Near-field thermophotovoltaics for efficient heat to electricity conversion at high power density

The study addresses how to directly convert heat to electricity efficiently at high power density using thermophotovoltaic (TPV) devices that leverage near-field (NF) radiative transfer. Conventional far-field TPVs, limited by the Stefan–Boltzmann law, have achieved high efficiencies but modest power densities because only propagating modes contribute. By contrast, in the NF (nanoscale gaps), evanescent modes can dominate energy transfer, enabling super-Planckian exchange and potentially higher power outputs. Despite extensive theoretical predictions, experimental progress has been limited by challenges in fabricating high-temperature-stable emitters, PV cells tailored to absorb above-band-gap (ABG) NF radiation while suppressing sub-band-gap (SBG) absorption, and maintaining parallelism and nanometer-scale gaps. The purpose of this work is to realize and study efficient NF TPV energy conversion between planar surfaces at room-temperature PV operation, quantify performance versus gap and temperature, and elucidate the underlying physical mechanisms.

Prior far-field TPVs have reported efficiencies up to ~30% at emitter temperatures around 1450 K but are power-density limited by blackbody radiation. Theoretical and computational works since early predictions have suggested that NF TPV can exceed blackbody limits and achieve high performance via evanescent mode contributions and spectral engineering (e.g., surface modes, thin-film reflectors, photonic crystals, graphene/plasmonic concepts). Experimental NF TPV demonstrations have shown enhanced power relative to far field but with low efficiency and power density: earlier systems with Si emitters and InAs or InGaAs cells reported efficiencies below ~1% and power densities up to ~120 W/m^2 at ~1040 K; another sphere-plane study used cryogenically cooled InSb cells to achieve high cell efficiency but with low overall system efficiency due to cooling overhead. These results highlight the need for high-temperature-stable emitters, spectrally selective PV cells, and precise, parallel, nanoscale gaps to advance NF TPV.

Devices: The emitter is a monolithic, heavily boron-doped (~3×10^19 cm^−3) silicon cantilever featuring a circular mesa (150 µm diameter) recessed from a surrounding ring (rec) by 15 µm within a 190 µm diameter island. Two stiff Si beams (each 20 µm wide, 270 µm long, 45 µm thick) provide electrical resistance (R_emitter ≈ 80 Ω), thermal conductance (~400 µW K^−1) and mechanical stiffness (~2 kN m^−1). Joule heating via a bipolar voltage across the beams raises the mesa temperature up to 1270 K. A 10-nm AlN layer conformally coats the emitter, providing electrical insulation and a diffusion barrier. The PV cell is a thin-film InGaAs/InP heterostructure (100 nm n+ InGaAs / 100 nm n+ InP / 1000 nm n InGaAs, E_g = 0.75 eV / 200 nm p+ InP) epitaxially grown on InP and transferred to a Si handle wafer with 2 µm Parylene-C and 400 nm Au bottom contact; the top and bottom Au serve as contacts, with the bottom Au acting as a back surface reflector (BSR) to recycle SBG photons. The PV active area is circular (190 µm diameter) under a 20 µm wide Au contact ring. AFM and dark-field microscopy confirm smooth, clean, planar surfaces (mesa peak-to-peak roughness ~1 nm; PV active surface ~4 nm). Experimental setup and alignment: The emitter and PV cell are mounted in a high-vacuum chamber (~1 µTorr) on a custom nanopositioning platform enabling lateral alignment and tip/tilt control (~6 µrad resolution). Coarse parallelization is achieved optically with a 50× objective, limiting cross-surface nonparallelism to ~80 nm. After heating, in situ parallelization is iteratively refined by approaching to contact, measuring the PV maximum power (P_MPP), withdrawing, and adjusting tip/tilt to maximize P_MPP, achieving ~15 nm deviation over the 150 µm mesa. Gap control and contact detection: The PV cell is approached from an initial ~7 µm gap toward the emitter using a feedback-controlled piezoelectric actuator, with large steps (~800 nm) then fine steps (~2 nm). A laser beam reflected from the emitter backside onto a segmented photodiode monitors deflection; a small 4 kHz AC modulation (~2 nm) on the piezo provides an optical AC signal. Sudden simultaneous changes in the optical signal, emitter resistance (rapid cooling upon contact), and PV short-circuit current indicate mechanical contact. Electrostatic effects are minimized by a bipolar drive keeping the mesa potential near ground. Temperature measurement: Emitter temperature T_emitter versus Joule power is calibrated using ultra-high-vacuum scanning thermal microscopy (UHV-STHM) with a temperature-sensing probe in contact with the heated emitter. Uncertainty is ±27 K at 1270 K and ≤±10 K at lower temperatures. Electrical measurements: At each gap size, the PV I–V curve is recorded using a Keithley 2401 sourcemeter to extract short-circuit current (I_sc), open-circuit voltage (V_oc), fill factor (FF), and maximum power point (P_MPP = I_sc·V_oc·FF). Example at 930 K: FF ≈ 0.73 at 100 nm. Modeling: Radiative transfer and PV output are modeled using fluctuational electrodynamics with a numerically stable scattering-matrix formalism, approximating laterally infinite multilayers. Total radiative heat transfer (Q_RHT) from emitter to PV layers (including mesa and recessed regions) is calculated spectrally; the photon flux into the active layer yields photocurrent, which is inserted into a PV diode model to compute I–V and P_MPP. Spectral decomposition distinguishes ABG and SBG contributions; transmission functions versus photon energy and parallel wavevector quantify propagating and evanescent modes.

- Record NF TPV performance with room-temperature PV cells: power density ~5 kW/m^2 at T_emitter = 1270 K and gaps <100 nm, with measured system efficiency η ≈ 6.8% (η = P_MPP / Q_RHT). - At 930 K, as gap decreases from 7 µm to 70±2 nm: P_MPP rises from ~2 µW (≥600 nm) to 14.8 µW at the smallest gap, an ~8× enhancement including recessed-ring contributions; considering only the mesa area, the NF enhancement is ~11×. Calculated Q_RHT increases from ~72 µW at 7 µm to ~1 mW at 70 nm. - I–V behavior shows increasing I_sc and moderately increasing V_oc with decreasing gap: at 930 K, I_sc increases from 9.8 µA (7 µm) to 56 µA (100 nm). Fill factor at 100 nm is ~0.73. - Temperature scaling: At 1270 K, I_sc increases by ~5× when reducing gap from 7 µm to ~100 nm. Between 1050 K and 1270 K, I_sc increases from ~30 µA to ~150 µA (at similar gaps). V_oc scales logarithmically with J_sc, consistent with PV diode behavior; experimental V_oc and J_sc agree with the model across temperatures and gaps. - P_MPP vs gap at multiple temperatures (810–1270 K): NF enhancement observed at all T; e.g., at 1050 K, P_MPP increases from ~7 µW (7 µm) to 41 µW (90 nm), ~6× enhancement. Nonmonotonic variations at 0.5–7 µm are due to interference effects, captured by the model. - Efficiency trends: η increases with temperature for all gaps; at 100 nm, η rises from ~0.5% (810 K) to 6.8% (1270 K). At fixed T, η initially dips at intermediate gaps (~400–500 nm) due to faster SBG increase than ABG, then rises at smaller gaps as ABG surpasses the blackbody limit. - Spectral insights (1270 K): At 100 nm gaps, ABG transfer to the active layer exceeds the blackbody limit, while SBG remains below blackbody due to the thin-film BSR. Estimated fractions at 100 nm: ~26% of Q_RHT absorbed in the active layer (P_AL), of which ~32% is extracted electrically; residual SBG absorption includes ~14% via low-frequency surface phonon-polaritons and ~55% in the Au BSR (0.074–0.74 eV range). - Theoretical vs experimental efficiency at 1270 K: model predicts η ≈ 8.3% vs measured ~6.8% (~18% difference), attributable to temperature uncertainty, dielectric property modeling, PV series/shunt resistances, or slight PV heating. - Compared to prior NF TPV reports, this work achieves >10× higher power density and ~6× higher efficiency at room-temperature PV operation.

The results directly demonstrate that placing a high-temperature, planar Si emitter within nanometer-scale gaps of a spectrally optimized thin-film InGaAs PV cell enables evanescent-mode coupling that strongly enhances ABG radiative transfer while maintaining SBG suppression through a back surface reflector. This broadband, wavevector-selective coupling increases I_sc and, with favorable diode characteristics, yields substantial gains in P_MPP even at lower emitter temperatures than required in comparable far-field TPVs. The observed efficiency dip at intermediate gaps is explained by the more rapid increase of SBG relative to ABG transfer as the gap shrinks from microns to hundreds of nanometers; at deep subwavelength gaps, ABG transfer dominates, increasing efficiency. Agreement between measurements and fluctuational electrodynamics modeling across temperatures and gap sizes validates the physical picture and design strategy. Remaining losses are dominated by SBG absorption in the Au reflector and thermalization within the active layer, suggesting clear engineering pathways for further improvements (e.g., alternative reflectors, air-gap cells, spectral shaping of emitters).

This work establishes a high-temperature-stable, parallel, planar near-field TPV platform that converts heat to electricity at record room-temperature power density (~5 kW/m^2) and substantially improved efficiency (~6.8%) using a doped-Si emitter at 1270 K and an InGaAs thin-film PV cell at gaps <100 nm. Comprehensive measurements vs temperature and gap, corroborated by rigorous modeling, elucidate the role of evanescent modes in enhancing ABG transfer while suppressing SBG. The approach provides a foundation for NF TPV nanotechnologies and points to clear routes for further gains: engineering emitters (e.g., photonic crystals, metamaterials) for spectral control at high temperature; optimizing PV architectures (e.g., air-gap cells, improved back reflectors) to suppress SBG and reduce nonradiative/ohmic losses; and advancing device planarity, stability, and integration for practical systems. Future work should investigate long-term emitter stability, protective coatings, pressure effects, and system-level designs for scalable power generation and waste heat recovery.

- Reported efficiency excludes heat losses from conduction and radiation from surfaces not facing the PV cell; thus, system-level efficiency may be lower. - Efficiency in the NF is limited by SBG absorption, especially in the Au back reflector; this reduces η relative to the far field and calls for improved reflector designs (e.g., air-gap cells). - Uncertainty in emitter temperature (±27 K at 1270 K) and material dielectric properties at high T contributes to ~18% discrepancy between measured and predicted η at the highest temperature. - PV device series/shunt resistances and possible slight PV heating can degrade FF and V_oc, impacting power output. - Only the mesa region enters the NF; the recessed ring remains in the far field, limiting total enhancement; full NF coverage would require different geometries. - Long-term stability of emitters at high temperature and under various protective coatings and pressures is not addressed; practical deployment requires durability studies. - Maintaining nanometer-scale gaps and parallelism over larger areas remains a technical challenge.