There are several methods of using an accelerator to perform scientific computing at high speed. In this article, I introduce a method using GPU. GPU is processor dedicated to image processing, but with the advent of CUDA, they are increasingly being applied to scientific computing. Also, the trend in supercomputer is GPU-equipped machines, which are increasing every year. Some programming models for using GPU will be introduced, and some examples will also be given. Finally, I will also touch on an example of using machine learning, a recent trend, and a quantum computer simulator for future GPU computing.

Low-energy excitation in a quantum Hall system is plasmons at the edge. Coupled plasmon modes in co- and counter-propagating edge channels of integer and fractional quantum Hall systems are experimentally investigated by evaluating fractionalization of charge wave packets. Particularly, the charge mode is well quantized when tunneling between the channels is significant. The scheme can be used to clarify the edge modes in various edge states in topological materials including quantum spin Hall systems.

Galaxies are believed to have supermassive black holes at their center. When gas falls to a black hole, it releases a large amount of gravitational energy. This energy is converted to internal energies of the gas, which leads to powerful radiations from radio to X-rays. These systems are observed as active galactic nuclei. The gas surrounding the black holes consist of collisionless plasmas where non-thermal high-energy particles are naturally produced. High-energy particles inside accretion flows can produce high-energy gamma-rays and neutrinos via hadronuclear and photohadronic interactions. We demonstrated that the photons and neutrinos from accretion flows onto supermassive black holes can reproduce a wide range of cosmic high-energy backgrounds, including X-ray to MeV gamma-ray and TeV–PeV neutrino backgrounds. Also, we are performing simulation studies dedicated to the particle accelerations in accretion flows. We found that the particle accelerations by turbulence in accretion flows can be described by a diffusion equation in energy space. This field is rapidly growing owing to the development of experimental techniques and numerical simulation methods, and we may be able to understand the sources and production mechanisms of the cosmic high-energy particles near future.

Radiation detection is a common tool covering all fields of physics and was started simultaneously with the discovery of X-rays and radiation at the end of the 19^th century. The principle of radiation detection has changed little since then, relying on the measurement of charges generated through ionization in a material by radiation or of photons generated by luminous phenomena in a material, such as scintillation. We have contrived a new method for radiation detection by measuring the reaction topology of a particle interaction with matter, such as Compton scattering of gamma rays, using a micro-pattern gas detector (MPGD). Here we applied this method to the detection of thermal neutrons using the reaction ³He+n→³H+p (Q=756 keV), and developed the micro Neutron Imaging Detector (μNID). Based on event counting, the μNID gives us the Time of Flight with sub-μs time resolution, a fine position resolution of 100 μm, and an excellent background rejection, easily satisfying the requirements for a neutron imaging detector at the J-PARC Pulsed Neutron Source. The development of the μNID is described and application results from J-PARC are introduced.