Uncooled IR is like "listening at room temperature"; cooled IR is "hearing a pin drop in a silence at minus two hundred degrees". What's commonly called a cooled detector is strictly an Integrated Detector Cooler Assembly (IDCA / IDDCA). It is not a single chip but a micro cryogenic lab — sensing material, readout circuit, vacuum dewar and cryocooler — sealed in one metal can, holding the focal-plane array at 77–150 K for extremely low dark current and very high SNR. Understand the IDCA, and you understand why high-end IR is expensive, hard and unavoidable.
01 An IDCA is not a "detector", it's a "detector + fridge"
An IDCA stacks at least four layers: ① Focal-Plane Array FPA — sensing material + ROIC readout, turning IR photons into electrical signals; ② Dewar — a vacuum thermal shell isolating the cold end from the outside; ③ Cryocooler — Stirling/pulse-tube/Joule-Thomson, pulling the FPA down to 77–150 K continuously; ④ Structure and window — IR-transparent window, cold shield, package flange, connectors. Uncooled VOx works on power-up; an IDCA first "cools down" and only when the temperature curve settles does the image become clean. Its essence is trading a micro cryogenic lab for extremely low dark current and very high SNR.
02 The sensing layer: three materials, three technical fates
① HgCdTe (MCT): the "full-spectrum aristocrat" of IR. A tunable-bandgap II-VI ternary. Mercury telluride (HgTe) is a semi-metal, cadmium telluride (CdTe) a wide-gap semiconductor; tuning the Cd fraction x shifts the response from 1 μm to 30 μm, covering SWIR, MWIR (3–5 μm), LWIR (8–14 μm) and even VLWIR. QE often exceeds 80%, with high absorption and fast response — the "gold standard" for space remote sensing, IRST and premium seekers. The cost is real: weak Hg-Te bonds, brittle material, poor large-area epitaxial uniformity, low LWIR yield and high cost, and extreme sensitivity to temperature and defects — so MCT is not "impossible to make" but "extremely hard to make large and stable".
② InSb: the "image-quality champion" of MWIR. An intrinsic III-V photon material with stable covalent bonds and mature process; in the 3–5 μm band it offers low dark current, good uniformity, clean images and easy large formats — the "worry-free" choice for industrial thermography, gas analysis and MWIR defense. The shortcoming is clear: the band is essentially locked to MWIR; extending to LWIR needs lower temperature and thicker absorber layers, less flexible than MCT.
③ Type-II Superlattice (T2SL): III-V process chasing MCT's bandgap. Typically InAs/GaSb grown alternately at the nanometer scale: layer thickness sets the bandgap, so the band is designable; III-V epitaxy gives better uniformity and yield than MCT; barrier structures (nBn / BIRD) suppress dark current; operating temperature can be pushed toward HOT (High Operating Temperature, 150 K and beyond). It's like "borrowing InSb's process skeleton to learn MCT's bandgap freedom". T2SL is moving from lab to production in LWIR, dual/multi-color, space and small-SWaP scenarios.
03 How the FPA turns "light" into "image"
The sensing material only generates photocarriers; what makes it an image is the ROIC (Readout Integrated Circuit): a tiny charge-processing unit per pixel performing integration → sampling → noise reduction → serial readout, turning the weak currents of millions of pixels into a digital frame stream. The FPA's difficulty isn't "sensing" but "uniform readout" — pixel response consistency, dark-current suppression, full-well capacity, frame rate and NETD, plus thermo-mechanical matching to the cryogenic environment. That's why big players rarely outsource the FPA: epitaxy, flip-chip, passivation, interconnect and test each hide years of trial-and-error data no one can replicate by buying equipment.
04 The dewar: not a bottle, a vacuum thermal-isolation system
The dewar is often mistaken for a "thermos", but it does three jobs: vacuum insulation (cutting gaseous conduction), structural support (holding FPA, cold finger and window on one optical axis), and optical sealing (the IR window must be highly transmissive yet vacuum-tight). Common housings: Kovar (Fe-Ni-Co) for CTE-matched, high-reliability packages; stainless steel for strength; aluminum for light weight in airborne payloads; glass-to-metal seals for some research parts. Window by band: MWIR uses germanium (Ge) most, then sapphire or zinc selenide; LWIR uses ZnSe, ZnS, sapphire depending on band; an AR coating is standard to swallow window reflection. The dewar's pitfalls are leak rate, cold deformation, frosting and stress cracks — so it's usually co-developed with the FPA, rarely fully outsourced.
05 The cryocooler: three ways to make 77 K
① Stirling. A high-pressure helium compression-expansion cycle; fast start (cold in minutes), small and SWaP-friendly, suited to handheld, pods, munitions and vehicle tactical gear; the cost is moving reciprocating parts, so vibration, wear and life are the engineering焦點(diǎn). Thales' RM series and Ricor's rotary Stirling (e.g. K508) are classics.
② Pulse Tube. The "no-cold-head-moving-part" version of Stirling: no piston at the cold end, heat exchanged by oscillating pressure waves, extremely low vibration and long life, suited to satellites, space remote sensing and precision science; the cost is slightly lower efficiency, high design門檻 and expense — in space, "one fewer moving part" matters ten times more than "a bit cheaper".
③ Joule-Thomson (J-T). High-pressure gas throttling expansion, cooling in seconds, extremely simple and cheap, suited to one-shot/short missions (missile seekers, emergency SAR, brief space observation); the downside is consumable working gas, not for long cycles. Honeywell, Sumitomo and Eaton have deep J-T micro-cooler experience.
06 Why "cold" matters: the life-and-death line of signal and noise
LWIR photons carry very low energy; at room temperature the semiconductor itself generates carriers furiously, and dark current drowns the target signal. Cooling to 77 K drops generation-recombination noise exponentially and brings dark current to acceptable levels, enabling weaker targets, longer range and higher frame rates. A quick comparison:
| Type | Operating temp | Typical strength | Typical weakness |
|---|---|---|---|
| Uncooled VOx | Room temp | Cheap, small, instant-on | Limited sensitivity/frame rate |
| MCT cooled | 77–110 K | Full band, very high D* | Expensive, brittle, low yield |
| InSb cooled | 77–100 K | Stable MWIR image quality | Weak LWIR |
| T2SL cooled | 100–200 K (HOT) | Designable, uniform | LWIR dark current still hard |
The IDCA's "cold" is not showing off — it is the prerequisite for a narrow-gap photon detector to survive.
07 The material supply chain: from by-product ore to wafer to battlefield
Take MCT: upstream is surprisingly "geological". Tellurium (Te) comes mainly from copper and lead-zinc by-product ores, scarce; cadmium (Cd) is a zinc-ore by-product; mercury (Hg) is doubly sensitive on both resource and environmental fronts; then refining → polycrystal synthesis → MBE/MOCVD epitaxy → wafer → FPA. China dominates refined Te and Cd supply globally; Europe's Umicore handles high-grade distribution and secondary refining; wafer foundries like IQE and IntelliEPI specialize in epitaxy for smaller players. So IR is not purely a "chip business" but a four-dimensional competition of "mineral leverage + epitaxy process + vacuum packaging + cryocooler reliability".
08 Industry landscape: vertical integration vs horizontal supply
Two camps. Vertical integrators grow their own epitaxy, make the FPA, package the dewar and tune the cryocooler — Lynred, SCD, Teledyne, and domestic players like Guide, Guangzhi and Guoke Tiancheng. Horizontal suppliers combine wafer foundry + ROIC house + outsourced cooler (Thales / Ricor / AIM), common among smaller detector makers and system integrators. The trend: MCT holds high-end full-spectrum; InSb holds the mature MWIR base; T2SL grabs HOT, multi-color, space and small-SWaP growth; the cryocooler evolves from "bought part" to "core system IP". Domestically, the past two years moved from "making single pieces" to a closed chain of "material — FPA — dewar — cryocooler — core".
09 The future: the end of IDCA is removing one layer of the "fridge"
The next evolution is clear: ① HOT-ification — from 77 K to 150–230 K, shrinking the cooler; ② multi-color — MWIR/LWIR/VLWIR on one chip, against decoys and camouflage; ③ thin absorber + photonic structures — guided-mode resonance and plasmonics compensating absorption, relying less on material thickness; ④ solid-state cooling — TEC plus localized deep cooling, reducing mechanical wear; ⑤ wafer-level vacuum packaging — turning the IDCA from "handmade art" into "semiconductor product". But the baseline won't change: as long as narrow-gap photon detection of LWIR is used, "cold" remains — only later it's less brutal and the fridge is lighter.