A set of standard enthalpies of formation for benchmarking, calibration, and parametrization of electronic structure methods

Ciosłowski, Jerzy; Schimeczek, Michael; Guang, Liu; Stoyanov, Vesselin

doi:10.1063/1.1321306

Cited by 114 publications

(122 citation statements)

References 129 publications

Supporting

Mentioning

117

Contrasting

Unclassified

Order By: Relevance

“…However, five of these atom types appear in three or less compounds, specifically lithium, beryllium, sodium, aluminum, and boron. The so-called X1/07 data set [31][32][33][34][35][36][37] includes 393 molecules and covers a wider range of different chemical structures and bonding situations than that of the G3/99 set 14 (c.f. Table 1).…”

Section: Setting Up the Neural Networkmentioning

confidence: 99%

Accurate prediction of heats of formation by a combined method of B3LYP and neural network correction

Wu¹,

Xu²

2008

J Comput Chem

View full text Add to dashboard Cite

Recently, we proposed the X1 method which combines the B3LYP/6-311+G(3df,2p)//B3LYP/6-311+G(d,p) method with a neural network correction for an accurate yet efficient prediction of heats of formation (Wu and Xu, J Chem Phys 2007, 127, 214105). In this contribution, we discuss in detail how to set up the X1 neural network. We give examples, showing how to apply the X1 method and how the applicability of X1 can be extended. The overall mean absolute deviation of the X1 method from experiment for the 488 heats of formation is 1.52 kcal/mol compared with 9.44 kcal/mol for the original B3LYP results.

show abstract

Section: Setting Up the Neural Networkmentioning

confidence: 99%

Accurate prediction of heats of formation by a combined method of B3LYP and neural network correction

Wu¹,

Xu²

2008

J Comput Chem

View full text Add to dashboard Cite

show abstract

“…The ccCA methods were tested on a small subset of the G2-2 and G3/99 molecules as well as C͑NO 2 ͒ 3 and adamantane from the set of 600 molecules compiled by Cioslowski et al 41 Benchmarking of the ccCA methods on larger training sets will be published in due course, but we believe that merits and deficiencies of the ccCA methods will be evident in the following discussion.…”

Section: Specific Test Casesmentioning

confidence: 99%

“…6͒, and G3/99 ͑Ref. 8͒ test sets, and two large systems from the 600-molecule training set proposed by Cioslowski et al 41 The performance of the ccCA will be examined and compared to experiment, Gn methods ͑with and without their respective HLCs͒, and other theoretical studies.…”

Section: Introductionmentioning

confidence: 99%

“…Other model chemistries, such as the multicoefficient correlation method ͑MCCM͒ of Truhlar and co-workers, optimize parameters to approach the CBS FCI correlation energy. [38][39][40] Attempts to parametrize aspects of density functional theory 41 ͑DFT͒ and perturbation theory 42 ͑MP2͒ energy computations to attain accurate enthalpies of formation have also been attempted.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

The correlation consistent composite approach (ccCA): An alternative to the Gaussian-n methods

DeYonker

Cundari

Wilson

2006

The Journal of Chemical Physics

307

327

View full text Add to dashboard Cite

An alternative to the Gaussian-n ͑G1, G2, and G3͒ composite methods of computing molecular energies is proposed and is named the "correlation consistent composite approach" ͑ccCA, ccCA-CBS-1, ccCA-CBS-2͒. This approach uses the correlation consistent polarized valence ͑cc-pV XZ͒ basis sets. The G2-1 test set of 48 enthalpies of formation ͑⌬H f ͒, 38 adiabatic ionization potentials ͑IPs͒, 25 adiabatic electron affinities ͑EAs͒, and 8 adiabatic proton affinities ͑PAs͒ are computed using this approach, as well as the ⌬H f values of 30 more systems. Equilibrium molecular geometries and vibrational frequencies are obtained using B3LYP density functional theory. When applying the ccCA-CBS method with the cc-pVXZ series of basis sets augmented with diffuse functions, mean absolute deviations within the G2-1 test set compared to experiment are 1.33 kcal mol −1 for ⌬H f , 0.81 kcal mol −1 for IPs, 1.02 kcal mol −1 for EAs, and 1.51 kcal mol −1 for PAs, without including the "high-level correction" ͑HLC͒ contained in the original Gn methods. Whereas the HLC originated in the Gaussian-1 method as an isogyric correction, it evolved into a fitted parameter that minimized the error of the composite methods, eliminating its physical meaning. Recomputing the G1 and G3 enthalpies of formation without the HLC reveals a systematic trend where most ⌬H f values are significantly higher than experimental values. By extrapolating electronic energies to the complete basis set ͑CBS͒ limit and adding G3-like corrections for the core-valence and infinite-order electron correlation effects, ccCA-CBS-2 often underestimates the experimental ⌬H f , especially for larger systems. This is desired as inclusion of relativistic and atomic spin-orbit effects subsequently improves theoretical ⌬H f values to give a 0.81 kcal mol −1 mean absolute deviation with ccCA-CBS-2. The ccCA-CBS method is a viable "black box" method that can be used on systems with at least 10-15 heavy atoms.

show abstract

“…This criterion for inclusion within the training sets is based only on firm experimental data and is not influenced by other theoretical studies or by conclusions made from the computation of Gn energies. Cioslowski et al 86 created their own benchmark set of 600 enthalpies of formation, many of which are included in the G3/99 test set.…”

Section: Comments On Some Of the Molecules Contained In The G3/99mentioning

confidence: 99%

The correlation-consistent composite approach: Application to the G3/99 test set

DeYonker

Grimes

Yockel

et al. 2006

The Journal of Chemical Physics

149

167

View full text Add to dashboard Cite

The correlation-consistent composite approach ͑ccCA͒, an ab initio composite technique for computing atomic and molecular energies, recently has been shown to successfully reproduce experimental data for a number of systems. The ccCA is applied to the G3/99 test set, which includes 223 enthalpies of formation, 88 adiabatic ionization potentials, 58 adiabatic electron affinities, and 8 adiabatic proton affinities. Improvements on the original ccCA formalism include replacing the small basis set quadratic configuration interaction computation with a coupled cluster computation, employing a correction for scalar relativistic effects, utilizing the tight-d forms of the second-row correlation-consistent basis sets, and revisiting the basis set chosen for geometry optimization. With two types of complete basis set extrapolation of MP2 energies, ccCA results in an almost zero mean deviation for the G3/99 set ͑with a best value of −0.10 kcal mol −1 ͒, and a 0.96 kcal mol −1 mean absolute deviation, which is equivalent to the accuracy of the G3X model chemistry. There are no optimized or empirical parameters included in the computation of ccCA energies. Except for a few systems to be discussed, ccCA performs as well as or better than Gn methods for most systems containing first-row atoms, while for systems containing second-row atoms, ccCA is an improvement over Gn model chemistries.

show abstract

A set of standard enthalpies of formation for benchmarking, calibration, and parametrization of electronic structure methods

Cited by 114 publications

References 129 publications

Accurate prediction of heats of formation by a combined method of B3LYP and neural network correction

Accurate prediction of heats of formation by a combined method of B3LYP and neural network correction

The correlation consistent composite approach (ccCA): An alternative to the Gaussian-n methods

The correlation-consistent composite approach: Application to the G3/99 test set

Contact Info

Product

Resources

About