“Pure shift” 1H NMR, a robust method for revealing heteronuclear couplings in complex spectra

We investigate the utility of “pure shift” techniques in revealing heteronuclear couplings in complex 1H NMR spectra. The results show the technique to be a robust and valuable complement to the standard 1H spectrum, and an attractive alternative to heteronuclear decoupling since the technique is independent of the size of the heteronuclear couplings and the chemical shift range(s) of the heteronuclei involved. We highlight some possible artefacts, and the subtle effects due to the presence of 13C nuclei in otherwise symmetric molecules when bilinear rotational decoupling (BIRD) elements are present in the pulse sequence.


Introduction
Since the seminal discovery by Fried and Sabo in 1953 that the replacement of a single hydrogen atom in cortisol by a uorine atom increased the pharmaceutical activity by an order of magnitude, 1 the use of uorine in medicinal chemistry has become widespread, to the point where around 25% of human and veterinary pharmaceuticals and agrochemicals currently in use contain at least one uorine atom. At rst sight this statistic may appear surprising given that there are very few naturally occurring organouorine compounds 2 and the majority of the few that are known have high toxicity, but it is a result of the ability of uorine to modify the pK a of neighbouring groups, affect lipophilicity, hydrogen bonding and other binding interactions and, in many cases, block pathways leading to rapid metabolisation of the drug thereby increasing its bioavailability.
Many of these effects can be traced to the particular steric and electronic properties of the C-F bond. 3 The importance of uorine in the modern pharmaceutical industry has been the subject of a number of reviews, 4,5 and the same unique properties have seen its use spread to the wider arena of synthetic chemical biology. 6 These developments have enjoyed a symbiotic relationship with the wider developments in synthetic organouorine chemistry over the same period from a specialist niche area for a few intrepid workers 50 years ago to a mainstream discipline in modern synthetic chemistry. 7,8 In view of this, the ability to characterise uorinated organic molecules routinely has become much more important in recent years and NMR plays a key role in this, particularly, for example, in distinguishing between possible constitutional isomers in multiply substituted aromatic and hetero-aromatic rings with one or more uorine substituents. There are a considerable number of well-established NMR experiments that are potentially useful here including two-dimensional homonuclear 1 H experiments (such as COSY and NOESY) and twodimensional heteronuclear 1 H-13 C and 1 H- 19 F experiments (such as HSQC, HMBC and HOESY). The experiments involving manipulation of 1 H and 19 F simultaneously impose particular NMR hardware requirements (the ability to generate two "highband" frequencies simultaneously and preferably a probe that can be tuned to 1 H and 19 F simultaneously). While spectrometer systems that meet these requirements have become much more common in recent years, they are by no means ubiquitous. Furthermore, routine two-dimensional experiments involving uorine are oen rendered problematic, both by the wide range of uorine chemical shis 9 and by the variability of uorine coupling constants. 10 Regardless of whether the experimenter has the hardware and the knowledge to run the aforementioned experiments, the reality is that the experiment that really plays a pivotal role in the day-to-day decision making process is the simple onedimensional 1 H spectrum. For better or for worse, this is the experiment that non-specialists use right away to check reaction outcomes and to decide whether to continue with a synthetic procedure or to ask for technical assistance in running more sophisticated NMR experiments. In theory, the presence in 1 H spectra of splittings due to 1 H- 19 F coupling should improve the decision-making process, but in reality these are oen hidden under a mass of homonuclear ( 1 H-1 H) splittings. Very early on it was realised 11 that a possible solution to this problem was to eliminate the latter from 1 H spectra, as this would reveal the 1 H-19 F coupling, while obviating both the need for a uorine channel and the need for a skilled operator. The alternative approach, decoupling uorine while observing proton, is technically challenging, both because of the dispersion of the uorine signals and the variability of the uorine couplings. Although oen ignored, these should be taken into account when setting the decoupler. 13 Unfortunately, suppressing homonuclear couplings in a broadband and efficient manner has proven to be one of the most intractable problems in NMR history. Only recently have practical techniques able to produce good quality results in reasonable amounts of time been developed. These techniques have been produced primarily to address the problem of limited spectral dispersion (peak overlap) in 1 H-NMR but, as we shall see, some can be used equally well to solve the problem at hand. These techniques are the ideal addition to the simple 1 H spectrum and even provide a useful complement to the 19 F one, if that is available, as they can be run in a few minutes by non-specialists, even under automation, without the need for time consuming calibrations and without the need for uorine pulsing capabilities. Furthermore, they are not limited to the case of uorine; they can also be used to reveal couplings to elements such as phosphorous, platinum, rhodium, tin, thallium, mercury and silver.
Several schemes have been proposed to collapse homonuclear multiplets into singlets. These have been variously described as "broadband homonuclear decoupling" and "pure shi" techniques, as well as by other names. The question is which technique or techniques to use for the task at hand. The ideal technique should produce phase sensitive results in reasonable amounts of time and deal well with strong coupling. The old 2D J-resolved experiment used to reveal heteronuclear splittings fails in both areas, although attempts have been made to address both problems. 14, 15 One practical experiment that deals well with strong coupling and produces good quality data in a matter of minutes is a hybrid experiment 16 that uses the BIRD rotation proposed by Pines et al. 17 and the chemical shi sampling scheme of Zangger and Sterk 18 extended by Morris and Nilsson. 19 The pulse sequence used is shown in Fig. 1 and is discussed more extensively in ref. 16. The combination of BIRD and hard proton 180 rotations refocuses the evolution under the homonuclear coupling but allows the chemical shi to be sampled. A chunk of data lasting 1/sw 1 is acquired, where sw 1 is an integer submultiple of sw, and t 1 is incremented in larger steps 1/sw 1 , typically of several tens of ms. In principle a classic Zangger-Sterk pulse sequence can be also used, but the presence of second-order, strongly coupled signals, very common in aromatic systems, made us favour the former. In reality they complement one another.
In order to analyse the adequacy of the technique, several aspects need consideration: rstly, the ability of the technique to eliminate homonuclear splittings leaving only the heteronuclear ones; secondly, its robustness; thirdly, its potential limitations.

Experimental
All spectra were recorded in CDCl 3 at 298 K using a Varian VNMRS-600 spectrometer. The pulse sequence used to obtain the pure shi spectra and the macro needed to assemble the data can be found in the ESI. † The carbon pulses used were 125 ms long Broadband Inversion Pulses (BIP). 12 The experiment was optimised for a typical 1 J CH ¼ 160 Hz. Deviations from this value lead to signal attenuation. Between 32 and 64 increments were used, depending on the desired resolution, while the number of transients varied from 2 to 8. sw 1 was set to 50 Hz and experimental times varied from 3 to 15 minutes. Typical concentration ranged from 10 to 100 mM. Spin-system simulations were performed using the spectral simulation soware package in the Varian VNMR 6.1c soware, 20 which is an implementation of the LAOCOON program. 21

Results and discussion
The ability of the technique to eliminate homonuclear splittings leaving only heteronuclear ones is exemplied in Fig. 2. The 1-uorobenzene 1 H spectrum (Fig. 2C) is moderately complex, yet there is no difficulty in identifying the heteronuclear splittings in the pure shi spectrum (Fig. 2B), so resolution and clarity  have been gained. The observed couplings of 9.1 and 5.7 Hz are consistent with the relevant protons being ortho and meta to the uorine. The third signal does not show any resolved coupling, as should be the case for a signal located in a para position with respect to the uorine. In the second example, Fig. 3, a similar situation arises, but this time the presence of the proton-uorine coupling helps identication of the uorinated ring and the position of the uorine atom (para) with respect to the second ring. The proton-uorine coupling constants measured independently in the 1 H pure shi and 19 F spectra in Fig. 2 agree to within experimental error (AE0.1 Hz) but it has to be borne in mind that the BIRD pure shi experiment deals with protons attached to 13 C while the uorine one mainly deals with protons attached to 12 C, so minor differences may be due to the fact that the two spectra are predominantly of different isotopologues (see discussion below).
The two examples presented so far do not pose particularly challenging problems, and indeed their structures were both known in advance and were chosen for illustrative purposes. The next two samples, however, were real unknown cases submitted by industrial users for structure conrmation. The rst unknown sample, Fig. 4, exemplies how the technique complements the 19 F spectrum. In this case, the proton-uorine couplings cannot easily be determined from the uorine spectrum but they can be easily found from the pure shi spectrum. In other cases where uorine multiplets are well resolved, these couplings could, in principle, be determined directly from the uorine spectrum but because couplings to multiple protons or to other uorine nuclei may be present, the analysis is oen far from trivial. In such cases, the pure shi spectrum would greatly facilitate analysis.
The second of the "unknown" samples, Fig. 5, is a uoronitrobenzene, but its substitution pattern was unknown. The fact that there are four distinct proton signals shows that it must be either 2-uoronitrobenzene or 3-uoronitrobenzene. In this case, the signicant feature is the fact that the apparent doublet signal at about 7.9 ppm in the normal 1 H spectrum is essentially unchanged in the pure shi spectrum, showing that the proton involved has little 1 H-1 H coupling. This immediately identies the sample as 3-uoronitrobenzene, since the isolated proton must be in the 2 position.
To assess the robustness of the experiment, it was run under automation over a few months using samples submitted to Durham University's NMR service. The vast majority of the samples analysed contained aromatic units displaying second order spectra. In all cases the same parameters were used, illustrating the fact that the experiment can be run by nonexperts. The results show that the pulse sequence generally performs well, even in the presence of strong coupling. The most common artefacts are sidebands, marked with an asterisk in the examples presented. Such artefacts have not been reported previously, but seem to be as common as second order spectra. Although undesirable, their importance is small once experimenters become familiar with them. This type of artefact only becomes problematic when the experiment is used to reveal heteronuclear splittings of spins with low natural abundance, such as 29 Si, or when several species of very dissimilar concentrations are present. The origin of the artefacts is still under investigation but is almost certainly a consequence of strong coupling. Because the wavefunctions of strongly coupled spins mix, the BIRD element partly inverts the coupled partner bound to 12 C, not fully refocusing its proton-proton coupling interaction with the 13 C-bound partner. As a result, there is additional evolution during t 1 that translates into these   sidebands. The effect seems to be related to similar problems reported for HETCOR and related experiments. 22 A second class of artefact, previously reported in ref. 18, is a negative sideband caused by the decoupling method. Such artefacts are typically a few percent of the main signal and are oen too weak to see. They can be recognised by their sign and because they are separated by multiples of sw 1 Hz. An example can be seen in Fig. 4B.
The relative resilience of BIRD-based pulse sequences where spin systems showing strong coupling are concerned arises from the liing of 1 H near-degeneracy caused by the one-bond 1 H-13 C coupling present in the minority of molecules observed when using BIRD. Spin systems that are strongly coupled in per-12 C species will oen be weakly coupled in a mono-13 C isotopologue, and even where a mono-13 C system is still strongly coupled this will usually apply only to one of the two 13 C spin states (i.e. in general at most one of the two 13 C satellites in the 1 H spectrum will be strongly coupled). Conversely, in some otherwise symmetric systems, the presence of a 13 C spin will break the symmetry, introducing strong coupling that would not otherwise intrude. This is exemplied by the case of 1,4-diuorobenzene. The reason for this effect (previously unreported in the literature relating to homonuclear broadband decoupling but commonplace elsewhere) lies in the fact that the spin system of interest is not that of Fig. 6B, but rather that of Fig. 6A, where the presence of 13 C breaks the symmetry, making all of the nuclei inequivalent and introducing strong coupling. This results in unwanted features, as shown in Fig. 6A. To emphasise the point, the sample was subjected to another pure shi technique (Zangger-Sterk) that does not require the selection of protons coupled to 13 C. The results are basically identical to the parent 1 H spectrum, as expected.
The effects of 13 C can be illustrated by simulating spectra with all homonuclear couplings set to zero. Simulation of the spectrum expected on the basis of a single chemical shi for the four hydrogen atoms, the 1 H-1 H coupling constants and the 1 H-19 F coupling constants set to the values determined by Wray et al. 23 (Fig. SI-1 †) shows reasonably good agreement with the normal 1 H spectrum, conrming that some other factor is affecting the pure shi spectrum. We need only concern ourselves here with 13 C nuclei that are directly bonded to 1 H, and the symmetry of the molecule means that we need only consider a single 13 C site. We do not need to consider heteronuclear 13 C-1 H coupling, since its effects are largely eliminated by the heteronuclear decoupling during acquisition, but we do need to consider the isotope shis introduced by the presence of a single 13 C nucleus in the molecule. These effectively render the 4 1 H environments chemically inequivalent, and the same is true for the two 19 F environments. The isotope shis involved are small (typically 1 to 10 ppb), but have been determined previously both for uorine 24 and for proton, 25 and were included in the simulation. The fact that the isotope shis are small means that both the 1 H-1 H and 19 F-19 F couplings now give rise to strong second order effects in the spectra observed. A simulation of the conventional 1 H spectrum of the spin system including the isotope shis due to the presence of a single 13 C in the molecule and all the 1 H-1 H, 19 F-19 F, and 1 H-19 F couplings is shown in the ESI. † It should be noted that signals such as OH and NH are eliminated, as the pulse sequence selects only those protons coupled to 13 C. In addition, non-equivalent geminal protons are not decoupled from one another, as they are both bonded to the same carbon. They typically appear as pairs of doublets since couplings to third partners are eliminated. In cases where strong coupling is not a pressing problem or when methylene signals are important, Zangger-Sterk or Pell-Keeler 15 sequences can be used instead. In addition, some recently described ultrafast pure shi BIRD pulse sequences should be considered. 26,27 Conclusions The pure shi BIRD experiment is a robust, simple to run pulse sequence that both simplies 1 H spectra and reveals heteronuclear couplings. Minor artefacts are produced that may be obtrusive in cases where coupling to isotopes of low natural abundance is of interest, or where mixtures with very different concentrations are concerned, and complications can arise where the presence of 13 C breaks symmetry. 1 J. Fried and E. Sabo, J. Am. Chem. Soc., 1953, 75, 2273-2274.