optics, too, can be changed like software
For generations, the tools we send into space have been fixed in their purpose — built for one task, blind to all others. A collaboration between KAIST and MIT has quietly challenged that assumption, producing an optical chip that changes its sensory function through electrical command alone, without any physical alteration. Announced in July 2026 and published in Nature Communications, this reconfigurable mid-infrared device embodies a broader philosophical turn: that even hardware, like thought itself, need not be rigid to be reliable. In the constrained and unforgiving environment of space, the ability to adapt without replacement may prove as consequential as the missions it serves.
- Every time a satellite's mission shifted, engineers faced the costly, time-consuming burden of designing and launching entirely new optical hardware — a cycle the field had accepted as unavoidable.
- The core technical obstacle was the 'sneak-path' problem, where electrical current bled into unintended pixels and corrupted the chip's ability to control light with precision.
- By embedding a silicon PIN diode into each pixel of a phase-change metasurface, the KAIST-MIT team gave every element its own gatekeeper, achieving independent pixel control for the first time.
- The chip's phase-change material holds its configured state without continuous power — a critical advantage in space, where every watt of electricity is a resource too precious to waste.
- A 6-by-6 pixel prototype survived over 16,700 switching cycles, thirteen times more durable than prior attempts, and was built using standard silicon photonics processes that make scaling to thousands of pixels feasible.
- The technology is already advancing toward operational space sensors, with applications spanning launch-vehicle diagnostics, space-station thermal monitoring, and optical communications — all from a single reconfigurable platform.
For decades, changing a satellite's mission meant changing its hardware. New optical filters had to be designed, fabricated, and launched. The cycle was expensive and inflexible. Now, a team led by Professor Hyun Jung Kim at KAIST, in collaboration with Professor Juejun Hu's group at MIT, has broken that pattern with a single optical chip capable of shifting its function on command — performing thermal imaging one moment and infrared camera work the next, all through electrical signals alone.
Announced by KAIST on July 14th and published in Nature Communications on July 7th, the device is the first transmissive mid-infrared spatial light modulator built on a scalable, electrically addressable metasurface. Its ultrathin optical structure controls light pixel by pixel, switching the intensity of transmitted infrared light between two states per pixel. The key innovation was solving the long-standing 'sneak-path' problem — electrical current leaking into unintended pixels — by embedding a silicon PIN diode into each pixel to act as a precise gatekeeper.
At the heart of the chip is GSST, a phase-change material that shifts its light-transmitting properties when it receives an electrical pulse — and crucially, holds that state even after power is removed. In space, where electrical power is scarce, a chip that stays configured without continuous current is not a convenience but a necessity.
The team demonstrated the concept with a 6-by-6 pixel array, achieving over 16,700 switching cycles — roughly thirteen times more durable than previous comparable technology. Because the chip is fabricated using standard silicon photonics processes, scaling to hundreds or thousands of pixels is considered straightforward.
The implications are broad. One reconfigurable platform could serve as a spectrometer, a thermal imager, an infrared camera, or an optical communications device — reprogrammed as mission needs evolve rather than replaced. Kim and her colleagues call this vision 'software-defined sensors.' The collaboration, rooted in a partnership Kim began while at NASA in 2018, has grown into a full research framework at KAIST spanning material development, chip design, fabrication, and space testing. What began as a technical problem has become a conceptual shift: optics, like software, can be reconfigured rather than replaced.
For decades, when a satellite's mission changed, so did its hardware. New optical filters had to be designed, fabricated, and installed. New sensors meant new equipment. The cycle was expensive, time-consuming, and inflexible. Now, researchers at KAIST and MIT have demonstrated something that breaks that pattern entirely: a single optical chip that can shift its function on command, performing the work of a thermal imaging sensor one moment and an infrared camera the next, all through electrical signals alone.
On July 14th, KAIST announced the breakthrough. A team led by Professor Hyun Jung Kim from the Department of Aerospace Engineering, working with Professor Juejun Hu's group at MIT, had created the first transmissive mid-infrared spatial light modulator based on what's called a scalable two-dimensional, electrically addressable metasurface. The name is technical, but the implication is simple: optics that behave like software.
The device works by controlling light on a pixel-by-pixel basis using an ultrathin optical structure called a metasurface—patterns so small they're measured in fractions of a human hair's width. When an electrical signal arrives, each pixel can switch the intensity of transmitted infrared light between two states. The researchers achieved something that had eluded the field: the ability to control each pixel independently, without interference from neighboring elements. They solved a persistent problem called the "sneak-path" effect, where electrical current would leak into unintended pixels, by embedding a silicon PIN diode into each pixel to act as a gatekeeper, allowing current only where it was meant to go.
The material at the heart of the device is GSST, a phase-change compound of germanium, antimony, selenium, and tellurium. When it receives an electrical pulse, its light-transmitting properties shift. Crucially, it holds that state even after power is removed—a nonvolatile characteristic that matters enormously in space, where electrical power is scarce and precious. A satellite doesn't need to keep feeding current to maintain the chip's configuration. Once set, it stays set until commanded to change.
The team demonstrated the concept with a 6-by-6 pixel array, successfully generating desired optical patterns and proving the device could survive more than 16,700 switching cycles—roughly thirteen times more durable than previous attempts at similar technology. Because the chip was fabricated using standard silicon photonics manufacturing processes, scaling it up to hundreds or thousands of pixels is theoretically straightforward. The same techniques that produce computer chips can produce these optical devices.
The implications ripple outward. A single platform could serve as a thermal imaging system for monitoring a launch vehicle's health, a spectrometer for analyzing light by wavelength, an infrared camera for observing space conditions, or an optical communication device. Instead of launching specialized hardware for each function, a spacecraft could carry one reconfigurable optical chip and reprogram it as needs shift. The technology represents what Kim and her colleagues call "software-defined sensors"—hardware whose capabilities change not through replacement but through reconfiguration.
This work builds on a collaboration that began in 2018 when Kim was a researcher at NASA. She brought that partnership with her to KAIST, and the two institutions have now established a full research framework spanning material development, chip design, fabrication, integration, and eventual space testing. Kim's team is already developing the technology toward operational space sensors, working under South Korea's Young Researcher Program to create ultra-precise systems for measuring launch-vehicle surface temperatures. A broader research center is exploring applications in space-station thermal monitoring, in-space manufacturing diagnostics, and optical communications.
The research was published in Nature Communications on July 7th. What began as a technical problem—how to make optical systems flexible enough to adapt to changing missions—has become a conceptual shift: the recognition that optics, like software, can be reconfigured rather than replaced. For space systems operating under severe constraints of weight, power, and cost, that shift could prove transformative.
Citas Notables
This research presents the foundation for an era of software-defined sensors, in which a single optical chip performs a variety of functions depending on the mission.— Professor Hyun Jung Kim, KAIST Department of Aerospace Engineering
By combining MIT's nanophotonics technology with KAIST's space sensor technology, we plan to develop this technology into an actual space system.— Professor Hyun Jung Kim
La Conversación del Hearth Otra perspectiva de la historia
Why does it matter that this chip can be reconfigured electrically? Couldn't engineers just design better sensors from the start?
Because space missions change in ways you can't always predict. A satellite launched for one purpose might need to serve another. Or a mission might discover something unexpected and need different sensing capabilities. Right now, that means hardware redesign—months of work, new fabrication, new launches. With this chip, you send an electrical command and the hardware transforms.
The material they used—GSST—holds its state without power. Why is that so important?
In space, every watt of electricity is rationed. Solar panels have limits. Batteries run down. If your optical system had to constantly consume power just to maintain its configuration, that's power stolen from other systems. GSST flips once and stays flipped. It's like a light switch that doesn't need to be held down.
They solved something called the "sneak-path" problem. What was actually going wrong before?
Imagine a grid of pixels. You want to activate one specific pixel with an electrical signal. But current takes the path of least resistance, so it leaks into neighboring pixels too. They all start switching when you only wanted one to switch. The PIN diodes act like one-way valves—current can only flow where you intend it to.
The chip survived 16,700 switching cycles. Is that a lot?
It's thirteen times more than previous technology could handle. For a satellite in orbit for ten or fifteen years, potentially switching its optical configuration hundreds of times, that's the difference between a system that fails mid-mission and one that doesn't.
What comes next? Is this ready for space?
Not yet. They've proven the concept works in the lab. Now they're building toward actual space qualification—testing it in conditions that mimic the radiation, temperature extremes, and vacuum of orbit. The goal is flight demonstration within a few years. But the foundation is solid.
If this works, what changes?
The economics of space systems shift. You launch fewer specialized instruments and more reconfigurable platforms. Missions become more adaptable. A single spacecraft can do the work that previously required multiple instruments. That's not just engineering—that's a different way of thinking about how space systems should be designed.