OptBench

Double Ended Saddle Search

For this benchmark we require simply a refined saddle point that is in between the starting and ending minima. The termination condition is that the norm of the gradient at the saddle point be less than 0.001. An example method to find that saddle point is (D)NEB + climbing image or (D)NEB + hybrid eigenvector following. The saddle point need not be a transition state directly separating the start and end minima.

LJ 38

A tar file containing the structures between which a saddle should be computed is here: coords.tgz

Algorithm <FCs> min(FCs) median(FCs) max(FCs) nfailed
optim

Comments: This method uses DNEB to get an image close to the saddle point, then uses hybrid eigenvector following method to optimize to that saddle point. This uses Rayleigh-Ritz minimization to find the lowest eigenvector and moves using the hybrid eigenvector following method. The Rayleigh-Ritz minimization uses a first order (forward finite differences) expansion.
Date: 06 Aug 2014 Contributor: Jacob Stevenson & Cheng Shang Code: wales-27058.tar.bz2 Input files: optim.tgz

131 66 120 260 0
eon-lanczos-improved1

Date: 29 Aug 2013 Contributor: Sam Chill Input files: eon-lanczos-improved.tgz

287 85 201 1623 0
pele

Date: 06 Aug 2014 Contributor: Jacob Stevenson Code: pele-a68ec5.tgz Input files: pele.tgz

293 107 222 782 0
eon-dimer1

Date: 22 Aug 2013 Contributor: Sam Chill Input files: eon-dimer.tgz

337 89 204 2079 0
eon-lanczos1

Date: 29 Aug 2013 Contributor: Sam Chill Input files: eon-lanczos.tgz

364 110 241 2794 0
optim-cl-string

Comments: This method uses climbing-string method to get an image close to the saddle point, then uses inexact newton method to accelerate the convergence. The final state is used as the inpute of the initial guessed saddle point for climbing-string.
Date: 06 Aug 2014 Contributor: Cheng Shang Code: wales-27058.tar.bz2 Input files: optim-cl-string.tgz

786 274 625 2792 3
eon-ci-neb1

Date: 22 Aug 2013 Contributor: Sam Chill Input files: eon-ci-neb.tgz

827 407 788 2347 7
eon-ci-swdneb1

Date: 22 Aug 2013 Contributor: Sam Chill Input files: eon-ci-swdneb.tgz

861 407 794 2187 5

Pt(111) Heptamer Island

A tar file containing the structures between which a saddle should be computed is here: pt-island-con.tgz

Algorithm <FCs> min(FCs) median(FCs) max(FCs) nfailed
optim

Comments: This method uses DNEB to get an image close to the saddle point, then uses hybrid eigenvector following method to optimize to that saddle point. This uses Rayleigh-Ritz minimization to find the lowest eigenvector and moves using the hybrid eigenvector following method. The Rayleigh-Ritz minimization uses a first order (forward finite differences) expansion.
Date: 06 Aug 2014 Contributor: Jacob Stevenson & Cheng Shang Code: wales-27058.tar.bz2 Input files: optim.tgz

103 59 98 194 0
eon-ci-neb-single1

Date: 23 Aug 2013 Contributor: Sam Chill Input files: eon-ci-neb-single.tgz

110 29 62 594 0
optim-6

Comments: This method uses DNEB to get an image close to the saddle point, then uses hybrid eigenvector following method to optimize to that saddle point. This uses Rayleigh-Ritz minimization to find the lowest eigenvector and moves using the hybrid eigenvector following method. The Rayleigh-Ritz minimization uses a first order (forward finite differences) expansion. Six images are used in DNEB
Date: 06 Aug 2014 Contributor: Jacob Stevenson & Cheng Shang Code: wales-27058.tar.bz2 Input files: optim-6.tgz

117 60 115 201 0
eon-dimer1

Date: 23 Aug 2013 Contributor: Sam Chill Input files: eon-dimer.tgz

182 74 182 322 0
eon-lanczos-improved1

Date: 29 Aug 2013 Contributor: Sam Chill Input files: eon-lanczos-improved.tgz

222 72 157 2039 8
eon-ci-neb1

Date: 23 Aug 2013 Contributor: Sam Chill Input files: eon-ci-neb.tgz

248 86 211 761 0
pele

Date: 06 Aug 2014 Contributor: Jacob Stevenson Code: pele-a68ec5.tgz Input files: pele.tgz

255 96 244 782 2
eon-ci-swdneb1

Date: 23 Aug 2013 Contributor: Sam Chill Input files: eon-ci-swdneb.tgz

256 92 219 956 0
optim-cl-string

Comments: This method uses climbing-string method to get an image close to the saddle point, then uses inexact newton method to accelerate the convergence. The final state is used as the inpute of the initial guessed saddle point for climbing-string.
Date: 06 Aug 2014 Contributor: Cheng Shang Code: wales-27058.tar.bz2 Input files: optim-cl-string.tgz

319 34 286 1076 0
eon-lanczos1

Date: 29 Aug 2013 Contributor: Sam Chill Input files: eon-lanczos.tgz

325 95 208 2660 5
eon-ci-neb-51

Date: 23 Aug 2013 Contributor: Sam Chill Input files: eon-ci-neb-5.tgz

391 162 395 797 0

LJ 38 with one intermediate minimum

For this benchmark we require a connected path of minima and transition states between the starting and ending points. For this benchmark the starting and ending point are not directly connected. There exists a path with one intermediate minimum, however we do not require this path to be found. A successful benchmark is one which finds any path between the starting and ending minima. Because the final path is not specified, we require the number of transition states in the found path to be reported. (note: this benchmark should be separated from those above. perhaps in "double ended pathway search")

A tar file containing the structures between which a saddle should be computed is here: coords.tgz

Algorithm <FCs> min(FCs) median(FCs) max(FCs) <TSs> nfailed
optim

Date: 06 Aug 2014 Contributor: Jacob Stevenson & Cheng Shang Code: wales-27058.tar.bz2 Input files: optim.tgz

2885 665 2565 7138 5 0
pele

Date: 06 Aug 2014 Contributor: Jacob Stevenson Code: pele-a68ec5.tgz Input files: pele.tgz

25022 1069 20638 114490 7 3