GETTING THE HOT STRUCTURES K. F. Kelton and A. K. Gangopadhyay Department of Physics, Washington University, St. Louis, Missouri 63130 ABSTRACT The recent advances in levitation methods for materials processing now enable structural, thermo-physical property, and phase transition studies to be made on high temperature solids and liquids without container contamination. These studies have led to new insights into the liquid state and have revealed how local order in the liquid can dictate phase formation. In this article, levitation techniques are briefly discussed, focusing most on electrostatic levitation. Recent synchrotron studies of electrostatically-levitated undercooled Ti-Zr-Ni liquids are presented, which demonstrate that developing icosahedral short-range order in the liquid causes the nucleation of a metastable icosahedral quasicrystal instead of the stable tetrahedral Laves phase. In addition to providing the first experimental proof of a half-century-old hypothesis linking the order of the liquid with the nucleation barrier, these data raise new questions about the general applicability of the thermodynamic model assumed in the classical theory of nucleation. The combination of electrostatic levitation and synchrotron high-energy diffraction also allows rapid and accurate determinations of phase diagrams for high temperature materials. This is demonstrated using Ti-Fe-Si-O as a case study. This new technique, then, is of practical as well as basic importance. INTRODUCTION The local structural order in liquids and the similarity to the order in crystal phases that form from the liquid during freezing are fundamental questions. Since Fahrenheit’s studies of the crystallization of water, for example, it has been recognized that under the right conditions liquids can be maintained in a metastable state below the equilibrium melting temperatures for extended periods of time [1]. This ability to undercool (or supercool) the liquid demonstrates the existence of a barrier to the phase transition. The initial formation of small regions of the ordered phase, i.e. nucleation, involves a stochastic thermally-activated transition over this barrier that is typically analyzed within the classical theory of nucleation [2]. In this theory, the nucleation barrier arises from the high interfacial energy between the liquid and small clusters of the crystal phase, presumably due to a significant difference in the local atomic structures of the two phases. “Homogeneous nucleation” is the more fundamental of the two types of nucleation, occurring randomly in space and time. “Heterogeneous nucleation,” is more common, however, catalyzed at specific sites in the initial phase. Until the middle of the last century, a failure to significantly undercool liquid metals was taken as evidence that the local atomic structures of the liquid and crystal phases were similar. In 1952, however, Turnbull showed that this was due to the tendency for heterogeneous nucleation in metallic liquids [3]; if the catalytic sites in the liquid were removed, liquid metals could also be undercooled, sometimes to as low as 2/3 of their melting temperature before crystallization (Figure 1). To explain this Frank postulated an icosahedral packing in the liquid, consisting of 20 Copyright ©JCPDS - International Centre for Diffraction Data 2005, Advances in X-ray Analysis, Volume 48. 1