How the James Webb Space Telescope’s infrared detectors actually work, why they almost didn’t, and what their engineering lineage tells us about the limits of observation
Marcia Rieke spent the better part of two decades worrying about photons. As the principal investigator for NIRCam, one of the key near-infrared cameras aboard the James Webb Space Telescope, Rieke oversaw the development of detector arrays that needed to register individual packets of light that had been traveling for over 13 billion years. The photons arriving at Webb’s focal plane from the earliest galaxies carry so little energy that the detectors themselves had to be cooled to roughly 37 kelvin (about minus 393 degrees Fahrenheit) just to keep their own thermal noise from drowning out the signal. Every warm atom in the detector vibrates, and every vibration risks mimicking a photon that never existed. The engineering challenge was not simply building a camera. It was building a camera so quiet that the universe’s faintest whispers could be heard above the instrument’s own breathing.
That fundamental tension between signal and noise defines everything about how JWST works. And understanding how the telescope’s infrared detectors were designed, tested, nearly abandoned, and ultimately perfected reveals something that goes beyond Webb itself: every act of observation, in astronomy and beyond, is bounded not by what exists in the universe but by the physical limits of the instrument doing the observing. The history of JWST’s detectors is the clearest case study I know for a principle that shapes all of science — that what we can know is ultimately determined by what we can build, and that every generation’s engineering frontier becomes the next generation’s observational ceiling.
The Physics of Seeing in Infrared
Visible light occupies a remarkably narrow slice of the electromagnetic spectrum: wavelengths between roughly 0.4 and 0.7 microns. Infrared light extends from about 0.7 microns out to around 1,000 microns, though JWST’s instruments operate in a more targeted range from 0.6 to 28.5 microns. The reason NASA committed to infrared observation for Webb was both cosmological and practical. Light from the universe’s earliest structures has been stretched by the expansion of space itself, a process called redshifting, which pushes ultraviolet and visible light emitted billions of years ago into infrared wavelengths by the time it reaches us. If you want to see the first galaxies, you need infrared eyes.
But infrared astronomy has a problem that visible-light astronomy does not. Everything warm emits infrared radiation. Your body does. The telescope structure does. Earth’s atmosphere is saturated with it. This is why ground-based infrared telescopes have always fought against overwhelming background noise, and why JWST had to be placed at the second Lagrange point, approximately 1.5 million kilometers from Earth, shielded by a five-layer sunshield made of kapton thinner than a human hair.
The sunshield handles the macro problem of thermal management. But it only creates the conditions under which the detectors can function. The detectors themselves had to solve the micro problem: converting individual infrared photons into electrical signals with extraordinary precision. This is where the observational limit first asserts itself — not in the cosmos, but in the crystal lattice of a semiconductor.
Mercury Cadmium Telluride: The Material That Made It Possible
The heart of JWST’s near-infrared instruments (NIRCam, NIRSpec, NIRISS, and the Fine Guidance Sensor) is a compound semiconductor called mercury cadmium telluride, or HgCdTe. This material has a remarkable property: by adjusting the ratio of mercury to cadmium in the crystal lattice, engineers can tune the detector’s sensitivity to different infrared wavelengths. More mercury shifts the sensitivity toward longer wavelengths. More cadmium shifts it shorter. The same base material can be customized to detect light from 0.6 microns all the way out to about 5 microns, which covers the full near-infrared range Webb needs.
The detectors are arranged in a hybrid architecture. The HgCdTe photosensitive layer sits on top of a silicon readout integrated circuit (ROIC). Each pixel in the HgCdTe layer is connected to a corresponding pixel in the silicon layer via a tiny indium bump bond, essentially a microscopic solder joint. When an infrared photon strikes the HgCdTe, it generates an electron-hole pair. The freed electron migrates through the bump bond to the silicon circuit below, which accumulates charge and measures it. One photon, one electron, one count.
NIRCam uses ten of these detector arrays, each with 2048 × 2048 pixels, covering short-wavelength and long-wavelength channels plus wavefront sensing. The total pixel count across the near-infrared instruments is in the tens of millions. Every single one of those pixels had to work reliably at cryogenic temperatures for a planned mission lifetime of at least ten years.
That’s where the trouble started.
The Decade of Detector Development Hell
HgCdTe detectors were not new when JWST’s design was being finalized in the early 2000s. The material had been used in military infrared sensors and in ground-based astronomical instruments for years. But those applications either operated at warmer temperatures, tolerated higher noise levels, or required far fewer pixels. JWST needed something that had never existed: large-format HgCdTe arrays with extremely low noise floors, operable at 37 kelvin, with fewer than one dead pixel per thousand, and with stability sufficient to survive launch vibrations and years of continuous operation in space.
Teledyne Imaging Sensors won the contract to develop these detectors. The company’s H2RG (Hawaii-2RG) sensor became the workhorse detector for JWST’s near-infrared instruments. The ‘Hawaii’ designation refers to the original development lineage at the University of Hawaii’s Institute for Astronomy, and ‘2RG’ indicates the 2048 × 2048 pixel format with reference pixels and guide mode capability.
Getting the H2RG detectors to meet JWST specifications took more than a decade of iteration. Early batches had unacceptable rates of hot pixels that registered false signal even in the dark. Some arrays showed excess dark current at the required operating temperatures. The indium bump bonds occasionally failed to make proper contact, creating dead spots. And a persistent problem called “persistence” caused bright sources to leave ghost images in subsequent exposures as trapped charge slowly leaked out of the detector material. Each of these problems required its own engineering campaign — changes to the HgCdTe growth process, exhaustive cryogenic screening of every array, optimization of detector architecture and voltage biases, and the development of sophisticated algorithms to model and subtract residual artifacts.
The total cost of JWST’s instrument development, including the detectors, contributed significantly to the telescope’s substantial budget overruns. The project was originally estimated at roughly $1 billion and ultimately cost approximately $10 billion. The detectors were not the sole driver, but the difficulty of making them work was representative of a broader pattern: JWST required technologies that did not exist when the project started, and developing those technologies proved harder and more expensive than anyone initially projected. That gap between what the science required and what engineering could deliver is itself a kind of observational limit — not a physical one, but an institutional and economic one that shapes what questions a generation of astronomers gets to ask.
MIRI: A Different Detector for a Different Challenge
The Mid-Infrared Instrument, MIRI, faces a fundamentally different detection challenge than the near-infrared instruments. MIRI operates from 5 to 28.5 microns, wavelengths where HgCdTe detectors become less effective and where the photons carry even less energy per particle. For this range, MIRI uses arsenic-doped silicon (Si:As) detectors, which work through a process called impurity band conduction. Arsenic atoms are implanted into a silicon crystal at precise concentrations. When a mid-infrared photon strikes the detector, it excites an electron from the arsenic impurity level into the silicon conduction band, generating a measurable signal.
These detectors require far colder operating temperatures than HgCdTe — approximately 7 kelvin (about minus 447 degrees Fahrenheit), compared to NIRCam’s 37 kelvin. JWST’s passive cooling from the sunshield gets the telescope’s cold side down to about 40 kelvin. Getting MIRI the rest of the way demands an active mechanical cooler: a Joule-Thomson refrigerator combined with pulse-tube pre-cooling, developed by Northrop Grumman and the Jet Propulsion Laboratory. This cryocooler represented another single-point failure risk for the mission. If it failed, MIRI would be blinded.
MIRI’s three 1024 × 1024 pixel arrays are smaller than NIRCam’s because mid-infrared detector fabrication is more difficult and yields are lower. This dual-technology approach — HgCdTe for near-infrared, Si:As for mid-infrared — is one of the underappreciated engineering choices in JWST’s design. A single detector technology could not cover the full wavelength range. Two separate development programs, two sets of fabrication challenges, two thermal environments on the same spacecraft. The complexity was staggering, and it illustrates how the range of observable wavelengths is not set by what light the universe produces, but by what detector materials we can fabricate and cool.
Pushing Against the Noise Floor
Every astronomical instrument ultimately runs into the same wall: noise. There is no such thing as a perfectly quiet detector. Even in the cold vacuum of L2, shielded from the Sun, cooled to within a few degrees of absolute zero, JWST’s detectors still have noise floors. Read noise (the uncertainty introduced each time you measure the accumulated charge), dark current (false signal from the detector material itself), and cosmic ray hits (high-energy particles that occasionally strike the detector and create bright streaks) all set hard boundaries on what JWST can see. These are not temporary engineering shortcomings. They are consequences of quantum mechanics, thermodynamics, and the particle environment of space. They are, in the most literal sense, the limits of observation for this instrument and this era.
The telescope’s designers dealt with these limits through a combination of hardware optimization and observing strategy. Webb uses a technique called “up-the-ramp sampling,” where each pixel is read non-destructively multiple times during an integration. Instead of a single measurement at the end of an exposure, the electronics sample the accumulating charge at regular intervals and fit a line to the resulting ramp. The slope of that line gives the photon flux, and the multiple samples reduce the effective read noise well below what a single measurement would yield. Cosmic ray hits show up as sudden jumps in the ramp and can be flagged and removed.
This means JWST’s effective sensitivity is better than the raw detector specifications might suggest. But it also means the telescope’s data pipeline is extraordinarily complex. Raw data from Webb undergoes extensive processing before it becomes the images that appear in press releases and scientific publications in journals like Nature. Flat-field corrections, dark current subtraction, persistence removal, cosmic ray rejection, optical distortion correction, and absolute flux calibration all happen before an astronomer ever analyzes the data.
What we see from JWST is not raw observation. It is observation processed through layers of engineering knowledge and algorithmic sophistication. The detectors set the fundamental physical floor, and then the data pipeline pushes as close to that floor as mathematics allows. But neither hardware nor software can push below it. That floor is real, and it defines the boundary of what this telescope — the most powerful ever built — can and cannot tell us about the universe.
The Engineering Lineage: From Spy Satellites to Deep Space
The technology behind JWST’s detectors did not emerge from pure astronomical research. HgCdTe infrared detectors were originally developed for military applications, primarily thermal imaging and missile tracking. The U.S. defense establishment invested billions of dollars over decades in infrared sensor technology, and that investment created the industrial base and the materials science knowledge that NASA later drew upon.
Teledyne’s Hawaii detector series began in the 1990s as a collaboration between Rockwell Science Center and the University of Hawaii, adapting military-grade HgCdTe technology for astronomical use. The H1R (1024 × 1024) and H2RG (2048 × 2048) sensors were successively larger and more capable versions designed specifically for space telescopes and ground-based instruments. The H2RG is now used not only in JWST but in numerous ground-based instruments worldwide and is baselined for the Nancy Grace Roman Space Telescope as well.
This lineage matters because it reveals something about how breakthrough scientific instruments actually get built. JWST’s detectors were not invented from scratch by astronomers. They were adapted from a decades-long defense technology program, refined through collaborations between national labs, universities, and commercial semiconductor manufacturers, and pushed to performance levels that no single funding stream would have supported. The detector development cost hundreds of millions of dollars across all these programs, spread over 30-plus years.
When I wrote about the Voyager spacecraft and their continued operation decades beyond their design life, one theme that emerged was how engineering decisions made in the 1970s constrained and enabled discoveries that nobody anticipated. The same dynamic applies to JWST’s detectors. Choices made about HgCdTe crystal growth techniques, pixel architectures, and readout electronics in the 2000s now determine what galaxies we can and cannot see, what exoplanet atmospheres we can and cannot characterize, and what questions about the early universe we can and cannot answer. The detector is not a passive window. It is an active filter that shapes the knowable universe.
What the Detectors Have Already Shown Us
The proof of the detectors’ quality is in the science they’ve enabled. JWST has confirmed galaxies at redshifts above 14, meaning we are seeing objects as they existed roughly 300 million years after the Big Bang. These detections were only possible because the H2RG arrays achieved their target noise performance and because the cryogenic cooling systems maintained stable operating temperatures over months of observation.
MIRI’s Si:As detectors have enabled direct imaging of exoplanets in the mid-infrared, a capability that was theoretical before launch. And the combined near- and mid-infrared coverage has allowed atmospheric characterization of exoplanets through transit spectroscopy, including the detection of specific molecules like sulfur dioxide in the atmospheres of hot Jupiters.
Even within our own solar system, the detectors’ performance has been striking. Recent observations of Saturn combined Webb’s infrared view with Hubble’s visible-light imaging to effectively slice through the planet’s atmosphere at multiple altitudes. Webb’s infrared sensitivity revealed clouds and chemical layers at different depths that Hubble simply cannot access, including distinct polar features emitting at wavelengths around 4.3 microns that may be linked to high-altitude aerosols or auroral activity. The Saturn observations also showed how JWST’s detectors render the planet’s water-ice rings as extremely bright features in infrared, while revealing fine structures like spokes and the thin F ring that appear differently between the two observatories.
As Space Daily explored in a recent piece on Webb’s W51 images and what they mean for star formation science, the telescope’s ability to peer through dust-laden regions of active star formation depends entirely on the infrared detectors’ sensitivity and dynamic range. Visible-light telescopes see opaque clouds. Webb’s detectors see through them.
The Remaining Vulnerabilities and the Next Observational Ceiling
JWST’s detectors are performing well, but they are not immune to degradation. Cosmic ray damage slowly degrades detector pixels over time. The rate of degradation in space is higher than on the ground because JWST sits outside Earth’s magnetic field, which would otherwise deflect many charged particles. The mission team monitors detector health continuously and updates the bad-pixel maps used in data processing.
MIRI has already shown some sensitivity changes that the operations team has had to manage, though these have been within acceptable parameters so far. The cryocooler has performed as designed, but it contains moving parts, and any mechanical system in space carries some risk of eventual failure. If MIRI’s cryocooler degrades, the mid-infrared capability would be the first thing lost. The near-infrared detectors are more resilient because they depend on passive cooling, but they still accumulate radiation damage. Current projections suggest JWST has enough fuel and system margin to operate well beyond its five-year design life, potentially for 20 years or more. Whether the detectors will still meet science-grade performance thresholds at the end of that extended timeline is an open question.
And this is where the story of JWST’s detectors connects most directly to the future. Every major space telescope eventually pushes up against the physical limits of its detection technology. Hubble’s CCDs set one observational ceiling. JWST’s HgCdTe and Si:As arrays set another, much deeper one. But a ceiling it remains. There are galaxies too faint, atmospheres too tenuous, and signals too feeble for even Webb’s detectors to register. The universe does not end where JWST’s sensitivity does, but our knowledge of it, for now, does.
The question for the next generation of space telescopes, including the Habitable Worlds Observatory currently in conceptual development, is what detector technology comes after this. Superconducting detectors, microwave kinetic inductance detectors (MKIDs), and transition-edge sensors are all candidates for future missions. Each offers potential advantages in sensitivity, energy resolution, or wavelength coverage. But none has been demonstrated at the scale and reliability required for a flagship space mission. The development timeline for JWST’s detectors — more than 15 years from initial concept to flight-ready hardware — suggests that whatever technology powers the next great space telescope will need to enter serious development soon if it’s going to be ready by the 2040s.
What the Detectors Teach Us
When I started covering the intersection of space and technology in the early 2010s, the conversation about JWST was almost entirely about cost overruns and schedule delays. The telescope was routinely cited as an example of NASA’s inability to manage large projects. And there was truth to those criticisms. The project blew through its budget, consumed resources that could have funded other missions, and was nearly cancelled by Congress in 2011.
But the detector story offers a different frame. Much of what made JWST expensive was the fact that the required technology literally did not exist when the project was approved. The H2RG detectors had to be invented, tested, failed, improved, tested again, failed differently, improved again, and eventually qualified for spaceflight. That process is slow and costly by nature. You cannot schedule invention. You can fund it, staff it, and create the conditions for success, but you cannot put a breakthrough on a Gantt chart.
The space industry has spent the past decade learning to build things faster and cheaper, largely through the influence of SpaceX and the broader commercial space movement. Reusable rockets, standardized satellite buses, and software-defined spacecraft have all accelerated development timelines and reduced costs. These are real and important gains. But JWST’s detectors represent the other kind of space engineering: the kind where the physics is at the boundary of what’s possible, where commercial off-the-shelf components don’t exist, and where the development timeline is measured in decades rather than sprint cycles. Both kinds of engineering matter. Both are needed.
The photons arriving at JWST’s focal plane from the first galaxies carry about one electron-volt of energy each. The detectors that catch them are the product of 30 years of materials science, semiconductor fabrication, cryogenic engineering, and relentless testing. They almost didn’t work. They cost more than anyone planned. And they are now showing us the universe in ways that no other instrument in human history has managed.
But here is the thing that the detector story makes inescapable: what they show us is not the universe as it is. It is the universe as filtered through the specific material properties of mercury cadmium telluride and arsenic-doped silicon, processed through algorithms designed by humans, bounded by noise floors set by quantum mechanics and thermodynamics. Every image from Webb is a collaboration between the cosmos and the instrument. The galaxies provide the photons. The detectors decide which photons become knowledge and which remain silence. That boundary — between signal and noise, between the detectable and the invisible — is not a failure of engineering. It is the permanent condition of observation itself. And the most important thing JWST’s detectors teach us is that pushing that boundary back, even slightly, requires the kind of slow, expensive, failure-prone work that no one wants to fund until they see what it reveals.
Photo by Akbar Nemati on Pexels
The post How the James Webb Space Telescope’s infrared detectors actually work, why they almost didn’t, and what their engineering lineage tells us about the limits of observation appeared first on Space Daily.