(Nanowerk News) Focused Ion-Beam (FIB) milling is a nanoscale direct-writing manufacturing technique in which the removal of material from a target surface is induced by a focused ion beam. It is a popular and versatile approach to create structures on the order of around 10–100 nm, especially in the semiconductor industry.
The work product of any FIB milling process is the result of a complex function of beam current, spot size, scan pattern, target material properties and design geometry, specifically the aspect ratio of the pattern.
Given this great complexity, the problem for researchers and industrial users alike is to develop a comprehensive analytical model that describes the physical processes involved in milling. The lack of such a model invariably requires time-consuming trial and error testing to determine optimal process parameters to achieve the intended result of a given milling operation at a given target.
In new work published in nano letters (“Deep Learning Assisted Focused Ion Beam Nanofabrication”) researchers show that deep learning can be used to simulate the post-fabrication appearance of FIB-milled structures in the 2D projection of a scanning electron microscope image, which is a very good indicator of process accuracy and quality.
Since each prediction is generated on a millisecond time scale, the approach can be used in FIB manufacturing processes for repeatability and precision.
In their proof-of-principle study, the researchers trained a neural network to simulate a specific type of FIB milling task on a specific target medium while only varying the ion current and dosage (i.e. keeping all other system parameters constant).
They state that in practice the network would be trained for the respective task(s) (ie depending on the application context, such as in the characterization of semiconductor wafers or in nanofabrication for plasmonics research) on a relevant variety of targets Materials and with a full range of substrate and system metadata (e.g. film deposition methods, rates and thickness, crystal orientations, etc.; ion current, dosage, raster scan pattern, number of iterations, ion source, aperture age, etc.). ).
In this way, the network would gain an “understanding” of the complex relationships between the numerous sample and system parameters that affect the process results.
The authors conclude that there is considerable scope for functional improvement of FIB/SEM systems as integrated micro/nano fabrication and sample characterization platforms (i.e. fabrication and in situ diagnostics) through the application of machine learning methods seems.
For example, families of materials have similar physical properties derived from similarities in composition and atomic/molecular structure, and neural networks are very effective at discovering such patterns in complex, multidimensional data sets; Consequently, they can similarly “learn” that there are relationships between types of materials