Uncooled Poisson Bolometer for High-Speed Event-Based Long-wave Thermal Imaging

TL;DR

The paper addresses the challenge of high-speed, uncooled LWIR imaging by introducing the Spintronic Poisson Bolometer (SPB), which converts absorbed infrared energy into Poisson-counting switching events in a nanoscale spintronic sensor. Its asynchronous pixel-level readout yields a frame-free, high-temporal-resolution data stream, with a NETD around $100\ \mathrm{mK}$ and broadband sensitivity from $0.8$ to $14\ \mu\mathrm{m}$. The key contributions include a detailed event-based characterization via the Differential Count Rate (DCR) framework, demonstration of kilohertz-scale DCR up to $1{,}250$ Hz (vs $\sim$348 Hz for a commercial uncooled camera), and circuit-level simulations showing dynamic scene reconstruction with sub-microsecond latency in 16×16 arrays. The work highlights data sparsity, energy efficiency, and pathway to large SPB FPAs, enabling neuromorphic, edge-processing compatible thermal imaging in the MWIR/LWIR band.

Abstract

Event-based vision provides high-speed, energy-efficient sensing for applications such as autonomous navigation and motion tracking. However, implementing this technology in the long-wave infrared remains a significant challenge. Traditional infrared sensors are hindered by slow thermal response times or the heavy power requirements of cryogenic cooling. Here, we introduce the first event-based infrared detector operating in a Poisson-counting regime. This is realized with a spintronic Poisson bolometer capable of broadband detection from 0.8-14$μ\text{m}$. In this regime, infrared signals are detected through statistically resolvable changes in stochastic switching events. This approach enables room-temperature operation with high timing resolution. Our device achieves a maximum event rate of 1,250 Hz, surpassing the temporal resolution of conventional uncooled microbolometers by a factor of 4. Power consumption is kept low at 0.2$μ$W per pixel. This work establishes an operating principle for infrared sensing and demonstrates a pathway toward high-speed, energy-efficient, event-driven thermal imaging.

Figures (7)

• Figure 1: Conceptual Paradigm and Technology Landscape for LWIR Event-Based Sensing. (a) Comparison of a high-speed thermal target in a ballistic trajectory. The frame-based sensor (FBS) exhibits significant motion blur, whereas the event-based sensor (EVS) captures sharp, asynchronous temporal edges with sub-microsecond resolution. (b) Performance Benchmarking: Power consumption versus event/frame rate. EVS technologies are denoted by blue-outlined triangles and FBS by black-outlined squares. The spintronic Poisson (SP) bolometer (this work, thick blue outline) shows superior efficiency over conventional technologies. (c) Spectral Landscape: The SP-bolometer bridges the MWIR--LWIR gap. EVS solutions are identified by thick blue borders and FBS by black borders. Fill colors indicate operating temperature (red: 300 K; teal: 80 K). Data sources: Si CCD stevens2017recentonsemi_KAI04050, EVS Si son20174, InGaAs geum2024highlysensors_640CSX_2024, Prophesee SWIR jakobson2022eventprophesee_packaged2024sony_imx636, CelePixel chen2019livegallego2020event, cooled InSb sato2024midscd_pelican_d_640, cooled MCT lobre2025newflir_x8580sls_datasheet, analog microbolometer fusetto2023readoutlynred_pico640gen2_datasheet, and SP bolometer bauer2025exploiting.
• Figure 2: (a) Temporal response comparison between analog LWIR bolometers and the SP bolometer. The traditional analog response is effectively filtered by its high thermal inertia, whereas the SP bolometer resolves high-frequency dynamics by utilizing sub-nanosecond transition times. (b) Sub-nanosecond output transition demonstrating high-speed digital response. (c) Operating schematic of the SP bolometer at room temperature. Absorbed radiation generates a localized hotspot that propagates through the transduction layer, while a synthetic antiferromagnet (SAF) stabilizes the readout orientation. (d) SEM image of a single SP bolometer pixel. Inset: nanoplasmonic antenna.
• Figure 3: (a) Schematic of the NETD measurement setup for the SP-bolometer, including an ultra-stable blackbody source and a broadband ZnSe lens (f/0.5). (b) Measured count rate as a function of blackbody temperature, showing linear responsivity (dC/dT) used to calculate NETD = Cn/(dC/dT) = 102 mK, measured at $H_{app}$ = 0 Oe and $I_{bias}$ = 575 $\mu$A. (c) Spectral dependence of NETD across the mid-wave infrared (MWIR) and long-wave infrared (LWIR) atmospheric transmission windows, showing enhanced sensitivity in regions of high atmospheric transmission.
• Figure 4: Event-based characterization of the SP bolometer. (a) Schematic of the EVS characterization setup, comprising an ultra-stable blackbody source ($<1$ mK stability), a mechanical chopper for high-speed thermal modulation, an f/0.5 ZnSe IR lens, the SP bolometer, and a bias magnet. (b) Differential count rate (DCR) comparison between the SP bolometer and a conventional sensor at a 30$^\circ$C blackbody temperature. The x-axis represents the chopper-driven modulation frequency, and the y-axis shows the differential count $D$ over a fixed integration window $\tau = 1$ ms. (c) DCR of the SP bolometer as a function of event rate for blackbody temperatures of 25, 27.5, and 30$^\circ$C, demonstrating reliable detection up to a maximum event rate of 1,250Hz. (d) Benchmarking DCR using a commercial uncooled LWIR camera (FLIR A325sc, featuring a $VO_x$ microbolometer FPA flirA325sc) at 25, 27.5, and 30$^\circ$C. The FLIR sensor exhibits a maximum event rate of only 348Hz, constrained by its 8-12 ms thermal time constant and 60Hz internal readout electronics.
• Figure 5: Circuit architecture of the SP bolometer event-based sensing (EVS) readout. The signal chain begins with the SP Bolometer, which converts incident thermal flux into a voltage signal. This signal is then AC-coupled to a Low-Noise Amplifier (LNA) to isolate the transient response. This is followed by a Differentiator (High-Pass Filter) that computes the temporal change in intensity, $x(t + \Delta t) - x(t)$, effectively suppressing static background. A Comparator stage then performs threshold detection against positive ($+\Delta$) and negative ($-\Delta$) limits to trigger asynchronous Event+ and Event- pulses. Insets illustrate the ultra-fast switching transition ($t_{trans} \approx 1$ ns) and the specific Op-Amp and comparator configurations used to achieve high-speed neuromorphic readout.