Each pixel controls infrared light independently, with no moving parts
There are wavelengths of light that carry the world's secrets — leaking pipelines, drifting methane, the slow heat loss of a building — and for most of human history, reading them required cumbersome, expensive instruments. Researchers at MIT have now built a chip-scale device that tunes infrared light at the pixel level, with no moving parts, using the same manufacturing processes that produce the chips in everyday electronics. The work, published in Nature Communications, suggests that the ability to see what human eyes cannot may soon be as scalable and accessible as the smartphone itself.
- Infrared sensing has long been trapped behind a wall of bulk, cost, and mechanical complexity — useful only to those with the resources to deploy it.
- MIT's new chip dissolves that barrier by giving each microscopic pixel independent control over infrared light, allowing the device to shift focus and filter wavelengths without a single moving part.
- A crossbar architecture of copper wires and phase-change materials — materials that toggle between crystalline and amorphous states under precise heat — makes the pixel-level control both stable and scalable to millions of pixels.
- A working 6-by-6 prototype survived tens of thousands of switching cycles without degradation, signaling the device is durable enough for real-world deployment in environmental monitoring, thermal imaging, and military night vision.
- Because the technology fits within standard semiconductor foundry workflows, the path from laboratory prototype to industrial-scale production is less a leap than a natural next step.
Infrared light carries information our eyes will never see — gas leaks, atmospheric methane, heat bleeding from walls. The equipment capable of reading those signals has always been expensive, bulky, and mechanically complex. MIT researchers have now built something that changes that calculus.
The team created a chip-based optical device that dynamically controls how infrared light enters a camera, functioning as a tunable lens with no moving parts. Its key innovation is at the pixel level: each microscopic pixel can independently control infrared light, allowing the system to shift focus and detect different signals on the fly. The work, published in Nature Communications, was built using standard semiconductor manufacturing processes — the same techniques already used to produce computer chips — which means it can scale.
The device layers two perpendicular grids of copper wires over a specially treated silicon layer, beneath which sits a phase-change material that shifts between crystalline and amorphous states when heated. Current flowing through the wire crossings generates precise heat, triggering those structural shifts. Because the two states interact with infrared light differently, the system can selectively tune which wavelengths pass through and which are blocked. A diode at each pixel prevents current from bleeding into neighbors, preserving precision even as the array grows. Calculations suggest the architecture could handle millions of pixels without losing that control.
Led by Juejun Hu and first author Cosmin-Constantin Popescu, the team built a working 6-by-6 prototype that cycled through tens of thousands of switching events without degrading — a durability that matters for any device meant to operate in the field. The mid-infrared wavelengths it targets are already used to detect gas leaks and monitor Earth's atmosphere. But the chip's tunability opens further possibilities: environmental monitors hunting specific airborne compounds, more selective thermal imaging, improved night vision, and even optical computing, where metasurfaces encode computational weights directly into material structure.
The next phase is scaling up — more pixels, richer infrared data, and deeper integration into existing semiconductor foundries. That last part is not a constraint but an advantage: chip manufacturers bring decades of process control to the table. What began as a laboratory experiment is now pointed toward the same facilities that produce the chips in your phone.
Infrared light carries information our eyes will never see. A pipeline leaking gas, methane drifting through the atmosphere, heat bleeding from a building's walls—all of it invisible to human vision, all of it detectable by the right equipment. The problem has always been the same: the equipment is expensive, bulky, and difficult to aim. It requires moving parts, calibration, and space.
Researchers at MIT have now built something smaller. They've created a chip-based optical device that can dynamically control how infrared light enters a camera, functioning as a tunable lens with no moving parts. The innovation lies in the pixel level. Each microscopic pixel of the device can control infrared light independently, allowing the system to shift its focus and detect different signals on the fly. The work, published in Nature Communications, demonstrates that the approach can be manufactured using standard semiconductor processes—the same techniques factories already use to produce computer chips. That matters. It means this could scale.
The device works by layering two perpendicular grids of copper wires on top of a specially treated silicon layer. Below that sits a phase-change material, the kind that shifts between solid and liquid states when heated. When current flows through the wires at their crossing points, the silicon generates heat precisely at those locations. That heat triggers the phase-change material to switch between crystalline and amorphous structures. Those two states interact with infrared light differently. By controlling which pixels switch and which don't, the system can selectively tune which infrared wavelengths pass through and which get blocked or reflected. A diode selector at each pixel prevents electrical current from leaking into neighboring pixels, keeping the control precise even as the system scales upward. Calculations suggest the architecture could handle millions of pixels without losing that precision.
The team, led by Juejun Hu, MIT's John F. Elliott Professor of Materials Science and Engineering, and first author Cosmin-Constantin Popescu, built a working prototype in MIT.nano and at a semiconductor foundry. Their demonstration featured a 6-by-6 array of metasurface pixels. Testing showed the system could switch reliably on and off, and the researchers found the mesh architecture resilient enough to cycle through tens of thousands of switching events without degrading. That durability matters for any device meant to operate in the field.
The applications are broad. Mid-infrared detection—the wavelength this system targets—is already used to spot gas leaks and monitor Earth's atmosphere. Military and aerospace applications rely on it. But the tunable nature of this chip opens new possibilities. Environmental monitors could be configured to hunt for specific compounds in the air. Thermal imaging systems could become more dynamic, more selective. Night vision could improve. Popescu noted that organic molecules absorb light in the mid-infrared range, which means the system could be tuned to detect them—methane, propane, and countless others. The technology could also enable a different kind of computing altogether. Researchers have already shown that metasurfaces can emulate neural networks, encoding computational weights directly into the material's structure. When light passes through, it interacts with the metasurface in ways that produce computational results. A tunable system like this one could make such optical computing more flexible and powerful.
The next phase involves scaling up. The team is working to add more pixels to their array and build more robust versions capable of capturing richer infrared information. Hu emphasized that moving from prototype to production requires integration into existing semiconductor manufacturing processes. That's not a limitation—it's an advantage. Chip foundries have decades of process control and efficiency built into their operations. By fitting this technology into those established workflows, the researchers can leverage that expertise and move toward industrial production. The work was supported by the U.S. Air Force, the National Science Foundation, the National Research Foundation of Korea, and the Draper Scholar Program. What began as a laboratory experiment could soon become something you can manufacture at scale, in the same facilities that produce the chips in your phone.
Notable Quotes
This could give us more information as we study space, or help with environmental protections where you want to monitor for specific compounds in the atmosphere.— Cosmin-Constantin Popescu, first author and MIT PhD student
The key innovation is this crossbar architecture, which creates a scalable way to increase the pixel-level switching of metasurfaces. It's the first time anyone's implemented it.— Juejun Hu, MIT's John F. Elliott Professor of Materials Science and Engineering
The Hearth Conversation Another angle on the story
Why does it matter that this uses a crossbar architecture borrowed from display manufacturing?
Because it solves a problem that's plagued researchers for years. If you want to control light at the pixel level, you need to address each pixel independently. Most approaches require wires running to every single pixel, and routing all those wires becomes a nightmare. The crossbar—two perpendicular grids—lets you address millions of pixels with far fewer wires. It's elegant, and it's proven to work at scale in displays. Using it here for the first time shows the approach actually works for phase-change metasurfaces.
What makes mid-infrared detection special compared to other wavelengths?
It's the sweet spot for seeing what matters. Humans can't see it, but heat signatures live there. More importantly, most organic molecules absorb light in that range—methane, propane, all sorts of chemicals. So if you're trying to detect a gas leak or monitor atmospheric pollution, mid-infrared is where the information lives. It's already used in defense and aerospace. This technology just makes it tunable and compact.
The paper mentions optical computing. How does a tunable infrared chip enable that?
Metasurfaces can encode information into their structure—essentially storing computational weights the way a neural network does. When light passes through, it interacts with the material and produces a result. A static metasurface can only do one computation. But if you can tune each pixel independently, you can reconfigure the computation on the fly. You're not just detecting light anymore; you're using light itself to perform calculations.
Why did durability testing matter so much to the researchers?
Because a prototype that works once is not a device. They needed to know the material could switch between states tens of thousands of times without degrading. If the phase-change material breaks down after a few hundred cycles, it's useless in the field. The fact that it held up through extensive testing suggests it could actually be deployed in real applications.
What's the path from here to something people actually use?
Integration into semiconductor manufacturing. The researchers didn't invent the crossbar architecture—displays have used it for years. But they're the first to show it works for pixel-level control of phase-change metasurfaces. Now they need to scale the pixel count and work with chip foundries to make it part of their standard processes. Once it's embedded in existing manufacturing workflows, you can produce it at scale, at cost, with quality control. That's when it moves from research to reality.