Since the Nobel prize winning invention of the scanning tunnelling microscope (STM) by Gerd Binnig and Heinrich Rohrer in 1981, it has been associated with some of the most inspirational and elegant experimental works in physics and chemistry. Arguably the instrument enables the ultimate imaging and control of matter on surfaces; by exploiting the rules of quantum mechanics one can image and even manipulate individual atoms, going far beyond traditional imaging techniques. It also, unfortunately, can be one of the most frustrating techniques used by the nanoscientist because the instruments operation is inherently reliant on the quality of a probe scanning (under feedback) just a few Angstroms (one ten millionth of a millimetre) above the surface of inquiry. Many difficult hours are spent manually altering the microscopes control parameters to coerce the arrangement of atoms, or atom, at the probes apex into a configuration that yields, retains, and accurately reproduces atomic resolution images.

Over some 30 years the community has made a number of gradually improving manual methods to mechanically sharpen the probes apex (e.g. electrochemical etching, electron and ion bombardment, and explosive delamination) which must then be followed by a human operator ís fine tuning of the imaging control parameters, which can also affect the probes structure. During the probe (and thus image) optimization stage these control and conditioning parameters are adjusted based on the operatorís experience (usually acquired over the course of a PhD) rather than arising from a well-defined metric (or set of metrics). We have addressed these longstanding problems of probe optimisation and metrics by using a Machine Intelligence approach that combines Machine Vision (including classification with universal similarity metrics) and a Cellular Genetic Algorithm. Our automated software control system varies the microscope's imaging parameters and evolves toward a good image (thus a Beat the Nano-machine held at the University of Nottingham (regulated by the institutes professional standards) between the computer system and human contestants, the computer system, overall, achieved best image and in a shorter time. The system regularly obtains better quality images than the first published results of the prototypical tip-sample combination; a PtIr probe scanning a highly oriented pyrolytic graphite (HOPG) sample. The system is so reliable that commercial licenses are being pursued (software of this nature cannot be patented under UK law). Additionally, the system could obtain different probe types by selecting different target images (as the image is the convolution between surface and probe). These results showed apparent phenotypic plasticity (same imaging parameters, different image), an important result and has obvious potential in the analysis and interpretation of SPM image generation.

This work has been published in R.A.J. Woolley, J. Stirling, A. Radocea, N. Krasnogor, and P. Moriarty. Automated probe microscopy via evolutionary optimization at the atomic scale. Applied Physics Letters, 98(25),253104, 2011 and constitutes a crucial stepping stone towards resolving a 30 years old problem: the automation of imaging and atom-by-atom materials fabrication.

The tantalising promise of creating nano-enabled devices that provide faster, safer, smaller and cheaper products, even quantum computers, relies on the ability of the researcher to image, interrogate and manipulate matter at the nanometre length scale. The scanning probe microscope is an obvious instrument to use in this area but since its invention has suffered greatly from the rate limiting step of probe optimisation. Our work presents a novel solution using a genetic algorithm to solve this long standing problem. Moreover, the control system repeatedly surpasses the performance of a human operator. Of particular relevance is the fact our system is the only method in the world of optimising a scanning probe without a human operator, hence, we are not only human-competitive but even human expert-independent.

This enabling technology has the potential to reach into thousands of research labs around the world, saving countless hours of operator time and thus greatly improved efficiency. Critically if we are to realise atomically precise engineering and manufacture based on scanning probe technology we believe we have created a technology cornerstone.

Finally, of particular importance is that the genetic algorithm controls a real world instrument, ergo, a robot whose actuators can manipulate individual molecules and even single atoms. In essence, when the system optimises a probe it is evolving a nanostructure toward a predefined functionality and also learning the ideal parameters to use that nanostructure.