Understanding cellulose pyrolysis via ab initio deep learning potential field

Comprehensive and dynamic studies of cellulose pyrolysis reaction mechanisms are crucial in designing experiments and processes with enhanced safety, efficiency, and sustainability. The details of the pyrolysis mechanism are not readily available from experiments but can be better described via molecular dynamics (MD) simulations. However, the large size of cellulose molecules challenges accurate ab initio MD simulations, while existing reactive force field parameters lack precision. In this work, precise ab initio deep learning potentials field (DPLF) are developed and applied in MD simulations to facilitate the study of cellulose pyrolysis mechanisms. The formation mechanism and production rate of both valuable and greenhouse products from cellulose at temperatures larger than 1073 K are comprehensively described. This study underscores the critical role of advanced simulation techniques, particularly DLPF, in achieving efficient and accurate understanding of cellulose pyrolysis mechanisms, thus promoting wider industrial applications.

A cyclodextrin-based reagent for cis/trans-geometrical isomers separation by mobility measurements and chemical calculations

Identification of cis/trans-carbon-carbon double-bond (CC) isomers remain challenging. Herein, a simple and rapid method for the separation and analysis of cis/trans-maleic acid (MA) and aconitic acid (AA) using Trapped Ion Mobility Spectrometry (TIMS) was developed. α-, β-, γ-cyclodextrin (CD) were served as the separation reagent, slight difference in mobility separation was obtained by [CD-MA/AA-H]^-. Specially, with the addition of divalent metal ion (G²⁺) as coordination metal ion, the separation effect was much increased by [CD-MA/AA + G-H]⁺, and α-CD has better mobility separation effect than β-/γ-CD. Moreover, chemical calculations revealed the binary and ternary complexes are in the inclusion forms, and microscopic interactions between cis/trans-MA/AA, CDs, and G²⁺ are somewhat different that making their mobility separation. Finally, quantifications of cis/trans-isomers were analyzed in food samples, with good linearity (R² > 0.99) and recoveries obtained from 87.25 % to 100.73 %.

Improving the accuracy of GIPAW chemical shielding calculations with cluster and fragment corrections

Ab initio methods for predicting NMR parameters in the solid state are an essential tool for assigning experimental spectra and play an increasingly important role in structural characterizations. Recently, a molecular correction (MC) technique has been developed which combines the strengths of plane-wave methods (GIPAW) with single molecule calculations employing Gaussian basis sets. The GIPAW + MC method relies on a periodic calculation performed at a lower level of theory to model the crystalline environment. The GIPAW result is then corrected using a single molecule calculation performed at a higher level of theory. The success of the GIPAW + MC method in predicting a range of NMR parameters is a result of the highly local character of the tensors underlying the NMR observable. However, in applications involving strong intermolecular interactions we find that expanding the region treated at the higher level of theory more accurately captures local many-body contributions to the N15 NMR chemical shielding (CS) tensor. We propose alternative corrections to GIPAW which capture interactions between adjacent molecules at a higher level of theory using either fragment or cluster-based calculations. Benchmark calculations performed on N15 and C13 data sets show that these advanced GIPAW-corrected calculations improve the accuracy of chemical shielding tensor predictions relative to existing methods. Specifically, cluster-based N15 corrections show a 24% and 17% reduction in RMS error relative to GIPAW and GIPAW + MC calculations, respectively. Comparing the benchmark data sets using multiple computational models demonstrates that N15 CS tensor calculations are significantly more sensitive to intermolecular interactions relative to C13. However, fragment and cluster-based corrections that include direct hydrogen bond partners are sufficient for optimizing the accuracy of GIPAW-corrected methods. Finally, GIPAW-corrected methods are applied to the particularly challenging NMR spectral assignment of guanosine dihydrate which contains two guanosine molecules in the asymmetric unit.

Accurate fragment-based 51-V chemical shift predictions in molecular crystals

Nuclear magnetic resonance (NMR) spectroscopy plays a crucial role in determining molecular structure for complex biological and pharmaceutical compounds. NMR investigations are increasingly reliant on computation for mapping spectral features to chemical structures. Here we benchmark the accuracy of fragment-based ⁵¹V chemical shielding tensor calculations using a training set comprised of 10 biologically and pharmaceutically relevant oxovanadium complexes. Using our self-consistent reproduction of the Madelung potential (SCRMP) electrostatic embedding model, we demonstrate comparable performance between fragment methods and computationally demanding cluster-based techniques. Specifically, fragment methods employing hybrid density functionals are capable of reproducing the experimental ⁵¹V isotropic chemical shifts with a training set rms error of ~9 ​ppm, representing a 20% improvement over traditional plane wave techniques. We provide training set-derived linear regression models for mapping the absolute shieldings obtained from computation to the experimentally determined chemical shifts using four common density functionals; PBE0, B3LYP, PBE, and BLYP. Finally, we establish the utility of fragment methods and the reported regression parameters examining four oxovanadium structures excluded from the training set including the tetracoordinate oxovanadium silicate [Formula: see text] , VO¹⁵NGlySalbz which contains redox-active ligands, and the solid-state form of the common ⁵¹V NMR reference compound VOCl₃.