Wednesday, April 5, 2017

Single bond correlations in HMBC spectra

HMBC spectra are designed to show correlations between hydrogens and heteronuclei separated by two or more bonds, however, single bond correlations are often present in HMBC spectra as well. Like the multiple bond correlations in HSQC spectra discussed in the previous post, single bond correlations in HMBC spectra can be confusing if you are not expecting them, but once aware that they may be present they are fairly easy to identify.

The figure below shows an expansion of the HMBC spectrum of cholesteryl acetate (blue) overlaid with the HSQC spectum (red). Long range HMBC correlations can be seen at 5.39,74.0 (H6,C3), 5.39,139.6 (H6,C5) and 4.62,139.6 (H3,C5). Two single bond HSQC correlations are present at 5.39,122.6 (H6,C6) and 4.62,74.0 (H3,C3). On either side of both HSQC correlations are two HMBC peaks separated by 150 Hz, the single bond HMBC correlations.

The HMBC experiment is designed to record long range correlations using scalar coupling. Multiple bond scalar coupling is much weaker than single bond coupling so in the HMBC the problem is how to exclude the single bond peaks. One solution is not to bother excluding them but to make them easily distinguished from the long range correlations. Recording the HMBC without heteronuclear decoupling during acquisition causes the single bond correlations to be split by the large 1JCH coupling, typically ~140Hz. Of course the nJCH coupling (0-10 Hz) is present too, but because this is similar in magnitude to the peak width, splitting due to nJCH is generally not observed. Thus, pairs of peaks separated by ~140 Hz appearing either side of HSQC correlations are almost certainly single bond HMBC correlations.

Modern versions of the HMBC experiment include J-filters to remove the single bond correlations. The three fold low pass J-filter in the sequence1 used by the Facility's standard HMBC parameter set works quite well and removes most single bond correlations, but not all. In crowded regions of your HMBC, particularly around strong methyl resonances, be aware that those weak peaks may be single bond correlations.

1. Julien Furrer
"A robust, sensitive, and versatile HMBC experiment for rapid structure elucidation by NMR: IMPACT-HMBC"
Chem. Commun. 2010 46(19):3396-3398


  1. Always informative, Brendan. As you wrote, 1J(CH) is sometimes observed in HMBC (esp. if one 'digs down' in the contour levels) but it's worth mentioning, here, that sometimes in HSQC experiments, the cross peaks disappear altogether for reasons related: the magnitude of 1J(CH). For example, when recording HSQC of terminal acetylenes, the expected CH correlation is missing because 1JCH ~ 250 Hz - far from the generalized 140 Hz that is nominally selected in the default experiment. Same thing goes for strained rings (e.g. cyclopropane, aziridine, even epoxides) where |1JCH| is much higher than 140 Hz

  2. Exactly. In both HMBC and HSQC experiments a "typical" value is used for the coupling one wants to detect, and if your system is not typical you may not observe the correlation. But, if you suspect you have an atypical molecule you can easily change the parameters to try and observe the correlation and prove your suspicion.