GC-MS discrimination / poor reproducibility

[benzech]

Hi, I was wondering if anyone had some suggestions on my problem. I analyze refrigerant gas (HFCs, CFCs, etc.) by GC-MS for impurities. I have a standard of 23 impurities dissolved in refrigerant at ~5 ppm each which I am using to establish method precision (interinjection precision, n=6). The impurities are also refrigerant-type halogenated compounds with boiling points ranging -50C to 50C. I am sampling out of the liquid phase into stainless steel sample lines leading to a pre-heater box and then to a gas valve with 500 ml sample loop. I purge the sample loop out thoroughly with the sample. The loop injects to a split detector with a high split 225:1 ratio. A high split appears to be needed to maintain peak shape and prevent detector saturation from the main peak. The liner used has glass wool and no taper. Narrow bore 0.18 mm ID capillary column -> GCMS monitoring several SIM channels I am seeing a lot of variability, 10-15% rsd for peak area of the impurities. I'm still looking for patterns but commonly second injection's response is always lower than the 1st, with the lower boiling / early eluting compounds affected more-so than the higher boiling, later eluting ones. When I tried straight a liner with no glass wool the RSD was still high but there was no differing response related to boiling point discrimination that I could see. Any suggestions to improve reproducibility of response are welcome

[cjm]

Have you tried injecting a single concentration multiple times, and if so, how does the variability look? If it's still high, it may be the sample introduction. 225x split seems very high. I'm not sure our instruments would easily handle that type of flow. Do you notice internal standard response variability, or is it just calculated concentration variability? How's the health of the MSD? Filaments on their way out can act strange when doing back to back analyses.

[James_Ball]

Have you tried reducing your sample loop to 100ml and dropping the split to about 50:1? That should result in about the same amount on column with less time needed to transfer the sample to the inlet which should help peak shapes. Also calculate the amount if analyte on column and see if you are exceeding what a 0.18mm column can handle. You may need to use a 0.25mm at the same flow rates for increased capacity.

What is the starting oven temperature?

Do you have the ability to cryo cool the oven if it is starting above ambient?

[benzech]

Thanks for the replies. I am using a cryo for this. The method is a variation of an EP monograph for 134a gas. On a second instrument I have the split at 75 and I am able to get slightly better results. There is still an overall high RSD (~5-10%+) I tested the RSD of the gas valve with a standard of 0.8% air in helium (3 reps) measuring oxygen. It was 2-3% RSD. Please explain further: how could a smaller loop with a smaller split ratio help in theory? Thanks

[James_Ball]

Thanks for the replies. I am using a cryo for this. The method is a variation of an EP monograph for 134a gas. On a second instrument I have the split at 75 and I am able to get slightly better results. There is still an overall high RSD (~5-10%+) I tested the RSD of the gas valve with a standard of 0.8% air in helium (3 reps) measuring oxygen. It was 2-3% RSD. Please explain further: how could a smaller loop with a smaller split ratio help in theory? Thanks

The larger the loop, the longer it takes to transfer the sample to the column, which can lead to broadening of the early eluting peaks. Also if you have too high split flow you might have differences in the flow characteristics within the inlet which could give better or worse efficiency of the analyte transfer to the column, such as the difference in laminar and turbulent flows. Turbulence introduces eddy spots and that can cause tailing. The narrower the band of sample you can introduce in a flow inlet like this the better your peak shapes will be.

[benzech]

Thanks for that response. To be clear, I have no problem at all with peak shape, even the early eluting ones. I only have variable area counts from replicate injections. Considering that, would it be reasonable to suspect some turbulence or eddies in a high split flow to result in variance in the amount of total sample that gets transerred onto the column? What about boiling point discrimination? I seem to be seeing a bit of both... I am trying a smaller loop coupled with a lower split ratio in today's experiment and I will report back with the findings.

Thanks! -B

[James_Ball]

I'm still looking for patterns but commonly second injection's response is always lower than the 1st, with the lower boiling / early eluting compounds affected more-so than the higher boiling, later eluting ones. When I tried straight a liner with no glass wool the RSD was still high but there was no differing response related to boiling point discrimination that I could see.

Thanks! -B

Does the response continue to fall with succeeding replicates or is it more random?

Do you cycle the system with a blank before the first standard injection?

Just trying to imagine the system and if there is some reason some part of the flow path could be colder on the first injection making it have a higher response.

Also sampling from liquid refrigerant, is there any headspace above the liquid? Could a change in headspace when sample is withdrawn cause a change in temperature or partial pressures above the liquid causing the low boilers to move into headspace from the liquid?

[benzech]

I have changed my loop from 500 ul to 50 ul and lowered my split ratio down to 20:1. My % RSD has improved quite a lot and is generally lower than 10% now. In some compounds less than 5%. I do still get an occasional random "spoiler" injection in which the areas of low boiling compounds are high and areas of high boiling compounds are low (or vice versa, but more commonly as stated). The bad injection comes at a rate of about 1 every four injections. I would love to be able to figure out why I might get that random spoiler injection, at least to narrow it down to the inlet vs. gas valve. I see no trends otherwise beyond a slight overall downward trend in response that is acceptable. Thanks to James for helping me get this far! -B

[Peter Apps]

How are you transferring the vapourized sample to the inlet - presumably a heated line. Are you sure that there are no cool spots, and how is the transfer line connect to the inlet ? - via a needle through the septum or connected to the carrier gas supply line ?

Peter

[benzech]

I have the liquid phase of the gas connected with stainless tubing that passes through a preheater box (150C) and then into the gas valve with is 150C. I let the liquid gas vaporize and purge through the loop for several minutes then shut off flow and immediately inject. When the valve switches the gas in the loop is carried by the carrier flow into the injector.

I have noticed a trend that suggests purging longer caused the responses of my higher boiling compounds to go down. I could not make sense of that but it occured to me one other event was also longer at the same time - the GC oven was equilibrating at initial conditions (cryogenic) longer... So perhaps the equilibration period is not long enough to allow the cold GC oven to come to equilibration with the hot injector sitting on top of it. The injector temp probe is holding steady but maybe I'm still getting a cool spot at the bottom of the injector that is taking down the responses of the high boilers. next time I run repeatability I was thinking I may try increasing the equilibration time from 30 sec to 2 min to test this hypothesis.

[James_Ball]

I have changed my loop from 500 ul to 50 ul and lowered my split ratio down to 20:1. My % RSD has improved quite a lot and is generally lower than 10% now. In some compounds less than 5%. I do still get an occasional random "spoiler" injection in which the areas of low boiling compounds are high and areas of high boiling compounds are low (or vice versa, but more commonly as stated). The bad injection comes at a rate of about 1 every four injections. I would love to be able to figure out why I might get that random spoiler injection, at least to narrow it down to the inlet vs. gas valve. I see no trends otherwise beyond a slight overall downward trend in response that is acceptable. Thanks to James for helping me get this far! -B

Is the valve on the injection a 2 position valve that switches back and forth, or is it a rotary valve that only rotates in one direction but is configured so that it is in load every other advance and inject every other advance? (such as an 8 port valve configured to load a loop)

If it always advances then it could have one position on the valve rotor that is contaminated/leaking causing the problem, which can come from a scratch or deposit on the rotor face. If it is switching back and forth, then maybe you get an incomplete movement ever so often.

[Peter Apps]

Are the tubes between heater box and valve and valve and split inlet also heated ?, insulated or what ? You are probably right that 30S is nowhere near long enough for temperatrues to stabilise. Does the flow form your sampling line supply all the split and carrier flow ?

Peter

[benzech]

OK, I decided to do an experiment to determine the effect of GC equilibration time (at initial conditions) before the injection starts. I performed replicate injections of my standard at 0 min, 0.5 min, 1 min and 2 min. Hypothesis was that the area of high boilers should go down with increasing equilibration time, to a point, then level out, based on the thinking that the GC cryo oven is pulling heat away from the bottom of the inlet as it cools, the inlet calls for more heat to compensate, until the push-pull comes to a thermal equilibrium.

My hypothesis was confirmed when I ran the experiment. areas went down from 0 min to 0.5 min and from 0.5 min to 1 min, and leveled out between 1 min and two min. This tells me that one minute should be enough, but I may select 1.5 min out of an abundance of caution.

I will repeat my repeatability experiment tomorrow running 6 reps using the 1.5 min equilibration time and see how it impacts RSD.

Thanks all

[James_Ball]

OK, I decided to do an experiment to determine the effect of GC equilibration time (at initial conditions) before the injection starts. I performed replicate injections of my standard at 0 min, 0.5 min, 1 min and 2 min. Hypothesis was that the area of high boilers should go down with increasing equilibration time, to a point, then level out, based on the thinking that the GC cryo oven is pulling heat away from the bottom of the inlet as it cools, the inlet calls for more heat to compensate, until the push-pull comes to a thermal equilibrium.

My hypothesis was confirmed when I ran the experiment. areas went down from 0 min to 0.5 min and from 0.5 min to 1 min, and leveled out between 1 min and two min. This tells me that one minute should be enough, but I may select 1.5 min out of an abundance of caution.

I will repeat my repeatability experiment tomorrow running 6 reps using the 1.5 min equilibration time and see how it impacts RSD.

Thanks all

This is interesting, glad you found that effect from temperature.

Is the bottom of your lnlet insulated? If not you may want to add some insulation there. I know some people run without the insulation cup on the bottom of the inlet which can cause similar problems with needle injected samples.