[curiouscat]

In trying to identify well resolved unknown peaks on a liquid sample GC (mostly hydrocarbon + O + Cl) what are the relative merits of GC-MS versus NMR? I do have some idea of the expected class of molecular targets.

Does NMR always need pure cuts to be prepared? I mean, in the case of a GCMS the GC does this separation job; what's the analogous separative stage for a NMR?

In terms of ease of interpretation or uniqueness of a hit is there a tradeoff? How does one select GCMS vs NMR for general sample identification work.

[Archer]

NMR does not have to be pure, it is very helpful if it is but provided that the impurities do not interferre with the sample peaks it is not a problem.

Usually it is prudent to run GC-MS first as this provides an idea of purity and some idea of what the impurities are.

I don't have an NMR but I do have access to one, I can usually get by with GCMS alone provided that reference materials are available.

Aliphatic hydrocarbons are notoriously tiresome by ¹H NMR, particualrly if they are long chains (even with 500MHz instruments). All Upfield signals, all in the same area. If your impurities are all similar to your compound of interest then NMR may not give you the information you need (unless you are looking for a specific diagnostic peak which is upfield from the C-H's).

Analogous stage would be preparative chromatography. The nice thing about GCMS is you only a tiny sample whereas NMR you need (ball park) 1mg / C of sample for a decent ¹³C spectrum.

There are LC-NMR systems available but these require expensive mobile phases and only provide ¹H NMR data.

[curiouscat]

It might be relevant that a lot of the targets of interest are in the 0.1% to 5% range.

I guess not a major constraint since I could always flash out some lights to concentrate the molecules of interest, within limits.

[discodermolide]

Usually for NMR one uses pure individual components. Purified by chromatography or crystallisation.
GC-MS gives you the MWt. of each component but no structural info. as is the case for NMR.
Some GC-MS packages come with a large compound database so it may spit a structure out at you.
Ideally one uses them in tandem.
For general identification the GC-MS with database may be an option for you.

[curiouscat]

@Archer: Thanks for the advice!

Sample quantity, luckily, is not a problem. We've plenty of sample.

Pure cuts are another story. Those are not available.

Reference samples are also hard to get. We only have access to a few.

When you write that for an NMR " impurities do not interfere with the sample peaks" how does one quantify interference. In the sense, if my sample shows 5 unknown peaks on a GC at the 1% range, is it feasible to inject a unseperated sample into a NMR and expect identification of all?

[curiouscat]

@disco: Thanks!

GC-MS gives you the MWt. of each component but no structural info. as is the case for NMR.

In addition the the MW, doesn't the m/z peak pattern count as structural info?

Or is it too close to call?

Our experience with " compound databases" in the past hasn't been great. They had too many spurious matches.

[Archer]

When you write that for an NMR " impurities do not interfere with the sample peaks" how does one quantify interference. In the sense, if my sample shows 5 unknown peaks on a GC at the 1% range, is it feasible to inject a unseperated sample into a NMR and expect identification of all?

Non-interferring peaks means that they are resolved from the peaks of interest.

1% should not be too much of an issue even if they are not resolved even 5% in total may not cause you too many issues. Sometimes it is best to run it and find out.

In addition the the MW, doesn't the m/z peak pattern count as structural info?

It will give some structural info (fragmentation patterns etc) but it is of limited use in, for example, determining positional isomers on aramatic rings etc.

[DrCMS]

When you write that for an NMR " impurities do not interfere with the sample peaks" how does one quantify interference. In the sense, if my sample shows 5 unknown peaks on a GC at the 1% range, is it feasible to inject a unseperated sample into a NMR and expect identification of all?

No.  If you have a sample made up from many different materials all at low levels the NMR will just be a mess of peaks. 

I thing for your application you either go for GC-MS (you may well need accurate mass MS or GC-MS-MS to get sufficient structural info to assign each compound without a library fragmentation match) or you attempt some other separation technique before giving purer cuts a go by NMR. 

[curiouscat]

Thanks @DrCMS.

Out of curiosity, has a GC-NMR never been an instrumental possibility? Giving the same separative advantages a GC gives an MS?

[discodermolide]

@disco: Thanks!

GC-MS gives you the MWt. of each component but no structural info. as is the case for NMR.

In addition the the MW, doesn't the m/z peak pattern count as structural info?

Or is it too close to call?

Our experience with " compound databases" in the past hasn't been great. They had too many spurious matches.

The ionisation pattern will point you in a direction leading towards the structure, but it will not give you all you need. NMR is the tool for that, combine the two and you should get what you want to know.

[discodermolide]

Thanks @DrCMS.

Out of curiosity, has a GC-NMR never been an instrumental possibility? Giving the same separative advantages a GC gives an MS?

GC-NMR good idea, but I think the sample amounts would be too small for the NMR. LC-NMR was also available, but I'm not sure of its status these days.

[curiouscat]

In terms of functional group identification, what's a good instrumentation strategy.

i.e. Say I've a 2% by product peak on the GC and it may or may not contain a Cl and I want to know if it does. Or similarly to know if it has an OH group or not at all.

Is there any preferred strategy here? In this case, not aiming for full structure elucidation but only pragmatically some smaller goals.

Does FT-IR help? Or is GC-MS / NMR assisted full structure resolution the only way to go.

[discodermolide]

GC-MS will certainly provide the isotopic content, that's characteristic of each element, like Br or Cl.
Depending upon ionisation mode differentiation between functional groups may be difficult. I still say you need an NMR somewhere in the picture for accurate functional group identification. IR is rather limited for various carbonyls, OH and aromatics for C-Cl or C-F I don't think IR gets you very far.
.

[Archer]

When you write that for an NMR " impurities do not interfere with the sample peaks" how does one quantify interference. In the sense, if my sample shows 5 unknown peaks on a GC at the 1% range, is it feasible to inject a unseperated sample into a NMR and expect identification of all?

No.  If you have a sample made up from many different materials all at low levels the NMR will just be a mess of peaks. 

I disagree, I have run crude samples, the proton may have a lot of baseline material but the major compound will be 95% of the sample. This may still provide useful data, particularly in HSQC and HMBC experiments where the attenuation can be adjusted to eliminate the small peaks.

It is all very dependent on your substrate and the related impurities though.

[Corribus]

Out of curiosity, has a GC-NMR never been an instrumental possibility? Giving the same separative advantages a GC gives an MS?

I would think NMR would be too slow as a back-end detector for GC, and LC would be problematic because you'd need a large volume of deuterated eluent.  (Plus all the engineering problems with the need for a rapidly spinning sample.)