Peak response irreproducibility - 7890A/5975C - LVI/PTV

[cjm]

I am hoping there's someone here who has experience with large volume injections using a programmable temperature vaporization inlet and can shed some light on the problems I'm having. I am currently running PCB homologs by method 680 using a 7890A/5975C with a LVI/PTV using CO2 for cryo-cooling the inlet. I am seeing extremely inconsistent results injecting the same standard numerous times. Internal standard response varies between 40,000 and 500,000, as well as extreme peak-splitting in some cases. I have leak-checked everything, before a run, during a run, after a run, and have only caught the system leaking twice, once at the column nut going to the transfer line, and the second was at the Gerstel septumless head. The second leak was so bad there was virtually no column flow...both of these were relatively easy fixes.

I am using a multi-baffled liner with no wool, 2x5uL injections with a 3second delay between injections, solvent flush (hexane) @ 40mL/min, inlet initial temp 30C. Months ago this very same method proved successful for another analyst in charge of the instrument. I just can't see what might be wrong. There must be something very very wrong with the system, and I'm so new to the gerstel PTV it could be something so simple...I hope it is, and I hope I discover it. There's only ONE thing I found in literature that could account for the peak-splitting I'm seeing, and that's a small amount of liquid hexane making it's way to the head of the column..but if this is the case..what could have changed?? I am now testing this morning with a small amount of wool in the liner to see if this improves..or just CHANGES anything..

[R13]

Do You inject manually? Injection speed/needle position might greatly influence results.

[Peter Apps]

We need the full operational details - gas flows, times, split ratios, column size, type and temperature programme etc etc, down to what type of needle you have on the syringe.

Seeing an actual chromatogram would be a huge help, instruction for posting them are in a sticky.

why do you do 2 x 5ul injections rather than 1 x 10 ul ?

Peter

[cjm]

Well, I may have solved my problem...here's the thing..I'm injecting 2x5uL injections because 1x10uL causes liquid hexane to exit the bottom of the liner. Autosampler 7693 with a 25uL syringe.

None of the conditions have changed since December, when the instrument was running great....that's what concerns me the most. I have since changed columns, from an agilent hp5 to a restek rtx-5sil ms w/10m integra guard. I also put a small amount of wool in the liner, thinking that liquid hexane may still be exiting the bottom of the liner during the solvent vent timeframe. I also tweaked the cryo conditions for the inlet, but barely. I will post more info when I have some extra time, and I will try and post chromatograms.

[Peter Apps]

Thanks for the feedback. Diagnosing from the cure suggests that you had liquid hexane getting into the column directly (unlikely because the end of the column should be well above the bottom of the inlet) or you had hexane condensing onto the column (very likely if the column start temperature was lower than the inlet temperature).

Why on earth are you using a 25 ul syringe to do 5 ul injections ? Use a 10 ul; it will be 2.5 times more repeatable.

Peter

[cjm]

So I'm still having trouble, some injections work great, others are terrible with peak splitting, mass discrimination, I'll try and give as much info as possible, and I'm working on including chromatograms. We've virtually ruled out the MS as a source for my problem, it appears to be the injection and nothing else, I just cannot get the parameters set in such a way to create reproducible results. I'm willing to experiment....

I'm injecting a PCB Homolog solution in hexane. To meet detection limits we need to inject 10uL.

GC: 7890A with 5975C MS with 7693 autosampler (10uL syringe) LVI/PTV inlet (Gerstel) with a multi-baffled liner, small amount of wool (internal volume 150uL).
Column: Rtx-5Sil MS 30m 0.25mmID 0.25um with 10m integra guard; constant flow 1mL/min
Injection volume: 5uL x2 with 3 second delay @ 750 uL/min
Inlet: 30C for 0.3min, 300C/min to 300C hold 20min
PTV Solvent vent: 100mL/min @ 0 psi until 0.3min
Purge flow to split vent: 100mL/min @ 3.2min
Oven: 50C for 3.5min, 20C/min to 160C hold 1min, 8C/min to 310C hold 5min

Zoomed view of portion of Simdata:


..and same standard injected right after:

[Peter Apps]

Hi Christian

Try increasing the time that the inlet is at 30C with solvent purge on, but decrease the solvent purge flow - at 100 ml/min you get dramatic cooling from the solvent evaporation that actually slows the evaporation down. You can also probably increase the inlet temperature to 40C and see if it helps.

For trouble shooting drop the second 5 ml injection - the peaks are plenty big enough, and we can get back to 10 ul later.

NB if you have an MS pulling vacuum at the end of the column you have some flow into it even with the inlet pressure nominally at zero. This should not be a problem as long as the inlet temperature is lower than the column because you will not get solvent condensing in the retention gap.

How far apart are the two apices of the split peaks ?

Peter

[cjm]

OK thanks. That's something good to try. Here's dftpp from earlier in the day, and one later last night...


good:


and bad:


So what changes from the good run to the bad? Is liquid still getting to the bottom of the liner in one injection and not the other?

[Peter Apps]

Hi Christian

Is the area of a good peak the same as the total area of a split peak ?

Peter

[cjm]

2.4million on split peak, 2.6million on good peak. area seems consistent.

[cjm]

I have changed the following: vent flow from 100 to 50, added delay of 3seconds between injections, inlet holding for 0.6min, solvent vent for 0.5min. I still need to do 10uL injection to get through a current job I'm working on...

[Peter Apps]

Short-cuts can make for long delays; first troubleshoot, then run the samples ! Go to 1 x 5 ul injections and if there is reliably no splitting then the problem becomes to inject 10 ul in one go. Which is soluble, as long as the times, flows and temperatures have been set right on the basis of the 5 ul injections.

Peter

[cjm]

I will push through today best I can to get samples completed. I will take your advice and try one 5uL injection and try and replicate peak shapes, and then I'll go from there. The changes I made seem to be working well for now, I just need to make it through today...thanks again Peter.

[cjm]

I am now injecting 1x5uL injections of a 20ppb calibration standard for method 680 (pcb homologs). I'll be injecting about 20 times to see if peak shape holds. It looks promising so far. My other gc has same equipment and I'm analyzing for all 209 pcb congeners (i.e. nightmare) but that requires even lower detection, we are running a 1ppb standard in sim. That requires 20uL injection. I am playing with that one trying one 20uL as opposed to our current 4x5uL. Peter, the changes I described I made the other day seemed to improve results dramatically. I hope I can fine-tune the parameters to allow a single 10uL injection.