the hot side temperature is lowered to 151 °C and the average open circuit voltage also reduces to 101.2 mV and the temperature difference is lowered to 46 K due to increased air cooling. At the medium wind speed of 3 m/s, the hot side temperature is further reduced to 149 °C and the average output voltage and the temperature difference become 96.43 mV and 43.9 K, respectively. At the fast wind speed of 4 m/s, the hot side temperature remains almost the same as that of the medium wind at 149 °C, perhaps within the uncertainty level, but the reduced average open circuit to 91.91 mV indicates that the temperature difference is reduced further to 41.8 K. It is obvious that when forced convection is applied to the legs, the top temperature is reduced with increased heat sink and this reflects as a drop temperature difference as well. These values closely align with the simulation results that are discussed below.

FIG. 4.

Open-circuit voltage, top surface temperature, and calculated temperature difference across TE legs for varying wind speed in real time for the developed STEG under 10 sun concentration.

FIG. 4.

Open-circuit voltage, top surface temperature, and calculated temperature difference across TE legs for varying wind speed in real time for the developed STEG under 10 sun concentration.

Close modal

We also performed finite element simulation to analyze the experimental data further. Figure 5 presents the simulated results of the top surface temperature and the open-circuit voltage of the STEG as a function of the air convective heat transfer coefficient along with the experimental data. In the simulation, the convective heat transfer coefficient h was varied to replicate different wind speeds. The heat transfer coefficient corresponding to each of the different wind speeds used in the experiments was found from the best fit of the open circuit voltage data. As a result, we find that the corresponding h steadily increases from 25 W/m2 K at 2 m/s wind to 27 W/m2 K at 3 m/s and then to 29 W/m2 K at 4 m/s. These heat transfer coefficient values are in good agreement with theoretical and experimental values from literature.26–31 Here, we assumed 8 W/m2 K for no wind condition and at the top surface where the plate is shielded from the wind. The experimental top surface temperatures are also in reasonable agreement with the simulation as shown in Fig. 5, within the uncertainty range of ±15 °C, indicating the reliability of the simulation model. Overall, our numerical results indicate that the voltages and temperatures match closely, and importantly, all data points fall within the error bar region, indicating that the observed trend is consistent with the level of uncertainty associated with our measurements in the experimental results.

FIG. 5.

Theoretical and experimental top surface temperatures and open-circuit voltage with varying convective heat transfer coefficients at the sidewall of TE legs. The experimental open circuit voltages are matched to those of simulation to find the corresponding heat transfer coefficient for each wind speed (no wind, 2, 3, and 4 m/s).

FIG. 5.

Theoretical and experimental top surface temperatures and open-circuit voltage with varying convective heat transfer coefficients at the sidewall of TE legs. The experimental open circuit voltages are matched to those of simulation to find the corresponding heat transfer coefficient for each wind speed (no wind, 2, 3, and 4 m/s).

Close modal
This simulated curve for the open-circuit voltage in Fig. 5 allows us to predict the trend of power output change with wind speed. The maximum power output of a STEG can be obtained from the open-circuit voltage under the electric load matching condition as
P max = ( V o c ) 2 4 R ,
(4)
where R is the internal resistance of the STEG. Typically, R does not change much with temperature. Hence, the power output follows ∼ ( V o c ) 2. From Fig. 5, we find that the power output steadily decreases with increasing wind speed from 0 to 4 m/s and beyond (h varying from 8 to 35 W/m2 K). This result indicates that air convection cooling at the sidewall of TE legs is detrimental to the power output performance of the STEG. The simulation also shows that the maximum power output is possible when the heat transfer coefficient h is very small at ∼3 W/m2 K. Such a low heat transfer coefficient may be possible when the TE legs are insulated from the surrounding air with a low thermal-conductivity filler such as a polymer epoxy.

As a summary, we found that the corresponding heat transfer coefficient steadily increases with increasing wind speed, and air convection cooling at the sidewall of TE legs is detrimental to the power output performance of the STEG. Furthermore, we demonstrated that the maximum power output is possible when the heat transfer coefficient is very small. Such a low heat transfer coefficient may be possible when the TE legs are insulated from the surrounding air with a low thermal-conductivity filler such as a polymer epoxy. Overall, the study provides a valuable insight into the design and optimization of STEGs for improved power generation efficiency.

We also tested the impact of partial thermal insulation of the TE legs on the performance of a thermoelectric generator (STEG). An ethylene-vinyl acetate adhesive [3M™ hot-melt adhesive 3747, thermal conductivity of ∼0.32 W/(m K)] was applied at the hot side of TE legs as a filler in between TE legs for thermal insulation with varying thicknesses of the epoxy layer as schematically shown in the inset of Fig. 6. The filler was used to allow the wind to cool down only the lower part of the STEG, while keeping the hot-side temperature higher, with the goal of achieving a higher temperature difference across the TE legs.

FIG. 6.

Open-circuit voltage of the STEG with varying wind speed over time for different hot-side insulation layer thicknesses.

FIG. 6.

Open-circuit voltage of the STEG with varying wind speed over time for different hot-side insulation layer thicknesses.

Close modal

As shown in Fig. 6, at no wind condition, voltage outputs for different thicknesses remain similar in the range from 170 to 180 mV due to the insignificant change in temperature difference between the hot side and the cold side. The modules with no polymer filler, i.e., l = 0 cm, reach the steady state at the earliest due to the smallest overall heat capacity and show the largest steady voltage output. In different wind conditions, the module with a polymer filler keeps the voltage high due to its heat insulation at the hot side. As the insulation thickness increases from 0 to 4 cm, the voltage steadily increases and reaches the highest value at 4 cm, nearly doubling the voltage compared to that of 0 cm over the entire range of wind speed. As the insulation thickness further increases, the voltage decreases, which may indicate that the cold-side temperature also increases, thereby reducing the net temperature difference across the TE legs. Optimal insulation thickness is, thus, required to maximize the power output.

Overall, these results suggest that partial thermal insulation of the TE legs can significantly improve the performance of a thermoelectric generator, particularly, in windy conditions. The findings have important implications for the design and optimization of thermoelectric generators, and future studies may investigate the effect of different types of thermal insulation materials and the optimal insulation thickness for different operating conditions.

We performed further simulation of STEG for efficiency optimization with the TE leg length and heat transfer coefficient at the sidewalls. Two efficiencies can be defined here. The TE efficiency is the ratio of the power output to the heat input to the TE legs, while the system efficiency is defined as the ratio of the power output to the total solar irradiation to the system. The heat input to the TE legs is smaller than total solar irradiation due to heat losses by convection and out-radiation off the hot side plate. Therefore, the system efficiency is always lower than the TE efficiency.

Figure 7 shows the theoretical calculations of both the TE efficiency and the system efficiency of the fabricated STEG with varying air convection heat transfer coefficients and TE leg lengths. For all the TE leg lengths from 2 to 10 cm, the maximum efficiency occurs at a heat transfer coefficient less than 8 W/m2 K, which is below the natural air convection limit. This result confirms our experimental results with insulating fillers that suppressing sidewall air convection is desired to achieve a high STEG efficiency. As shown in Fig. 7, the optimal heat transfer coefficient for maximum efficiency decreases with the increasing length, which implies that longer wire-type TE legs would need a larger portion of their sidewalls to be insulated with low thermal-conductivity fillers for better thermal insulation, while shorter legs with length less than 2 cm may not need a filler, as long as the external wind is blocked from blowing through the legs, so that natural convection is ensured because the maximum efficiency is achieved at the natural air cooling condition with h ∼ 8 W/m2 K.

FIG. 7.

(a) Thermoelectric efficiency and (b) system efficiency with varying side-wall convective heat transfer coefficients for different leg lengths (L) of the metallic TE materials (chromel and constantan).

FIG. 7.

(a) Thermoelectric efficiency and (b) system efficiency with varying side-wall convective heat transfer coefficients for different leg lengths (L) of the metallic TE materials (chromel and constantan).

Close modal

The same study has been extended for state-of-the-art Bi2Te3 materials32,33 with varying TE leg lengths and heat transfer coefficients. As shown in Fig. 8, the maximum efficiency occurs at h less than 8 W/m2 K as similarly shown with metallic materials. By using these Bi2Te3 materials, our design could reach a maximum power density of 16.5 mW/m2 and a system efficiency of ∼2.8%. This could be reached with a leg length of 4–5 cm. Despite the promising results in terms of power density and system efficiency achieved with our design, the integration of state-of-the-art TE materials is likely to be challenging due to the need for high aspect-ratio legs and the associated difficulties in fabricating and characterizing these materials.

FIG. 8.

(a) Power density and (b) system efficiency with varying side-wall convective heat transfer coefficients for different leg lengths (L) of STEGs with state-of-the-art Bi2Te3 alloy TE materials.

FIG. 8.

(a) Power density and (b) system efficiency with varying side-wall convective heat transfer coefficients for different leg lengths (L) of STEGs with state-of-the-art Bi2Te3 alloy TE materials.

Close modal

These findings suggest that optimizing the length of the TE legs and controlling the sidewall heat transfer coefficient can significantly enhance the performance of STEGs, which could have important implications for developing efficient and cost-effective energy harvesting technologies for a wide range of applications.

In this study, we investigated the impact of sidewall air cooling on the power output of thermoelectric generators (TEGs) made of high aspect-ratio, V-shaped p/n TE couples without additional heat sinks. We observed that sidewall cooling lowered both the hot-side and cold-side temperatures as well as the temperature difference across the TE legs, resulting in reduced power and efficiency. We tested partial thermal insulation on the hot side of the TE legs using a low thermal-conductivity filler to increase the temperature difference across the TE legs. The module with no polymer filler under no wind condition produced the highest power density of 230 μW/cm2 and maximum system efficiency of 0.023% under 10 suns due to the optimized convective heat transfer to the ambient air, maximizing the end-to-end temperature difference across the elements. When the length of the epoxy layer is 4 cm, it keeps the voltage and power output nearly the same as no-wind performance even in a windy environment. The maximum system efficiency is found to be ∼0.07% after optimizing the leg length. The low efficiency is mainly due to the low ZT of the metallic materials used. However, with the low-cost metallic TE materials and without the need for expensive heat sinks, the system cost can be significantly lower than those of conventional solar TEGs, implying the potential practical application of the proposed STEG with cost benefits.

It is noted that our results may be applicable only to specific materials we used in our experiments. With different TE materials, an extensive analysis of the side-wall heat transfer might be necessary to optimize the STEG design for maximum power output or efficiency under different air convection conditions. According to our simulation, significant efficiency improvement is expected by replacing the materials with state-of-the-art Bi2Te3 alloys with ZT ∼ 1 to achieve a maximum power density of 16.5 mW/m2 and a system efficiency of ∼2.8% by optimizing the leg length with the same fill factor. It is noted, however, that most of the state-of-the-art TE materials are rigid and fragile, so they cannot be easily made in high aspect-ratio legs. An efficient heat sink at the cold side would still be needed to create sufficiently large temperature differences across the TE legs and increase the power output.

These findings have important implications for the development of future solar thermoelectric generators. The ability to increase power output by reducing sidewall cooling is especially relevant for devices where convection cooling can have a significant impact on performance. The use of a low thermal-conductivity filler on the hot side of the TE legs also offers an approach to improve the power output in windy environments, without the need for additional heat sinks.

This work was supported by the National Science Foundation (NSF) under Grant No. 1905571.

The authors have no conflicts to disclose.

Xinjie Li and Thiraj Mohankumar, are equally contributed to this work.

Xinjie Li: Formal analysis (equal); Methodology (lead); Visualization (equal); Writing – original draft (equal). Thiraj Mohankumar: Formal analysis (equal); Validation (lead); Visualization (equal); Writing – original draft (equal). Je-Hyeong Bahk: Conceptualization (lead); Resources (lead); Supervision (lead); Writing – review & editing (lead).

The data that support the findings of this study are available from the corresponding author upon reasonable request.

1.
L. L.
Baranowski
,
G. J.
Snyder
, and
E. S.
Toberer
,
Energy Environ. Sci.
5
,
9055
(
2012
).
2.
X.
Wen
,
J.
Ji
,
Z.
Song
, and
Z.
Li
,
Appl. Therm. Eng.
212
,
118594
(
2022
).
3.
Y.
Özcan
and
E.
Deniz
,
Eng. Int. J. Eng. Sci. Tech.
40
,
101362
(
2023
).
4.
Z.
Sun
,
D.
Luo
,
R.
Wang
,
Y.
Li
,
Y.
Yan
,
Z.
Cheng
, and
J.
Chen
,
Energy
256
,
124667
(
2022
).
5.
W.
He
,
L.
Tao
,
L.
Han
,
Y.
Sun
,
E.
Pietro Campana
, and
J.
Yan
,
Renew. Energy
179
,
96
(
2021
).
6.
N. S.
Lewis
et al.,
Report of the Basic Energy Sciences Workshop on Solar Energy Utilization
, U.S. Department of Energy, April 18–21,
2005
.
7.
M.
Telkes
,
J. Appl. Phys.
25
,
765
(
1954
).
8.
H. J.
Goldsmid
,
J. E.
Giutronich
, and
M. M.
Kaila
,
Sol. Energy
24
,
435
(
1980
).
9.
R.
Rush
, Solar Flat Plate Thermoelectric Generator Research, Tech. Doc. Rep. Air Force AD 605931 (General Electric Corp.) (
1963
).
10.
C. L.
Dent
and
M. H.
Cobble
, in
Proceedings of the 4th International Conference on Thermoelectric Energy Conversion
(IEEE,
1982
), pp.
75
78
.
11.
S. A.
Omer
and
D. G.
Infield
,
Sol. Energy Mater. Sol. Cells
53
,
67
(
1998
).
12.
R.
Amatya
and
R. J.
Ram
,
J. Electron. Mater.
39
,
1735
(
2010
).
13.
D.
Kraemer
,
B.
Poudel
,
H. P.
Feng
,
J. C.
Caylor
,
B.
Yu
,
X.
Yan
,
Y.
Ma
,
X.
Wang
,
D.
Wang
,
A.
Muto
,
K.
McEnaney
,
M.
Chiesa
,
Z.
Ren
, and
G.
Chen
,
Nat. Mater.
10
,
532
(
2011
).
14.
D.
Kraemer
,
Q.
Jie
,
K.
Mcenaney
,
F.
Cao
,
W.
Liu
,
L. A.
Weinstein
,
J.
Loomis
,
Z.
Ren
, and
G.
Chen
,
Nat. Energy
1
, 16153 (
2016
).
15.
G.
Chen
,
J. Appl. Phys.
109
, 104908 (
2011
).
16.
P.
Li
,
L.
Cai
,
P.
Zhai
,
X.
Tang
,
Q.
Zhang
, and
M.
Niino
,
J. Electron. Mater.
39
,
1522
(
2010
).
17.
P.
Sundarraj
,
D.
Maity
,
S. S.
Roy
, and
R. A.
Taylor
,
RSC Adv.
4
,
46860
(
2014
).
18.
P.
Sundarraj
,
R. A.
Taylor
,
D.
Banerjee
,
D.
Maity
, and
S. S.
Roy
,
J. Phys. D: Appl. Phys.
50
,
015501
(
2017
).
19.
J.
Xiao
,
T.
Yang
,
P.
Li
,
P.
Zhai
, and
Q.
Zhang
,
Appl. Energy
93
,
33
(
2012
).
20.
S.
Lv
,
W.
He
,
Z.
Hu
,
M.
Liu
,
M.
Qin
,
S.
Shen
, and
W.
Gong
,
Energy Convers. Manag.
196
,
69
(
2019
).
21.
D.
Luo
,
R.
Wang
,
Y.
Yan
,
Z.
Sun
,
W.
Zhou
, and
R.
Ding
,
Renew. Energy
180
,
1266
(
2021
).
22.
R.
He
,
G.
Schierning
, and
K.
Nielsch
,
Adv. Mater. Technol.
3
,
1700256
(
2018
).
23.
A.
Guan
,
H.
Wang
,
H.
Jin
,
W.
Chu
,
Y.
Guo
, and
G.
Lu
,
Rev. Sci. Instrum.
84
,
043903
(
2013
).
24.
P. L.
Singh
,
R. M.
Sarviya
, and
J. L.
Bhagoria
,
Appl. Energy
87
,
541
(
2010
).
25.
B.
Sundqvist
,
J. Appl. Phys.
72
,
539
(
1992
).
27.
P.
Tikuisis
and
R. J.
Osczevski
,
J. Appl. Meteorol.
41
,
1241
(
2002
).
28.
29.
Z.
Brodnianská
and
S.
Kotšmíd
,
Energies
13
,
973
(
2020
).
30.
S. W.
Churchill
and
M.
Bernstein
,
J. Heat Transfer
99
,
300
(
1977
).
31.
P. A.
Sipple
and
C.
Passel
, in
Proceedings of the American Philosophical Society
(American Philosophical Society,
1945
), Vol. 89, pp.
177
199
.
32.
B.
Poudel
,
Q.
Hao
,
Y.
Ma
,
Y.
Lan
,
A.
Minnich
,
B.
Yu
,
X.
Yan
,
D.
Wang
,
A.
Muto
,
D.
Vashaee
,
X.
Chen
,
J.
Liu
,
M. S.
Dresselhaus
,
G.
Chen
, and
Z.
Ren
,
Science
320
,
634
(
2008
).
33.
X.
Yan
,
B.
Poudel
,
Y.
Ma
,
W. S.
Liu
,
G.
Joshi
,
H.
Wang
,
Y.
Lan
,
D.
Wang
,
G.
Chen
, and
Z. F.
Ren
,
Nano Lett.
10
,
3373
(
2010
).