# How Chips Use On-Chip Triboelectric Generators for Energy Harvesting

An episode of [The Hardware Podcast with Fexingo: Chips, Devices, and Electronics Engineering Conversations](https://invisiblenetwork.ai/podcast/1896799982), 22 Jul 2026, 7:31.

Audio: https://audio.fexingo.com/business/the-hardware-podcast/episode-0126.mp3

In this episode, Lucas and Luna explore how on-chip triboelectric generators are enabling energy harvesting from mechanical motion—vibration, typing, footsteps—to power sensors and IoT devices without batteries. They dive into the physics of the triboelectric effect, the specific challenge of integrating nanoscale generators onto silicon, and a real-world case from a 2026 paper out of the University of Michigan achieving 3.2 microwatts per square millimeter from vibration. The hosts also discuss commercial use cases from industrial vibration monitoring to self-powered wearables. #TriboelectricGenerator #EnergyHarvesting #IoT #SelfPoweredSensors #VibrationEnergy #Nanogenerator #UniversityOfMichigan #ChipDesign #Semiconductor #PowerManagement #Wearables #IndustrialIoT #Microelectronics #CleanEnergy #Technology #FexingoBusiness #BusinessPodcast #HardwarePodcast Keep every episode free: buymeacoffee.com/fexingo

## Transcript

Lucas: So you're in a factory with thousands of vibration sensors all running on coin-cell batteries that die every few months. Someone has to physically swap them, which costs time and money. What if the sensor just powered itself from the vibration it's already measuring?
Luna: That's the dream, right? Self-powered IoT. But we've been hearing about energy harvesting for years, and it's still mostly niche.
Lucas: It is. But there's a specific approach that's been quietly making progress inside the chip itself, not an add-on harvester. It's called the on-chip triboelectric generator. And it's actually starting to deliver useful power at the microwatt level.
Luna: Triboelectric—like the static shock you get from a doorknob?
Lucas: Exactly that. The same physical principle. When two different materials come into contact and then separate, electrons transfer from one to the other, building up a voltage. If you channel that into a circuit, you get a current pulse.
Luna: So we're basically putting a miniature static-electricity generator onto a chip.
Lucas: Precisely. Researchers have been working on triboelectric nanogenerators—TENGs—for about a decade. But most of them were macroscopic: a little device you glued onto a machine. The new push is to fabricate the TENG directly into the silicon die, using standard MEMS processes.
Luna: And that's hard because…?
Lucas: Several reasons. First, the triboelectric effect is surface-driven—you need a large contact area relative to the chip footprint. Second, the materials that work best, like fluorinated polymers, aren't standard CMOS materials. And third, you've got to mechanically isolate the moving parts so they actually vibrate without breaking.
Luna: So how are they solving that?
Lucas: A team at the University of Michigan published a paper earlier this year—I think it came out in March 2026—where they built a cantilever structure out of silicon, coated with a thin layer of a polymer called FEP, fluorinated ethylene propylene. The cantilever vibrates at its resonant frequency, and as it contacts a fixed electrode, it generates charge.
Luna: What kind of power are we talking?
Lucas: They reported 3.2 microwatts per square millimeter of chip area from a vibration source at 50 hertz—that's typical for industrial machinery. A square-millimeter chip can power a basic temperature sensor and a Bluetooth Low Energy beacon once every few seconds.
Luna: That's actually useful. It's not running a phone, but for sensor nodes it's plenty.
Lucas: Right. And the key is that it's on-chip. You're not adding a separate harvester module. The MEMS structure is part of the same silicon die that carries the sensing and radio circuitry. So the bill of materials shrinks, assembly is simpler, and reliability goes up.
Luna: What about real-world deployments? Is anyone commercializing this?
Lucas: There's a startup out of Singapore called Vibron that's been sampling an industrial vibration sensor powered entirely by a chip-scale TENG. They're targeting predictive maintenance on motors and pumps. They claim a 10-year lifetime with no battery, just from the machine's normal vibration.
Luna: That's a game changer for factories. They'd save on battery replacement, but also on wiring—because these could be wireless.
Lucas: Exactly. And it's not just industrial. There's work on using body motion to power wearables. Imagine a smartwatch that charges itself from your arm swing. Or a pacemaker that tops up from the heartbeat.
Luna: Heartbeat—that's a much lower frequency, though. Like one to two hertz. Can the cantilever tuned for 50 Hz still work?
Lucas: It can't—you'd need a different mechanical design. But researchers at UCLA published a flexible TENG in 2025 that wraps around a blood vessel and generates power from the pulse. It's not on-chip yet, but it points in that direction.
Luna: So the challenge is really about mechanical resonance matching. You need to tailor the structure to the vibration source.
Lucas: Exactly. And that's why the on-chip approach is clever—you can design an array of cantilevers with different resonant frequencies on the same die, covering a range of conditions.
Luna: So a single chip could harvest from a 10 Hz footstep or a 100 Hz motor.
Lucas: In principle, yes. There's a paper from IMEC last year showing a four-cantilever array that covers 20 to 200 Hz. They got a peak power of 5 microwatts at 80 Hz. It's not huge, but it's enough for intermittent sensing.
Luna: What about the other end of the spectrum? Very low vibrations, like in a home?
Lucas: That's tougher. Ambient vibrations in a typical office are around 0.1 G or less. Most TENGs need at least 0.5 G to start producing meaningful power. So for now, these are best for industrial or automotive environments where vibration is strong and predictable.
Luna: And what about durability? If the cantilever is physically moving millions of times, does it wear out?
Lucas: That's actually one of the surprising results. The Michigan team tested their device for 100 million cycles and saw only a 10 percent drop in output. The polymer coating is surprisingly robust. And because it's on silicon, you can add a thin passivation layer to protect against debris.
Luna: So it's a viable long-term solution. That's impressive.
Lucas: It really is. Now, one thing I want to touch on—and this ties into why we can have this conversation without an ad break—is that we deliberately keep this show ad-free. We think a focused, honest tech conversation shouldn't be interrupted by a sponsor spot for a VPN or a mattress.
Luna: Absolutely. And if that resonates with you, supporting the show on Buy Me a Coffee is what keeps it that way. That link is buy me a coffee dot com slash fexingo.
Lucas: No pressure—just if you find yourself getting value out of episodes like this, it's a nice way to keep the lights on.
Luna: Yeah. So back to the tech: where do you see this heading in the next, say, three years?
Lucas: I think we'll see the first commercial chips that combine a TENG harvester, a sensor, and a radio in a single package, maybe as soon as late 2027. The power density is still a bit low for continuous streaming, but for event-driven sensing—like 'send an alert when the temperature spikes'—it's already there.
Luna: And that's a huge chunk of the IoT market. Most sensors don't need to stream; they just need to report anomalies.
Lucas: Exactly. And if you can make those sensors battery-free and wireless, you suddenly enable dense sensor networks that are too expensive to maintain otherwise.
Luna: There's also the environmental angle. Removing millions of coin cells from landfills is a real benefit.
Lucas: Absolutely. So the next time you walk through a factory, think about all those vibrations going to waste. In a few years, they might be powering the very sensors that keep the machines running.

Shows are listed from Apple Podcasts' charts and directory, and Podcast Index where it is used. Every show and episode belongs to its publisher and plays from the publisher's own feed; this site hosts none of them.