13 two-dimensional materials identified for FETs, some new

Swiss scientists have had a computational look at two-dimensional materials, like graphene, to see which will make the best transistors.

From 100 candidate compounds, 13 show promise – in some cases more promise than the predicted trend of silicon finfets.

The team, from ETH Zurich and EPFL (École Polytechnique Fédérale de Lausanne), modelled current versus voltage characteristics from first-principles on the Piz Daint supercomputer, combining density functional theory and quantum transport theory on devices with gate lengths from 5nm to 15nm.

The 100 candidates were picked from 2018 work by the EPFL team in 2018, when Piz Daint sifted through 100,000 materials to find 1,825 from which 2-d layers of material could be obtained.

Selection from 1,800 to 100 was based on which mono-layers of atoms were most likely to build into fets.

Piz Daint first determined the atomic structure of the materials using density functional theory (DFT). “They then combined these calculations with a so-called quantum transport solver to simulate the electron and hole current flows through the virtually generated transistors,” according to the Swiss National Computer Centre where PizDaint resides. It used a quantum transport simulator developed by Mathieu Luisier and his team at ETH Zurich – the underlying method was awarded the Gordon Bell Prize in 2019.

Being so thin (usually <1nm), 2-d materials lend themselves to conductivity modulation from a single surface gate on one side.

“Although all 2-D materials have this property, not all of them lend themselves to logic applications,” said Luisier. “Only those that have a large enough band gap between the valence band and conduction band.”

Without a large-enough bandgap, tunnelling effects will cause excessive leakage.

The project aim was to fint 2d materials that could supply 3mA/μm, both as n-type transistors (electron transport) and as p-type transistors (hole transport) – with channels as short as 5nm without impaired switching behaviour.

“Only when these conditions are met can transistors based on two-dimensional materials surpass conventional silicon finfets,” said Luisier.

Of the 13 materials which met the criteria, some are already known – ‘black’ phosphorus and hafnium disulphide, for example. Others are completely new, according to Luisier: Ag₂N₆ or O₆Sb₄, for example.

“We have created one of the largest databases of transistor materials thanks to our simulations. With these results, we hope to motivate experimentalists working with 2-D materials to exfoliate new crystals and create next-generation logic switches,” said Luisier.

The work is published in ACS Nano as ‘Materials for ultra-scaled field-effect transistors: Hundred candidates under the ab initio microscope‘.

Piz Daint, named after a Swiss mountain, is the 10th most powerful computer in the world, and the third most powerful in Europe (after two in Italy) according to the Top500 list. The UK’s Met Office computer is 32 on the list, after two more from Italy, two from France and one from Germany. The fastest computer anywhere at the moment is Arm-based – Supercomputer Fugaku in Japan.