A research group comprising members from a number of UK universities has developed a method that slashes the variation in the behaviour of semiconductor hetrostructure devices made using molecular beam epitaxy
The team is making Aspat diodes, used in microwave detection. They consist of a 10-monolayer 100% pure AlAs barrier buried in a GaAs n-i-n multilayer structure with an asymmetric doping profile from the top surface to the substrate.
Using a rapid offline calibration technique, the group first grows a sacrificial coarse calibration layer consisting of a superlattice and a doping staircase. This is then assessed using X-ray diffraction and C-V profiling so that the growth conditions can be fine-tuned.
Professor Michael Kelly from Surrey University's School of Electronics, Computing and Mathematics said: "A variation of one monolayer can lead to a current variation of a factor of four. Conventional growth methods produce devices accurate to 11 monolayer. Our method achieves an accuracy of 10.2 monolayers."
Once the coarse calibration structure has been fabricated and analysed, a batch of 20 precision devices can then be created using the same machine.
Prof Kelly said: "For every one wafer you end up throwing away, 20 precise, uniform, ones can be grown."
The team believes that the method can be refined further.
"Engineers want devices to have less than 115% absolute current variation about a specified value," said Prof Kelly. "So far, we are at about 20 to 30%. The calibration structure can be improved and we think we already have the data to achieve this."
As well as the team members from the University of Surrey, Dr Mohamed Missous from the University of Manchester, Institute of Science and Technology, has played a large role by growing the semiconductor layers. The devices were processed at Cambridge University and calibration tests were performed at Loughborough Surface Analysis.
The technique is not just applicable to microwave devices, but any devices which rely on electron tunnelling.
Marconi, which did some of the initial tests, is interested in the final product, and the team expects to have achieved the required precision levels within two years.