VI. RESEARCH DIRECTED BY D. M. HOLBURN, B. C. BRETON AND N. CALDWELL
A. Computational Techniques for the SEM
B. SEM based undergraduate projects for the M.Eng degree
A. Computational Techniques for the SEM
In 1995, Nicholas Caldwell, a graduate of the Computer Laboratory, arrived as a PhD student under the supervision of Holburn and Breton. Caldwell had learned of the research through a chance email from a friend who happened to work at the instrument company. He was eventually funded by a CASE award with LEO Electron Microscopy.
The overall aim of Caldwell’s research was to improve the ease of use of the SEM through more effective automation. This included the development of “AutoSat”, which automatically saturated tungsten filaments to either first (false) peak or second peak (saturation). AutoSat ramped up the filament current, measuring the image intensity at each 10 mA interval, and tracing the intensity curve to locate the first peak and continuing onwards to identify the second peak (if necessary). Failure to detect a singular first peak indicated an electron-optical misalignment, which was rectified by invoking an automatic alignment routine. AutoSat resulted in more accurate saturation and longer filament lifetime.
The major thrust of Caldwell’s work was in creating knowledge-based (expert) systems to perform instrument fault diagnosis and optimal operation. The first task resulted in an expert system (First A.I.D.), which encapsulated service engineer knowledge regarding hardware faults (main electron-optics, subsystems, and X-ray accessories), computer and software problems, and typical user mistakes. Work by Gopal Chand into web-based remote microscopy was incorporated and extended in First A.I.D. to provide a facility to remotely detect and correct selected software faults and maladjusted parameters.
Caldwell ’s second task involved the modelling of SEM operation by expert microscopists. This knowledge model was transformed into a rule-based expert system (XpertEze) to perform appropriate initialisation of key instrument parameters according to sample type, desired magnification, and detector in use. Further optimisations for detecting charging, improving effective resolution, shading and punch, and applying appropriate levels of noise reduction for the intended use of resulting SEM images, were also designed and implemented.
Caldwell was awarded his PhD in 1998, continuing as a postdoctoral worker (funded jointly by the Isaac Newton Trust and LEO) under Holburn and Breton. His research efforts have focused on the ongoing development of web-based systems for instrument diagnosis and control, and further applications of artificial intelligence in microscopy.
The triumvirate of Holburn, Breton, and Caldwell have jointly supervised several graduate students. Christopher Batten (1999/2000) received the M.Phil. for his research into autofocusing and astigmatism correction using various sharpness measures. Batten investigated the use of variable stepsize and Fibonacci search algorithms to locate maximum sharpness, creating new algorithms for full and fine autofocusing as well as real-time autofocusing to detect and correct image defocusing. He also showed that reduced domain median filters could mitigate limited bandwidth distortions, and developed an interpolation method based on a model of variance as a function of defocus. Malavika Chandra (2002) investigated techniques for measuring resolution based on the edges of gallium arsenide wafers. Hui Peng (2002/2003) experimented with Fourier-based techniques for automatic beam alignment of tungsten filaments.
B. SEM based undergraduate projects for the M.Eng degree
The shift to a four-year M.Eng. course at the Department with a final-year project provided the triumvirate with a steady stream of undergraduate students to undertake microscopy-related research and development.
David Wood (1996/1997) implemented a web-based remote control user interface for the SEM using Java applets. Andrew Stewart (son of Gary Stewart) (1997/1998) explored the use of 'ImageObjects' from Synoptics to perform image processing and analysis for the SEM.
Software to teach microscopy through simulation became an important theme with Siva Rubakumar (1997/1998) developing a very basic emulation of the SEM. This was followed by the 1998/1999 Virtual SEM (VSEM) projects of Paul Robertson (who constructed a Java-based simulation of the then current LEO software incorporating QuickTime movies to mimic operations such as focusing, changing magnification, and so on) and John Thompson (who built a basic virtual reality interactive model of the SEM column). Dan Lamping (2000/2001) explored the use of image processing to dynamically generate image sequences from key images rather than storing dozens of movies in the VSEM for all potential simulations of parameter interactions. Geoff Martin (2001/2002) designed and implemented the interactive VSEM Encyclopaedia to teach microscopy concepts and principles. Hui Peng (2002/2003) experimented with Fourier-based techniques for automatic beam alignment of tungsten filaments.
Laurence Brown (2000/2001) and Rob Cole (2001/2002) developed tools to perform automatic and remote diagnosis of the principal circuit boards of the current generation of LEO SEMs. These relied on the ability of newer instruments to measure rather than estimate the actual values of set parameters.Recently, Michael Goodwin and Tom Young (2002/2003) investigated the use of artificial neural networks for image analysis of semiconductors, textiles and particles. Young investigated both neural network and conventional image processing techniques for detecting cracks in wear particles (debris from jet engine lubricants), producing an efficient and robust solution for this application. Most recently, Robert Foreman (2004/2005) has been researching Fourier-based and other techniques for the automatic beam alignment of tungsten filaments.
Batten, C.F. (2000). Autofocusing and astigmatism correction in the scanning electron microscope. M.Phil. Dissertation, University of Cambridge.
Brown, G.L. (2001). Networked Diagnostics for a Scanning Electron Microscope. M.Eng. Dissertation, University of Cambridge.
Caldwell , N.H.M. (1998). Knowledge-based engineering for the scanning electron microscope. Ph.D. Dissertation, University of Cambridge.
Chandra, M. (2002). Investigation, Optimisation and Automation of Resolution Measurement in the SEM. NST Part III Report, University of Cambridge.
Cole, R.A.J. (2002). Networked Diagnostics for Scanning Electron Microscopes. M.Eng. Dissertation, University of Cambridge.
Goodwin, M.K.P. (2003). Application of Neural Networks. M.Eng. Dissertation, University of Cambridge.
Lamping, D.R. (2001). The Virtual Scanning Electron Microscope. M.Eng. Dissertation, University of Cambridge.
Martin, G.C. (2002). Virtual-Reality Scanning Electron Microscope: An Online Encyclopaedia of the SEM. M.Eng. Dissertation, University of Cambridge.
Peng, H. (2003). An Investigation into Image Processing Algorithms for SEM Automation. CPGS Dissertation, University of Cambridge.
Robertson, R.P. (1999). Virtual Scanning Electron Microscope. M.Eng. Dissertation, University of Cambridge.
Rubakumar, S. (1998). Virtual Scanning Electron Microscope. M.Eng. Dissertation, University of Cambridge.
Stewart, A.W. (1999). Image Processing and Analysis for Control of the Scanning Electron Microscope. M.Eng. Dissertation, University of Cambridge.
Thompson, J.S. (1999). Virtual Scanning Electron Microscope. M.Eng. Dissertation, University of Cambridge.
Wood, D.J.G. (1997). Remote Control of a Scanning Electron Microscope over the Internet using Java. M.Eng. Dissertation, University of Cambridge.
Young, T.C.W. (2003). Neural Networks for Automatic Feature Detection in Scanning Electron Microscope Images. M.Eng. Dissertation, University of Cambridge