Quantitative Prediction of Vertical Ionization Potentials from DFT via a Graph-Network-Based Delta Machine Learning Model Incorporating Electronic Descriptors

Abstract

While accurate wave function theories like CCSD­(T) are capable of modeling molecular chemical processes, the associated steep computational scaling renders them intractable for treating large systems or extensive databases. In contrast, density functional theory (DFT) is much more computationally feasible yet often fails to quantitatively describe electronic changes in chemical processes. Herein, we report an efficient delta machine learning (ΔML) model that builds on the Connectivity-Based Hierarchy (CBH) schemean error correction approach based on systematic molecular fragmentation protocolsand achieves coupled cluster accuracy on vertical ionization potentials by correcting for deficiencies in DFT. The present study integrates concepts from molecular fragmentation, systematic error cancellation, and machine learning. First, we show that by using an electron population difference map, ionization sites within a molecule may be readily identified, and CBH correction schemes for ionization processes may be automated. As a central feature of our work, we employ a graph-based QM/ML model, which embeds atom-centered features describing CBH fragments into a computational graph to further increase accuracy for the prediction of vertical ionization potentials. In addition, we show that the incorporation of electronic descriptors from DFT, namely electron population difference features, improves model performance well beyond chemical accuracy (1 kcal/mol) to approach benchmark accuracy. While the raw DFT results are strongly dependent on the underlying functional used, for our best models, the performance is robust and much less dependent on the functional.

Leveraging DFT and Molecular Fragmentation for Chemically Accurate pKa Prediction Using Machine Learning

“…Figure 12 shows a bar plot of the 10 most important CBH fragments, ranked by their feature permutation score. It is interesting to note the prevalence of fragments featuring amines (13,36,16,72,25,41), alcohols, and carboxylic acids (2,10,25). This aligns with our expectations since N and O are the most abundant heavy atoms (excluding carbon) in the data set (see Figure 3).…”

“…To address this demand, several recent studies have demonstrated the successful integration of quantum mechanical (QM) calculations and machine learning (ML) techniques for highly accurate physicochemical property prediction. The adaptation of ML as a viable tool in the quantum chemist’s toolbox has led to numerous applications in materials discovery, catalysis, drug design, etc. − When it comes to p K a prediction, one QM/ML model by Hunt et al used semiempirical features along with radial basis functions to obtain commendable performance on the SAMPL6 and Jensen data sets . Similarly, Lawler et al used features derived from DFT with a kernel ridge regression model to achieve a low mean absolute error (MAE) of 0.60 on oxoacids.…”

Approaching Coupled Cluster Accuracy with Density Functional Theory Using the Generalized Connectivity-Based Hierarchy

“…For example, we pointed out earlier that the aromatic units (e.g., phenyl groups) were left intact in our p K a studies . In a more recent study, we obtained slightly better performance from coarse-graining other functional groups such as nitro groups, sulfoxides, nitriles, etc . This is an active topic of ongoing research.…”

MQM 2022: The 10th Triennial Conference on Molecular Quantum Mechanics

“…They demonstrate the effectiveness of their method compared to other approaches using L- and M-edge XAS for transition metal and actinide compounds. In the machine-learning (ML) domain, coupled-cluster-level accuracy from DFT may be possible using well-designed ML models, according to work by Maier, Collins, and Raghavachari at Indiana University . Using an efficient Δ-ML model with a systematic molecular fragmentation approach, they report such accuracy for vertical ionization potentials.…”