Stepping up theoretical investigations of ultrashort and intense laser pulses with overdense plasmas. Combining particle-in-cell simulations with machine learning and big data

Over the past decade, advances in the laser technology brought about an increase in the maximum achievable laser intensity of six orders. At the same time, the pulse duration was considerably shortened. The interaction of such ultrashort and intense laser pulses with solid targets and dense plasmas is a rapidly developing area of physics. Hence, a growing interest in characterizing as accurately as possible the phenomena of absorption and reflection that occur during this interaction. Particle-in-Cell (PIC) simulations have traditionally been known to be one of the most important numerical tools employed in plasma physics and in laser-plasma interaction investigations. However, PIC codes are subject to non-physical behaviours such as statistical noise, non-physical instabilities, non-conservation, and numerical heating. Secondly, they require considerable computational resources. This paper proposes a novel approach by combining PIC simulations with machine learning in order to derive optimal laser-plasma interaction scenarios for particular given laboratory experiments. Over 2TB of interaction data consisting of PIC output and also of available literature data have been processed using Hadoop and Apache Mahout, respectively. The combination is a reliable tool for estimations of electron temperatures, plasma densities, parametric instabilities, offering valuable insights on potential interaction phenomena.