Comparing the ISO-recommended and the cumulative data-reduction algorithms in S-on-1 laser damage test by a reverse approach method

• Published in: Review of scientific instruments Vol. 89; no. 1; pp. 013104 - 13112
• Main Authors: Zorila, Alexandru, Stratan, Aurel, Nemes, George
• Format: Journal Article
• Language: English
• Published: United States American Institute of Physics 01.01.2018
• Subjects: Algorithms; Computer simulation; Distribution functions; Fluence; Goodness of fit; Laser damage; Normal distribution; Optical materials; Probability distribution; Probability distribution functions; Reduction; Scientific apparatus & instruments
• ISSN: 0034-6748; 1089-7623; 1089-7623
• DOI: 10.1063/1.4989930

We compare the ISO-recommended (the standard) data-reduction algorithm used to determine the surface laser-induced damage threshold of optical materials by the S-on-1 test with two newly suggested algorithms, both named “cumulative” algorithms/methods, a regular one and a limit-case one, intended to perform in some respects better than the standard one. To avoid additional errors due to real experiments, a simulated test is performed, named the reverse approach. This approach simulates the real damage experiments, by generating artificial test-data of damaged and non-damaged sites, based on an assumed, known damage threshold fluence of the target and on a given probability distribution function to induce the damage. In this work, a database of 12 sets of test-data containing both damaged and non-damaged sites was generated by using four different reverse techniques and by assuming three specific damage probability distribution functions. The same value for the threshold fluence was assumed, and a Gaussian fluence distribution on each irradiated site was considered, as usual for the S-on-1 test. Each of the test-data was independently processed by the standard and by the two cumulative data-reduction algorithms, the resulting fitted probability distributions were compared with the initially assumed probability distribution functions, and the quantities used to compare these algorithms were determined. These quantities characterize the accuracy and the precision in determining the damage threshold and the goodness of fit of the damage probability curves. The results indicate that the accuracy in determining the absolute damage threshold is best for the ISO-recommended method, the precision is best for the limit-case of the cumulative method, and the goodness of fit estimator (adjusted R-squared) is almost the same for all three algorithms.