-
Neural-network-assisted in situ processing monitoring by speckle pattern observation
Authors:
Shuntaro Tani,
Yutsuki Aoyagi,
Yohei Kobayashi
Abstract:
We propose a method to monitor the progress of laser processing using laser speckle patterns. Laser grooving and percussion drilling were performed using femtosecond laser pulses. The speckle patterns from a processing point were monitored with a high-speed camera and analyzed with a deep neural network. The deep neural network enabled us to extract multiple information from the speckle pattern wi…
▽ More
We propose a method to monitor the progress of laser processing using laser speckle patterns. Laser grooving and percussion drilling were performed using femtosecond laser pulses. The speckle patterns from a processing point were monitored with a high-speed camera and analyzed with a deep neural network. The deep neural network enabled us to extract multiple information from the speckle pattern without a need for analytical formulation. The trained neural network was able to predict the ablation depth with an uncertainty of 2 \micron, as well as the material under processing, which will be useful for composite material processing.
△ Less
Submitted 19 June, 2020;
originally announced June 2020.
-
Fast algorithm for calculating two-photon absorption spectra
Authors:
Yoshiyuki Kurokawa,
Shintaro Nomura,
Tadashi Takemori,
Yoshinobu Aoyagi
Abstract:
We report a numerical calculation of the two-photon absorption coefficient of electrons in a binding potential using the real-time real-space higher-order difference method. By introducing random vector averaging for the intermediate state, the task of evaluating the two-dimensional time integral is reduced to calculating two one-dimensional integrals. This allows the reduction of the computatio…
▽ More
We report a numerical calculation of the two-photon absorption coefficient of electrons in a binding potential using the real-time real-space higher-order difference method. By introducing random vector averaging for the intermediate state, the task of evaluating the two-dimensional time integral is reduced to calculating two one-dimensional integrals. This allows the reduction of the computation load down to the same order as that for the linear response function. The relative advantage of the method compared to the straightforward multi-dimensional time integration is greater for the calculation of non-linear response functions of higher order at higher energy resolution.
△ Less
Submitted 9 December, 1998;
originally announced December 1998.
-
Calculating the linear response functions of non-interacting electrons by the time-dependent Schroedinger equation
Authors:
Toshiaki Iitaka,
Shintaro Nomura,
Hideki Hirayama,
Xinwei Zhao,
Yoshinobu Aoyagi,
Takuo Sugano
Abstract:
An O(N) algorithm is proposed for calculating linear response functions of non-interacting electrons in arbitray potential. This algorithm is based on numerical solution of the time-dependent Schroedinger equation discretized in space, and suitable to parallel- and vector- computation. Since it avoids O(N^3) computational effort of matrix diagonalization, it requires only O(N) computational effo…
▽ More
An O(N) algorithm is proposed for calculating linear response functions of non-interacting electrons in arbitray potential. This algorithm is based on numerical solution of the time-dependent Schroedinger equation discretized in space, and suitable to parallel- and vector- computation. Since it avoids O(N^3) computational effort of matrix diagonalization, it requires only O(N) computational effort where N is the dimension of the statevector.
This O(N) algorithm is very effective for systems consisting of thousands of atoms, since otherwise we have to calculate large number of eigenstates, i.e., the occupied one-electron states up to the Fermi energy and the unoccupied states with higher energy.
The advantage of this method compared to the Chebyshev polynomial method recently developed by Wang (L.W. Wang, Phys. Rev. B49, 10154 (1994);L.W. Wang, Phys. Rev. Lett. 73, 1039 (1994)) is that our method can calculate linear response functions without any storage of huge statevectors on external storage. Therefore it can treat much larger systems.
△ Less
Submitted 5 May, 1997; v1 submitted 26 March, 1997;
originally announced March 1997.
