Does Autotune Change Response Factor?

[jvtruebl]

I've been trying to calibrate this method for 2 years. Put briefly, I'm using PFPH to derivatize formaldhyde and thermally desorbing it into a 450 gc 200 ms. I'm using methanol as a dilution solvent for the HCHO because I want to look at nanomolar concentrations.

I put the diluted HCHO/methanol (5 microliters in total) into a glass bulb which evaporates the solution, and then passes it onto my PFPH soaked desorption tube.

First of all, the methanol reacts with the PFPH for some reason, giving a large blank size. Second of all, last night I ran 4 samples, two of them were blanks and 2 of them were 50 nanomoles of HCHO. The blanks were slightly lower than the spiked samples, so I figured I could just subtract the blank off. Today I came in hoping to replicate my results from yesterday. I autotuned the machine, and ran a 50 nanomole sample. Suddenly the response is 20 megacounts. This is roughly 5 times bigger than yesterday's response.

My question: is this change in day to day response factor due to the autotuning of my GC/MS or is it something else? I'm totally at a loss as to why I can't get the same results from one day to the next.

[orcicdejan]

OK, I never run the HCHO-PFPH reaction myself, but I'll try to give some suggestions...

1. "MeOH reacting with PFPH"
Is it possible that your MeOH contains traces of HCHO (maybe try running parallel reaction with MeOH of some other manufacturer)? Or, maybe, some oxidation occurs (MeOH -> HCHO)?

2. "The blanks were slightly lower than the spiked samples, so I figured I could just subtract the blank off". Not a good idea. Generally, your blank signal shouldn't be more than 10 % of sample.

3. Yes, tuning can change the response, due to changes in electron multiplier voltage, mass axis assignments, and, I guess, other ion optics' settings will also affect the amount of ions reaching the detector. However, I wouldn't expect 5x increased signal, unless your instrument was previously really out-of-tune (no pun intended!). We do have some problems with our MS (LC-MS/MS system), with sensitivity slowly drifting (decreasing), so we tend to use freshly tuned instrument, and *always* recalibrate after the tune. If you don't want to tune/recalibrate so often, maybe you could use some ISTD to compensate for sensitivity change (but with molecular weight close to that of your compound).

[Don_Hilton]

With your samples dissolved in methanol - what chance is there if oxidation of that methanol as the samples stand over night? And that oxidation would result in further reaction of formalehyde with PFPH?

Can you move away from methanol as a solvent?

To asses the stability of the mass spectrometer, one would use a compund that is known to be stable - like HCB or OFN.

Also, response ofhte GC/MS system can change due to changes in inlet conditions. How often do you change inlet liner and septum - and does that seem to relate to sensitivitiy changes?

[James_Ball]

If tuning the instrument changed the electron multiplier settings then it could definitely change your sensitivity. Do you autotune every day? If you do then compare the multiplier settings and counts of the base m/z to see how they compare from day to day. If they vary quite a bit it could be your problem.

If you want to just verify your tune and sensitivity each day before analyzing samples I would inject a solution of 4-Bromofluorobenzene in Methanol( maybe try 50-100ppm) to verify your mass assignments have not drifted. Much better than running the autotune each day.

[jvtruebl]

Thanks to both of you for your responses.

I wondered about the possibility of MeOH oxidizing to form HCHO, as well. But according to literature, that doesn't occur until temperatures around ~400C, and even then you need a catalyst. In fact, I've been running the same sample over a period of days to see what happens. The signal response has actually been decreasing over time. (Although, as you said that could be due to a loss of sensitivity.)

I change the septum and bakeout the system every week, so I don't think that's the problem. Right now, I'm very worried about doing another autotune, as I just created a 3 pt. calibration curve (.88 R^2, which is the best I've gotten) and I'm afraid I'll change the sensitivity of the machine.

As far as moving away from methanol: I started the experiments using Hexane, but formalin (which has methanol in it) is immiscible with Hexane. Every time I tried to take 2 ul from that solution, I got a different amount of HCHO because the solution was not very well mixed. I'm not aware of any other solvents that would work with this method, although I am relatively knew to GC/MS.

[jvtruebl]

Quick question, theoretically, if I were to autotune the machine again, but set the electron multiplier setting back to what it was pre-tune, would I retain my sensitivity to the analyte?

[James_Ball]

Quick question, theoretically, if I were to autotune the machine again, but set the electron multiplier setting back to what it was pre-tune, would I retain my sensitivity to the analyte?

It would depend on how much the other tune settings changed.

Depending on which instrument you are using it can change several settings simply by running the autotune procedure. With the Agilent systems the autotune can change Electron Multiplier, Repeller, Ion Focus, Entrance Lens, Entrance Lens Offset, and peak widths ( controlled by AMU Gain and AMU Offset). It would hold the filament, Emission Current, and Electron Voltage the same while varying the others. Not sure what other manufacturers do when autotuning.

[orcicdejan]

@jvtruebl

I wouldn't dismiss oxidation so easily. What is your sample? Are there any oxidants that could be expected?
As for signal decrease, depending on nature of the sample, polymerization or further oxidation might occur.

3 point calibration curve, and r squared of 0.88? I would suggest adding 2 more points, and (especially since you're doing derivatization, so you're introducing additional errors) preparing each in e.g. triplicate. Otherwise, I'm not sure how reliable your results will be...

As for the solvents, is there any method for MeOH purification (formaldehyde removal)? Hexane is a bit too non-polar solvent, have you tried with somethin more polar, say, dichloromethane? Or some alcohol less probable to contain formaldehyde, e.g. EtOH or iPrOH?

[jvtruebl]

My sample is pure methanol mixed with formalin. The solution is placed inside of a 250ml glass bulb and ultra pure nitrogen is passed over it. The glass sampling bulb is held at 90C to aid in the volatilization. Aside from that, there is some hexane and of course the derivatizing agent PFPH on the Tenax. I don't know enough to say if that woudl cause oxidation, but maybe someone here does?

I was assuming polymerization was occurring, as well. The only reason I was hesitant to believe that was because my formaldehyde solution is in methanol, which isn't that what is used to stabilize formalin in the first place?

My three point calibration curve has each concentration injected in triplicate, but not each concentration prepared in triplicate. Does that make a difference?

I haven't tried dichloromethane..perhaps I'll give it a look. Is there any other reason methanol would be interfering so much?

[orcicdejan]

"My three point calibration curve has each concentration injected in triplicate, but not each concentration prepared in triplicate. Does that make a difference?"

Oh, yes. The purpose of replicates is compensation of random errors (those we usually express by standard deviation). If the injection is the main source of such errors, then repeated injection is OK. However, I would expect the derivatization to be the critical step (i.e. the step giving the highest contribution to the total variance), thus the need for sample prep in triplicate. I know it's more time & reagent consuming, but could really improve your calibration.

By the way, have you tried running the same reaction in parallel, using MeOH from some other manufacturer (or, at least, of different quality)? There is still a possibility your MeOH contains traces of HCHO. Even the highest-quality solvents come with some impurities.

[jvtruebl]

Oh, yes. The purpose of replicates is compensation of random errors (those we usually express by standard deviation). If the injection is the main source of such errors, then repeated injection is OK. However, I would expect the derivatization to be the critical step (i.e. the step giving the highest contribution to the total variance), thus the need for sample prep in triplicate. I know it's more time & reagent consuming, but could really improve your calibration.

By the way, have you tried running the same reaction in parallel, using MeOH from some other manufacturer (or, at least, of different quality)? There is still a possibility your MeOH contains traces of HCHO. Even the highest-quality solvents come with some impurities.

Hmm..Just to be clear, the derivatization does not occur until AFTER it has been swept onto the desorption tube. So to restate, I make a .03% solution of formaldehyde in methanol. I then inject 1ul of that into my glass bulb, where it volatilizes and is swept onto the tube. This is where derivatization happens. Then I run it through thermal desorption/gc/ms. I do that 2 other times, using the same .03% formaldehyde-methanol solution. So unless I'm just really bad at making solutions (which is possible), I don't see why I would need to make 3 solutions as well. Correct me if I'm wrong though!

I can definitely try some other methanol. I am home now, but before I left I tried a sample blank with DCM instead of methanol, perhaps it will make less of an interference. One of the things that is really peculiar to me is that all the literature that uses this method cites the use of methanol of hexane as solvents. Granted, when calibrating, they allow the derivatization to occur in liquid phase, and THEN spike the desorption tube....I also double checked the GC/MS system conditions and everything looks fine. You were right though, some of my old runs had an electron multiplier voltage either higher or lower than what it's at now, which could explain the variance then. But I still have no way of explaining the current lack of precision..


EDIT- I should also mention I'm using a non-polar column. I've heard methanol is really tricky with nonpolar columns because instead of forming a nice slick, it elutes in globs due to insolubility. If you have any thoughts on that theory, I'd be glad to hear them.

[orcicdejan]

My bad. If the reaction occurs after the injection, then repeated injections (instead of separately prepared samples) are OK. Have you tried analyzing additional calibration levels, just to be sure whether your points deviate due to random errors or non-linearity?

[jvtruebl]

No worries. I was just making sure. So the DCM run gave an even bigger interference than the Methanol.

I've decided to attempt calibration at a higher concentration, somewhere around 75 nanomoles instead of the desired 25.

Since I'm using a tenax desorption tube, do you know of any kind of compound I can use that will test the sensitivity of the system? It's been a few days since I injction my 75nm concentrations, and I want to inject some 100nm concentrations now, but am not sure if the sensitivity of the machine has gone down.

[Don_Hilton]

- For reproducability of the instrument - go back to a liquid injection of a compund that will not change in solution. Is the instrument set up such that you can make a liquid injection through a septum wiht the desorption device installed? If so, a solution of HCB, HFN, or BFB can be used to confirm signal strength for a known quantity of material introduced to the GC. If you can not make a liquid injection - then add the 1 microliter of sample into a desorbtiontube and then, with minimum time to allow evaporative losses, get the sample onto the instrument. Divide the world in half and determine if the problem is in the left half or the right half. Then take the half with the problem and cut that in half again... Isolate the problem. (We do that with analytes!)
- keep in mind that part of the objective of tuining is maintanance of consistancy in the instrument. If your instrument has a step in the tune that adjusts the electron multiplier to give a specific number of counts for the tune compund, there is a step that should make the responsivness of the instrument to be close from one tune to the next. However this is for the tune compund and other compunds will be affected differently - and this works for instruments where the source of the tune compund is effectively constant (such as introduced as vapors off of liquid,thorugh a frit).
- The sensitivity of the instrument is affected by all of the factors adjusted. So a tune followed by an adjustement of EM voltage will not necesarily bring the instrument back to the same sensitivity.
-As far as methanol and a non-polar column. If in the adsorbtion process the methanol is volatialized and vented, there will not be enough to become a problem. If the methanol is trapped in the desorbtion tube, you need to have the column above the boiling point of methanol at the start of the run. The problem with methanol is that f it condenses on the head of the column it soes not form a band that will focus analytes. It forms beads that spread along the column, defocusing the analytes.
- Reproducability of evaporation and transfer. Can you place the 5 uL of solution directly into the desorption tube and then draw the gas nitrogen through? This eliminates the question of losses to the large flask used for evaporation of the sample.
- And, are the desorption tubes home made? Do you use one for several injections or change with each injection? Aplication of the liquid allows the reaction to occur in liquid phase - a place where we like to run reactions. If you are reacting vapors against a solid phase, this can be a more difficult place to run a reaction, flow rate and surface characteristics that would not be important for a drop of liquid sitting in the desorption tube become important.