An Optoelectronic Thermometer Based on Microscale Infrared-to-Visible Conversion Devices

Accurate, spatially and temporally resolved temperature sensing is vital for industrial, environmental, and healthcare applications. Conventional thermoelectric/resistive sensors can be susceptible to electromagnetic interference, particularly during MRI. Optical sensors overcome these limitations by enabling remote, minimally invasive, EMI-immune measurements with high resolution. Luminescence-based thermometry can use intensity, wavelength, linewidth, and lifetime as transduction signals. Infrared thermography is non-contact but surface-limited and emissivity-dependent; cavity-based sensors can achieve high precision but often require complex, stationary spectrometers; liquid crystal thermography can be sensitive to ambient lighting and viewing geometry. Upconversion-based approaches that convert NIR (650–950 nm) to visible emissions reduce autofluorescence, improve tissue penetration, and offer convenient visible readouts, making them attractive for biomedical sensing. Optoelectronic NIR-to-visible upconversion devices based on semiconductor heterostructures have recently shown linear response, fast dynamics, and low excitation power, making them promising for implantable microscale sensing. This study investigates the temperature-dependent photoluminescence of such devices, elucidates the sensing mechanisms, demonstrates spatial mapping with device arrays, and validates in vitro and in vivo thermal sensing with fiber-integrated probes.

• Optical thermometry modalities include readouts based on luminescence intensity, wavelength shift, peak width, and decay lifetime.
• Infrared imagers provide non-contact, spatial temperature maps but are limited to surface temperatures and affected by emissivity.
• Optical cavity-based sensors can yield spectrally resolved, high-precision measurements but often need complex spectrometric setups, limiting biomedical deployment.
• Liquid crystal thermometry shows temperature-dependent color but is sensitive to ambient lighting, viewing angles, and polarization.
• Lanthanide-based upconversion nanoparticles provide multiple emission bands with strong temperature dependence and have been applied to biomedical thermometry, reducing autofluorescence and improving penetration.
• Semiconductor optoelectronic upconversion devices (NIR-to-visible) have emerged with linear response, fast dynamics, and low excitation power, suitable for microscale and implantable applications.

• Device architecture: Integrated optoelectronic upconversion device comprising a GaAs-based double-junction photodiode (low bandgap) in series with an InGaP LED (large bandgap). Epitaxial growth on GaAs substrate with a sacrificial interlayer, followed by lithographic definition and epitaxial lift-off to yield thin-film microscale devices (~300 × 300 µm²).
• Optical excitation and spectral measurement: NIR steady-state excitation in the 770–830 nm range at ~40 mW cm⁻² (verified not to cause additional photothermal effects in tissue). Red upconverted emission recorded by a fluorescence microscope coupled spectrometer. Measurements conducted over 25–90 °C (materials and encapsulants stable in this range). Ten devices characterized for statistical validation.
• Temperature sensitivity analysis: Extracted intensity-temperature sensitivity (relative) and peak wavelength-temperature sensitivity (absolute). Compared measurements with theoretical calculations using diode detailed balance and the Varshni relation for bandgap temperature dependence. Evaluated spectral noise and SNR within ±5 nm of the emission peak and derived temperature resolution from standard deviation of normalized intensity and peak wavelength at representative temperatures (27, 48, 65, 90 °C).
• Component-level studies: Fabricated individual microscale InGaP LEDs and GaAs double-junction photodiodes with identical epitaxial structures and similar geometries. Measured LED EL spectra at 25–90 °C under 20 µA injection to assess intensity and peak shifts. Measured photodiode EQE spectra versus temperature to analyze absorption edge shifts and peak changes due to junction photocurrent mismatch.
• Circuit operating point analysis: Recorded I–V characteristics of the InGaP LED and the GaAs photodiode (under 810 nm, 40 mW cm⁻²) at 25–90 °C. Determined operating points from intersection of I–V curves to assess temperature-dependent current/voltage within the series circuit and their impact on LED emission.
• Dynamic and spatial mapping: Performed cyclic heating/cooling (26–38 °C) and step temperature changes, monitoring PL dynamics. Fabricated large arrays (~2 cm²) containing ~1500 devices for spatially resolved thermal mapping; acquired PL images under uniform and nonuniform heating (e.g., hot airflow) and quantified temperatures using intensity–temperature calibration.
• Fiber integration and calibration: Employed epitaxial lift-off and transfer printing to mount encapsulated devices on quartz fiber tips. Implemented a Y-shaped fiber system with an ~810 nm NIR excitation LED and a spectrometer on separate branches; included a co-localized thermocouple for calibration. Assessed device stability in phosphate-buffered saline (PBS) up to 30 days at room temperature and mechanical robustness. Demonstrated in vitro monitoring of human exhalation temperature and reported correlation with thermocouple. In vivo deep-brain sensing capability described using the fiber-coupled probe.

• Temperature-dependent emission: Under 770–830 nm excitation (~40 mW cm⁻²), the upconverted red emission peak redshifts from ~625 nm to ~637 nm as temperature increases from ~25 to 90 °C; emission intensity decreases with temperature.
• Sensitivities: Intensity–temperature relative sensitivity ~1.5% °C⁻¹; spectrum–temperature absolute sensitivity ~0.18 nm °C⁻¹. Measurements across 10 devices show good agreement with detailed balance and Varshni-based calculations.
• Noise and resolution: Peak-window SNR (>±5 nm) above 15 at 27, 48, 65, and 90 °C. Derived temperature resolutions (intensity-based): 0.01, 0.06, 0.14, 0.49 °C at 27, 48, 65, 90 °C, respectively. Wavelength-based resolutions: 0.04, 0.08, 0.11, 0.22 °C at the same temperatures.
• Mechanism: Spectral shift primarily governed by InGaP bandgap narrowing with temperature. Intensity drop arises from efficiency reductions in both the InGaP LED (EL sensitivity ~1.2% °C⁻¹) and GaAs photodiode (EQE decrease and slight peak redshift due to photocurrent mismatch in the double junction) and from temperature-dependent circuit operating points (I–V intersection).
• Tunability: Intensity sensitivity depends on excitation conditions and device structure; can be tuned by changing excitation intensity.
• Temporal response: Upconversion devices exhibit ultrafast PL dynamics with ~20 ns decay, suitable for temporally resolved sensing.
• Spatial mapping: Large arrays (~2 cm², ~1500 devices, ~300 µm pitch) enable spatially resolved thermal imaging; hot airflow produces nonuniform temperature maps and emission extinguishes when local surface temperature exceeds ~95 °C.
• Fiber-integrated sensing: Fiber-coupled devices stable in PBS up to 30 days; portable Y-fiber system enables in vitro human exhalation monitoring with good agreement to a thermocouple (R² = 0.90). In vivo deep brain temperature monitoring demonstrated with the fiber-integrated sensor.

The study demonstrates that integrated semiconductor upconversion devices can act as precise optoelectronic thermometers, addressing the need for EMI-immune, minimally invasive, and remotely readable temperature sensing, particularly in biomedical contexts. By leveraging NIR excitation within the biological transparency window, the devices reduce autofluorescence and enable deeper tissue access while providing visible emissions that are easy to detect. The temperature dependence of both emission intensity and wavelength is well explained by semiconductor bandgap narrowing and device-circuit operating conditions, validating the physical basis for sensing. Absolute spectral shift offers robust calibration independent of external collection factors, while intensity provides a high-sensitivity relative metric subject to setup conditions. The ultrafast emission dynamics allow tracking of rapid thermal changes, and lithographically defined arrays enable fine-grained spatial mapping, with resolution limited by device pitch rather than thermal radiation wavelength. The fiber-integrated format facilitates practical deployment in portable systems, achieving accurate in vitro monitoring (exhalation) and demonstrating feasibility for in vivo deep-brain thermometry, with particular advantages in electromagnetically harsh environments (e.g., near MRI).

This work introduces a thin-film, microscale, optoelectronic NIR-to-visible upconversion thermometer that exhibits strong, predictable temperature dependence in both emission intensity (~1.5% °C⁻¹) and peak wavelength (~0.18 nm °C⁻¹), underpinned by semiconductor band structure changes and circuit-level effects. The devices offer ultrafast response (~20 ns), compatibility with array-based spatial mapping, and integration with fiber optics for minimally invasive, portable sensing. Demonstrations include spatial temperature mapping with large arrays, in vitro monitoring of human exhalation with high agreement to thermocouples, and in vivo deep-brain temperature sensing. Future work could focus on enhancing sensitivity and stability, miniaturizing device pitch for higher spatial resolution, optimizing excitation conditions, expanding biocompatible encapsulation and long-term in vivo performance, and integrating with clinical imaging modalities for multi-modal diagnostics.

• Relative intensity readout depends on external factors (excitation power, spectrometer efficiency, device geometry), whereas the spectral shift is an absolute measure tied to the InGaP bandgap.
• Operating points and intensity sensitivities depend on device structure and excitation conditions; sensitivity can vary with excitation intensity.
• The demonstrated spatial resolution of arrays is currently limited by device pitch (~300 µm); further lithographic optimization is required for finer mapping.
• In dynamic tests, observed rise/decay times (>10 s) were limited by the external heating plate response rather than intrinsic device speed.
• Emission can be extinguished when local temperatures exceed ~95 °C under certain conditions, defining an upper practical operating range for the demonstrated setup.