35 A good review has recently been published in European Journal of Organic Chemistry (well, five months ago) in which structural elucidation with NMR spectroscopy is discussed. (DOI: 10.1002/ejoc.200700966)
While X-Ray crystallography often provides the “last word” on absolute stereochemistry and atom placement (though one can easily confuse functional groups like amides and esters or, far more rarely, isocyanates and thiocyanates) a far easier method is NMR spectroscopy as sample preparation is infinitely easier and interpretation of data can be done with only a modicum of knowledge. The problem most people have is a rather infantile approach to determining which experiments to run and, even more, what experiments are out there. It is, for instance, completely possible to determine structures as complex as vitamin B12 with NMR and, using total synthesis, confirm your structural hypothesis (and growing a crystal of something isolated in miligram quantities is frustrating if at all possible.)
So, behold, Eugene Kwan and Shaw Huang have written an easily understood microreview to help the student through a structural predicament.
The general procedure is this: skeletal connectivity is deduced by homonuclear and heteronuclear correlation spectroscopy (COSY and then HSQC or HMBC) followed by stereochemical determination by NOE correlations between protons – the provided coupling constants will then guide you to your overall structure.
I have been thinking about producing a series of posts on this review with a more comprehensive description but I’m hesitant for two reasons – firstly, the review is well written and I would just be needlessly reproducing something that could just as easily be read, granted with fewer liberties in curse words. Secondly, I am not an expert at structural elucidation with NMR spectroscopy and wouldn’t do the subject justice, but I’ll tell you what I typically run when I have a “problem” compound.
Firstly, I determine the molecular formula via high-res Mass Spec or, if that fails, elemental analysis. Then, I run -not walk- to the NMR, usually the 600, and drop in a sample. It’s a Varian, so I have to curse it for a few seconds. I then set up seven experiments to be run sequentially. The Varian does this for me, which is the least the piece of shit could do. I come back about 24 hours later and all 7 experiments are pretty much done, assuming the C13 doesn’t take more than two or three hours.
We should be familiar with 1-3. TOCSY is nice because it gives correlations to protons in the same spin system, which help distinguish individual rings or alkyl chains. HSQC and HMBC are HetCorr experiments which provide C-H and C-C-H connectivities respectively. Both of these are very useful if you’re unsure which fragment of your molecule a particular substituent is on OR if you’re unsure if a proton is connected to a carbon at all (in my case, amides give me no indication that they are attached to a heteroatom, though I or anyone in my position could do an exchange experiment with D2O to resolve most of those problems – but that could cause problems too, such as hydrolysis of sensitive compounds and exchange of acidic C-H compounds). Taken together, these 7 experiments should be enough to give at least the skeletal arrangement of even a significantly complex small molecule (as oppose to proteins).
Solving the structure usually starts with an analysis of spectra 1, 2, 6 and 7, which gives me a rough idea of connectivity. After I get a skeleton sketched out, I can use 3, 4 and 5 to give good assignments for the 1D spectra and make conclusions about connectivity. I then celebrate and hope no one proves me wrong with a crystal structure.
Keep in mind that doing these experiments aren’t my normal routine since 99.9% of the time I know what I’m making and I would hope you do too. The spectra of a few compounds, however, are clearly unanticipated or unclear or difficult to assign, so at the very least a second experiment is done so my assignments on my 1D NMR spectra aren’t just guesses based off ChemDraw.
I suggest anyone who seriously is going to publish a spectrum with assigned peaks to do at least one or more of these experiments to make sure your assignments are correct.
Eugene E. Kwan, Shaw G. Huang (2008). Structural Elucidation with NMR Spectroscopy: Practical Strategies for Organic Chemists European Journal of Organic Chemistry, 2008 (16), 2671-2688 DOI: 10.1002/ejoc.200700966
Back to top
I have to admit that I am a bit rusty in the NMR interpretation skillz. Not only does the NMR dept acquire the spectra for us, but they interpret it as well
For a good list of experiments, complete with example spectra, look at Acorn.com.
ACORN? OMG Milo is a terrorist!
No… just a socialist… I want to redistribute ALL your smarts…er… I mean cash….
That should be Acornnmr.com….
Our NMR tech’s favourite job is cursing Varian. The software is so buggy at the moment.
Thanks for the heads up on this paper.
5. ROSEY -> ROESY?
Thats a good list of experiments if you can get the spectrometer time. I would swap TOCSY and ROESY in order of importance, but it depends on your molecules.
Also, there are rumours that Varian’s next processing software version will be Windows based. They must enjoy making problems for their customers.
Windows? Seriously?! Oh fucking gross. How do you convert a magnet for Bruker probes?
I always liked the *nix environments for NMR software, it made me feel a little bit smarter than the average person…. especially the old GE I used that ran on a Sparc Station.
Why ROESY and not NOESY? As I understand it the results are pretty much the same for small molecules (with short correlation times), but NOESY is twice as sensitive as ROESY (because of some coherence pathway stuff). Or am I remembering something wrong here?
For intermediate-sized molecules (1000-2000 Da), the NOE might have zero intensity (small molecules give positive NOEs relative to the diagonal in NOESY while ROESY gives negative NOEs and the intensity decreases to zero in the 1000-2000 Da region). ROESY corrects this by putting the NOEs in a rotating frame.
I’m not following you. Are you saying that what Christian is saying is the opposite of true? ROESY is more sensitive than NOESY?
It just depends on the conditions and the analyte, like Sparky said. I don’t think one is inherently superior to the other, but ROESY does work better in some situations. ROESY tends to have more artifacts than NOESY, thus the “more sensitive” comment.
This has to do with molecular tumbling, correlation times, and the like. The confusing thing is that the ROESY pulse sequence and the TOCSY pulse sequence are idemtical, and I never understood why.
I’ve only ever done peptides (14-37 aa) and NOESY was par for the course.
crap, next time I’ll read all the comments first before chiming in.
Excimer has this pretty much right. I would only add that this is a function of rotational correlation time, not size per se, so altering your temperature or solvent viscosity can conceivably bring you into a better sensitivity range for either experiment.
NOESY may give the same or equally useful data, I haven’t tried it, but I’m working with molecules around or above 1500 daltons, so I use ROESY. See above for why.
you should replace C13 with an APT, same amount of time, more information
Same could be said for an HSQC with edited multiplicities.
I envy your situation where you can get ~24hours on a 600 in a row, both financially and instrument time-wise. I DO believe that NMR structural assignment is an underutilized technique.
Growing crystals is an art and CAN be difficult with just mg of some greasy compound. However, in many routine applications you can get the entire 3D structure in ~1-2 hours for (at least at my institution and my former (graduate) institution) far less $$ than 24hrs of 600 MHz NMR time. FAR less. Especially if you can throw a p-bromophenyl ester or whatever on the molecule.
I’m just trying to say that every experiment has to be planned carefully, something which is often ignored in analytical experiments like NMR, x-ray, MS, etc. Far, far too often I see scientists using techniques by rote. (is that your molecule in the HRMS or a H-bonded dimer/hydrate/whatever of starting material, did you close to form the heterocycle or just fragment the oligomer/polymer and only see the monomer as it breaks up, etc etc etc).
A day on the 600 here will cost about $110. A full crystal on one of the X-Ray instruments costs about $250 which is a flat fee regardless of how long you diffract – there’s an additional cost of $200 to have someone solve it for you. (Getting unit cells and whatnot is free, consequently.) I’ve never run a crystal for less than 18 hours – I just don’t see why you wouldn’t want as much redundancy as you can get. We have a great setup for both X-Ray and NMR, and if I had the crystals I would certainly get an X-Ray structure BUT, even with the latest software from Bruker (we beta test the software for them for their crystal data solution programs) I’ve had exactly one crystal “solve” with autostructure. If there’s any disorder in solvent or multiple occupancy, not even the most advanced Bruker software solves it for shit. Therefore, most labs here are going to pay up about $450 for a single crystal structure making the 600 easily the cheapest method for structural determination.
I don’t really know how it works at other places. I never submitted crystals at my undergrad and the highest field instrument we had there was a 500 and it was always booked. I think the 400s there would have cost about $50 a day but it would have been impossible to get 24 hours on one of them without a month’s notice since you could only run long experiments on Sundays.
Your NMR rates are fantastic! I’m envious. Our NMR rates are higher (but lower overnight, so a 12-hour battery of NMR experiments would be fine). Our x-ray rates are collection time-based. NMR really is best at a lot of problems (aka “sporting methods”). I’m still shocked that your rates are so low AND your machines are relatively free like that. Sweeeeeet.
As for x-ray I can perform my own structural determinations, given an instrument, so the only fees are for instrument time. Auto-structure and “hitting enter” in x-ray solution programs are exceptionally dangerous methods of “data analysis”.
Another random x-ray thing: A MoO (multiplicity of observation) of ~10 is PLENTY for most small-molecule data acquisitions. 5 is good enough in many cases if you don’t really care about having Fourier-truncation artifacts by heavy atoms.
Varian threatening Wincrap could be in response to Brucker’s triumphant diagonal arrow key. One would need pour an NMR tube’s contents into the probe to get the NMR lady all screaming ballistic and such (U/Victoria undergrad lab). Hitting the Brucker keyboard diagonal arrow accomplishes much the same endpoint without losing your sample.
Crystal structures can be exactly elegantly wrong for receptor sites and catalysts. NMR structure determination outputs configuration but only averaged solution conformation. NMR structures are never exact, though real world useful. Ask yourself, punk, do you want a publication or a patent?
I’d like to see a poll for who prefers Bruker vs Varian, especially if they’ve worked with both. Brukers always worked great for me, while the Varians I use now crap out about a quarter of the time through some random software problem.
My two cents….
Bruker – Never had a problem running solution state spectra on Brukers, whether for small molecules or proteins. Then again, I wasn’t doing anything unusual, all of the pulse programs I used were already installed and ready to go. Basically ran everything at room temperature or a bit below, no crazy temperature dependence studies.
Only used a Bruker solid state spectrometer twice for a protein sample, again, no problems (although the experiments turned out to be a bit more of a wash than I had hoped due to other reasons). I really loved the air-driven rotor handling setup (didn’t need to slide the probe out to put a new sample in – heaven!).
Varian – never used a solution state spectrometer from Varian, but have a good bit of experience using solid state spectrometers (my old lab has three dedicated Varian ssNMR spectrometers). I will say that most of my grief in grad school was due to these spectrometers (one in particular), but I suspect it was mostly due to the fact I would spend weeks at a time on the spectrometer, and therefore I got to see it through good times and bad. When it was working well, everything ran smoothly, my standards looked pretty, and my sample data was decent. Other times, it could have been the VT stack that got jostled out of place, or one of the channels was fluky due to a bad connection in the probe, or an amplifier wasn’t giving me the RF fields I needed, or I had to debug a pulse program I wrote (ugh). And I had to take the damn probe out to change samples, so if I was in the colder regime of a temperature-dependence study, I’d have warm the probe and bore back up so I could avoid condensation. I would occasionally have software problems, but not as often as it sounds like the above commenter does.
To be fair, though, some of our probes were getting a little up there or infrequently used (I think I was the first person to even attempt deuterium NMR on this one probe since the late Clinton Administration, didn’t pan out since I couldn’t get the capacitors set up just right in time before I had to graduate) and we didn’t have a full-time staff engineer like other NMR labs, so we had to make do with in-lab repairs and occasionally sending things to outside vendors when it got too involved for us. The results from studies using Varian’s new probes (the BioMAS and Fast/VeryFastMAS ones) look fantastic though, from what I’ve seen, and I’m sure are an improvement over the T3 and APEX probes I’ve mostly used.
Anyway.
1. Not sure how the newer versions are, but I run ChemDraw Ultra v. 7.0, and the 1H-NMR predictions are terrible. 13C-NMR predictions are much more accurate.
2. We have a 600, and it’s reserved for biochemists (I think there are 10 in our department, and only 1 or 2 run NMR). Us organikers are limited to the 500 (at best) and it will only run 1 experiment; I wish it could run 7. You can usually get reasonably good 13C w/ 1000 scans (~45 min) in 10-20 mg of compound on the 500.
3. I’m a fan of running individual NOE difference experiments because you can line them up over the 1D spectrum to see which signals “pop up.”
Keep in mind, NMR is good for organic molecules. Not everyone is an organic chemist, and hence NMR isnt the end all be all technique. I prefer x-ray crystallography myself.
Yay for crystal structures, that’s the gold standard in my (inorganic) book.
Thanks for the nice comments about our article. I wanted to make a few points. If you’re faced with a complicated structure, don’t bother getting a 1D 13C until you’re sure you need it. It wastes a ton of spectrometer time, and you hardly get any useful information you couldn’t get elsewhere. Also, run HSQC/HMBC before bothering with NOESY/ROESY, TOCSY, etc. There’s no point in trying to assign stereochemistry or selectively irradiate things before you even know which signal is what. COSY is heavily overused because people are just familiar with it. If you’re going to run it, use COSY-45 with a decent number of increments, and don’t symmetrize it. This will clean up the diagonal and sometimes, you’ll be able to tell which peaks are vicinal or geminal couplings from which way they’re slanted. Also, if you see a peak appear only on one side of the diagonal, it’s a clue that it might be a long-range coupling.
I don’t pretend NMR will solve your every problem in life. It’s just a cool technique. I find that in a lot of cases, the OEM NMR software is very unfriendly. You should consider using the software produced by ACD Labs. It’s sort of expensive, but it’s really easy to use, does a good job, and has some handy features like automatic processing, kinetics capabilities, and converting a 1D 1H into journal report format.
All good points. I run COSY-45, not the old fashioned kind, and I always obtain a C13 because it’s something that must go into my thesis for every novel compound I create per PI policy but I have yet to find it necessary for structural determination when I have all the 2D data handy.
I can’t comment on OEM NMR software since there are very few data stations to work on that aren’t connected to an instrument. Our department has a site license for ACD labs, so everyone has a copy of version 10 or 11 installed on their laptops or desktops which means NMR processing is always done away from the instruments and out of the NMR lab. So… I really don’t know how to manipulate the spectra much on the OEM software other than putting peaks in and integrating and phasing…
ACD labs software is great, but maybe better for pharma or high throughput applications due to the cost. For a moderate organic lab, an old mac and iNMR can be better and only cost the same as some new packets of tubes. The new version is also available at a reduced price for October.
I would suggest it is perfectly acceptable to quote C13 chemical shifts from 2D correlation spectra. If anything, I should think a 2D spectrum would give more detailed information than a 1D one, so it should be considered better, assuming it’s of good quality.
also – dqf cosy has a significant advantage over cosy in that sign of the signal is conserved, so identifying multiplets becomes a lot easier b/c the alternating colors… it takes more time to run the sample, but saves so much time in interpreting
DQF-COSY is really for determining H,H coupling constants. It’s fine if you have a decent amount of sample (it’s less sensitive than COSY by a factor of 2) and quite a lot of time. You need a lot of time because you need a lot of increments to get enough resolution in T1 to avoid the cancellation of adjacent signals. There are published procedures on how to narrow the spectral window and computationally remove the aliased signals. In principle, that would let you “zoom in” on a particular multiplet to get its couplings, but it’s complicateed in practice. For most applications, it is unnecessary. It’s way easier to use some other technique, like 1D-TOCSY, to get the couplings. However, one case where it might come in handy is where the spectrum has a lot of singlets in it arising from methyl groups. They don’t show up in DQF-COSY, so they won’t swamp the spectrum.
Given your complaints related to a lack of discussion about chemistry articles, it appears you should write about characterization techniques if you’re jonesing for some threads.
So it would seem!