Effect of Ti Micro-additions on the Structure of Al88-xY7Fe5Tix Glasses (x=0, 0.5) and finding the natural units in atomic packings

The present talk is divided in two parts. In the first part, we will talk about the micro-additions in Al88-xY7Fe5Tix based metallic glasses that shows interesting features both structurally and in their nucleation phenomenon when devitrified. In particular, 0.5% Ti micro-addition suppresses the nucleation phenomenon considerably making it a better glass. Earlier studies found that the Ti micro-addition does not change the local atomic structure around the Fe atoms, and changes it only slightly around the Y atoms. However, the local structure appears more ordered around the Al atoms. This intriguing fact makes it difficult to explain how a very small amount of Ti can play such a major role. In this talk we present the evidence of phase separation in Al88Y7Fe5 based on Local Electrode Atom Probe (LEAP) alternatively known as 3-D atom probe (for obtaining the coordinates of individual atoms) data. The Al-rich regions are possibly playing the role of precursors for nucleating α-aluminum. Constructing the pair distribution function for the crystallites in the TEM images of annealed glass revealed a characteristic distance which is probably related to the diffusion zone implying that diffusion plays a major role. Underneath this phase separation, there are sub-nanometer size pure-Al zones that are stochastically distributed spatially causing very high nucleation rate. This phase separation is considerably suppressed in AlYFeTi and the phase separation distance is considerably changed leading to much lesser nucleation rate.

Part II: 3-D atom probe (and also Reverse Monte Carlo Simulations that we study routinely to obtain atomic coordinates from diffraction study at synchrotron source) provides us wealth of data about this type of semi-randomized systems. Partitioning the free space in a system of (semi)randomized packing is of importance because of many reasons. It also helps to find the natural structural units of the system. The available methods, like the Voronoi tessellation, Delaunay triangulation etc. have been useful methods for partitioning the space, but have serious limitations to be used for identifying the structural units. To address this we will be proposing two methods: one based on geometrical information, the other will be based on energy. The first method is devised based on the mathematical formulation of natural neighborhood. The definition of the void cells (similar to the unit cells) will be derived in this talk. In the present form, the method produces non-unique results, but it will be shown that results produced by this method are statistically reproducible. Another more generic method will be discussed which is based on the construction of Potts model Hamiltonian and finding the minimum of the this Hamiltonian  and identifying the associated partition/clusters (or unit cells).Unlike the previous method, which uses geometric information extensively, this method does not require any geometric information explicitly and thus has wide applicability ranging from providing on-line suggestion (amazon.com) to identification of social network of a terrorist. This presentation will primarily focus on the application of this community detection scheme in 2-D randomized systems and Al88Y7Fe5. However, application of these methods in Al88-xY7Fe5Tix systems is still pending.