Towards in situ monitoring of YBCO T-c and J(c) via neural network mappingof Raman spectral peaks

Raman spectroscopy has proved promising for use as an in situ process monit or and feedback control sensor during pulsed laser deposition (PLD) of YBa2 Cu3O7-x(YBCO) thin films. Several engineering challenges must be solved bef ore Raman spectroscopy can be utilized in situ. Taken in the aggregate, dep osition conditions such as high temperature, low vacuum, and electromagneti c radiation emitted from the plume, lead to a hostile environment for in si tu Raman spectroscopy. This paper investigates the feasibility of applying Raman as a process-control sensor. Tests are conducted to determine Raman p robe integrity under processing conditions when positioned 5-10 mm above th e substrate. An empirical characterization of the black-body effects in thi s temperature region is presented to determine impact over the spectral reg ion of interest. Optical emission from the ablation plume, a series of disc rete spectral features (emission lines) across the visible and UV regions, is investigated to determine if the features coincide with any of the signi ficant peaks in the Raman spectra needed for in situ control. Additionally, a study of the effects of surface degradation of YBCO films on both the Ra man spectra and film superconductive transition critical temperature (T-c) is undertaken. A neural-net (NN) architecture is built, which maps T-c from reduced Raman spectral data and optimizes the number of nodes based upon t he generalization error. (C) 1998 Published by Elsevier Science Ltd. Ail ri ghts reserved.