Correlation of HPPMS plasma and coating properties using artificial neural networks

Abstract

The development of industrial coating processes for tool coatings by means of physical vapor deposition (PVD) is usually extremely complex. This is caused by the large number of necessary coating batches and associated coating analyses until suitable process parameters are found. Artificial neural networks (ANN) are basically capable of describing complex relationships between various characteristic process values. Hence, within the scope of this paper the capability of describing complex correlations was tested on the example of a reactive high power pulsed magnetron sputtering (HPPMS) (Cr,Al)ON process. Selected process parameters pulse frequency and process gas composition were chosen, since they exhibit strongly non-linear cause-effect relationships. The ANN was used in order to correlate selective results from efficient substrate-oriented plasma diagnostics and coating analyses. Regarding the plasma properties the Al/Cr ratio and the metal-to-gas ion flux ratio were considered. With respect to the coating properties the Al/Cr ratio and the universal hardness were examined. From the correlation of these results, conclusions on the process parameters for desired coating properties were deduced and successfully proven for the investigated HPPMS (Cr,Al)ON process. Hence, the ANN exhibits a great potential to supplement the fundamental understanding of PVD processes in order to contribute to a simplification of the development of industrial coating processes.