Thermoradiative energy conversion in infrared interband cascade InAs/GaSb/AlInSb/GaSb type-II superlattice diodes Open Access

Radiative heat transfer, driven by the temperature difference between an emitter and an absorber, can be utilized for thermal energy conversion thermodynamically described by the Carnot cycle. This principle is harnessed in photovoltaic devices, where solar radiation is absorbed by a semiconductor diode, converting light into electrical energy. A related but distinct application is found in thermoradiative (TR) devices, where a semiconductor diode in direct contact with the thermal source can generate electrical power by emitting radiation toward a cooler environment. The operating principles for TR energy conversion and efficiency limits have been previously outlined.^1,2 This technology offers a versatile platform for thermal-to-electrical energy conversion, finding applications from reclaiming waste heat to supporting power generation from radioisotopes in space missions.³ Moreover, TR cells can be integrated with photovoltaic (PV) systems to enhance overall conversion efficiency, offering a synergistic approach to energy harvesting and utilization. A theoretical analysis of the thermoradiative-photovoltaic (TR-PV) mode was reported previously.² 

Calculations of theoretical efficiency limits¹ suggest that an ideal diode operating in TR mode could achieve a power density as high as 54.8 W/m² when the diode temperature is 300 K while facing a 3 K cold reservoir. In realistic devices, non-radiative processes such as Shockley–Read–Hall (SRH) and Auger recombination will have a dramatic impact on energy conversion efficiency, restricting the open circuit voltage (V_OC) to far below the thermal voltage (V_th = kT/q), resulting in linear current–voltage (I–V) characteristics and consequently reducing the maximum achievable power well below theoretical predictions.⁴ The projected impact of Auger recombination on TR cells was reported for narrow bandgap semiconductors,⁵ indicating sharp limitations on performance at the desired temperature for TR cells. The detailed analytical models based on well-established material parameters projected a maximum power generation limit of ∼10 W/m² for type-II superlattices (T2SLs) and HgCdTe materials at a device temperature of 400 K.⁶ Further practical implementations are expected to yield power densities below 1 W/m², reflecting the significant impact of non-radiative Auger and SRH processes.⁶ 

Thermoradiative energy conversion has been experimentally demonstrated using commercial HgCdTe photodiodes. A radiative efficiency of 1.8% and an electrical power density of 2.23 mW/m² were reported at a temperature differential of 12.5 °C for a light-emitting diode emitting at a peak wavelength of 4.7 μm.⁴ The experiment utilized commercial HgCdTe photodiodes with fixed areas (0.1 × 0.1 mm²) operating at a fixed temperature with a variable scene temperature of 8–40 °C. While the results demonstrate power density dependence on the temperature differential, the critical influences of Auger and SRH non-radiative processes that depend on cell temperature are inaccessible in the measurements. Vurgaftman and Meyer⁶ have numerically shown that T2SL structures effectively suppress non-radiative Auger losses in TR cells, enhancing power density and conversion efficiency. However, experimental validation of these effects remains scarce. Our work bridges this gap, demonstrating T2SL-based TR cell performance across a range of cell temperatures to clarify the role of non-radiative processes, including elevated temperatures above room temperature. This experimental insight is essential for advancing the practical understanding of this energy conversion mechanism and optimizing TR cell performance under realistic operating conditions.

This report investigates the potential of thermoradiative energy conversion using an infrared (IR) type-II superlattice diode where Auger processes are not expected to be a limiting factor. Devices of variable areas were tested over a wide temperature range (50–120 °C), while the temperatures of the TR cell and its surroundings were independently regulated. The generated power from the experiments is also compared with the calculated predictions for T2SL structures.

The device structure is based on a narrow-bandgap interband cascade T2SL emitter, meticulously designed with a three-stage configuration featuring layer thicknesses of 570, 644, and 741 nm, respectively. Each period of the T2SL emitter is composed of a carefully engineered sequence of four layers: InAs (20 Å), GaSb (15 Å), Al_0.2In_0.8Sb (7 Å), and GaSb (15 Å). The emitter stages were initially designed with varying thicknesses to ensure current matching across all stages to function as a photovoltaic cell. Therefore, the multistage emitter might reduce the maximum current density in the TR cell operating mode. The complete structure of the TR device is illustrated in Fig. 1 with top and bottom electrodes, including the heat source to control the TR cell temperature. The respective top-view microscopic image of a fabricated circular mesa device is shown to the right of the figure. The device has an effective emitter area of 7.07 × 10⁻⁴ cm², where a half-metal ring is deposited on the side as an electrode.

The quantum efficiency (QE) of the T2SL emitters is strongly influenced by both the emitter thickness and the operating temperature. It has been observed that at elevated temperatures, the QE decreases significantly for T2SL thicknesses in the range of 0.8–2.0 μm.⁷ Consequently, the emitter thicknesses in this design were kept below 0.8 μm to mitigate this effect. The hole barrier comprises seven InAs/AlSb quantum wells (QWs) with a total thickness of 47.4 nm, while the electron barrier consists of three GaSb/AlSb QWs with a combined thickness of 23.3 nm. These barrier layers are characterized by larger bandgaps (>0.5 eV) relative to the emitter layers, effectively confining free carriers within the emitter regions, thereby increasing the probability of radiative recombination. The entire material structure was grown on a p-type GaSb substrate utilizing molecular beam epitaxy (MBE) to ensure precise layer composition and thickness control.

In a traditional single p–n junction cell, power generation for solar or thermophotovoltaic operation occurs via net minority carrier generation producing a forward bias and reverse current flow, while TR operation occurs via net minority carrier recombination producing a reverse bias and forward current flow. The T2SL device in this work utilizes an interband cascade structure with three stages, incorporating absorption regions between electron and hole barriers with broken-gap tunnel junctions connecting each stage. The energy band diagram and current transport mechanisms for thermophotovoltaic operation are detailed in Ref. 9. Analogous to the single p–n junction case, TR cells operate in reverse with each stage acting as a net radiative emitter with carrier flow in the opposite direction and reverse bias developed in the power-generating regime. For a comprehensive understanding of the material growth, device design, and energy band structure, readers are referred to previous detailed reports.^8,9

A series of circular and square mesa TR devices were fabricated using standard microfabrication procedures, including contact photolithography, chemical etching, metal evaporation, and liftoff. The phosphoric and citric acid solution (H₃PO₄/C₆H₅O₇/H₂O₂/H₂O = 15 ml/20 g/15 ml/120 ml) was used as wet etch chemistry for device mesa isolation. The average etch rate is 500 nm/min for InAs/InAsSb/AlAsSb and 450 nm/min for GaSb. A 200 nm SiO₂ passivation layer was deposited via plasma-enhanced chemical vapor deposition (PECVD) to reduce leakage currents. The SiO₂ layer was deposited after the mesa etch, covering both the surface and perimeter sidewalls to reduce perimeter surface recombination. Ti/Au (50/200 nm) layers were deposited by e-beam evaporation for both n-type (top layer) and p-type contacts (bottom layer). Optical and electrical measurements were performed to assess emission properties and analyze diode behavior. Finally, the devices were wire-bonded for TR performance characterization at elevated temperatures.

Optical emission was measured using photoluminescence spectroscopy with a 1064 nm continuous-wave (CW) laser excitation and a VERTEX 80v Fourier transform infrared (FTIR) spectrometer from Bruker Optics. Spectral response measurements were done using a Thermo Fisher Scientific (iS50R) FTIR spectrometer, optical chopper, and SR860 500 kHz DSP lock-in amplifier (Standard Research Systems). Temperature-dependent I–V measurements were taken using a closed-cycle cryogenic probe station from Lake Shore Cryotronics and a semiconductor parameter analyzer.

The energy conversion experiment was conducted using a specially designed apparatus at Advanced Cooling Technologies (ACT). The system includes a heat sink chamber controlled with liquid nitrogen and TR cell heaters. The cell was mounted on a sample holder inside the cold chamber, and its position could be adjusted using a rod passing through the cover of the chamber. More details on the experimental setup can be found in a previous report.¹⁰ Ten thermocouples (TCs) were used to monitor and maintain temperatures at key locations of the cold chamber wall, dry nitrogen flow inside the chamber, and the backside of the thermoradiative cell.

The real-time temperature of the experiments was recorded and shown in Fig. 2. TC numbers 1–8 represent the cold chamber wall temperature in different clock positions. TC9 was used to measure the gas (i.e., dry nitrogen) temperature inside the cold chamber. TC10 was used to monitor the TR cell temperature while a coupled heater was controlled to increase the cell temperature. The measurements were taken at steady-state TR cell temperatures as indicated by numbers in Fig. 2. The surrounding cold chamber temperature was set to −100 °C, while the cell temperatures were varied at four distinct temperatures: 50, 76, 99, and 121 °C. At each temperature, a steady-state condition was ensured before recording the current-voltage characteristics using a Keithley Source Meter Unit (model 2450). The device was measured at a steady temperature state after maintaining a constant temperature for ∼30 min. Given the relatively thin device layers (thickness from emitter 1 to emitter 3 is ∼2 μm), the estimated temperature difference between emitter 1 and emitter 3 should be <1 mK. As a result, a uniform temperature distribution across the multistage emitter device is expected.

Temperature-dependent photoluminescence (PL) spectra are shown in Fig. 3. At low temperatures (80–100 K), the PL peak is observed around 2.62 μm, shifting to ∼2.68 μm as the temperature approaches room temperature. The relatively narrow PL spectra shape is expected from direct bandgap materials when radiative band-to-band transitions are dominant. The PL intensity diminishes at higher temperatures due to increased nonradiative recombination rates. The band edge of the TR cell is near 2.7 μm at room temperature. The spectral response of the TR cell at room temperature is shown in Fig. 4. The cutoff wavelength (∼50% point) similarly corresponds to the band edge location of 2.7 μm. The PL and spectral response demonstrate agreement in absorption and emission spectra for the TR cell. This reciprocity is an indication of well-behaved radiative diode performance. The spectral response peak is observed at 2.1 μm, corresponding to the original design as a short-wavelength infrared (SWIR) photodetector.⁸ The same structure was also used for thermophotovoltaic (TPV) applications, as reported previously.⁹ 

The current–voltage characteristics for devices with varying active areas were measured at room temperature, providing insight into diode performance. The dark I–V curves of the TR devices are presented in Fig. 5(a), demonstrating dark current densities ranging from 4.5 × 10⁻³ to 2.1 × 10⁻² A/cm² at 100 mV reverse bias. The dark current density dependence on the perimeter to area (P/A) ratio is shown in Fig. 5(b). The observed area-dependent dark current density indicates contributions from perimeter surface recombination. The behavior suggests that further attention to sidewall passivation would be expected to improve TR cell performance.

The values of (J_B) and (J_S) can be extracted from this relation via the intercept and the slope, respectively. The extracted values at −0.1 V bias are J_B = 5.5 × 10⁻³ and J_S = 1.6 × 10⁻⁵ A/cm², respectively. The bulk parameters have a major contribution to the dark current density. The surface current J_S increases with increasing reverse bias, which is attributed to trap-assisted tunneling in the higher electric field.¹² The bulk dark current density, J_B follows typical diode behavior according to diffusion and Shockley–Read–Hall processes.¹² 

The extracted (R₀A)_Bulk and ρ_Sidewall values are 12.42 Ω·cm² and 3.55 kΩ·cm, respectively. The sidewall resistivity is similar to (and approximately double) prior reports of devices fabricated from this material structure.⁸ 

Temperature-dependent dark I–V characteristics spanning 80–290 K are shown in Fig. 6. The dark current of the TR cell decreases with decreasing temperature as expected. A circular device with a diameter of 200 μm and a respective area of 3.14 × 10⁻⁴ cm² was chosen for this measurement. The dark current density at 290 K and 100 mV reverse bias is 7.62 × 10⁻⁵ A/cm² and decreased to 2.24 × 10⁻⁹ A/cm² at 80 K with the same bias voltage. These low dark current density values are in agreement with the literature values of dark currents associated with T2SL due to suppression of Auger recombination.¹³ Dark current shows no significant voltage dependency in the reverse bias regime across all temperature ranges, indicating no perimeter surface leakage current is dominant in the TR device. In addition, series resistance is not observed to be a significant factor in the forward bias regime. The activation energy E_a of 0.13 eV was calculated for temperatures ranging from 120 K and above, which corresponds to approximately one-third of the bandgap energy (E_g/3). This indicates that the generation–recombination (G–R) process dominates the carrier transport process, including shallow defect trap-assisted tunneling.

The TR cell was initially kept at room temperature (25 °C), with no detectable power generation. The TR cell temperature was increased to 50, 76, 99, and 121 °C, respectively, with measured I–V characteristics shown in Fig. 7 for a circular device with an area of 7.07 × 10⁻⁴ cm². The power generation from the second quadrant of the I–V characteristics is shown in the magnified area near 0 V in the figure inset. The I–V curves in the power-generating regions are linear as the diode operates at a smaller voltage than the thermal voltage. The current increases with increasing temperature, with an accompanying increase in power generation. The short-circuit current (I_SC) increases from 36.5 nA to 2.95 μA when the temperature of the TR cell rises from ambient (298 K) to 121 °C (394 K). The respective current density increases 80 times from 5.16 × 10⁻⁵ to 4.17 × 10⁻³ A/cm².

The power density generated over the range of I_SC and V_OC is shown in Fig. 8 for a circular device with an area of 7.07 × 10⁻⁴ cm². The maximum power (P_max) density is 0.18, 0.12, 1.30, and 2.87 mW/m² for TR cell temperatures of 50, 76, 99, and 121 °C, respectively. The generated power at 76 °C is slightly lower than the equivalent power at 50 °C due to the low produced V_OC at 76 °C shown in the inset of Fig. 7, where the source of this deviation from expected behavior will be described subsequently. We believe that the generated power would continue to increase at cell temperatures beyond the maximum cell temperature of 121 °C (394 K) allowed by the measurement setup. Theoretically, higher-bandgap materials are expected to be less efficient due to reduced radiative emission at moderate temperatures while providing the ability to facilitate higher temperature operation before performance is limited by Auger recombination. This work demonstrates among the highest operating temperatures reported for TR power generation and is in a regime where SRH recombination is dominant. The power density produced is similar to the report by Nielsen et al.⁴ for narrower-bandgap material operating at lower cell temperatures. Increasing TR cell power density will likely require an intermediate regime that optimizes non-radiative pathways (Auger vs SRH-limited behavior) and the desired operating temperature. The data in Fig. 10 illustrates models based on detailed balance and SRH pathways, while performance will depend significantly on intrinsic non-radiative recombination parameters for particular material technologies employed.

Multiple devices with different areas were measured to analyze the energy conversion characteristics of the fabricated TR cells and to illustrate observed variability in measured performance. The full mesa area is assumed without considering shading from the top contact metal. The half-ring front contact (Fig. 1) was designed to minimize shading, and is ∼10% of the mesa area, which is a factor that can be optimized in future device development. The characteristics of the measured devices with increasing TR cell temperatures are shown in Fig. 9, for (a) short-circuit current density, (b) open-circuit voltage, and (c) maximum generated power density. Device 1 and Device 2 have the same size and area of 7.07 × 10⁻⁴ cm², while Device 3 has a larger area of 2.5 × 10⁻³ cm². The short-circuit current density (J_SC) and generated power density demonstrate expected increasing trends with increasing TR cell temperature. The larger device (Device 3) has a smaller short-circuit current density and corresponding smaller maximum power density in comparison to the smaller area devices. The reduced short-circuit current density for the larger device is unexpected given the higher R₀A, requiring further investigation. Possible explanations for this include non-radiative pathways that are not reflected in the dark current characteristics, area-dependent optical outcoupling, and area-dependent minority carrier transport. There is significant variability in open-circuit voltage, without obvious dependence on temperature. The combination of low values of V_OC (<1 mV), the relatively small temperature range studied, and the expected logarithmic dependence of V_OC on temperature likely all contribute to the observed variability in V_OC.

A comparison (Fig. 10) was performed of measured maximum power densities to predictions based on prior models,^6,14 and previously published parameters for the T2SL.^15,16 The maximum power density predicted by detailed-balance¹⁴ assuming perfect external luminescent efficiency, presents the upper-performance limit. A comparison between the experimental data of this work and the ideal case suggests an external luminescent efficiency of ∼2%. This also highlights that further improvements in external luminescent efficiency will significantly enhance the power output of TR cells, underscoring its importance for future advancements in this field. The shaded regions of the figure show expected power density ranges calculated using the model by Vurgaftman and Meyer⁶ for both planar photon-extraction geometry and full photon-extraction over a hemisphere. These ranges correspond to different assumed values for SRH lifetime and the resulting impact of non-radiative losses. In this temperature range for the T2SL bandgap energy, the TR performance is SRH-limited with negligible impact from Auger processes. Assuming SRH lifetimes for these T2SL structures in the range of 10–100 ns (consistent with prior results for InAs/GaSb T2SLs^17–19), the measured maximum power density based on the material parameters assumed would seem to exceed predicted values for photon extraction in a planar geometry while approaching values for full hemispherical photon extraction. While the mesa device and passivation/anti-reflection coating may provide some improvement over a planar geometry, we would expect photon extraction to be closer to expectations for a planar geometry given the device diameter. Given the very strong dependence of the model prediction on assumed values for T2SL parameters, non-radiative processes, and photon extraction, it is difficult to directly identify the most significant source of variation in reaching an accurate agreement between the model and experimental results. The three-stage cascaded device structure may also offer performance advantages that are not directly accounted for in the model. Further studies to measure parameters independently (e.g., cells with variable bandgap energy, cells with differing photon extraction approaches, single junction vs cascaded designs, etc.) are needed to resolve this issue to further understand and develop models to predict power generation in TR cells. Based on the experimental studies in this work, the achievement of substantial increases in power generation is expected for narrower bandgap materials to increase radiative emission while ensuring the management of non-radiative Auger processes.

We have experimentally demonstrated the TR energy conversion using a narrow-bandgap interband cascade InAs/GaSb/AlInSb/GaSb T2SL with an effective bandgap of 0.39 eV near room temperature. Our TR cell achieved a power density of 2.9 mW/m² at a cell temperature of 121 °C with a device area of 7.07 × 10⁻⁴ cm². The T2SL device under study demonstrates the expected TR cell power generation in a regime where efficiency is limited by non-radiative SRH recombination. The findings establish a strong foundation for future development, particularly in exploring materials with smaller bandgaps and strategies to reduce non-radiative recombination to enhance TR radiation and overall energy conversion efficiency. These advancements will be crucial for leveraging TR technology in various applications. These initial experimental results highlight the promise of TR devices for applications in waste heat recovery and energy generation in extreme environments, paving the way for further advancements in this emerging technology.

We acknowledge the assistance of Jonathan Murray and Phil Martin in building the thermoradiative measurement system at ACT. The work at UD is based on work supported by the National Science Foundation under Grant No. ECCS-2317609. This article has been authored by an employee of National Technology & Engineering Solutions of Sandia, LLC under Contract No. DE-NA0003525 with the U.S. Department of Energy (DOE). The employee owns all rights, title, and interest in and to the article and is solely responsible for its contents. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this article, or allow others to do so, for United States Government purposes. The DOE will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan: https://www.energy.gov/downloads/doe-public-access-plan.

The authors have no conflicts to disclose.

Md Toriqul Islam: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (lead); Writing – review & editing (equal). Sheikh Alimur Razi: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). Nuha Ahmed Babikir: Data curation (equal); Methodology (equal); Writing – review & editing (equal). Jianjian Wang: Data curation (equal); Methodology (equal); Writing – review & editing (equal). Rui Q. Yang: Methodology (equal); Writing – review & editing (equal). John F. Klem: Methodology (equal); Writing – review & editing (equal). Jamie D. Phillips: Conceptualization (equal); Funding acquisition (equal); Supervision (equal); Writing – review & editing (equal).

The data that support the findings of this study are available within the article.