My thesis work will focus on studying the switchable optical properties of rare-earth metal hydrides and the underlying mechanism for the metal-insulator transitions in these materials. The goal is to apply the state-of-the-art computational techniques in electronic-structure calculations to the hydrogen-metal system and to gain a more systematic understanding of the anomalous and interesting properties observed.

Being the lightest element, hydrogen easily diffuses through the metal host to form new compounds. These hydrides can be stabilized at various compositions and the change in the electronic, structural, and mechanical properties of the parent metal upon absorbing hydrogen is often dramatic. In addition to playing an important role in the development of renewable energy technologies, the metal-hydride systems exhibit many unique physical properties which are as yet not firmly understood. A number of interesting phase transitions as a function of hydrogen concentration and temperature have been observed. The complexity of hydrogen interaction with materials is particularly interesting and challenging in both technological applications and basic research.

One of the recent discoveries in the hydrogen-metal system is the metal-insulator transition observed in rare-earth hydrides that exhibit switchable optical properties. The rare-earth metals (such as Y, La, and Sc) are able to absorb a large amount of hydrogen, up to three atoms per metal atom. An unusual reversible transition exists between the concentration of two and three, which is characterized by a metal-insulator transition and a change of the optical properties. The origin of this transition is still under debate.

We plan to start from the microscopic quantum theory of electrons and use appropriate approximations to obtain numerical results for this specific problem. In particular, we are using the GW method to calculate the quasiparticle energy in the material, which uses results from standard LDA calculations of density functional theory as input. These are the state-of-the-art methods in electronic structure calculations. In the past year, we rewrote part of the codes so that calculations on materials with a metallic band structure and fractional filling could be done. We also eliminated intermediate disc storage and I/O. By altering the parallel implementation we were able to reduce the overall memory consumption per processor by a factor of 1.5.

Most of our computations to date have been carried out on the Cray-T3E (Mcurie) at NERSC. Our LDA code makes heavy use of the ScaLAPACK library , part of the ACTS toolkit. Shared memory calls are used in the FFT section of the code. For further development of this code we would be interested in any ACTS work on distributed fast-Fourier transforms. The codes which require and extend the LDA results, one computes dielectric screening for example, incorporate operations involving large arrays (tens of millions of components) that are difficult to parallelize via a data parallel algorithm. A combined task and data parallel solution may be needed. Previous calculations using this theoretical method have only been done for smaller systems. We are curious about the results of this method for larger systems and perhaps other ACTS tools can help us achieve them in the most efficient way.