Related Experiment Videos
Direct calculation of atomic coordinates from diffraction intensities: space group P1.
1Laboratory for the Structure of Matter, Naval Research Laboratory, Washington, DC 20375-5000.
Summary
This study introduces a novel method to improve convergence in least-squares minimization for nonlinear systems. The technique enhances accuracy in crystal structure determination from x-ray scattering data.
Area of Science:
- Crystallography
- Computational Chemistry
- Applied Mathematics
Background:
- Least-squares minimization is crucial for solving nonlinear systems.
- Determining crystal structures from x-ray scattering data presents convergence challenges.
- Existing methods often struggle with false minima and limited convergence range.
Purpose of the Study:
- To present a new method for extending the convergence range of least-squares minimization.
- To apply and validate this method for direct crystal structure determination.
- To address and overcome limitations of current minimization techniques.
Main Methods:
- Altering the minimization function to reduce false minima.
- Dynamically modifying false minima characteristics by varying data sets and parameters.
- Utilizing a modified least-squares approach for nonlinear system optimization.
Main Results:
- Successful convergence to a global minimum for a 30-atom crystal structure from random initial coordinates.
- Demonstrated efficiency with convergence achieved in minutes on a Cray XMP/216.
- Observed decreased success rates for more complex 40-atom structures, indicating scalability challenges.
Conclusions:
- The presented method effectively enhances convergence in least-squares minimization for specific problems.
- The approach shows promise for direct crystal structure determination from x-ray data.
- Potential applicability to other mathematical systems warrants further investigation.