Table 2. Number of Macroiterations to Converge the Molecular Geometry Optimization of the Molecular Structures of the Baker-TS Test Suite Using Conventional RS-RFO and RVO Supported by GEKa.
| this
work |
ref (20)b |
RVO vs GPR | ||||
|---|---|---|---|---|---|---|
| reaction | RS-RFO | RVO | rmsd | L-BFGS | GPR | rmsd |
| 1: HCN ⇌ HNC | 12 | 13 | 0.000 | 22 | 18 | 0.000 |
| 2: HCCH ⇌ CCH2 | 13 | 12 | 0.000 | 24 | 20 | 0.000 |
| 3: H2CO ⇌ H2 + CO* | 32 | 36 | 0.985 | 59 | 103 | 0.557 |
| 4: CH3O ⇌ CH2OH | 7 | 7 | 0.000 | 18 | 15 | 0.000 |
| 5: ring opening cyclopropyl* | 24 | 11 | 0.842 | 42 | 37 | 0.842 |
| 6: ring opening bicyclo[1.1.0]butane (TS 1) | 19 | 13 | 0.000 | 30 | 28 | 0.001 |
| 7: ring opening bicyclo[1.1.0]butane (TS 2) | 23 | 13 | 0.000 | 54 | 48 | 0.001 |
| 8: 1,2-migration β-(formyloxy)ethyl | 36 | 28 | 0.001 | 87 | 93 | 0.005 |
| 9: butadiene + ethylene ⇌ cyclohexenec* | 116 | 75 | 0.009 | 89 | 122 | 1.005 |
| 10: s-tetrazine ⇌ 2HCN + N2 | 9 | 7 | 0.000 | 15 | 21 | 0.000 |
| 11: trans-butadiene ⇌ cis-butadiene | 9 | 6 | 0.000 | 32 | 30 | 0.001 |
| 12: CH3CH3 ⇌ CH2CH2 + H2 | 11 | 9 | 0.000 | 24 | 16 | 0.000 |
| 13: CH3CH2F ⇌ CH2CH2 + HF | 11 | 7 | 0.000 | 20 | 15 | 0.001 |
| 14: vinyl alcohol ⇌ acetaldehyde | 15 | 13 | 0.000 | 19 | 26 | 0.001 |
| 15: HCOCl ⇌ HCl + CO | 10 | 8 | 0.000 | 12 | 12 | 0.000 |
| 16: H2O + PO3– ⇌ H2PO4– | 32 | 28 | 0.002 | 64 | 74 | 0.001 |
| 17: CH2CHCH2–O–CHCH2 ⇌ CH2CHCH2CH2CHO | 26 | 21 | 0.002 | 98 | 73 | 0.008 |
| 18: SiH2 + CH3CH3 ⇌ SiH3CH2CH3 | 17 | 17 | 0.001 | 44 | 38 | 0.008 |
| 19: HNCCS ⇌ HNC + CS | 17 | 13 | 0.000 | 25 | 19 | 0.001 |
| 20: HCONH3+ ⇌ NH4+ + CO | 13 | 9 | 0.000 | 21 | 19 | 0.000 |
| 21: rotational TS in acrolein | 21 | 12 | 0.000 | 49 | 47 | 0.001 |
| 22: HCONHOH ⇌ HCOHNHO | 11 | 9 | 0.000 | 23 | 20 | 0.000 |
| 23: HNC + H2 ⇌ H2CNH | 15 | 10 | 0.000 | 21 | 18 | 0.000 |
| 24: H2CNH ⇌ HCNH2 | 16 | 13 | 0.000 | 25 | 18 | 0.000 |
| 25: HCNH2 ⇌ HCN + H2* | 48 | 42 | 0.482 | 254 | 30 | 1.197 |
a
The root mean square displacement (rmsd, in Å) between the final structures is shown in the third numerical column. As a reference the GPR and L-BFGS results of Denzel and Kästner, using DFT, are listed. Finally, the rmsd between the RVO and GPR optimized structures is presented. Highlighted in bold are cases where the difference between the first two columns is larger than 1 iteration. An asterisk marks cases where RVO clearly converges to a different local minimum from RS-RFO and/or GPR.
b
See Table S2 in ref (20).
c
RS-RFO and RVO optimizations performed with symmetry constrained to Cs.