Novel technique developed to grow semiconductors on silicon
Startup company AmberWave Systems, based in Salem, USA, has developed a novel technique to grow semiconductors, such as germanium, gallium arsenide and indium phosphide, on silicon. Researchers hope the method will lead to faster and smaller transistors, and cheaper lasers and photonic devices. By etching trenches about 500nm deep through a silicon oxide insulation film on a silicon wafer, the team has epitaxially grown non-silicon semiconductor materials through chemical vapour deposition.
Dr Anthony Lochtefeld, Vice President of Research at AmberWave Systems, explains, ‘Electrons can move much faster in various non-silicon semiconductor materials. However, only silicon substrates are suitable for mass manufacture [due to their] mechanical and electrical properties. This means that key materials like germanium, indium antimonide and gallium arsenide need to be integrated onto the silicon wafer.’
Faster and smaller transistors are the key to extending Moore’s Law – this chip industry axiom predicts that doubling the number of transistors in an integrated electronic circuit every two years will improve performance.
Furthermore, by depositing materials such as indium phosphide and germanium, which emit or detect light, on silicon, manufacturers could combine the light handling and electronic functionalities of photonic devices in one chip. This would in turn reduce the costs of packaging multiple chips in a module and of module-level integration and chip interconnection.
‘Growing epitaxially gallium nitride on silicon is reasonably well established,’ comments Professor Christopher Snowden, Vice Chancellor of the University of Surrey, UK, and a specialist in semiconductor materials and devices.
‘Other combinations have been explored – for example, gallium arsenide on silicon – but the economics of this combination do not stack up. ‘[Overcoming] lattice mismatch and strain are key parameters. The use of trenches to negate the effect of dislocations would appear to be innovative and interesting. It would be good to compare the outcome and economics of the AmberWave approach with whole wafer epitaxy.’
In the AmberWave technique, defects in the crystal lattices are trapped at the vertical sidewalls of the trenches, allowing high quality non-silicon semiconductors to be grown above the defect-trapping region. But this is only achievable if the height of the trench is the same as or greater than its width – hence the method is called ‘aspect ratio trapping’ (ART).
Lochtefeld explains, ‘The atoms of the non-silicon semiconductors are typically much larger than silicon atoms. This leads to mismatch problems when you try to combine them [and] defects in the crystal lattices, which degrade device performance and reliability.
‘Thermal expansion and contraction of silicon is [also] different compared to that of non-silicon materials. This leads to cracking of the [latter] during large temperature changes [during] device manufacture.’
According to AmberWave, silicon oxide film growth of just 0.5 microns thick can act as a buffer for lattice mismatch, while low epitaxial thickness can reduce the stresses and eliminate the cracking caused by thermal expansion coefficient differences.
The team has successfully grown germanium and gallium arsenide, and initial results for indium phosphide are promising. By collaborating with universities in the USA, the company aims to develop prototypes of germanium and III-V transistors by the end of this year.
But planarisation of the device needs to be addressed. ‘Aspect ratio trapping does not produce flat surfaces,’ says Lochtefeld. ‘We are exploring chemical mechanical polishing of epitaxial ART regions so that standard device fabrication techniques can be used. [Also] because the trench depth needs to be about the same as the width, a single region produced by ART is more than enough for transistors [but] not large enough for lasers. We are working on variations of the technique to overcome this limitation.’
Snowden adds, ‘AmberWave’s local growth in defined areas may allow a better combination of silicon and non-silicon technologies if it can develop the technique to achieve flexible manufacturing and design.’