Jan 28

Google’s neural network beats humans at the ancient game of Go

A neural net system built by Google has beaten the European champion in the Chinese game of Go, winning five out of five games and crossing a new threshold for machine intelligence.
In a recent paper published in Nature,  DeepMind researchers revealed how the system was constructed and how it was able to succeed where decades of previous Go systems have failed. Go has long been considered one of the hardest games to automate, making the new DeepMind system particularly interesting for artificial intelligence researchers. The game of Go has long been viewed as the most challenging of classic games for artificial intelligence owing to its enormous search space and the difficulty of evaluating board positions and moves. Here we introduce a new approach to computer Go that uses ‘value networks’ to evaluate board positions and ‘policy networks’ to select moves.

They used a neural networks trained by combinations of supervised learning from human expert games, and reinforcement learning from games of self-play. “Without any lookahead search, the neural networks play Go at the level of state-of-the-art Monte Carlo tree search programs that simulate thousands of random games of self-play”, they said.

Jan 09

New hints on the metallization of Hydrogen

Recently scientist from the Edinburgh’s Centre for Science at Extreme Conditions  have come close to creating a long-sought new state for hydrogen. The group used a set-up called a diamond anvil cell to compress its sample of molecular hydrogen. This apparatus is essentially two gems that have been placed in opposition to each other. Their polished tips, comparable in size to the width of a human hair, are made to press into a cavity containing the sample.

In their experiments, the scientists are able to achieve in excess of 350 gigapascals (3.5 million atmospheres) at room temperature. These pressures are not dissimilar to what would be experienced at the centre of the Earth. The big squeeze on the molecules of hydrogen gas turns them first into a liquid and then into a solid.


As the pressure gets ever more intense, the atoms in the hydrogen molecules pack closer and closer together,
and the electrical conductivity in the crystalline material increases.
Ultimately, the hydrogen atoms should stack so efficiently that their electrons become shared -just as in a metal. However, the team does not quite see this phase, but rather something that is probably just short of it. The work puts new constraints on where the full metallic hydrogen phase might exist:  possibly below 450 gigapascals at room temperature.

The ambient temperature is very significant, because if metallic hydrogen can ultimately be produced this way it opens the door potentially to a new type of perfect (zero resistance) conductor – a material to boost the performance of next-generation computers. “It’s been predicted that metallic hydrogen could be a room-temperature superconductor, which is still yet to be achieved with any material…
Scientists are also fascinated by metallic hydrogen because they think it may account for a large fraction of the internal composition of planets such as Jupiter.
Link to the original work published in Nature (letter) on January 7th, 2016.
BBC reportage.


Dec 07

Yet another superconductor above 100 K ?

Following the recent discovery of very high temperature conventional superconductivity in sulfur hydride (critical temperature Tc of 203 K). Experimentalist have shown results on the covalent hydride phosphine (PH3) which also exhibits a high Tc > 100 K at pressure P > 200 GPa. However, neither the exact composition nor the crystal structure of the superconducting phase have been conclusively determined. In our latest work (December 7, 2015) the phase diagram of P-H was extensively explored by means of ab initio crystal structure predictions. The results did not supported the existence of thermodynamically stable PHx compounds, which exhibit a tendency for elemental decomposition at high pressure even when vibrational contributions to the free energies are taken into account. Although the lowest energy phases of PH1,2,3 display Tc’s comparable to experiments, it remains uncertain if the measured values of Tc can be fully attributed to a phase-pure compound of PH. Link to our work here. This work is featured as en Editor’s Suggestion published in  Rapid. Comm. Phys Rev. B (2016).


Aug 27

Superconductivity under high pressure at minus 70 degrees Celsius

Researchers at the Max Planck Institute for Chemistry in Mainz and the Johannes Gutenberg University Mainz observed that hydrogen sulfide becomes superconductive at minus 70 degree Celsius —when the substance is placed under a pressure of 1.5 million bar. This corresponds to half of the pressure of the earth’s core. With their high pressure experiments the researchers in Mainz have thus not only set a new record for superconductivity— their findings have also highlighted a potential new way to transport current at room temperature with no loss.

The work was published in Nature Materials Journal (link) on August 17, 2015. The first original results were released in Cond. Mat. (arxiv) back on December 2014.


Soon after the discovery in the H-S system, we predicted another system with fairly similar characteristics: transition temperatures of the order of 100 K could be reached in principle in H-Se system at lower pressures. This research is publicly available in arxiv and it was first show on 26 January 2014.

Mar 27

Superconductivity in group-IV honeycomb materials

Our work has just been published in Physical Review B  !

disilicidesIn this theoretical work, we studied in detail the electron-phonon superconductivity of honeycomb MX_2 layered structures where X is one element of group IV (C, Si, or Ge) and M is an alkali or an alkaline-earth metal.  Among the studied compositions we predict a novel compounds energetically stable and showing relatively high transition temperatures of superconductivity:  7 K in RbGe_2 , 9 K in RbSi_2, and 11 K in SrC_2. All these compounds feature a strongly anisotropic superconducting gap. Our results show that despite the different doping levels and structural properties, the three families of materials fall into a similar description of their superconducting behavior. This allows us to estimate an upper critical temperature of about 20 K for the class of intercalated group-IV structures, including intercalated graphite and doped graphene.


Sep 15

Thesis defense


Bonjour a tous! Je suis très content de pouvoir inviter toutes les personnes intéressantes que j’ai rencontré ici, à Lyon et de manière générale en France à deux événement très importants dans ma vie. D’abord, la soutenance publique de ma thèse, et deuxièmement le fameux pot de thèse. La soutenance de thèse se dérouleront dans la salle de conférence de la bibliothèque universitaire de Lyon 1 sur le campus de la DOUA le mardi 18 septembre 2012 à 14:00 heures. Et après, le pot de thèse (si tout se passe comme il faut) sera dans un autre bâtiment vers, 16:30. Merci a la Pequetshina!

Aug 27

Kick-off ! Fysik-Aztek


Kick-off ! I am delighted to welcome you to my webpage.

Officially opened on Monday 27 august of 2012 at 09:00 local time (Paris).

» Newer posts