As space missions travel farther from Earth, spacecraft must increasingly be able to process and store their own data. Soon, artificial intelligence (AI) could be the primary tool for handling this growing volume of information. NAND flash memory is the current state-of-the-art technology used to store these massive amounts of data, offering storage capacities in the terabit range. It’s the same technology used in laptops, smartphones, and data centers. Ensuring NAND’s reliability in space is critical as these systems increasingly rely on high-density, low-power storage.
But the radiation in harsh space environments can significantly degrade data stored in NAND flash memory. To counteract this, Georgia Tech researchers have developed a new form of NAND flash memory that can both handle AI and withstand extreme radiation.
This technology uses ferroelectricity, which is when certain materials can hold a permanent, spontaneous electric charge, called polarization. In a recent Nano Letters paper, the researchers show that NAND flash memory made with ferroelectric materials can withstand radiation levels up to 30 times higher than more conventional NAND flash memory.
“If you send traditional flash memory to space, the radiation interacting with flash memory’s trapped electric charge can easily corrupt the data,” said Asif Khan, Associate Professor, School of Electrical and Computer Engineering (ECE). “In contrast, ferroelectric NAND flash storage does not store data as trapped electrical charge but rather stores it as polarization in the material. And polarization is very resilient to radiation effects.”
The insight that NAND flash-compatible ferroelectric memory could withstand high amounts of radiation surprised the researchers. Ferroelectricity in hafnium oxide — the silicon-compatible material that makes this memory possible — was discovered just 15 years ago, and Khan’s lab has been determining its capabilities for the past decade. The team knew ferroelectricity was radiation-tolerant but not exactly how tolerant when implemented in NAND flash architectures.
Lance Fernandes, an ECE Ph.D. student and the paper’s first author, built the ferroelectric NAND memory chips in Georgia Tech’s cleanroom, then sent the chips for radiation testing to collaborators at Pennsylvania State University. Those tests revealed just how extreme the technology’s tolerance could be.
The Penn State researchers’ testing showed that ferroelectric flash technology can sustain radiation as high as 1 million rads (radiation absorbed doses) — the equivalent of 100 million X-rays — making it 30 times more durable than traditional memory. This is well within the radiation-tolerance threshold for most spacecraft: Low-Earth orbit satellites require a tolerance of 5 – 30 kilorads, geostationary orbits need 100 – 300 kilorads, and deep space missions top out at 1 million rads.
Here is an exclusive Tech Briefs interview, edited for length and clarity, with Khan.
Tech Briefs: What was the biggest technical challenge you faced while developing this new form of NAND flash memory?
Khan: One of the biggest technical challenges was achieving long-term stable data retention while still maintaining NAND-compatible performance. In simple terms, it is not enough for the memory to just store data — it must reliably hold that data for long periods without drifting or being lost, especially under harsh conditions like radiation exposure.
To solve this, we developed a novel laminated ferroelectric gate stack, where multiple ultra-thin material layers work together to improve both performance and stability. This structure enabled the memory device to achieve the desired performance and strong data retention needed for practical NAND flash applications, while also making it resilient to radiation. In many ways, getting the device stable and reliable before even performing the irradiation experiments was one of the key breakthroughs of the work.
Tech Briefs: Can you explain in simple terms what the testing at Penn State entailed please?
Khan: At Penn State, the chips were exposed to high levels of gamma radiation to simulate the harsh radiation conditions found in space. The goal was to test whether the memory devices could continue operating reliably after exposure to radiation similar to that produced by cosmic rays and solar flares during deep-space missions.
Tech Briefs: Do you have any set plans for further research/work/etc.? If not, what are your next steps?
Khan: The most exciting next step is to actually fly our chips. Ground-based gamma testing is the standard way to qualify space electronics, but the real radiation environment in orbit is more complex — protons, heavy ions, single-event upsets — and the only way to fully validate the technology is to expose it to that environment directly. We are exploring CubeSat and hosted-payload opportunities for an in-space demonstration.
In parallel, we are working on the engineering challenge of scaling our laminated ferroelectric stack into commercial 3D vertical NAND geometries, where memory cells are stacked hundreds of layers high. We are partnering with industry on this, since the manufacturing know-how for high-layer-count 3D NAND lives in the major memory companies. If those efforts succeed, the same kind of solid-state drive that sits inside a laptop today could one day fly inside a spacecraft headed to Europa or Enceladus — taking AI-grade storage along for the ride.
Tech Briefs: Is there anything else you’d like to add that I didn’t touch upon?
Khan: Two things. First, this result is genuinely the product of a multi-institution collaboration spanning Georgia Tech, Penn State, Colorado State, and the University of Modena and Reggio Emilia in Italy. Lance Fernandes, the paper's first author and a Ph.D. student in my group, held the whole effort together.
Second, the bigger picture: Artificial intelligence is moving into space — onto satellites, rovers, and deep-space probes — and AI workloads need a lot of nonvolatile storage. The radiation sensitivity of conventional NAND flash has quietly been one of the bottlenecks on that vision. If ferroelectric NAND can scale, it removes a key obstacle to running AI directly on spacecraft.
Tech Briefs: Do you have any advice for researchers aiming to bring their ideas to fruition?
Khan: My advice would be this: If you see something that others don't yet see, stick with it. Passion is what carries you through — there will be people who doubt you, who tell you the idea won't work, or who simply don't share your conviction. That is normal, and it is often the strongest signal that you are onto something real. The biggest breakthroughs in science and engineering have almost always come from researchers who held onto an idea long after the consensus told them to move on. The win rarely comes on the first try — it comes after you have pushed the idea long enough, pivoted when the evidence demanded it, learned from each failed version, and kept going. Trust your own eyes, do the work, and let the results speak for themselves.
