All-Optical Photonic Crystal Bolometer with Ultra-Low Heat Capacity for Scalable Thermal Imaging

TL;DR

The paper tackles the challenge of fast, high-sensitivity LWIR thermal imaging without cryogenic cooling or wired electronics by proposing an all-optical, photonic-crystal bolometer with ultra-low heat capacity. It demonstrates a heterogeneously integrated pixel that couples broadband LWIR absorption in pyrolytic carbon pillars to a high‑Q silicon PhC cavity, whose thermo-optic shift is read out optically, yielding D^* ≈ 1.1×10^7 Jones and τ_th ≈ 27 μs at ambient pressure. The work emphasizes that readout noise currently limits performance and identifies a thermo-refractive-noise (TRN) floor (~0.064 fm/√Hz) that could be reached with improvements in Q and optical collection, projecting potential gains >25× toward D^* ≈ 3×10^8 Jones. The architecture promises scalable, high-speed, uncooled thermal imaging with a wireless readout path and compatibility with scalable manufacturing, paving the way for large-format FPAs and real-time LWIR imaging in diverse applications.

Abstract

High-speed thermal imaging in the long-wave infrared (LWIR) is critical for applications from autonomous navigation to medical screening, yet existing uncooled detectors are fundamentally constrained. Resistive bolometers are limited by electronic noise and the parasitic thermal load of wired readouts, while state-of-the-art nanomechanical resonators typically rely on vacuum packaging to maintain the mechanical $Q$ needed for sensitivity. Here, we introduce and demonstrate an uncooled thermal detector that addresses these challenges via an all-optical transduction mechanism. The heterogeneously integrated pixel is engineered for minimal thermal mass, combining pyrolytic carbon absorbers for broadband LWIR absorption, hollow zirconia structures for ultra-low-conductance thermal isolation, and a silicon photonic crystal cavity that serves as a high-$Q$ optical thermometer. Operating at ambient temperature and pressure, we measure a specific detectivity of $1.1\times10^{7}$ Jones and a thermal time constant of $27~μ\mathrm{s}$, corresponding to a speed that surpasses typical high-sensitivity uncooled technologies by an order of magnitude. Based on this detectivity measurement, which is limited by the noise floor of the external optical detection electronics, a physics-based model predicts a $>25\text{-fold}$ performance enhancement to its fundamental thermorefractive-noise-limited value ($3\times10^{8}$ Jones). The optical readout remains functional across ambient and vacuum environments. We expect this architecture to provide a general route toward scalable, high-performance thermal imaging systems.

Figures (14)

• Figure 1: Conceptual framework for an optically-probed bolometer array.a, A generalized bolometer model. Incident radiation is absorbed ($P_{\text{abs}}$) by a thermal mass ($C_{\text{th}}$) at temperature $T$, which is coupled to a thermal sink ($T_0$) via a thermal conductance ($G_{\text{th}}$). The resulting temperature change ($\delta T$) is transduced by one of several mechanisms. b, Principle of all-optical transduction (this work). The $\delta T$ alters the optical path length of a generic cavity via thermal expansion ($\delta L$) and the thermo-optic effect ($\delta n$), producing a measurable shift ($\delta\lambda$) in its resonance wavelength ($\lambda_0$). c, Conceptual architecture for a wire-free focal-plane array (FPA). A structured LWIR scene (input "P") is imaged onto the FPA, creating a pattern of heated pixels ($\delta T$). A separate probe source (bottom left) illuminates the array via a beam splitter. The probe light (green beams) reflects off the FPA pixels; the reflected light (purple beams) from heated pixels is modulated, carrying the signal $r_{i,j}(t)$, which represents the complex cavity reflection coefficient of the pixel. This reflected signal is directed to the all-optical wireless readout plane (detector array), transducing the thermal image into an optical image.
• Figure 2: Device architecture and operating principle of the photonic crystal bolometer.a, 3D schematic of the heterogeneously integrated pixel. The device consists of an array of pyrolytic carbon nanopillars that function as LWIR absorbers, a silicon L4/3 PhC cavity for the optical readout, and hollow zirconia support that provide thermal insulation from the substrate. b, Scanning electron micrograph of a fabricated device. Scale bar, $5~\mu \mathrm{m}$. c, Lumped-element thermal model of the bolometer pixel. We model the device as a thermal RC circuit where the suspended pixel has a total heat capacity $C_{\text{th}}$ and is connected to a thermal bath $(T_0)$ through parallel thermal paths for conduction $(R_{\text{cond}})$, convection $(R_{\text{conv}})$, and radiation $(R_{\text{rad}})$. d, Simulated (finite-difference time-domain) mode profile of a L4/3 PhC cavity. e, Measured optical resonance on the final device, centered at $1568.83~\mathrm{nm}$ and showing a quality factor of $3.9 \times 10^4$.
• Figure 3: Photothermal responsivity characterization.a, Measured reflection spectra of the PhC cavity resonance for increasing absorbed optical power, from $0$ to $2.5~\mu \mathrm{W}$. The experimental data (dots) are fitted to Lorentzian lineshapes (solid lines), revealing a systematic red-shift with applied power. b, Extracted resonance detuning $(\Delta \lambda)$ as a function of absorbed power. The linear fit (dashed line) confirms the expected thermo-optic response, and its slope yields the DC photothermal responsivity of the device.
• Figure 4: Thermal dynamics and bandwidth of the PhC bolometer.a, Measured time-domain response to a $1~\mathrm{kHz}$ square-wave modulated heating laser. The homodyne voltage (purple trace), which tracks the pixel's temperature, is fitted (dark purple trace) to our linearized lumped-element, single-pole thermal RC model, yielding a time constant of $\tau_{\text{th}} = 27~\mu \mathrm{s}$. b, Calculated frequency-dependent responsivity. It follows a first-order low-pass filter response, with a $3~\mathrm{dB}$ bandwidth of $f_{\mathrm{3dB}} \approx 5.9~\mathrm{kHz}$ determined from the measured time constant.
• Figure 5: Optical readout calibration and noise characterization.a, Calibration of the wavelength-to-voltage transduction. The homodyne voltage signal $(v_h)$ is measured as the probe laser is swept across the cavity resonance (dots), showing excellent agreement with a fit to a dispersive theoretical model (solid purple line). The right axis shows the calculated transduction gain $(dv_h/d \lambda)$. The vertical line marks the optimal operating point $(\lambda_{\text{op}})$ of maximum gain. b, Wavelength noise amplitude spectral density of the readout. The spectrum is characterized by technical noise at low frequencies and a white noise plateau at higher frequencies. The dashed line indicates the measured instrument-limited noise floor of $1.2~\mathrm{fm}/\sqrt{\mathrm{Hz}}$.