dorsal/arxiv
View SchemaEnhancing structure relaxations for first-principles codes: an approximate Hessian approach
| Authors | J. M. Rondinelli, B. Deng, L. D. Marks |
|---|---|
| Categories | |
| ArXiv ID | physics/0608160 |
| URL | https://arxiv.org/abs/physics/0608160 |
| DOI | 10.1016/j.commatsci.2007.01.004 |
| Journal | Comput. Mater. Sci. 40 (2007) 345-353 |
Abstract
We present a method for improving the speed of geometry relaxation by using a harmonic approximation for the interaction potential between nearest neighbor atoms to construct an initial Hessian estimate. The model is quite robust, and yields approximately a 30% or better reduction in the number of calculations compared to an optimized diagonal initialization. Convergence with this initializer approaches the speed of a converged BFGS Hessian, therefore it is close to the best that can be achieved. Hessian preconditioning is discussed, and it is found that a compromise between an average condition number and a narrow distribution in eigenvalues produces the best optimization.
{
"annotation_id": "e5da4187-2b63-46c0-8c48-5b91621f0c26",
"date_created": "2026-03-02T18:01:10.961000Z",
"date_modified": "2026-03-02T18:01:10.961000Z",
"file_hash": "9fd7d4b5847548d647c005d9ca2383ea4ec988aca27fc6ce34de75c3b9fa1ff4",
"private": false,
"record": {
"abstract": "We present a method for improving the speed of geometry relaxation by using a\nharmonic approximation for the interaction potential between nearest neighbor\natoms to construct an initial Hessian estimate. The model is quite robust, and\nyields approximately a 30% or better reduction in the number of calculations\ncompared to an optimized diagonal initialization. Convergence with this\ninitializer approaches the speed of a converged BFGS Hessian, therefore it is\nclose to the best that can be achieved. Hessian preconditioning is discussed,\nand it is found that a compromise between an average condition number and a\nnarrow distribution in eigenvalues produces the best optimization.",
"arxiv_id": "physics/0608160",
"authors": [
"J. M. Rondinelli",
"B. Deng",
"L. D. Marks"
],
"categories": [
"physics.comp-ph"
],
"doi": "10.1016/j.commatsci.2007.01.004",
"journal_ref": "Comput. Mater. Sci. 40 (2007) 345-353",
"title": "Enhancing structure relaxations for first-principles codes: an approximate Hessian approach",
"url": "https://arxiv.org/abs/physics/0608160"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "5507f72c-3c59-4a3d-a100-23a66a6bf1b0",
"id": "arXiv Dataset IDs",
"type": "Model",
"variant": "snapshot-2026-03-01",
"version": "0.1.0"
},
"user_id": 1000002
}