Atomera (Nasdaq: ATOM) will present work on 4F² DRAM at SISPAD 2026, the semiconductor process and device simulation conference, held this year in Kumamoto, Japan. The paper, “Enabling 4F² DRAM Junctionless Vertical Channel Transistors with Novel Oxygen Inserted Si Substrates,” is scheduled for September 28 in Japan (the evening of September 27 in California). Staff engineer Albert Lu is presenting.
Kumamoto is a fitting venue. It’s where TSMC built its first Japanese fab, and it’s become a symbol of Japan’s push back into chipmaking.
Why 4F² matters
DRAM makers are running out of room to shrink. Today’s cells use a 6F² layout, with the transistor lying flat next to the capacitor. A 4F² cell stands the transistor on end, as a vertical channel transistor (VCT), and puts the capacitor on top. That cuts cell area by about a third without shrinking the lithography, which is why Samsung, SK hynix and China’s CXMT are all working on it for the next generations of DRAM.
Vertical transistors bring their own problems. They need enough drive current, they have to avoid the floating body effect that makes the transistor behave unpredictably, and they have to leak little enough that the cell holds its charge between refreshes.
Atomera’s pitch
Atomera licenses MST, a technology that inserts very thin oxygen layers into the silicon. The paper argues that MST substrates make a junctionless VCT design workable and meet all three requirements. Atomera says the approach simplifies the process flow, cuts selective epitaxy steps and offers a simpler, cheaper path to 4F² than current approaches.
A second poster, from principal engineer Dr. Ilya Rumyantsev, models leakage and breakdown in GaN-on-silicon power devices. It reports that MST-treated devices show lower leakage at low voltage and a delayed current rise at high fields compared with plain silicon.
The caveat
This is simulation work at a modeling conference, not a customer win. Atomera has spent years licensing MST with small revenue, and investors have heard promising technical results before. What would change the story is a memory maker committing MST to a 4F² production flow.
Still, the target is right. Whoever makes 4F² cheap to build will matter a great deal to the memory market later this decade.