Limits...
Accurate and efficient representation of intra ­ molecular energy in ab initio generation of crystal structures. I. Adaptive local approximate models

View Article: PubMed Central - HTML - PubMed

ABSTRACT

1172: The global search stage of crystal structure prediction (CSP) methods requires a fine balance between accuracy and computational cost, particularly for the study of large flexible molecules. A major improvement in the accuracy and cost of the intramolecular energy function used in the CrystalPredictor II [Habgood et al. (2015 ▸). J. Chem. Theory Comput., 1957–1969] program is presented, where the most efficient use of computational effort is ensured via the use of adaptive local approximate model (LAM) placement. The entire search space of the relevant molecule’s conformations is initially evaluated using a coarse, low accuracy grid. Additional LAM points are then placed at appropriate points determined via an automated process, aiming to minimize the computational effort expended in high-energy regions whilst maximizing the accuracy in low-energy regions. As the size, complexity and flexibility of molecules increase, the reduction in computational cost becomes marked. This improvement is illustrated with energy calculations for benzoic acid and the ROY molecule, and a CSP study of molecule (XXVI) from the sixth blind test [Reilly et al. (2016 ▸). Acta Cryst. B, 439–459], which is challenging due to its size and flexibility. Its known experimental form is successfully predicted as the global minimum. The computational cost of the study is tractable without the need to make unphysical simplifying assumptions.

No MeSH data available.


Related in: MedlinePlus

Absolute difference between ab initio and LAM predicted intramolecular energies (kJ mol−1) based on a 5° scan, with LAMs computed based on (a) the coarse regular grid of Fig. 5 ▸(a). (b) The adaptive scheme of Fig. 5 ▸(b). The black triangles represent experimental conformations.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC5134761&req=5

fig7: Absolute difference between ab initio and LAM predicted intramolecular energies (kJ mol−1) based on a 5° scan, with LAMs computed based on (a) the coarse regular grid of Fig. 5 ▸(a). (b) The adaptive scheme of Fig. 5 ▸(b). The black triangles represent experimental conformations.

Mentions: The intramolecular energy contribution is also computed ab initio over the same range of degrees of freedom at 5° increments and shown in Fig. 6 ▸(c). Visual comparison of the three energy landscapes show that key qualitative features are captured by both LAM-based approximations. A more quantitative comparison is presented in Figs. 7 ▸(a) and (b), where the differences between the LAM approximation and the ab initio energies are computed at 5° intervals. The average absolute deviation for the regular coarse grid scheme is 0.75 kJ mol−1, while for the adaptive scheme it is 0.56 kJ mol−1. More importantly, it is evident that with the regular grid, there are many areas in which the error is more than 5 kJ mol−1, particularly at the edges of LAM validity. This can lead to the generation of a low-accuracy energy landscape during the global search, in which some structures are found to have unrealistically low or high lattice energy. Finally, it can be seen that in the areas surrounding the experimental structures (black triangles), improved accuracy is achieved.


Accurate and efficient representation of intra ­ molecular energy in ab initio generation of crystal structures. I. Adaptive local approximate models
Absolute difference between ab initio and LAM predicted intramolecular energies (kJ mol−1) based on a 5° scan, with LAMs computed based on (a) the coarse regular grid of Fig. 5 ▸(a). (b) The adaptive scheme of Fig. 5 ▸(b). The black triangles represent experimental conformations.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC5134761&req=5

fig7: Absolute difference between ab initio and LAM predicted intramolecular energies (kJ mol−1) based on a 5° scan, with LAMs computed based on (a) the coarse regular grid of Fig. 5 ▸(a). (b) The adaptive scheme of Fig. 5 ▸(b). The black triangles represent experimental conformations.
Mentions: The intramolecular energy contribution is also computed ab initio over the same range of degrees of freedom at 5° increments and shown in Fig. 6 ▸(c). Visual comparison of the three energy landscapes show that key qualitative features are captured by both LAM-based approximations. A more quantitative comparison is presented in Figs. 7 ▸(a) and (b), where the differences between the LAM approximation and the ab initio energies are computed at 5° intervals. The average absolute deviation for the regular coarse grid scheme is 0.75 kJ mol−1, while for the adaptive scheme it is 0.56 kJ mol−1. More importantly, it is evident that with the regular grid, there are many areas in which the error is more than 5 kJ mol−1, particularly at the edges of LAM validity. This can lead to the generation of a low-accuracy energy landscape during the global search, in which some structures are found to have unrealistically low or high lattice energy. Finally, it can be seen that in the areas surrounding the experimental structures (black triangles), improved accuracy is achieved.

View Article: PubMed Central - HTML - PubMed

ABSTRACT

1172: The global search stage of crystal structure prediction (CSP) methods requires a fine balance between accuracy and computational cost, particularly for the study of large flexible molecules. A major improvement in the accuracy and cost of the intramolecular energy function used in the CrystalPredictor II [Habgood et al. (2015 ▸). J. Chem. Theory Comput., 1957–1969] program is presented, where the most efficient use of computational effort is ensured via the use of adaptive local approximate model (LAM) placement. The entire search space of the relevant molecule’s conformations is initially evaluated using a coarse, low accuracy grid. Additional LAM points are then placed at appropriate points determined via an automated process, aiming to minimize the computational effort expended in high-energy regions whilst maximizing the accuracy in low-energy regions. As the size, complexity and flexibility of molecules increase, the reduction in computational cost becomes marked. This improvement is illustrated with energy calculations for benzoic acid and the ROY molecule, and a CSP study of molecule (XXVI) from the sixth blind test [Reilly et al. (2016 ▸). Acta Cryst. B, 439–459], which is challenging due to its size and flexibility. Its known experimental form is successfully predicted as the global minimum. The computational cost of the study is tractable without the need to make unphysical simplifying assumptions.

No MeSH data available.


Related in: MedlinePlus