Efficient Neighbor Search Considering Motion

Hi. I'm using MATLAB to analyze the results of a molecular simulation of silica done on LAMMPS. I'm currently using rangesearch() to determine the bonds between atoms (I call atoms within a certain distance bonded). However, this can be very computationally intensive and can take a long time to run. I was wondering if I could use the fact that the atoms are moving to determine the nearest neighbors based on the nearest neighbors in the previous frame. Would there be a way of doing this, and would it make it more computationally efficient? Thanks.

Answer 1

[Pratyush]

Hi Ephraim,

To improve computational efficiency in analyzing molecular simulations for determining bonded atoms, you can utilize the movement of atoms between frames to limit the search for nearest neighbors. Here's a summarized approach:

1. Initial Frame Analysis: Perform a comprehensive search using "rangesearch()" to identify bonded atoms based on a distance threshold.
2. Subsequent Frames: For each new frame, instead of searching the entire space, only search within a smaller subset around each atom's previous position. This subset is defined by the distance threshold plus an additional buffer based on the maximum expected atom movement.
3. Update Bonds: Update the list of bonded atoms based on distances in the current frame, focusing on the reduced set of potential neighbors.
4. Periodic Full Searches: Occasionally, conduct a full search to correct any errors that might accumulate from using the reduced search space. The frequency of these full searches depends on the dynamics of the simulation.

Advantages:

• Significantly reduces the number of distance calculations needed, enhancing computational efficiency.
• Maintains a balance between accuracy and performance by dynamically adjusting the search radius and incorporating periodic full searches for corrections.

Considerations:

• The method involves a trade-off between computational speed and the potential for missing dynamically forming bonds.
• The optimal frequency for correctional full searches varies with the simulation's characteristics.

This strategy leverages the temporal continuity in atom positions across frames to optimize the identification of bonded neighbors, resulting in a more efficient analysis of molecular dynamics simulations.