Understanding non-covalent interactions in larger molecular complexes from first principles

Abstract

Non-covalent interactions pervade all matter and play a fundamental role in layered materials, biological systems, and large molecular complexes. Despite this, our accumulated understanding of non-covalent interactions to date has been mainly developed in the tens-of-atoms molecular regime. This falls considerably short of the scales at which we would like to understand energy trends, structural properties, and temperature dependencies in materials where non-covalent interactions have an appreciable role. However, as more reference information is obtained beyond moderately sized molecular systems, our understanding is improving and we stand to gain pertinent insights by tackling more complex systems, such as supramolecular complexes, molecular crystals, and other soft materials. In addition, accurate reference information is needed to provide the drive for extending the predictive power of more efficient workhorse methods, such as density functional approximations that also approximate van der Waals dispersion interactions. In this perspective, we discuss the first-principles approaches that have been used to obtain reference interaction energies for beyond modestly sized molecular complexes. The methods include quantum Monte Carlo, symmetry-adapted perturbation theory, non-canonical coupled cluster theory, and approaches based on the random-phase approximation. By considering the approximations that underpin each method, the most accurate theoretical references for supramolecular complexes and molecular crystals to date are ascertained. With these, we also assess a handful of widely used exchange-correlation functionals in density functional theory. The discussion culminates in a framework for putting into perspective the accuracy of high-level wavefunction-based methods and identifying future challenges.

Recent developments in symmetry‐adapted perturbation theory

“…An important example of such a benchmark SAPT(DFT) study is the calculation of interaction energies for the S12L database of large complexes . For systems of this size, a computation of unambiguous benchmark values exceeds our current algorithmic and computational capabilities, and the original S12L benchmark data (obtained by back‐correcting experimental association free energies) as well as high‐level local CC and quantum Monte Carlo interaction energies differ by up to several kcal/mol . In this case, it is highly advantageous to have other data available that combine reasonable accuracy with physical insight via the SAPT(DFT) energy decomposition.…”

“…Noncovalent interactions are not only ubiquitous, but they are also absolutely necessary to describe many physical, chemical, and biochemical phenomena, ranging from thermodynamics of nonideal gases to spectroscopy and scattering cross sections of interstellar complexes to the performance and selectivity (including enantioselectivity) of molecular catalysts to the secondary, tertiary, and quaternary structure of proteins, the double‐helix structure of DNA, and the interaction between enzymes and their substrates (or inhibitors). Thus, theoretical and experimental investigations of noncovalent interactions are published in very large quantities, and various aspects of these interactions have been reviewed in this journal and elsewhere …”

Finding chemical concepts in the Hilbert space: Coupled cluster analyses of noncovalent interactions

“…As a note of caution, one should be aware that the remarkable accuracy achieved by dispersion‐corrected mean field approaches for standard benchmark sets is the result of careful parametrization using available high quality benchmark values on relatively small systems, typically obtained from accurate first principles calculations. Such methods include the “gold standard” CCSD(T) and its local variants as well as Quantum Monte Carlo approaches …”

“…On the other hand, their main limitation is the heavy reliance on benchmark studies for parametrizing, assessing the accuracy and in some cases re‐parametrizing the exchange‐correlation functional. Considering that the large majority of reliable benchmark studies for NCIs focus on small systems with a few tens of atoms, one might question the reliability of these methods on computational predictions for large systems . Fortunately, with the development of fast and accurate first principles methods like the modern local CCSD(T) variants, it is becoming possible to put these methods to a quantitative test for large systems …”

Toward a less costly but accurate calculation of the CCSD(T)/CBS noncovalent interaction energy

“…For the DLPNO‐MP2 and DLPNO‐CCSD(T/T0) calculations for the study of noncovalent systems, it is necessary to use the tight PNO thresholds because the LoosePNO settings and even the NormalPNO settings maybe exclude the weakly interacting electronic states which play important roles in the correlation energy calculations of the noncovalent systems . For the isolated DLPNO‐MP2 or DLPNO‐CCSD(T/T0) calculations, the case is simple and just use the recommended tight PNO thresholds or even the very tight PNO thresholds.…”