College of Science

Center for Simulation and Modeling
(formerly known as Computational Materials Science Center)

SAMP: Structure-Adaptive Materials Prediction

Overview | Highlight | Cloud Computing  | People |


Faculty: Estela Blaisten-Barojas, Iosif I. Vaisman, Daniel B. Carr

Research scholars: D. A. Carr, M. M. Lach-hab, Shujiang Yang, Weixiao Ji, Qi (Jason) Xing

  • Zeolite Structure 1


    ICSD 323.

    Orange Fish
  • Zeolite Structure 2


    ICSD 40128.

    Sea Turtle
  • Zeolite Structure 3


    ICSD 40134.

    Red Coral
  • Zeolite Structure 4


    ICSD 40136.

    Coral Reef
  • Zeolite Structure 5


    ICSD 40137.

    Blue Fish
  • Zeolite Structure 6


    ICSD 40138.

    Yellow Fish
  • Zeolite Structure 7


    ICSD 40139.

  • Zeolite Structure 8


    ICSD 100095.

    Small Fish

During this project we developed machine learning models that allow to classify zeolite crystals according to their framework type. Zeolites are microporous crystalline materials with highly regular framework structures consisting of molecular-sized pores and channels. The characteristic framework type of a zeolite is conventionally defined by combining information on its coordination sequences, vertex symbols, tiling, and transitivity information. Our categorization model, the Zeolite Structure Predictor (ZSP), is based on the Random Forestâ„¢ algorithm and uses a nine-dimensional feature vector including topological descriptors obt ained by computational geometry techniques, together with selected physical and chemical properties of zeolite crystals. Trained on the cr ystallographic structures of known zeolites from the Inorganic Crystal Structure Database, this model predicts the framework types of zeol ite crystals with 98% accuracy.

In addition we developed a cloud-based computing system in the Windows Azure cloud that allows users to use the ZSP model through a Web browser. This automated system permits a user to calculate the feature vector used by ZSP. The workflow is named "SAMPCloud" and integrates executables in Fortran and Python for number crunching with Weka for machine learning and Jmol for Web-based atomistic visualization in an interactive compute system accessed through the Web (see figure on left). SAMPCloud is robust, easy to us e, and open source. Communities of scientists, engineers, and students knowledgeable in Windows-based computing will find this new workflo w attractive and easy to be implemented in scientific scenarios in which the developer needs to combine heterogeneous software components.

SAMPCloud is open source. The source code can be downloaded here or at Sourceforge