Prof Eoin O’Reilly and his Tyndall colleagues aim to create energy-efficient devices for future internet platforms from metastable alloys, thereby leveraging existing mass-production approaches to new functionality and capabilities
Fig 1

CLICK TO ENLARGE Fig 1: Variation of conduction band edge energy and valence band energy in the (a) OFF state and (b) ON state of a (c) tunnelling Field Effect Transistor (T-FET). When a voltage is applied to the gate, the conduction band edge under the gate moves below the valence band edge in the source region, allowing electrons to tunnel from the source valence band to the channel conduction band, thereby switching the transistor on. Electrons at the bottom of the conduction band have the same momentum as electrons at the top of the valence band in a direct gap semiconductor such as GeSn, but different momentum in an indirect gap semiconductor such as Si or Ge. The requirement for conservation of momentum then reduces the tunnelling rate in the indirect gap case compared to what can be achieved with a direct gap

The volume of internet data traffic has been doubling every two years, with the internet now estimated to consume about 5-10% of the electricity that we generate, giving a larger carbon footprint than the whole aviation industry. This points to a critical problem – the internet is already starting to make unsustainable global energy demands which, unless it becomes greatly more efficient, can only get worse.

Each time we ask Google a question, we access a giant data centre of highly interconnected computers. These transistors and interconnects are all consuming so much energy that the major design issue for data centres is how to manage the level of heat created by the computer components.

How can we deal with this problem? In spite of the enormous success of existing devices, there remain critical barriers to their further advancement to support the continued exponential growth of their application in everyday life.

Many approaches are being tried to overcome these barriers, ranging from the investigation of entirely new materials systems, such as graphene and other 2D material systems, through to the investigation of new device concepts such as tunnelling field effect transistors (T-FETs, Fig 1) and junctionless transistors for electronic device applications.

Prof Eoin O’Reilly and colleagues at Tyndall National Institute in Cork have recently received funding from Science Foundation Ireland (SFI) that targets to dramatically enhance the capabilities of the incumbent technologies. They plan to investigate metastable alloys, which combine new elements with well-established materials. This offers the opportunity to leverage existing mass-production approaches to new functionality and capabilities.

New functionality from incumbent devices

As transistors scale towards 10 nm feature sizes and below, the need for new device concepts, including T-FETs and optical interconnects present challenges for the silicon-germanium (SiGe) used in current technologies. SiGe is an indirect-gap semiconductor, which means that it cannot be used as an efficient light source, nor is it best suited for use in tunnelling devices.

Theoretical and experimental analysis shows that the incorporation of tin (Sn) into SiGe alloys can give a direct gap Group IV semiconductor, opening the opportunity to achieve direct gap optical sources and band-to-band tunnelling for the first time in a group IV alloy.

However, because of the chemical and size differences between tin atoms and Si and Ge, the incorporation of tin into SiGe alloys is not straightforward, requiring what are referred to as metastable growth conditions. The growth and demonstration of SiGeSn alloys and electronic devices is being pursued experimentally by Prof Justin Holmes, who is based in Tyndall and in the Department of Chemistry in University College Cork.

Fig 2

CLICK TO ENLARGE Fig 2: (a) Scanning electron microscopy image of GeSn nanowires and (b) Power-dependent photoluminescence of GeSn nanowires at 77 K. Spectrum for low power (P0 = 30 mW) enhanced by a factor of 300 for clarity

Prof Holmes was funded by SFI last year to develop Si-compatible, direct bandgap nanowires from group IV elements as a platform for energy-efficient electronic devices. Using non-equilibrium growth techniques, Prof Holmes and his team have already succeeded to grow GeSn nanowires containing up to 10% tin (Figure 2).

More importantly, they have also shown in a paper published in Nature Communications earlier this year that GeSn can behave as a direct gap semiconductor when as little as 6% tin is incorporated into the alloy. These results provide a very promising basis to now develop and demonstrate GeSn-based electronic and optical devices.


Multiscale simulation

The development and optimisation of new devices requires close interplay between theory and experiment. The exponential growth in electronic device characteristics over the last 50 years has been enabled through the use of T-CAD: technology computer aided design.

However, most device models are based on assuming bulk, average properties for the semiconductor materials being used in the device. This assumption is reasonable for device dimensions exceeding 25 nm, but becomes increasingly untenable as device dimensions shrink towards 10 nm, where a line of only about 50 atoms will take you from one side of the device to the other side. Atomistic details then become of central importance.

The description at a microscopic scale of selected critical regions of a device, down to its very basic atomistic ingredients, is a distinctive feature that new device simulators must include in order to capture details otherwise inaccessible. This issue is important for SiGe alloys, but becomes even more critical for GeSn, and for SiGeSn.

The properties of conventional alloys such as Si1-xGex vary smoothly with alloy composition, x, and so it is possible to treat the alloy as a ‘virtual crystal’, where each atom assumes the average SiGe properties. However, because tin has very different size and electronic properties compared to Ge or Si, this approach does not work for GeSn or SiSn alloys.

Fig 3

CLICK TO ENLARGE Fig 3: Multiscale simulation of electronic devices requires not just to have first principles models at an atomistic level (a) but to be able to treat critical device regions at an atomistic level (b) while using conventional continuum models (c) to then treat the full device

Rather, it is necessary to treat exactly the detail of the atomic structure of the alloy. Using first principles quantum mechanics approaches, it is possible to calculate exactly the electronic properties of supercells containing of order 100 to 1000 atoms. This is an excellent starting point, but is well short of the number of atoms (t 100,000) that are in even the smallest device (Figure 3).

Prof O’Reilly’s SFI project targets to develop models that will operate across a range of length scales, allowing to describe consistently the electronic properties from the atomistic level through to full device simulations. This builds on the methods which his group has established in the European Union FP7 project, DEEPEN.

In DEEPEN, Tyndall has led development of an integrated open source multiscale simulation environment, targeted at problems common to future nanoscale electronic and photonic devices. The new SFI project will take advantage of the work undertaken in DEEPEN, to coherently combine state-of-the-art existing methods with new methodologies, integrated within a multiscale framework spanning from first-principles to macroscopic continuum models.

Tin-based alloys: a bright future

While the SFI projects of Holmes and O’Reilly primarily target the benefits of SiGeSn alloys for electronic devices, the development of these materials offer several further opportunities.

The development of direct band gap SiGeSn alloys should allow for the first time efficient optical absorption and emission from Si-based alloys, finally paving the way to Si- and Ge-based light sources. SiGeSn can also provide the elusive 1 eV absorption material to optimise the efficiency of multi-junction solar cells grown on Ge.

Prof O’Reilly said: “I very much welcome the support that SFI are providing to my group and to Prof Holmes to investigate these novel semiconductor alloys. There are still many fundamental issues to understand regarding the growth and properties of these emerging alloys.

“We believe that they are very interesting alloys to investigate, not just because of their novel material properties but also because they offer a genuine opportunity to bring a paradigm shift in the capability of a wide range of existing semiconductor technologies.”

Prof Eoin O’Reilly, chief scientific officer, Tyndall National Institute, Cork O'RiordanTechelectronics,internet,research,Tyndall National Institute
The volume of internet data traffic has been doubling every two years, with the internet now estimated to consume about 5-10% of the electricity that we generate, giving a larger carbon footprint than the whole aviation industry. This points to a critical problem – the internet is already starting...