ANI-1ccx-gelu Universal Interatomic Potential and Its Fine-Tuning: Toward Accurate and Efficient Anharmonic Vibrational Frequencies

The Journal of Physical Chemistry Letters, Journal Year: 2025, Volume and Issue: unknown, P. 483 - 493

Published: Jan. 2, 2025

Calculating anharmonic vibrational modes of molecules for interpreting experimental spectra is one the most interesting challenges contemporary computational chemistry. However, traditional QM methods are costly this application. Machine learning techniques have emerged as a powerful tool substituting methods. Universal interatomic potentials (UIPs) hold particular promise to deliver accurate results at fraction cost methods, but performance UIPs calculating frequencies remains hitherto unknown. Here we show that despite known excellent representative UIP ANI-1ccx thermochemical properties, it fails due original unfortunate choice activation function. Hence, recommend evaluating new on an additional important quality test. To remedy shortcomings ANI-1ccx, introduce its reformulation ANI-1ccx-gelu with GELU function, which capable IR reasonable accuracy (close B3LYP/6-31G*). We also our can be fine-tuned obtain very some specific more effort needed improve overall and capability fine-tuning. The will included part universal updatable AI-enhanced (UAIQM) platform available together usage fine-tuning tutorials in open-source MLatom https://github.com/dralgroup/mlatom. calculations performed via web browser https://XACScloud.com.

Language: Английский

---

Improving the Reliability of, and Confidence in, DFT Functional Benchmarking through Active Learning

Journal of Chemical Theory and Computation, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 2, 2025

Validating the performance of exchange-correlation functionals is vital to ensure reliability density functional theory (DFT) calculations. Typically, these validations involve benchmarking data sets. Currently, such sets are usually assembled in an unprincipled manner, suffering from uncontrolled chemical bias, and limiting transferability results a broader space. In this work, data-efficient solution based on active learning explored address issue. Focusing─as proof principle─on pericyclic reactions, we start BH9 set design reaction space around initial by combinatorially combining templates substituents. Next, surrogate model trained predict standard deviation activation energies computed across selection 20 distinct DFT functionals. With model, designed explored, enabling identification challenging regions, i.e., regions with large divergence, for which representative reactions subsequently acquired as additional training points. Remarkably, it turns out that function mapping molecular structure divergence readily learnable; convergence reached upon acquisition fewer than 100 reactions. our final updated more challenging─and arguably representative─pericyclic curated, demonstrate has changed significantly compared original subset.

Language: Английский

---

X2‐PEC: A Neural Network Model Based on Atomic Pair Energy Corrections

Journal of Computational Chemistry, Journal Year: 2025, Volume and Issue: 46(8)

Published: March 18, 2025

ABSTRACT With the development of artificial neural networks (ANNs), its applications in chemistry have become increasingly widespread, especially prediction various molecular properties. This work introduces X2‐PEC method, that is, second generalization X1 series ANN methods developed our group, utilizing pair energy correction (PEC). The essence X2 model lies feature vector construction, using overlap integrals and core Hamiltonian to incorporate physical chemical information into vectors describe atomic interactions. It aims enhance accuracy low‐rung density functional theory (DFT) calculations, such as those from widely used BLYP/6‐31G(d) or B3LYP/6‐31G(2df,p) methods, level top‐rung DFT highly accurate doubly hybrid XYGJ‐OS/GTLarge method. Trained on QM9 dataset, excels predicting atomization energies isomers C 6 H 8 4 N 2 O with varying bonding structures. performance standard enthalpies formation for datasets G2‐HCNOF, PSH36, ALKANE28, BIGMOL20, HEDM45, well a HCNOF subset BH9 reaction barriers, is equally commendable, demonstrating good ability predictive accuracy, potential further achieve greater accuracy. These outcomes highlight practical significance elevating results lower‐rung calculations higher‐rung through deep learning.

Language: Английский

---

Enhancing the prediction of TADF emitter properties using Δ-machine learning: A hybrid semi-empirical and deep tensor neural network approach

The Journal of Chemical Physics, Journal Year: 2025, Volume and Issue: 162(14)

Published: April 8, 2025

This study presents a machine learning (ML)-augmented framework for accurately predicting excited-state properties critical to thermally activated delayed fluorescence (TADF) emitters. By integrating the computational efficiency of semi-empirical PPP+CIS theory with Δ-ML approach, model overcomes inherent limitations in key properties, including singlet (S1) and triplet (T1) energies, singlet–triplet gaps (ΔEST), oscillator strength (f). The demonstrated exceptional accuracy across datasets varying sizes diverse molecular features, notably excelling ΔEST values, negative regions relevant TADF molecules inverted S1–T1 gaps. work highlights synergy between physics-inspired models accelerating design efficient emitters, providing foundation future studies on complex systems advanced functional materials.

Language: Английский

---

Accurate and Affordable Simulation of Molecular Infrared Spectra with AIQM Models

The Journal of Physical Chemistry A, Journal Year: 2025, Volume and Issue: unknown

Published: April 14, 2025

Infrared (IR) spectroscopy is a potent tool for identifying molecular structures and studying the chemical properties of compounds, hence, various theoretical approaches have been developed to simulate predict IR spectra. However, based on quantum calculations suffer from high computational cost (e.g., density functional theory, DFT) or insufficient accuracy semiempirical methods orders magnitude faster than DFT). Here, we introduce new approach, universal machine learning (ML) models AIQM series targeting CCSD(T)/CBS level, that can deliver spectra with close DFT (compared experiment) speed GFN2-xTB method. This approach harmonic oscillator approximation frequency scaling factors fitted experimental data. While benchmarks reported here are focused spectra, our implementation supports anharmonic simulations via dynamics VPT2. These implementations available in MLatom as described https://github.com/dralgroup/mlatom be performed online web browser.

Language: Английский

---