Intelligent Optimisation and Control of Low Temperature Plasma Processes

Research Fellows:  Heather Phillips, Hua Meng, Lars Nolle, Jafar Al-Kuzee
Principal investigator:Adrian Hopgood
Co-investigators:Nick Braithwaite, Phil Picton
Project completed:September 2003

Project summary

Two consecutive projects were carried out in this area with funding from the Engineering & Physical Sciences Research Council (EPSRC). In the first (ref GR/J32213), artificial intelligence (AI) techniques were successfully used to control plasma deposition processes from pump-down to switch-off. The project used a mixture of rule-based, fuzzy-logic, and algorithmic approaches to control the plasma within the known operating parameter space.

The flexibility of the AI controller opened up the possibility of a second project (ref. GR/M71039/01) in which a system was built that could explore the parameter space automatically. The system thus became adaptable to new equipment and was able to find improved and previously unknown operating conditions. In particular, the close control afforded by this approach opened up the capability of exploiting the knife-edge conditions between plasma deposition and etching. Operating under these conditions allowed the production of sub-micron near-perfect vertical mesas and trenches in silicon. These structures form the basis of micro-electromechanical systems whose diverse applications include monolithic integrated circuits, digital bipolar and bipolar complementary metal-oxide-semiconductor devices, and silicon Bragg-Fresnel lenses for hard X-rays.

The second project introduced additional AI techniques, notably evolutionary algorithms and contour following. Our underpinning software architecture was re-designed and re-implemented as DARBS (Distributed Algorithmic and Rule-based Blackboard System), which has formed the basis of several subsequent projects.

Final report (Word format 2.4 MB)
Final report (PDF format 1.2 MB)
IGR form (Word format 171 kB)
Play movie of plasma deposition (17MB - requires QuickTime)

Return to Adrian Hopgood's page.

Valid HTML 4.01! Owner:
  Adrian Hopgood
  March 2019