In lab, 0.8-nanometer carbon film could replace two chip layers

On a 4-inch wafer, researchers at the National University of Singapore grew a carbon film thinner than a nanometer and spread it uniformly around patterned structures. The 0.8-nanometer layer could perform two jobs now handled by separate chip-wiring materials: insulating copper lines and stopping copper from escaping into the surrounding insulation.

The team, led by Barbaros Oezyilmaz, a professor at NUS, measured a dielectric constant — the k-value that indicates how strongly a material responds to an electric field — of 1.35 from 0.8 to 2.7 nanometers. Lower values reduce interference between neighboring signals, a growing problem as wiring gaps shrink below 10 nanometers. The film is amorphous, meaning its atoms have no repeating arrangement; NUS says that disorder limits how freely electrons respond to an electric field, helping keep the k-value low even at extreme thinness.

The second job is copper containment. Copper can migrate into insulation and create unintended electrical paths, so chip designs typically add a separate barrier, often tantalum nitride. In accelerated tests, the carbon film blocked copper ions. Its projected time to failure at a typical operating electric field exceeded the benchmark for a 10-year service life and was more than 100 times the corresponding projection for tantalum nitride.

The material also withstood electric fields of 28–31 megavolts per centimeter before losing its insulating ability. NUS measured its hardness at about 100 gigapascals, at least 10 times that of silicon dioxide. The team deposited it by chemical vapor deposition below 300 degrees Celsius, directly on silicon dioxide, copper and cobalt, including sidewalls and corners of patterned structures.

So what changes in practice? If the film works inside future manufacturing processes, combining insulation and copper-barrier functions could free fractions of a nanometer for wider copper lines. Wider lines have lower resistance, which could help data move faster while using less energy — especially in data-hungry artificial intelligence processors. For now, the finding is a laboratory materials result published in Nature Electronics on Aug. 18, 2026, not evidence of a chip already being manufactured or deployed.