Research School of Chemistry Machine VP
Co-Investigators John W White, Philip A Reynolds and Piotr A Wielopolski
Research School of Chemistry and
Australian National University Supercomputer Facility
Molecular Simulations of Superconducting Crystals RbxC60
Reaction of buckminsterfullerene, C60, with alkali metals can result in materials which remain superconducting to high temperatures (e.g. Rb3C60 Tc= 30K). Other compositions, alkali metals, and dopings of other molecules, change the electrical properties markedly. Our recent work has concentrated on linking the structure and dynamics of some of these materials to their superconductivity [1,2]. It has become clear that while the interaction of the potentially superconducting electrons with vibrations of the C60 ball is important, details of the structure are equally so. X-ray and neutron experiments show that the time-average of the structures are mostly disordered. Molecular simulation of the crystal structures offers a way of examining the time development of the structures which is not available experimentally. In particular we will attempt to simulate the relative positions and hopping motions of the rubidiums relative to the C60 balls, and the resulting charge changes on the balls themselves. Understanding this may provide clues about the mechanism of superconductivity in this class of materials.
What are the basic questions addressed?
Superconductivity in the RbxC60 system requires electron transport between C60 balls. Experimentally this is modified by rubidium content. We wish to know the arrangement of the rubidiums within the crystal, and the effect this arrangement has on the electron distribution on the C60 fragments, for varying rubidium contents. This information is not available experimentally. As a first approximation we will use a classical model which has been used successfully in the graphite intercalates.
What are the results to date and the future of the work?
Before we did any computation we imagined that the rubidiums would not be very ordered, and that they would move among the holes between the C60's in a somewhat random fashion, only correlating their motion sufficiently to prevent Rb-Rb overlap. We conceived of the charge on the C60 balls as merely modifying this by providing screening, even if this was quite large. That is to say the behaviour would resemble the Cs-graphite simulations.
The first simulations (of RbC60) showed very different behaviour. Realistic Rb+-Rb+ potentials and electrostatic forces produced a very ordered tetragonal structure. So ordered that it is difficult to reconcile with the (alleged) cubic time-average experimental structure. Each C60 seemed to interact with a single Rb+, producing a highly polarised RbC60 `dimer'. These `dimers' then become mutually orientationally order.ed -- there was no hopping and no appreciable disorder.
Further simulations showed that this conclusion was hardly dependent on the strength of the interactions. Changes of about a factor of ten in potentials change the structure little.
The departure of the principal investigator from ANU has slowed the pace of work, but it is now clear that there is an unexplained discrepancy between theory and experiment. Further investigation of starting configurations, boundary conditions, charge and potential models are required before we can definitively state that our understanding of the physically important factors here requires change.
What computational techniques are used and why is a supercomputer required?
We calculate the equilibrium molecular dynamics in an isokinetic canonical ensemble (N,V,T). The equation of motion is solved using a fourth-order predictor-corrector Gear scheme, and energy minimisation by a multidimensional steepest gradient method. Because of the large scope for vectorisation of the code in the most time-consuming routines (force calculations, energy minimisation) the VP2200 is ideal for this type of calculation.
The simulations require the numerical solution of the equations of motion of 128 or 256 rubidium ions in a fixed body centred tetragonal framework of C60 ions. Each C60 has a charge fixed so as to make the whole crystal electroneutral. This charge is confined to the surface of the C60 ball, but otherwise is free to respond to the motion of the Rb+ ions. The solution of this problem is computationally demanding because of the large number of long-range charge-charge interactions, and because of the need to iterativelydue to the release of pressure solve detailed charge distributions and energies at a given rubidium distribution.
Mo , R Durand, W K Fullagar, G Lindsell, P A Reynolds and J W White, Molecular Physics (1995), in press.
Mol C J Carlile, R Durand, W K Fullagar, P A Reynolds, F Trouw and J W White, Molecular Physics (1995), in press.
J W White, P A Wielopolski , Phys. Rev. B 36 (11), 6069 (1987).