Practical aspects of real-time pure shift HSQC experiments

Abstract

Pure shift NMR spectroscopy has become an efficient tool for improving resolution in proton NMR spectra by removing the effect of homonuclear couplings. The introduction of real-time acquisition methods has allowed the main drawback of pure shift NMR, the long experiment times needed, to be circumvented. Real-time methods use periodic application of J-refocusing pulse sequence elements, acquiring a single free induction decay, in contrast to previous methods that construct a pure shift interferogram by concatenating excerpts from multiple free induction decays. In the important heteronuclear single-quantum correlation experiment, implementing real-time pure shift data acquisition typically leads to the simultaneous improvement of both resolution and sensitivity. The current limitations of and problems with real-time pure shift acquisition methods are discussed here in the context of heteronuclear single-quantum correlation experiments. We aim to provide a detailed account of the technical challenges, together with a practical guide to exploiting the full potential of such methods.

Keywords: HSQC; NMR; homodecoupling; pure shift; real time.

Figure 1

Part of the (a) conventional and (b) real‐time pure shift heteronuclear single‐quantum correlation spectrum of a mixture of carbohydrates containing 80 mm of d‐glucose, 75 mm of d‐trehalose, 56 mm d‐raffinose, and 30 mm of α‐cyclodextrin in D₂O. The COSY‐type artefacts (indicated by grey stars), which are often seen in conventional HSQC spectra, are not visible in the pure shift heteronuclear single‐quantum correlation spectrum because their antiphase character causes these signals to cancel on homodecoupling

Figure 2

Data acquisition section of a real‐time pure shift heteronuclear single‐quantum correlation (HSQC) pulse sequence using bilinear rotation decoupling (BIRD) J refocusing. The HSQC part is not detailed here. Open and filled rectangles are 90° and 180° proton pulses, respectively. The diagonally and horizontally crossed rectangles are BIP broadband inversion pulses34 and heteronuclear decoupling using WURST4035 pulses, respectively, on the carbon channel. The directions of the chemical shift (δ) and J evolution are shown below the pulse sequence, using grey where evolution cancels within a given half cycle and red where it does not. The optimum value of τ is 0.5/(¹ J _CH). The gradient pulses and associated delays are optional (see text). The duration of the final chunk of data acquisition is halved (not shown) to match the first half chunk (τ_ch/2) of data. The timing of transmitter/receiver switching is depicted above the pulse sequence. Additional data points (occupying a time τ_dr) can be collected and discarded prior to Fourier transformation to reduce data corruption at the beginning of each data acquisition period (see Section 2.7). Note that the timing of the double spin echo is not disrupted by this because each period τ_dr is accommodated in one half of an echo period (e.g., τ₁ after the first half chunk and τ₂ before the first full chunk)

Figure 3

(left) F ₂ traces from the deshielded half of real‐time pure shift heteronuclear single‐quantum correlation spectra of 2,3‐dibromothiophene in deuterated dimethyl sulfoxide doped with chromium acetylacetonate (T ₁ = 1.2 s, T ₂ = 0.66 s) at 500 MHz and (right) simulations of the first increments of the experiments. Experiments a–i used 20, 40, 60, 80, 100, 200, 300, 400, and 500 ms gaps, respectively, between the acquired data chunks. The full line width at half height is shown at the right‐hand side of each trace. The increase in broadening is caused purely by T ₂ relaxation loss between successive data chunks, while the basic linewidth contains a significant contribution from pulse imperfections. The chunk duration was 25.6 ms, and the total t ₂ acquisition time of 0.4 s was the result of 16 consecutive J‐refocusing steps (i.e., n = 8). Gaussian weighting with a time constant of 0.2 s was used in both experiment and simulation.

Figure 4

Real‐time pure shift heteronuclear single‐quantum correlation spectra of 2,3‐dibromothiophene in deuterated dimethyl sulfoxide (doped with chromium acetylacetonate) zoomed to show only the CH neighbouring the sulphur atom. The spectra in columns a–d were acquired without (a) and with (b) coherence transfer pathway gradients and without (a, b) and with (c, d) an extended phase cycle (x, y on all pulses of the J‐refocusing element), respectively. The experiments along rows i–iv were acquired without (a, c) and with (b, d) MLEV‐16 chunk‐to‐chunk phase sequencing and with (a, b) and without (c, d) the second carbon inversion pulse in the bilinear rotation decoupling element. Traces plotted above the 2D spectra are taken at the carbon chemical shift of the relevant signal

Figure 5

F ₂ traces from the deshielded halves of real‐time pure shift heteronuclear single‐quantum correlation (HSQC) spectra of doped 2,3‐dibromothiophene in deuterated dimethyl sulfoxide (left), and simulations of the first increment of the experiments (right). The real‐time pure shift HSQC experiments a–e used bilinear rotation decoupling delays τ optimised for one‐bond couplings of 210, 190, 170, 150, and 130 Hz, respectively. The HSQC insensitive nuclei enhanced by polarisation transfer (INEPT) elements were kept at the optimum timing for the actual coupling constant of 190 Hz. The full line width at half height is shown at the right side of each trace. The broadening is caused by imperfect refocusing by the bilinear rotation decoupling element that effectively reduces the amount of magnetisation surviving successive data chunks

Figure 6

Doped water signal measured with single‐pulse excitation followed by (a) normal acquisition and (b–d) real‐time pure shift acquisition using a band‐selective pulse instead of the bilinear rotation decoupling element. MLEV phase sequencing and an eight‐step phase cycle (CYCLOPS and two‐step EXORCYCLE) were used (b) without gradient pulses, (c) with single axis z gradient pulses, and (d) with gradient pulses along +y (13 G cm⁻¹) and −x (6 G cm⁻¹) directions, where the amplitudes of the gradient pulses were set to cancel the field–lock disturbances caused. The change in peak position when only z gradient pulses were used was 0.3 Hz