molecularGSM

Official Resources

• Homepage: https://zimmermangroup.github.io/molecularGSM/
• Source Repository: https://github.com/zimmermangroup/molecularGSM
• Documentation: https://zimmermangroup.github.io/molecularGSM/
• License: As specified by Zimmerman group (open for academic use)

Overview

molecularGSM (Growing String Method) is a reaction path and transition state finding tool developed by the Zimmerman group. GSM is utilized in two main fashions: double-ended (DE), which requires a reactant and product pair, and single-ended (SE), which only requires a reactant and a driving coordinate specified in internal coordinates (angles, bonds, and torsions).

Unlike the nudged elastic band (NEB) and traditional string methods that require an initial guess for the reaction pathway, the growing string method begins as two string fragments associated with reactants and products that are grown separately until they converge. This eliminates the need for an initial pathway guess and significantly reduces the number of electronic structure force calculations needed. The single-ended GSM allows reaction path searches starting from a single structure, enabling automated systematic exploration of unknown reactivity.

Scientific domain: Computational chemistry, reaction path finding, transition state search
Target user community: Computational chemists studying reaction mechanisms

Theoretical Methods

• Growing string method (GSM)
• Double-ended (DE) and single-ended (SE) variants
• Delocalized internal coordinates
• Constrained optimization in hyperplane orthogonal to path
• Climbing image and eigenvector following for TS optimization
• Tangent vector constraint optimization
• Hessian estimation for convergence acceleration

Capabilities (CRITICAL)

• Transition state finding without initial pathway guess
• Double-ended GSM (reactant + product known)
• Single-ended GSM (reactant + driving coordinate only)
• Automated systematic reaction search (found 165 paths from 333 searches)
• Exact TS optimization via climbing image and eigenvector following
• Internal coordinate (bonds, angles, torsions) driving
• Growth phase (outside-in) and optimization phase
• Applicable to transition metal reactivity
• Kabsch alignment of endpoints

Inputs & Outputs

Input formats:

• Reactant and product structures (for DE-GSM)
• Reactant structure and driving coordinate (for SE-GSM)
• Internal coordinate specifications
• Electronic structure method selection

Output data types:

• Minimum energy pathway (MEP)
• Transition state structures (exact TS)
• Reaction path nodes (xyz files)
• Energy profiles along reaction path
• IRC (intrinsic reaction coordinate) paths

Interfaces & Ecosystem

• Programming language: C++/Python
• Electronic structure backends: Various (Gaussian, ORCA, etc.)
• GUI: GitHub Pages documentation with tutorials
• Related tools: NEB, dimer method, P-RFO optimization
• Platforms: Linux, macOS

Limitations & Known Constraints

• Requires electronic structure calculations at each node
• Single-ended GSM quality depends on driving coordinate choice
• Large molecules may require many nodes for accurate path
• Convergence can be sensitive to initial parameters
• Not all reaction types may be found in systematic search

Performance Characteristics

• Significantly fewer force evaluations than NEB or string method
• Growth phase only evaluates frontier nodes (not all images)
• Adaptive growth strategy more efficient for large systems
• Hessian estimates accelerate convergence
• Single-ended GSM can be more efficient than double-ended

Comparison with Other Codes

• vs NEB: GSM doesn't require initial pathway guess; fewer force evaluations
• vs String method: GSM grows adaptively from both ends; traditional string uses fixed images
• vs Dimer method: Dimer finds TS from single point; GSM finds full pathway
• vs NEB-TS: NEB-TS uses interpolation; GSM grows from endpoints

Best Practices

• Use DE-GSM when both reactant and product are known
• Use SE-GSM for systematic exploration of unknown reactivity
• Choose driving coordinates that correspond to the assumed reaction coordinate
• Use internal coordinates for better convergence
• Optimize highest-energy node to exact TS with eigenvector following
• Validate TS with IRC calculation

Verification & Sources

Primary sources:

1. GSM website: https://zimmermangroup.github.io/molecularGSM/
2. P.M. Zimmerman, J. Comput. Chem. 36, 601-611 (2015) - single-ended GSM
3. B. Peters et al., J. Chem. Phys. 120, 6 (2004) - original growing string method

Confidence: VERIFIED - GitHub repository and peer-reviewed publications confirmed