GCMS Sample concentration - How much is too much?

[Tarapan]

Hi all,

Our company just purchased a Shimadzu QP2010 Plus GCMS with split/splitless injector and a DB5-MS+DG Column (30m(L)×0.25mm(ID)×0.25mcm(film).

Part of our project involves identifying unknowns peaks with MS library. during training the instructor explained we should inject a high concentration of sample in order to have a fairly good identification match. Ive kept asking her How much is TOO much but she kept dodging the question or saying she will ask her company. literatures and internet didnt help either as they only talk about their lowest concentration they can detect.

Could anyone give us an advice on whats the maximum concentration of sample we should use for GCMS?

[thohry]

Hi all,

Our company just purchased a Shimadzu QP2010 Plus GCMS with split/splitless injector and a DB5-MS+DG Column (30m(L)×0.25mm(ID)×0.25mcm(film).

Part of our project involves identifying unknowns peaks with MS library. during training the instructor explained we should inject a high concentration of sample in order to have a fairly good identification match. Ive kept asking her How much is TOO much but she kept dodging the question or saying she will ask her company. literatures and internet didnt help either as they only talk about their lowest concentration they can detect.

Could anyone give us an advice on whats the maximum concentration of sample we should use for GCMS?

If you inject into the column too much, it may degrade the MSD. My experience is just inject 1ul of 200 - 1000ppm solution in split mode.

[Peter Apps]

"we should inject a high concentration of sample in order to have a fairly good identification match"

Sounds slightly odd. As long as the signal"noise ratio (i.e. peak height vs background) is high enough, the identifications do not get any better as the peak gets bigger. With both Agilent and Varian MSs I've had clean spectra give very close matches with fractions of a ng in the peak (from real samples, not test mixes) as long as the peak was on a clean baseline and/or I used baseline subtract to get rid of the background. If you have a complex sample with peak overlaps the bigger peaks will usually have cleaner spectra and better matches, but increasing the sample concentration increases the sizes of the interfering peaks so does not improve the matches.

As sample concentration increases, at some point the biggest peaks in the mass spectrum will be outside the linear range of the MS, and match quality will degrade.

Also, with more concentrated samples you get more concentration overloading of the column, more co-elution, dirtier spectra and poorer matches.

Peter

[Ron]

10 ng on column is plenty for a good mass spectrum. 100 ng is not too much, but is more than is needed. If you get up to 200 ng on column you are probably going to overload the column and saturate the detector. I have generated mass spectra that give good library matches at 20 pg on column when I had a very clean matrix.

Tuning is very important for library matching. You should be able to adjust the mass pattern during tuning. The ion ratios I like to use during tuning are

69 100%
131 45%
219 55%
414 2.5%
502 1.5%
614 0.6%

Your software may or may not allow setting all these ratios. The ratios above have worked for me, that does mean that this is the final answer.

[jh1]

Since your GC has a split injector you can always inject 1 uL of sample and if your peaks look horrible (due to overloading the column) then increase the split ratio in your method.

[Tarapan]

Peter Apps.
We've tested actual samples on our machines and for some reason with the same samples the library match does gets lower at lower concentrations (range of 1ng or less on column). But I suspect that our matrix is dirty and affects the quality of our peak spectra as you've explained.

Jh1
Column overloading and Detector Saturation are the two things we wish to avoid completely whenever possible and we plan to do this is by adjusting our "total" sample concentration.

which brings up another question concerning the detector signal on the TIC, since this project involves dealing with samples of unknown content with unknown concentration any advice whats the max intensity that we are allowed to have for each peak signal??

[Peter Apps]

Hi Tarapan

Are you doing baseline subtraction of any kind (different softwares call this different things) ?, and are your peaks resolved from their neighbours ?

Peter

[CE Instruments]

The maxium detectable injectable amount will be different for each compound and may or may not be greater than the amount required to cause peak fronting/tailing on your column. No one quotes maximum identifiable amount as the dynamic range of MS detectors is less than FID (typically only about 3 orders of magnitude) and almost everyone is interested in trace levels.
You too are interested in Trace levels and as you are looking for unknowns the suggestion is to run the most concentrated samples applicable. This is not quite the right idea as you need to run the samples that allow you to get a good detectable spectra for the lowest concentration required by your method for the unknowns in the sample. I think you should therefore choose the concentration of the sample to get between 10 and 100 times the manufacturer quoted specs for full scan (1pg on column s/n 10:1 ?? of OFN) You need to get enough of the lowest concentration of your least sensitive compounds onto the column to allow full scan identification. If this results in peak overload and poor spectra for the more concentrated peaks you can always run again at a greater dilution

[orcicdejan]

And don't forget one thing: the high match is not a guarantee that you have identified your compound, nor the low match necessary means your cmpd and the library entry are different. It's simply a mathematical comparison of your spectrum (with/without your baseline, with your tune etc.) and someone else's (again, with same comment).

But, basically, the intensity of peaks in your spectrum should be high enough to be distinguished from the noise, and not to be chopped of by the threshold, so as to have enough peaks to 'feed' your library search algorithm.

If you don't use only MS but some other information (e.g. retention) too, even worse spectra can be sufficient (if your chromatography peak has several MS peaks and the same retention time as the standard, it's quite reliable identification).

[Tarapan]

Hi Tarapan

Are you doing baseline subtraction of any kind (different softwares call this different things) ?, and are your peaks resolved from their neighbours ?

Peter

Hi peter
Yup I did background subtraction using the software. I can differentiate the peak from the background very clearly and there doesnt seem to be any underlying peaks.

Orcicdeja>
retention time is definitely one of the other parameters we will be using for ID. but from what i read so far retention index seems to be a better choice.

[Peter Apps]

Hi Tarapan

If you are substracting the baseline from a resolved peak I am surprised to hear that match quality declines from 1 ng downwards. I've never used a Shimadzu MS or software, maybe there are some tricks that you could try. Is the baseline substraction an automatic function or are you doing it manually peak by peak ? How about averaging the spectra over a few scans on either side of the peak apex ?

Peter

[orcicdejan]

Tarapan>
Well, retention indices provide you some flexibility (if you change chromatographic parameters, while keeping the same column type, your retention indices will remain the same, while retention times might change) and comparability with literature data, if available (e.g. Kovats indices for essential oils components). Also, the long-term changes in retention times (due to column degradation or trimming) will be compensated. So, in general, indices are better but, depending on your needs, retention times might suffice.

Peter Apps>
If the MS peaks intensity falls below threshold (or whatever this parameter is called at Shimadzu - I use Agilent), some ions will not be recorded, thus lowering the total score i.e. match. But, at <=1 ng? It is a bit strange. Bad tune / way too high threshold / something else significantly decreasing sensitivity?

[Tarapan]

Oricdejan>

What about thermal degradations? can that effect the match quality as well??

[orcicdejan]

Tarapan>

If thermal degradation occurs with your sample, but haven't occurred (or had, but to a lesser/higher extent) with the sample used for library entry creation, yes, it might influence the match. Depends on whether differences are great, and also on mass peaks 'importance' (if I recall correctly, not all the peaks are used for sample-vs-library comparison, nor all of them have the same weighting coefficients).

[Don_Hilton]

You may be able to export chromatograms to AMDIS (available from www.NIST.GOV) and from this program get deconvoluted spectra with background removed. This helps take care of coelutions and some background noise. It does not help with saturated masses resulting from too much material entering the mass spec. I would suggest that you inject at a concentration that seems to be appropriate and display spectra through the larger peaks. If it looks like the lower intensity peaks rise in relative abundance toward the center of the peak, the haier abundance masses are saturated and the sample must be run again, more dilute or with a higer split rate. Plan on multiple injections.