An automated multi-order phase correction routine for processing ultra-wideline NMR spectra

Abstract

Efficient acquisition of wideline solid-state nuclear magnetic resonance (NMR) spectra with patterns affected by large inhomogeneous broadening is accomplished with the use of broadband pulse sequences. These specialized pulse sequences often use frequency-swept pulses, which feature time-dependent phase and amplitude modulations that in turn deliver broad and uniform excitation across large spectral bandwidths. However, the resulting NMR spectra are often affected by complex frequency-dependent phase dispersions, owing to the interplay between the frequency-swept excitations and anisotropic resonance frequencies. Such phase distortions necessitate the use of multi-order non-linear corrections in order to obtain absorptive, distortion-free patterns with uniform phasing. Performing such corrections is often challenging due to the complex interdependence of the linear and non-linear phase contributions, and how these may affect the NMR signal. Hence, processing of these data usually involves calculating the spectra in magnitude mode wherein the phase information is discarded. Herein, we present a fully automated phasing routine that is capable of processing and phase correcting such wideline NMR spectra. Its performance is corroborated via processing of NMR data acquired using both the WURST-CPMG (Wideband, Uniform-Rate, Smooth Truncation with Carr-Purcell Meiboom-Gill acquisition) and BRAIN-CP (BRoadband Adiabatic Inversion Cross Polarization) pulse sequences for a variety of nuclei (i.e., 119Sn, 195Pt, 35Cl, 87Rb, and 14N). Based on both simulated and experimental NMR datasets, it is demonstrated that automatic phase corrections up to and including second order can be readily achieved without a priori information regarding the nature of the phase-distorted NMR datasets, and independently of the exact manner in which time-domain NMR data are collected and subsequently processed. In addition, it is shown that NMR spectra acquired at both single and multiple transmitter frequencies that are processed with this automated phasing routine have improved signal-to-noise properties than those processed with conventional magnitude calculations, along with powder patterns that better match those of ideal NMR spectra, even for datasets possessing low signal-to-noise ratios and/or affected by spectral artifacts.