yes don. NMR is consistently giving lower values than GC. the Column used for GC is DB-5 which is typical one used for such application. 

You are injecting your reaction samples into a hot GC inlet liner and assuming that this has no effect. You need to prove that injection of the reaction mixture into the heated GC injector is not causing this. Before injecting the sample of the reaction onto the GC try quenching the sample with something that will destroy, consume or remove the catalyst. Also try mixing the reactants without the catalyst and inject them on the GC in the same manner. You might find that either your GC inlet liner is contaminated with active catalyst or the heated zone of the GC inlet is enough to force the reaction! You could also lower the temperature of the GC inlet and see if this reduces the effect you observe.

Dear John the article is definitely useful and it is something I was thinking of changing the IS to something else to see if I get any positive effects. Bruce, you are certainly right that the product aminoalcohol are not the most stable compound and on gc it may be getting destroyed early on. I will certainly check the effect however under such distructive conditions the yield should be on the lower side and nmr should give higher values. However the case is exactly opposite. It is certainly worth to change Undecane as IS though. I am thinking of mesityelne as IS for both gc and nmr to see if I have any good effect now but any other suggestions are welcome.

I am saying the product is forming when the reaction mixture is being injected into the hot inlet of the GC, that is why the yield from the GC appears to be higher. Inject the starting materials without the catalyst and see if the product forms. I think undecane is a suitable internal standard unless you have a very stable analogue of the product available to use. Instead of Mesitylene you might try hexamethylbenzene as an IS since this compound only gives one NMR signal and is higher boiling (264C). 

ah..Understood. I always inject the positive and negative control, which means, without catalyst and just the solvent. Moreover, I have injected the sample after filtration through silica gel pad to remove the catalyst (That may have leached, it is not the case). On the other hand, I kept the solution after filtration for several hours and again injected to see if there was any activity due to keeping the solution at room temperature as well as leaching effect. But no change in GC values in such experiments.  

Do you think 4-MethylBenzophenone is a good choice for NMR IS? or you never heard off?? would you suggest any other IS?

How long is the dwell time on the NMR experiment? I typically used a dwell time of 10 T1's (or more) for the longest relaxation time proton with a 90 degree pulse to get quantitative integrations. It seems to me that if the NMR results are too low then the relaxation time for the experiment should be increased dramatically to exclude this possibility. To shorten the experiment times I would use a 30 degree pulse at 6 times the longest T1. 4-MethylBenzophenone would be a good IS for the NMR you will have to deal with a more complex aromatic pattern in proton NMR and 6 signals in 13C. Are the NMR samples filtered through the silica gel pad like the GC samples?

Yeah, that is a fantastic suggestion. I did already use extending the T1 slightly like usually, it is 1 sec and I used 1.5 but did not know that we should use 10T1. I will use that today to see if there are such discrepancies coming from relaxation time. But here is another thing about NMR. I have a standard sample and I used known values for IS and standard compound to back calculate under the identical NMR conditions to see If I can get the purity/yield using such standard conditions. It is almost same (plus minus 5%). Do you think this is an indication that relaxation time may be compatible with yield calculation? I don't know about interference coming from other compounds such as undecane and by-products forming in the reaction so I think it is worth doing 10T1 experiment anyways. Will update the results later. Thank you for your help.

A lot of this variation could even be due to the NMR tube itself it all depends on the field strength. I have found that it is worth spending the extra money to have NMR tubes that were made or qualified for high field experiments. A slight variation and tube camber or wall thickness goes a long way especially in the higher field strengths, this could account for some of the variability. If the spins aren't relaxed to ground state there is no telling what the actual signal strength is and hence what the integrations should be. Also be sure to open up the acquisition window to include all the proton signals that are present including COOH (up to 14ppm) and CHO (around 10ppm) or these resonances will be folded into the rest of the spectrum making for non-quantitative integration. You need to do the TI experiments to determine the relaxation times of all the protons if not at lest use 30 seconds of dwell time for a 90 degree pulse! I typically used 11 seconds for a 30 degree pulse and got away with that for quantitative integration 95% of the time.

Something else that should be considered: Are your NMR signals for your internal standard similar in integration to your sample? Comparing the integration of each species is all relative. So, if the signal for your internal standard is 10x your integration on your sample then you can expect a bit of error.

Also, I have a similar problem when analyzing sugars by NMR vs. UPLC. The NMR is typically erroneously high when I run the reaction at high catalyst loading or high sugar concentration due to byproducts having very similar spectra causing broadening of my baseline. Make sure that you don't need to do some manual baseline smoothing to get rid of a broad byproduct peak. I find that when I do this the results are still a little higher from NMR than UPLC, but they are much closer.

Another thought: How do you prepare the sample for NMR? Is it possible to run the reaction in deuterated solvent and do NMR directly on the reaction mixture or do you have to filter off catalyst, evaporate proteo solvent, and reconstitute in deutero solvent? I have found in my work with both carbon materials and some zeolites and zerconia materials that high quality NMR spectra can be taken with these species in solution provided the sample is not magnetic and you allow the solid material to settle to the bottom.

Dear All Thank you very much for the help and suggestions. The update from my side is that I have found a little step that I may have been not doing correctly. Now I stir the CDCl3 solution for 30 mins with stirrer using almost 2 ml of solution and use 0.7 ml of that solution. Earlier, I use to use the entire material in 0.7 ml only. The efficiency of mixing may be was a little problem. however, I use to sonicate the solution instead of magnetically assisted mixing. This has helped me improvised the yield by 5-7%. But there is still a lot of improvement possible. I did the calibration curve method to minimize the error coming from some interaction of product with solvent or any other invisible things. But single point analysis and calibration curve method are not fantastically different. in fact, they give almost the same results. I have optimized the ratios of material to be used. Now the difference in the GC yield and NMR yield is about 15-18% as opposed to 25% earlier. But sadly it is not enough. :( 

Your calibration curve should be made from independent weighing's of the standard not dilution of a single stock. This eliminates the error caused by weighing the standard compound, which is not accounted for in a simple dilution scheme. For standard curves I typically weigh out 5 different amounts of standard and dilute these all the same way for the levels in the standard curve.

Dear Bruce, Yes I have done the calibration curve using three different test point using individually weighed samples externally from the purified material (well not exactly pure but I know what the other impurities are and purity is 97%+3% solvent used for reaction). 

Are you using split or splitless injections on the GC? I would suggest trying splitless injections. There might be some bias when using an internal standard like undecane as an IS which behaves very differently than the analyte compound. Perhaps try another internal standard on GC with more similarity to the analyte.

Hi Bruce and all others. I have found the source of the error just now. First of all, thank you all for your kind and useful suggestions. I learned a lot from this experience. The problem was  my catalyst. Since I am using mesoporous material, there is about 10% of material, that is impossible (?) to wash out (even after 5 times vigorous washings and sonications). Since the Internal standard (for GC) was added at the start of the reaction, in principle the ratios and hence the yield was obtained as if all material was extracted (as per GC) but in reality, only 90% of the material was theoretically obtained. This has always lead to an error of 10% between GC and NMR. I thought I would share this information to everyone so that they will be benefited if they are working on mesoporous materials, or any such things. Once again I would like to thanks everyone who helped me with this and hope in future to get the same type of response from everyone. Hope that I would be helpful to others as well if the time arise and would like to say that at this time the question is answered and hence closed.

Thanks all