# Predicting Ti‐49 NMR Chemical Shift With New NMR‐DKH Basis Set

## TL;DR

This paper introduces a new computational method for predicting Ti-49 NMR chemical shifts using a novel basis set, achieving high accuracy with lower computational cost.

## Contribution

The development of a new NMR-DKH basis set for titanium and a predictive model for δ49Ti with high accuracy and low computational cost.

## Key findings

- A new computational protocol using the NMR-DKH basis set achieved a mean absolute deviation of 48 ppm for δ49Ti predictions.
- The proposed protocol showed excellent agreement with experimental data (R² of 0.9888) and outperformed relativistic methods in computational efficiency.
- A linear regression model was developed and validated with a mean absolute deviation of 48 ppm on an external test set.

## Abstract

In the present study, a computational protocol for predicting the Ti‐49 NMR chemical shift (δ49Ti) was constructed with our NMR‐DKH basis sets for all atoms, including titanium (Ti), which was developed in this work. Thus, computational protocols were proposed considering 55 different DFT functionals using nonrelativistic Hamiltonian. Besides, four‐components (4c) calculations employing the relativistic modified Dirac–Kohn–Sham Hamiltonian at GIAO‐4c‐BLYP/dyall.VDZ level was also considered. In the best protocol, the structures of the 41 Ti(IV) complexes studied, which cover a wide range of δ49Ti ranging from −1389 to +1325 ppm, were optimized at the BLYP/def2‐SVP/IEF‐PCM (UFF) level and the δ49Ti was calculated at the GIAO‐OLYP/NMR‐DKH/IEF‐PCM (UFF) level, both employing a nonrelativistic Hamiltonian. In this protocol, a mean absolute deviation (MAD) of only 48 ppm and a coefficient of determination (R
2) of 0.9888 were found, which represents an excellent agreement, and with lower computational cost, with the MAD of 62 ppm and R
2 of 0.9860 obtained with the relativistic full 4c Hamiltonian at GIAO‐4c‐BLYP/dyall.VDZ, indicating that the proposed protocol with the NMR‐DKH basis set is an excellent alternative for the study of Ti‐49 NMR. Additionally, a predictive model based on linear regression (δTicalc49=−1.0027×σcalc−1000.0) was developed, adjusted from 41 complexes and validated in an external set of nine Ti(IV) complexes, presenting a MAD of 48 ppm and confirming the robustness and extrapolation capacity of the proposed protocol.

A new NMR‐DKH basis set was proposed for Ti atom and the δ49Ti was studied in a set of 41 + 9 (calibration‐blue/validation‐red) Ti(IV) complexes, resulting in a mean absolute deviation of only 50 ppm with the GIAO‐OLYP/NMR‐DKH/IEF‐PCM(UFF)//BLYP/def2‐SVP/IEF‐PCM(UFF) protocol.