The large ADD following (l-ADDF) attempts to find the largest n ADDs and related paths. To distinguish the full search from l-ADDF, the full search may be called full-ADDF (f-ADDF). The larger ADD is often associated with the lower barrier and the lower product. The larger anharmonicity modifies the curvature along corresponding reaction coordinate from positive to negative the more rapidly, resulting in the earlier TS. According to the Hammond postulate, the earlier TS relates to the lower energy product. The Bell-Evans-Polanyi principle suggests that the more exothermic reaction has the lower TS. Combining these three postulates, a new principle can be proposed; the larger ADD relates to the more exothermic reaction through the lower TS.
In the early stage of the IOE procedure, as explained in Introduction of ADDF, small ADDs are not found as explicit local minima on the scaled hypersphere surface. Thus, large ADDs can be found efficiently by terminating the IOE procedure at an early stage. Although various ways may be considered to search for largest n ADDs, the GRRM17 program adopts the following algorithm: first, perform IOE starting from positive and negative directions of the softest mode until 3n ADDs are found, second, omit smallest n ADDs among the 3n, and third, follow the remaining 2n by PC-IOE until n ADDs among the 2n reach MINs or DCs. In other words, to find n lowest paths, find 3n ADDs by IOE, follow 2n ADDs by PC-IOE, and complete ADDF for n ADDs, where such a calculation is indicated as l-ADDFn.
An example of input using these parameters for (H2O)8 clusters is:
# ADDF/RHF/6-31G
0 1
O -0.000000000000 0.010468354749 0.754819362610 1
H 0.000000000000 0.771670165130 1.319188188547 1
H -0.000000000000 -0.747077150089 1.324086214974 1
O -0.000000000000 0.010468354749 0.754819362610 2
H 0.000000000000 0.771670165130 1.319188188547 2
H -0.000000000000 -0.747077150089 1.324086214974 2
O -0.000000000000 0.010468354749 0.754819362610 3
H 0.000000000000 0.771670165130 1.319188188547 3
H -0.000000000000 -0.747077150089 1.324086214974 3
O -0.000000000000 0.010468354749 0.754819362610 4
H 0.000000000000 0.771670165130 1.319188188547 4
H -0.000000000000 -0.747077150089 1.324086214974 4
O -0.000000000000 0.010468354749 0.754819362610 5
H 0.000000000000 0.771670165130 1.319188188547 5
H -0.000000000000 -0.747077150089 1.324086214974 5
O -0.000000000000 0.010468354749 0.754819362610 6
H 0.000000000000 0.771670165130 1.319188188547 6
H -0.000000000000 -0.747077150089 1.324086214974 6
O -0.000000000000 0.010468354749 0.754819362610 7
H 0.000000000000 0.771670165130 1.319188188547 7
H -0.000000000000 -0.747077150089 1.324086214974 7
O -0.000000000000 0.010468354749 0.754819362610 8
H 0.000000000000 0.771670165130 1.319188188547 8
H -0.000000000000 -0.747077150089 1.324086214974 8
Options
Temperature=500.0
LADD=5
NLowest=24
NRUN=24
EQOnly
UpDC=12
DownDC=12
Here, LADD=5, NLowest=24, and NRUN=24 indicate that 5 largest ADDs are followed from each EQ (l-ADDF5), the ADDF is applied to 24 lowest EQs in the current EQ list, and 24 random structures are considered as starting points, respectively. Without LADD, NLowest, and NRUN, the f-ADDF is applied to all EQs in the current EQ list, and a search is performed starting from the initial (input) structure, respectively. The last integer in each Cartesian coordinate designates a part to which the corresponding atom belongs, and this input is used for generation of initial random structures (see Part designation for random structure generation). TS optimizations will be skipped if EQOnly option is used, where the TS list is empty in this case. Temperature (in K) is related to the NLowest=24 option, and the l-ADDF5 is applied to 24 lowest EQs in terms of (harmonic) free-energy at eleven different temperatures, i.e., 0, 50, 100, 150, 200, 250, 300, 350, 400, 450, and 500 K, in this case with the Temperature=500.0 keyword.