| name | apt-cluster |
| description | Run DBSCAN cluster analysis on atom probe data. Use when the user wants to find clusters, precipitates, or perform cluster analysis. |
| argument-hint | [epsilon minPts] |
Run DBSCAN cluster analysis on the atom probe data. Use the MATLAB MCP server.
Prerequisites
The MATLAB workspace must contain:
pos — loaded atom probe data with ions allocated (has ion column)
If ions haven't been allocated, tell the user to run /apt-composition first.
Parameters
Parse $ARGUMENTS for epsilon and minPts values. If not provided, use defaults:
epsilon = 0.5 nm (neighborhood radius)minPts = 10 (minimum atoms in cluster)
If the user hasn't specified parameters, ask what they expect:
- For fine precipitates: epsilon ~0.3-0.5 nm, minPts ~5-15
- For larger precipitates: epsilon ~0.5-1.0 nm, minPts ~10-30
- These depend on the detection efficiency and atomic density
Execution
1. Run DBSCAN
[clusterIdx, clusterTable] = clusterDBSCAN(pos, epsilon, minPts);
Where epsilon and minPts come from the arguments or defaults.
2. Analyze results
clusterSizeAnalyse;
3. Report
Tell the user:
- Number of clusters found
- Size distribution (if available)
- Suggest adjusting epsilon/minPts if results don't look right
- Remind them that cluster analysis results depend heavily on parameter choice
- Suggest using
voronoiVolumeAnalysis as an alternative approach
Notes
- DBSCAN is GPU-accelerated when available
- For large datasets (>1M atoms), consider using an ROI first to reduce computation time
- The user may want to filter by specific ion species before clustering
