In the Right Light, It All Makes Sense
Matthias Imboden · Co-Founder and CEO · 4K-MEMS · September 29, 2026
Almost everything worth measuring — the sugar in your blood, the water in your tissue, the methane in a pipeline, the protein in a grain of wheat — announces itself in the infrared. The obstacle has never been the physics but the light source: broadband infrared has meant hot filaments and bulky cans, compact infrared has meant narrow-line LEDs and lasers that each see one thing.
The Cornerstone of Next-Generation Embedded Infrared Sensing
Overview
Switzerland’s 4K-MEMS SA has built a third option: a thermal light emitting device, or TLED, producing genuinely broadband infrared from a millimeter-scale surface-mount chip, made on standard eight-inch wafers. It is the component the company’s advisors describe as the cornerstone of next-generation embedded infrared sensing — the part that is set first and that everything else is built on. Co-founder and CEO Matthias Imboden explains what makes it hard to build, why the company sells an illumination module rather than a sensor, what an engineer should be skeptical about, and which applications will change first.
The Conversation
MICROTECH VENTURES: Start with the simplest version. What does 4K-MEMS make?
MATTHIAS IMBODEN: We make light, broadband infrared light, from a chip. Our device is a thermal light emitting device, which we call the TLED: a micro-machined structure we heat to above 2,000 kelvin inside a sealed cavity, radiating across roughly 0.7 to 5 microns. That range is the whole story. Nearly every molecule people want to detect has its absorption fingerprint inside it, and until now you could not put a source that wide into something so small. Our first demonstrator is 1.4 by 1.4 millimeters, solders onto a board like any other component, and is RoHS compliant. Future devices will be even smaller.
MICROTECH VENTURES: Broadband infrared is old technology — a filament in a glass bulb does it. What is genuinely hard about doing it on a chip?
MATTHIAS IMBODEN: Heat is hard. You hold a tiny structure above 2,000 kelvin, in vacuum, millimeters from silicon and solder that must stay cool and stable for years, switching on and off up to hundreds of times a second without drift or failure. Three things had to come together: a refractory material that can sustain millions of extreme heat modulations; a sub-micron-precision device design that heats the emitter plate to a uniform temperature while mitigating the thermomechanical stresses that would typically tear the system apart; and a manufacturable MEMS process that a typical foundry can run on eight-inch wafers, so the complexity is amortized through high-volume production. Any one alone is a research result. All three is a commercial product.
MICROTECH VENTURES: Why can’t LEDs or lasers simply be improved to cover the same ground?
MATTHIAS IMBODEN: It is physics and economics, not effort. The LEDs in today’s wearables run out above about 1 micron, but the interesting biochemistry happens at longer wavelengths. Mid-infrared LEDs are inefficient and expensive, and their narrow emission and high temperature sensitivity make them poorly suited for sensing applications. And because all of those are narrow-line, covering a spectrum means stacking ten or more, each with its own drive electronics, alignment, and calibration. We replace the stack with one part. Designers do not just need better LEDs; they cannot find the space to place enough of them to cover a wide sensing band.
MICROTECH VENTURES: Set the spectrum aside for a moment. What else does a chip-scale source do that a bulb cannot?
MATTHIAS IMBODEN: Two things a design engineer feels immediately. It switches roughly ten to a hundred times faster than a halogen bulb, so the signal can be modulated and pulled cleanly out of ambient light and long-term drift. And because a measurement takes only tens of milliseconds, the source spends most of its life off — energy goes into each reading instead of into keeping a filament hot, which is what makes battery-powered infrared sensing realistic in a watch or a meter that has to last years on a cell. The third thing is quieter but defines programs: it is a surface-mount part that goes down on the same pick-and-place line as everything else on the board. A bulb is challenging to integrate into an automated assembly system and leads to higher costs and lower yield. But that is not all: while bulbs generate a lot of light, almost all of it goes to waste. Our TLED interfaces effortlessly and efficiently with microlenses, which means that, at minimal cost, most of the light goes where you want it. It’s a smarter way to engineer optics.
MICROTECH VENTURES: Where do you make the biggest difference first?
MATTHIAS IMBODEN: Gas sensing is one of the first places we make a difference, because the pain is immediate and the regulation is real. Non-dispersive infrared gas sensing is roughly a $1.6 billion market growing about 14 percent a year, pushed by European air-quality rules, methane obligations, and hydrogen blending in the gas grid, which means meters should measure calorific value rather than volume. Those systems are built around bulb sources and are far too large. We have demonstrated a carbon dioxide sensor roughly ten times smaller by volume than commercial equivalents. That is not a percentage improvement; it is a different product category that will generate new use cases and market opportunities.
MICROTECH VENTURES: And beyond gas?
MATTHIAS IMBODEN: Three applications, in rough order of readiness. Portable spectroscopy for material identification, counterfeit detection, grain quality and plastics sorting, where the instrument has shrunk for a decade and the source is now the limiting part. Inline process control, in the chemical industry or pharma, which needs broadband line illumination for continuous quality control, ideally at every process stage. And health, the one that changes daily life. Hydration, lactate, glucose, and blood alcohol all sit in the short-wave infrared, unmeasurable in a wrist device today because nobody could put the right light in one. Hydration and lactate are the honest near-term targets; glucose is the prize, and deserves to be proven slowly.
MICROTECH VENTURES: You are deliberately an illumination module supplier rather than a sensor company. Why is that the right position?
MATTHIAS IMBODEN: Our target is a light engine, an application-agnostic module that generates broadband illumination on command. Our advantage is breadth, and building a finished product eliminates it. The moment we become a glucose company we are one application deep, having thrown away the other nine the same solution enables, and there is a graveyard of companies that raised very large sums trying to own the whole optical stack for one biomarker. Our customers know their application, algorithms, and regulatory path far better than we ever will. What they lack is the IR source.
MICROTECH VENTURES: Your advisors describe 4K-MEMS as the cornerstone of next-generation embedded infrared sensing. Is that a slogan, or does it carry specific claims?
MATTHIAS IMBODEN: It carries three, and we are ready to be held to each. Ubiquity: one emitter spans the near, short-wave and mid-wave infrared, so a single chip serves hydration, lactate, glucose, carbon dioxide, methane and calorific measurement. It collapses a stack of ten or more narrow-line LED sources into one part. Invisibility: at millimeter scale, wafer-level packaged and pick-and-place compatible, it integrates the way a resistor integrates rather than the way an instrument does; the customer’s product looks like their product. Inevitability: standard MEMS steps on a qualified foundry line, RoHS compliant, freedom to operate confirmed, and granted claims in the countries where these products are actually built. A cornerstone is not the part of a building anyone photographs. It is the part everything else is set against.
MICROTECH VENTURES: You describe a system that could be ubiquitous, and this is only feasible if it can be made in volume. Where does manufacturing stand?
MATTHIAS IMBODEN: We are fabless by design and industrializing with a high-volume MEMS foundry, standard processes on a qualified line rather than our own laboratory. An eight-inch wafer yields many thousands of devices, which is what makes consumer volumes work, and the part goes onto a customer’s board with existing pick-and-place equipment, no special handling, no optical bench. We hold eleven patent families with claims granted across the major manufacturing and consumer markets. These focus on system design and architecture rather than process, which gives us the flexibility to work with different foundries, scale production and build our products close to our clients.
MICROTECH VENTURES: What is the first objection a serious engineer raises?
MATTHIAS IMBODEN: Lifetime, and it is the right question. Anything held above 2,000 kelvin invites it, because a filament in a bulb is a consumable and nobody wants a consumable sealed inside a watch or bricked into a wall. Our answer is that most of a bulb’s failure modes belong to the bulb: a coiled wire, a glass envelope, a getter, hand assembly. Ours is a micro-machined structure, sealed in vacuum at the wafer level, pulsed rather than held on, and qualified against the same reflow and reliability standards as the rest of the board. We would rather prove it than assert it. We are developing a comprehensive quality-control program to validate target specifications and continuously improve the system at every node. Engineers who have been let down by an infrared source before are exactly the customers worth winning.
MICROTECH VENTURES: You announced a partnership with MantiSpectra this summer. What does that kind of relationship signal?
MATTHIAS IMBODEN: It signals how the ecosystem actually assembles. MantiSpectra builds a compact infrared spectral sensing chip; we build the source. Neither has to become the other to give a customer something complete. That is the pattern we want to repeat: emitter, detector, optics, control electronics, and algorithms from specialists excellent at one layer each, integrated into a module an OEM can design in. It is also the fastest way to be evaluated: rather than imagine what our chip enables, a company can look at a working system.
MICROTECH VENTURES: Say a company reads this and is interested. What happens next, concretely?
MATTHIAS IMBODEN: A technical call first, where we look at their spectral requirements, their optical path and their power budget and say plainly whether the part fits. If it does, samples and the preliminary datasheet follow, and where it helps we bring a working demonstrator so the conversation is about measured spectra rather than slides. From there most relationships become a joint development project scoped to one end product, combining their application knowledge with our source. The industrialization timeline partly defines the shared calendar. What we ask in return is a real specification: the band, the signal-to-noise the application needs, and the volume if it works. With that on the table, both sides can determine whether this becomes a design-in.
MICROTECH VENTURES: If this works, what is different in three to five years?
MATTHIAS IMBODEN: Infrared sensing stops being an instrument you buy and becomes something simply present, in a gas meter, a watch, a production line, a ventilation duct. The measurement moves next to the thing being measured, where it always belonged. I would like the reason to be that there was finally a source small enough, broad enough, and cheap enough to put there.
MICROTECH VENTURES: What are you looking for from the companies reading this?
MATTHIAS IMBODEN: A conversation with whoever owns the hard constraint: the advanced sensing team, the module architect, the engineer told to shrink the optical stack. We will bring a device and the measurements behind it. If it belongs in your product, we will both know quickly. In the right light, it all makes sense.
Note
Microtech Ventures is an advisor to 4K-MEMS SA. Technical specifications describe preliminary devices and are subject to change, and forward-looking statements about markets, applications and timing are not guarantees of future performance.