Lennard-Jones 38 transition-state separation

SystemLennard-Jones
Atoms38
Runs51
Cap250000 force calls

The main metric is Force calls. Scoring rules.

Charts

Success rate on Lennard-Jones 38 transition-state separation, in rank order. The rate is the share of runs that finished. 2 ranked rows did not record a success rate and are not drawn.
Success rate on Lennard-Jones 38 transition-state separation, in rank order. The rate is the share of runs that finished. 2 ranked rows did not record a success rate and are not drawn. PDF.
Dolan-Moré profile for Band / NEB on Lennard-Jones 38 transition-state separation. The curve is the fraction of starts whose cost is at most tau times the best finite cost in this family. Tau is 1, 2, 4, 8, and 16. An unsolved start stays in the denominator and does not cross a finite tau. Dolan, E. D. and Moré, J. J., Math. Program. 91:201-213, 2002. Colored lines are the six largest areas. The other curves are gray.
Dolan-Moré profile for Band / NEB on Lennard-Jones 38 transition-state separation. The curve is the fraction of starts whose cost is at most tau times the best finite cost in this family. Tau is 1, 2, 4, 8, and 16. An unsolved start stays in the denominator and does not cross a finite tau. Dolan, E. D. and Moré, J. J., Math. Program. 91:201-213, 2002. Colored lines are the six largest areas. The other curves are gray. PDF.
Dolan-Moré profile for Minimum mode on Lennard-Jones 38 transition-state separation. The curve is the fraction of starts whose cost is at most tau times the best finite cost in this family. Tau is 1, 2, 4, 8, and 16. An unsolved start stays in the denominator and does not cross a finite tau. Dolan, E. D. and Moré, J. J., Math. Program. 91:201-213, 2002. Colored lines are the six largest areas. The other curves are gray.
Dolan-Moré profile for Minimum mode on Lennard-Jones 38 transition-state separation. The curve is the fraction of starts whose cost is at most tau times the best finite cost in this family. Tau is 1, 2, 4, 8, and 16. An unsolved start stays in the denominator and does not cross a finite tau. Dolan, E. D. and Moré, J. J., Math. Program. 91:201-213, 2002. Colored lines are the six largest areas. The other curves are gray. PDF.
Force calls on each recorded start for Lennard-Jones 38 transition-state separation. The box is the quartiles of those counts. A blue point is a success and a vermillion cross is a recorded failure (768). The horizontal position is the recorded count. The vertical offset only keeps marks from covering one another.
Force calls on each recorded start for Lennard-Jones 38 transition-state separation. The box is the quartiles of those counts. A blue point is a success and a vermillion cross is a recorded failure (768). The horizontal position is the recorded count. The vertical offset only keeps marks from covering one another. PDF.
eOn on Lennard-Jones 38 transition-state separation. Each point is one recorded version of one method. Versions run left to right in numeric order. PAR-2 is omitted where that row has none. No line is drawn across a missing value.
eOn on Lennard-Jones 38 transition-state separation. Each point is one recorded version of one method. Versions run left to right in numeric order. PAR-2 is omitted where that row has none. No line is drawn across a missing value. PDF.
Performance profile for Lennard-Jones 38 transition-state separation00.51124816Cost ratioFraction of startsEon svn (r2549) · Dimer L-BFGSeOn (3.5.0) · dimereOn (develop 500350d9) · dimereOn (develop 500350d9) · energy-weighted CI-NEBCOMPASS (1.0.0) · COMPASSrgsaddle (0.2.2) · CI-NEB energy-weightedeOn (develop 500350d9) · doubly nudged CI-NEBEon svn (r2549) · Doubly nudged CI-NEBeOn (3.5.0) · doubly nudged CI-NEBeOn (develop 500350d9) · CI-NEB lj/cuteOn (develop 500350d9) · OCI-NEBeOn (develop 500350d9) · CI-NEB ljclustereOn (3.5.0) · CI-NEB lj/cutEon svn (r2549) · CI-NEBgeodesic (1.0.0) · CI-NEB-LBFGS geodesicrgsaddle (0.2.2) · CI-NEBrgsaddle (0.2.2) · OCI-NEBgeodesic (1.0.0) · CI-NEB geodesicrgsaddle (0.2.2) · IDPP CI-NEBrgsaddle (0.2.2) · S-IDPP CI-NEBASE (3.29.0) · NEB-FIREeOn (3.6.0) · LBFGSgeodesic (1.0.0) · NEB-FIRE geodesicLAMMPS (30 Sep 2026) · CI-NEB-fireLAMMPS (30 Sep 2026) · CI-NEB-quickminrgsaddle (0.2.2) · NEBSella (2.6.0) · Sella+NEBTSASE (1.0) · CI-NEBTSASE (1.0) · NEBeOn (develop 500350d9) · Lanczos improvedeOn (3.5.0) · Lanczos improvedEon svn (r2549) · Lanczos improved L-BFGSEon svn (r2549) · Lanczos L-BFGSeOn (3.5.0) · LanczoseOn (develop 500350d9) · Lanczos
Performance profile. Higher means more starts solved within that cost ratio. Colored lines are the six largest areas. Thin lines are the other 29.
  • eOn (develop 500350d9) · Lanczos improved
  • eOn (3.5.0) · Lanczos improved
  • Eon svn (r2549) · Lanczos improved L-BFGS
  • Eon svn (r2549) · Lanczos L-BFGS
  • eOn (3.5.0) · Lanczos
  • eOn (develop 500350d9) · Lanczos
Cost distribution for Lennard-Jones 38 transition-state separation0816502.784 to 1347.75: 71347.75 to 3612.73: 23612.73 to 9684.17: 09684.17 to 25959.1: 125959.1 to 69585.1: 469585.1 to 186528: 4186528 to 500000: 19502.784500000
Cost distribution of 37 ranked rows. Logarithmic bins, from 502.784 to 500000.

Performance profile SVG Cost and success SVG

Performance profile for Lennard-Jones 38 transition-state separation
Performance profile. Select the plot for the full SVG.
Cost and success for Lennard-Jones 38 transition-state separation
Cost and success. Select the plot for the full SVG.
Best cost for lj38_2ts_separationlj38_2ts_separation10²10³10⁴10⁵10⁶cost (force calls, log scale)Eon svn: 502.784Eon svn502.784eOn: 520.333eOn520.333OPTIM: 2885OPTIM2885Sella: 31447.3Sella31447.3pele: 52961.9pele52961.9COMPASS: 100831COMPASS100831rgsaddle: 266216rgsaddle266216geodesic: 412102geodesic412102ASE: 444244ASE444244TSASE: 451753TSASE451753LAMMPS: 451959LAMMPS451959
A shorter bar is a lower cost on a log scale shared by the codes of this problem.

Leaderboard

A shorter bar is a lower cost on the log scale for that problem. Select a column heading to sort.
RankCodeBestMedianSuccessPAR-2
1 Eon svn 502.784 681.784 100.0% 502.784
2 eOn 520.333 44040.5 100.0% 520.333
3 OPTIM 2885 2885 100.0% 2885
4 Sella 31447.3 31447.3 100.0% 31447.3
5 pele 52961.9 52961.9 94.1% 52961.9
6 COMPASS 100831 100831 80.4% 100831
7 rgsaddle 266216 417164 47.1% 266216
8 geodesic 412102 412455 17.6% 412102
9 ASE 444244 444244 11.8% 444244
10 TSASE 451753 475877 9.8% 451753
11 LAMMPS 451959 475980 9.8% 451959

The tables below list every ranked row, by method family and overall.

A run short of the full set, or a run that is not the stock method, stays off the ranked table.

Band / NEB

Band / NEB. One row is one code, one version, and one method. A lower cost is better. PAR-2 counts a finished run as the calls it used and a failed run as twice the cap. An archived row shows the average it published, and its PAR-2 applies that rule to the failures the entry records. Hessians is the mean analytic Hessian count. FD calls is that count times one force call per coordinate.
RankRowCostUnitHessiansFD callsRuns
1eOn (develop 500350d9) · energy-weighted CI-NEB15260.3force calls0051
2Sella (2.6.0) · Sella+NEB31447.3force calls0051
3eOn (develop 500350d9) · CI-NEB lj/cut38080.5force calls0051
4eOn (develop 500350d9) · CI-NEB ljcluster50000.5force calls0051
5eOn (develop 500350d9) · doubly nudged CI-NEB72306.9force calls0051
6eOn (develop 500350d9) · OCI-NEB95094.3force calls0051
7rgsaddle (0.2.2) · CI-NEB energy-weighted266216force calls0051
8eOn (3.5.0) · doubly nudged CI-NEB363362force calls0051
9Eon svn (r2549) · Doubly nudged CI-NEB373102force calls0051
10Eon svn (r2549) · CI-NEB402321force calls0051
11eOn (3.5.0) · CI-NEB lj/cut402340force calls0051
12rgsaddle (0.2.2) · CI-NEB402545force calls0051
13rgsaddle (0.2.2) · OCI-NEB402564force calls0051
14geodesic (1.0.0) · CI-NEB-LBFGS geodesic412102force calls0051
15geodesic (1.0.0) · CI-NEB geodesic412455force calls0051
16rgsaddle (0.2.2) · IDPP CI-NEB431764force calls0051
17rgsaddle (0.2.2) · S-IDPP CI-NEB441455force calls0051
18ASE (3.29.0) · NEB-FIRE444244force calls0051
19TSASE (1.0) · CI-NEB451753force calls0051
20LAMMPS (30 Sep 2026) · CI-NEB-quickmin451959force calls0051
21LAMMPS (30 Sep 2026) · CI-NEB-fire500000force calls0051
22TSASE (1.0) · NEB500000force calls0051
23geodesic (1.0.0) · NEB-FIRE geodesic500000force calls0051
24rgsaddle (0.2.2) · NEB500000force calls0051

Minimum mode

Minimum mode. One row is one code, one version, and one method. A lower cost is better. PAR-2 counts a finished run as the calls it used and a failed run as twice the cap. An archived row shows the average it published, and its PAR-2 applies that rule to the failures the entry records. Hessians is the mean analytic Hessian count. FD calls is that count times one force call per coordinate.
RankRowCostUnitHessiansFD callsRuns
1Eon svn (r2549) · Lanczos improved L-BFGS502.784force calls0051
2eOn (3.5.0) · Lanczos improved520.333force calls0051
3Eon svn (r2549) · Lanczos L-BFGS524.667force calls0051
4eOn (develop 500350d9) · Lanczos improved536.667force calls0051
5eOn (3.5.0) · Lanczos564.647force calls0051
6eOn (develop 500350d9) · Lanczos573.745force calls0051
7Eon svn (r2549) · Dimer L-BFGS681.784force calls0051
8eOn (develop 500350d9) · dimer3236.35force calls0051
9eOn (3.5.0) · dimer108356force calls0051

Connect / double-ended

Connect / double-ended. One row is one code, one version, and one method. A lower cost is better. PAR-2 counts a finished run as the calls it used and a failed run as twice the cap. An archived row shows the average it published, and its PAR-2 applies that rule to the failures the entry records. Hessians is the mean analytic Hessian count. FD calls is that count times one force call per coordinate.
RankRowCostUnitHessiansFD callsRuns
1OPTIM · DNEB + hybrid eigenvector following + Djikstra shortest path2885force calls0051
2pele · DNEB + hybrid eigenvector following52961.9force calls0051
3COMPASS (1.0.0) · COMPASS100831force calls0051
4eOn (3.6.0) · LBFGS343644force calls0051

Overall

One row is one code, one version, and one method. A lower cost is better. PAR-2 counts a finished run as the calls it used and a failed run as twice the cap. An archived row shows the average it published, and its PAR-2 applies that rule to the failures the entry records. Hessians is the mean analytic Hessian count. FD calls is that count times one force call per coordinate. The per-image mean divides the total by the number of moving images.
RowFamilyFamily rankUnitPAR-2PublishedMedian callsMean callsPer-image meanHessiansFD callsFailuresRuns
Eon svn (r2549) · Lanczos improved L-BFGSMinimum mode1force calls502.784418502.784100.55700crash 0, divergence 0, cap 0, wrong_stationary 051
eOn (3.5.0) · Lanczos improvedMinimum mode2force calls520.333422520.333104.06700crash 0, divergence 0, cap 0, wrong_stationary 051
Eon svn (r2549) · Lanczos L-BFGSMinimum mode3force calls524.667467524.667104.93300crash 0, divergence 0, cap 0, wrong_stationary 051
eOn (develop 500350d9) · Lanczos improvedMinimum mode4force calls536.667419536.667107.33300crash 0, divergence 0, cap 0, wrong_stationary 051
eOn (3.5.0) · LanczosMinimum mode5force calls564.647482564.647112.92900crash 0, divergence 0, cap 0, wrong_stationary 051
eOn (develop 500350d9) · LanczosMinimum mode6force calls573.745482573.745114.74900crash 0, divergence 0, cap 0, wrong_stationary 051
Eon svn (r2549) · Dimer L-BFGSMinimum mode7force calls681.784451681.784136.35700crash 0, divergence 0, cap 0, wrong_stationary 051
OPTIM · DNEB + hybrid eigenvector following + Djikstra shortest pathConnect / double-ended1force calls2885288525652885000 failed51
eOn (develop 500350d9) · dimerMinimum mode8force calls3236.355683236.35647.27100crash 0, divergence 0, cap 0, wrong_stationary 051
eOn (develop 500350d9) · energy-weighted CI-NEBBand / NEB1force calls15260.3243715260.33052.0700crash 0, divergence 0, cap 0, wrong_stationary 051
Sella (2.6.0) · Sella+NEBBand / NEB2force calls31447.33217831447.36289.4700crash 0, divergence 0, cap 0, wrong_stationary 051
eOn (develop 500350d9) · CI-NEB lj/cutBand / NEB3force calls38080.51425738080.57616.1100crash 0, divergence 0, cap 0, wrong_stationary 051
eOn (develop 500350d9) · CI-NEB ljclusterBand / NEB4force calls50000.51488740196.58039.3100crash 0, divergence 0, cap 2, wrong_stationary 051
pele · DNEB + hybrid eigenvector followingConnect / double-ended2force calls52961.9250222063825022003 failed, class not recorded51
eOn (develop 500350d9) · doubly nudged CI-NEBBand / NEB5force calls72306.92075752699.110539.800crash 0, divergence 0, cap 4, wrong_stationary 051
eOn (develop 500350d9) · OCI-NEBBand / NEB6force calls95094.33839765682.513136.500crash 0, divergence 0, cap 6, wrong_stationary 051
COMPASS (1.0.0) · COMPASSConnect / double-ended3force calls10083140314586.12917.22400crash 0, divergence 0, cap 9, wrong_stationary 151
eOn (3.5.0) · dimerMinimum mode9force calls1083565681200.33240.06700crash 0, divergence 11, cap 0, wrong_stationary 051
rgsaddle (0.2.2) · CI-NEB energy-weightedBand / NEB7force calls266216100006803.941360.7900crash 0, divergence 0, cap 27, wrong_stationary 051
eOn (3.6.0) · LBFGSConnect / double-ended4force calls3436444441240738.48147.6700crash 0, divergence 34, cap 0, wrong_stationary 051
eOn (3.5.0) · doubly nudged CI-NEBBand / NEB8force calls36336250024245.53849.10600crash 0, divergence 37, cap 0, wrong_stationary 051
Eon svn (r2549) · Doubly nudged CI-NEBBand / NEB9force calls37310250074283.76856.75300crash 0, divergence 38, cap 0, wrong_stationary 051
Eon svn (r2549) · CI-NEBBand / NEB10force calls40232150074385.63877.12500crash 0, divergence 41, cap 0, wrong_stationary 051
eOn (3.5.0) · CI-NEB lj/cutBand / NEB11force calls40234050024400.43880.08600crash 0, divergence 41, cap 0, wrong_stationary 051
rgsaddle (0.2.2) · CI-NEBBand / NEB12force calls402545100008309.861661.9700crash 0, divergence 0, cap 37, wrong_stationary 451
rgsaddle (0.2.2) · OCI-NEBBand / NEB13force calls402564100008470.331694.0700crash 0, divergence 0, cap 38, wrong_stationary 351
geodesic (1.0.0) · CI-NEB-LBFGS geodesicBand / NEB14force calls412102100008401.491680.300crash 0, divergence 0, cap 38, wrong_stationary 451
geodesic (1.0.0) · CI-NEB geodesicBand / NEB15force calls412455100008769.921753.9800crash 0, divergence 0, cap 38, wrong_stationary 451
rgsaddle (0.2.2) · IDPP CI-NEBBand / NEB16force calls431764100008882.941776.5900crash 0, divergence 0, cap 41, wrong_stationary 351
rgsaddle (0.2.2) · S-IDPP CI-NEBBand / NEB17force calls441455100008880.21776.0400crash 0, divergence 0, cap 41, wrong_stationary 451
ASE (3.29.0) · NEB-FIREBand / NEB18force calls4442444000138362.87672.5600crash 0, divergence 0, cap 45, wrong_stationary 051
TSASE (1.0) · CI-NEBBand / NEB19force calls451753100009690.351938.0700crash 0, divergence 1, cap 43, wrong_stationary 251
LAMMPS (30 Sep 2026) · CI-NEB-quickminBand / NEB20force calls451959998299821996.400crash 0, divergence 45, cap 0, wrong_stationary 151
LAMMPS (30 Sep 2026) · CI-NEB-fireBand / NEB21force calls50000081208114.781622.9600crash 0, divergence 36, cap 0, wrong_stationary 1551
TSASE (1.0) · NEBBand / NEB22force calls500000100009720.221944.0400crash 0, divergence 3, cap 44, wrong_stationary 451
geodesic (1.0.0) · NEB-FIRE geodesicBand / NEB23force calls5000001000010000200000crash 0, divergence 0, cap 51, wrong_stationary 051
rgsaddle (0.2.2) · NEBBand / NEB24force calls500000100009773.861954.7700crash 0, divergence 3, cap 45, wrong_stationary 351