Research

Current research

Electronic structure on graphics processing units (GPUs)
I developed an approach to perform DFT calculations efficiently using GPUs. The code, based on a real-space discretization, is one of the fastest DFT implementations available.
Massive parallelization
I have worked in the parallelization of electronic structure calculations on supercomputing platforms, in particular of the real-time TDDFT approach. The resulting implementation can scale to 100,000 cores.
Development of the Octopus electronic-structure code
I am one of the main developers of Octopus, an open source package for electronic-structure simulations used by many research groups. I have implemented several components and coordinated the development effort.

Past research

Compressed sensing for atomic simulations

I have shown that compressed sensing, a method to optimize the amount of samples required to reconstruct a signal, can also be applied to numerical simulations for reduced computational cost and increased precision.

The gap problem in density functional theory (DFT)
I proposed a new approach to address the asymptotic-limit problem in DFT and to predict the fundamental gap. The method gives accurate energy levels for atomic and molecular systems.
Dynamic Sternheimer equation
I developed an new approach for frequency-dependent linear and non-linear response in time-dependent density functional theory (TDDFT). The method was applied to the calculation of different properties publications.
Modified Ehrenfest molecular dynamics
My collaborators and I proposed a new method for ab initio molecular dynamics based on the Ehrenfest approach that is more efficient for large systems than Car-Parrinello and Born-Oppenheimer methods.