Imec of Leuven, Belgium, which does R&D for all the major DRAM makers, has shown the way forward for scaling capacitors in DRAMs. A finding by Imec researchers allows the stored charge to be increased while leakage levels remain acceptable.

Imec’s finding is that the effective electron tunnelling mass in the dielectric film of a DRAM metal-insulator-metal (MIM) capacitor is a crucial parameter for the further scaling of DRAM.

This finding results from a study of the impact of the effective tunnelling mass on the intrinsic leakage current in MIM structures. Values for the effective tunnelling masses were extracted by both experiments and calculations for two dielectric film candidates, titanium oxide (TiO) and strontium titanate (STO).

The continuous downscaling of the DRAM MIM capacitor cell – one of the three main components of a DRAM memory cell – has so far been enabled by employing dielectrics with ever increasing dielectric constants.

This way, the amount of charge stored in the capacitor could be enhanced and the leakage currents could be kept below a specified limit. Future vertical DRAM integration schemes however bring about another challenge, namely the development of capacitors with maximum specified physical thickness, and this will be in direct conflict with maintaining leakage currents.

Imec researchers have now shown that, when the insulator thickness is restricted, DRAM scaling may be not limited by the ability to achieve sufficiently high dielectric constants, but by finding a dielectric material with sufficiently high effective tunnelling mass.

The approach taken by the researchers stipulates that the ultimate leakage is limited by the intrinsic direct-tunnelling mechanism, which depends on the dielectric layer’s physical thickness, the metal/dielectric barrier and the effective tunnelling mass.

In particular, the researchers investigated the impact of the effective tunnelling mass on the intrinsic leakage and determined which particular properties of the dielectric material affect the effective tunnelling mass.

Values for the effective tunnelling mass were extracted by both experiments and calculations for two dielectric films: TiO and STO. For the experimental part, films were deposited on 5nm Ru/10nm TiN/Si(100) 300mm diameter wafers. The resulting orientation was (110) for both materials.

The dielectric film physical thicknesses were ~12nm and ~9nm, and relative dielectric permittivities ~80 and ~90 for TiO and STO, respectively. The effective tunnelling mass was extracted by exploiting the non-ideal property of the films, namely the trap-assisted leakage.

The theoretical effective tunnelling mass was deduced from first principles calculations of the complex band structure of both materials.

The study shows that the effective tunnelling mass is a critical parameter for further DRAM MIM capacitor scaling. For example, simulations revealed that with higher effective tunnelling mass, a thinner dielectric film is sufficient to achieve the specified leakage.

The next step is therefore to find dielectric materials with sufficiently high effective tunnelling mass. This search can be facilitated by the first-principle calculations proposed in this work.

The results further suggest that the material crystal orientation impacts the effective tunnelling mass and might become essential to meet the leakage specifications of a given physical film thickness.