I am trying to assign my compound by proton NMR where glycerol is one of the starting materials. To assign, I was looking for an appropriate proton NMR explanation (splitting tree with coupling of geminal and vicinal proton) for glycerol. I found one explanation which looks not appropriate because the vicinal coupling constant has shown to be greater than geminal (attached below). I would appreciate your kind help. Thank you very much.
If you look at other spectra (on the internet) you will see that the coupling Ha to Hc is different to the coupling Hb to Hc and that will allow you to find the coupling Ha to Hb which will be the same in both multiplets.
You should also be careful about magnetic inequivalance in thsi compound.
David Stephenson , thank you for your answer. I found another spectra (attached below, compound is exactly not same) from internet (Wisconsin University). They showed that the coupling of Ha to Hc or Hb to Hc is same and geminal coupling constant is higher than vicinal which is opposite to the spectra I attached before with the question.
Here, you mentioned the other spectra "on the internet", would please give me some [[email protected]]? Thank you again.
The 12 Hz (or rather -12 Hz) coupling is a very typical value for the geminal coupling 2J(Ha,Hb), as confirmed unambiguously also by the strong roof effect in the Ha and Hb multiplets. There is no doubt about that.
The 4 Hz splittings you are showing is 3J(Ha,Hc). As David Stephenson noted, the values of 3J(Ha,Hc) and 3J(Hb,Hc) might be different (though not necessarily so).
Measure carefully the splittings in your Hb multiplet and compare them to those in the Ha multiplet. You must find the 12 Hz values exactly the same, while the 4 Hz values could be slightly different in the two multiplets.
PS: I agree with the the Wisconsin Uni spectrum interpretation, including the geminal 2J(A,B), except for its sign. However, it is a very different molecule and the fact that H_M and H_N have accidentaly the same shift makes it even more special.
It is going to be very similar to 1,2,3-trichloropropane. Where the two bond coupling is (-)11.8 and the three bond couplings are 4.0 and 6.9 Hz.
CHENERY,D.H. ET AL. J.MOL.STRUCT. 26, 189 (1975)
Dr. Stan Sykora , thank you very much explanation of the splitting coupling constants. As you said, the geminal should have around (-)12 Hz and for vicinal it is around 4 Hz. In my spectra (attached below), I have found geminal as 11 Hz and for vicinal around 5 and 5.8 Hz for two multiplets.
Dr. David Stephenson , thank you very much for the reference and the coupling constants. this reference helps me to understand the conformation.
Would you please have a look to my few questions:
(a) I found that the integration values do not match in between doublets or multiplets although there is a similarity. Why this happen? (please see the given spectra with relative integration values)
(b) Does a splitting tree always have to start with a higher coupling constant (geminal)?
(c) Is there any possibility for Ha or Hb to form pentet? Please have a look to a new spectra where it looks very confusing.
Thank you very much.
I'm not sure where you are getting your spectra from (or what is the operating frequency).
I have done a simulation using -11,5.8 and 5.0 Hz and chemical shifts of 3.55,3.64,3.77 ppm (at 400 MHz). Any distortions are due to the strong couplings.
The spectrum at 200 MHz is also included now we have very strong coupling.
Dear Dr. David Stephenson , thank you very much for the simulations. Your simulated spectra at 400 MHz is same is mine (400 MHz). Does the splitting in 200 MHz give nonet, pentet and pentet from left for the protons in C2, C1 and C3 respectively?
For all my NMR measurements (attached before), I have used DMSO-d6 and 400 MHz. Only the difference in between two spectra is the glycerol from different source (Company).
Normally the integration for inner peaks are higher, I wonder why the integration values do not match in between doublets of doublet (Ha/Hb) or multiplets (Ha and Ha) !!!
Thank you very much.
The distortions are due to the coupling being of a similar magnitude to the chemical shift differences (in Hertz). The integration will still be 1:2:2 if you integrate the entire coupling pattern. Generally the inner parts of the pattern get stronger and the outer parts weaker due to mixing of the wave-functions caused by the coupling interaction. There is also magnetic inequivalence in this compound which may be showing up at 200 MHz (the spectrum is an AA'BB'C) because the two A protons are coupled to different B protons.
These distortions are called second and higher order effects, which cause the line frequencies and intensities to deviate from the generally expected but actually only approximate first order rules. In other words: frequency differences do not represent coupling constants and line intensities do not follow Pascal's triangle anymore. This rather common phenomenon is obstructing the complete and accurate interpretation of NMR spectra even at higher field. This also becomes noticeable in the frequent use of ‘multiplet’ annotations, which are essentially confessions of not fully understanding the NMR data. Fortunately NMR spectroscopy is based on a sound mathematical (quantum mechanical) theory providing the tools to solve this problem. Accurate chemical shifts and coupling constants can be obtained by calculating the NMR spectrum and iteratively minimizing the difference between the calculated and experimental one. Please let me know if you are interested in doing this yourself with your glycerol example.
I am very sorry for late reply. Dr. David Stephenson, the integration pattern is true. Are you indicating the glycerol for being an AA`BB`C system, it showed some distortion or extra peaks (Fig. uploaded before as Glycerol_4.png) which is also dependent on the magnetic field (although my spectrum was obtained at 400 MHz ). According to Dr. Hans J. Reich's (Univ of Wiscosin) analysis, this is possible.
Dr. Matthias Niemitz, thank you very much for your kind answer. I am very interested to know more about the appropriate calculation. This is very interesting that " frequency differences do not represent coupling constants and line intensities do not follow Pascal's triangle anymore". However, this is what I do normally. I would appreciate your kind help in this regard.
Thank you very much.
I recommend performing a spin simulation calculation based on the measured spectrum. For example, using MNOVA software. In the era of electromagnetism, not only chemical shifts and spin coupling constants read from apparent spectra, but also spin simulation calculations were regarded as important. In the era of superconducting magnets, it has been neglected. However, with AA'BB'C, accurate values cannot be obtained without spin simulation. It will be necessary for the research of sugar chain chemistry which is an important theme in the future.
These days NMR instruments have such a high magnetic field that it has become unusual to encounter strongly coupled protons. As such it can be surprising when they are encountered for the first time. Except for simple cases, AB pattern for example, he only reliable way to sort out the mess is to do a simulation.
Even better to use an iterative program to find the chemical shifts and couplings (is LAOCOON still a thing?)
Please review Article Quality Control of Therapeutic Peptides by 1 H NMR HiFSA Sequencing
for more information. The supporting info contains a list of software packages that have different quantum mechanical capabilities. Some of these packages also feature the iterative optimization (fitting) of chemical shifts and coupling constants to match the experimental spectrum. This method is not only superior when it comes to the complete interpretation of second and higher order spectra but also helps to resolve severe overlap, which is a challenge often also at higher field.
In publishing the attached paper, the MNOVA simulation program was used to change the chemical shift and the spin coupling constant little by little to see the spectral change. It was confirmed that if the proton signal is not more than 0.04ppm, the 600MHz device will be complicated. In the case of a sugar chain or glycan, the difficulty in analysis is higher than that of a peptide chain because the chemical shift of protons cannot be separated.
Manually changing chemical shift and coupling constant is tedious. Manually assigning (as in LAOCOON and similar programs) is tedious as well. Today's computational power enables full automation of this process.
Why? Are you also shimming manually or are you even afraid of loosing your job? Expert jobs are hardly in danger because of automation. Quite in the contrary: the more NMR automation, the more NMR spectra are produced, the more experts are needed. Wouldn't you agree that experts should spend their time more efficiently where the automation fails rather than iterating manually, which quickly becomes unfeasible in more complicated cases anyway.
Jun Uzawa , I am the one who coded the Mnova simulation package and in the past, indirectly, I have also kind of influenced what Matthias Niemitz is doing, I think (this is after all a small world and there are not so many old-timers able to do that properly). So, let's say that I am competent enough, and quite a few people are routinely using my code. And I say that yes, automation is a must today, the markets (pharmas and the likes) require it; they almost crave for it. Yet, to some extent, I understand you as well. It is fun to do the interpretations and calculations oneself, and remaine amazed when the complex calculated spectrum really perfectly matches the experimental one, despite the presence of unremovable strong couplings (it makes some of the "modern" sceptical discussions about quantum mechanics totally foolish). It is also important that some people (like you) learn this trade because it is apparently in the danger of extinction.
Yet, I also side with Matthias Niemitz and I know that he is right.
Like you, I was myself worried for quite a few years that through my work I might be making highly trained people in industrial environments "redundant" and in academic environments diseducated. Once computers do some job, who is willing to laboriously do the same, right?
But today I see that I was wrong and there was no need to worry. Automation opens new avenues to research in previously unexpected areas, and it also soon hits its own limits - the "fauna" of real molecules and their ensembles is much richer and fuller of surprizes than what we have thought and so eventually the number of people who come back and spontaneously try to (re)learn about spin systems like AA'BB'C is actually higher than in the past.
First of all, the simple answer seems to have made a misunderstanding. I am not completely denying automation. But I don't think "full autometion" will solve the difficult problems we face. In the study of glycans, we found new knowledge when conformation analysis by supercomputer and probability calculation were combined with analysis by NMR. Again, good results have been obtained when several calculation parameter settings are entered correctly. While studying this result, research has been developed by taking NMR spectra with various measurement methods. Next, now, the X-ray diffraction method (for example, Spring-8) is available if good crystals that are small enough are obtained. If it is a very small amount of sample that does not crystallize, it is the age of Mass Spectrometry. So, considering the advantages of NMR, I believe that the information of three-dimensional structure and dynamics in solution can be obtained simultaneously. In addition, NMR allows fairly detailed analysis of mixtures. And the samples (such as oligosaccharides) that change spectrum with a slight change in temperature will require manual analysis with the help of automation. In short, my current view is that proper use and combination of automation and manual is necessary.
With "full automation" I meant that all relevant chemical shifts and coupling constants can be fully automatically extracted from the corresponding NMR data in many cases but not necessarily in every single case. Therefore NMR experts will always be needed and as Stan Sykora put it: automation opens new avenues to research in previously unexpected areas and therefore is no threat to experts but rather a benefit because it makes their work more efficient = valuable. NMR spectral analysis is a combinatoric problem. In such a situation the human brain is superior in its ability of quickly reducing the number of possible solutions, whereas computers are superior in the tedious effort of going through large numbers of trial and error tests that might still be required to finally solve a problem. Therefore we probably all agree with Jun Uzawa that a combination of manual and automated approaches is necessary. So we all are on the same page. Thank you very much for this inspiring discussion!
Thank you for agreeing to my view. For further understanding, I will attach some summary of the Annual Meeting of the NMR Society of Japan and the PowerPoint used for the presentation. This is an oral presentation in Japanese that has not yet been published. Molecular Dynamics was the result of calculations using a supercomputer and was very useful. We also believe that a one-dimensional analysis method that carefully selects protons or carbon is also useful for reducing measurement time and improving digital resolution.
Thank you very much for your time. I did not expect to get help from such expert and great persons. I am extremely grateful to you Dr. David Stephenson , Dr. Stan Sykora , Dr. Jun Uzawa , and Dr. Matthias Niemitz . I will study the articles, files, and simulation you recommended. Thank you again.
You are indeed lucky to have people like Stan Sykora and Matthias Niemitz contributing - they know their stuff! My observation about "automation" is that the main approaches are mechanical and iterative to converge to a solution. Humans tend to use a very different approach which is based on pattern recognition. After enough years of looking at NMR spectra, you pick up the patterns intuitively. We also get bored with iteration very quickly!
The other thing that humans do is take advantage of their knowledge of context. There is a whole load of (often orthogonal) learning which helps humans reach a solution. Is that a likely structure? Have I seen anything like this before? Given the chemistry, is this a likely outcome?
Having been involved in the "automation" piece for many years, I still think that there is room for a radical departure from iteration (or deeper levels of theory) to pattern-matching / deep learning.
John Hollerton, you are absolutely right. I would love to have your distinguished expertise for training a deep learning algorithm but would this algorithm still work with benchtop data as well? Depending on the field strength, patterns can look very different. The nice thing about chemical shifts and coupling constants is that they are highly independent of the magnetic field. Once established with adequate care, they are practically natural constants for a given compound under given conditions. Knowing all relevant chemical shifts and coupling constants of a compound allows calculating its NMR spectrum at any field for visual and computational comparison with the experimental data. Even more important, chemical shifts and coupling constants are by far the smallest, intrinsically self-consistent set of parameters that completely describe the respective NMR data. This is a considerable advantage for computational approaches because the degree of freedom in the solution space (e.g. the number of possible assignments) increases rapidly with increasing data complexity leading to a growing number of more and more ambiguous solutions. Just an example: describing the seven lines in the NMR spectrum of an ethyl group (a triplet and a quartet) in a peak list requires 14 parameters (7 frequencies and 7 intensities) but only two chemical shifts, one coupling constant and one population (intensity). And this is only at high field (first order, no overlap). At low field you can expect to see more lines, where you might even wonder, which belong to the ethyl group and which do not. But the chemical shifts and coupling constants are still the same. Awesome, isn’t it? The advantage of such small and intrinsically self-consistent parameter sets on the stability and accuracy of computational approaches for example in the quantitation of mixtures (qNMR) is huge. Together with the field independence, such computational approaches have the potential to increase the efficiency of NMR similar to the impact FFT or cryo probes had before.
Thank Stan for the interesting question. I had met several of these problems in my career. I think that this is one of the so called " deceptively simple " spectra. The term was created by Anna Laura Segre many years ago.
The cause I believe is due to the nature of the negative sign of the geminal coupling constant. I verified often that in the TOCSY spectrum of Glycine the crosspeak was negative with respect to the other couplings. Thus in ther calculation of spin system it is easy to put the negative sign to a coupling constant but it not so easy to explain which it means physically.
Strange discussion - about automation and all that stuff - whereas the question was about the explanation of the coupling pattern and the "splitting tree" in the figure... The explanation of the splitting given there is clearly wrong; the 12 Hz splitting is due to coupling H(a)-H(b), as easily seen from the line intensities ("roof effect"), and the smaller splitting is due to coupling H(a)-H(c). Coupling between magnetically inequivalent protons is obviously too small to induce perceptible effects of higher order on the splitting pattern...
Wolfgang Baumann gives the correct answer to the original question. The discussion somewhat diverged from that and included the signal of H(c), which shows higher order effects that can only be explained by quantum mechanical spectral analysis (QMSA). It should also be mentioned that the splittings measured from the line distances in the signals of H(a) and H(b) are only first order approximations and not the actual coupling constants. The latter can only be retrieved by QMSA as well.
As others stated, 12 Hz is J_AB
The shifts of H_A and H_B should be corrected according to the rules for AB-Signals. If a strong roof effect is observed, the shifts will not be in agreement with those extracted using spectrometer at other frequencies.
The j values should be comparable.
Sebastian Jung, line distances represent J values reasonably well only in so called first-order spectra and when you know which distances represent J values and which not. But with increasing second and higher order effects line distances do NOT represent the J values anymore. They actually can diverge to an extent that the J-values are not accessible through line distances anymore even roughly. Luckily you still can get very accurate J-values in most cases by performing a classical quantum mechanical spectral analysis (QMSA).
For H_C I agree, that strong coupling is obfuscating all J values.
The AB-Signal can however be analyzed if the frequencies are corrected as here:
http://anorganik.uni-tuebingen.de/klaus/nmr/spinsystems/index.php?p=ab_gen
For J_AC there is 4 Hz measured twice. Within this accuracy there will be no changes in J. The difference in shielding of H_A an H_C is sufficient, that this can be interpreted as 1st order.
I have found data easier to calculate the multiplicities. I obtained a better shimming and I have used in Bruker TopSpin a function lb -3, gb 0.15, and after the gfp processing it rendered the file attached, where we can see Ha-Hb 11.9 Hz; Ha-Hc 6.4Hz; Hb-Hc 4.4 Hz; and almost perfectly Hc is a triple triplet of 6.4 and 4.4 Hz.
- Badges
- Science topic
- Similar topics
- Instrumentation Engineering
- Instruments