Headspace Equilibrium: Choice of Diluent

[4meljones]

I am working with headspace GC (Agilent equipment) and I have a question about choice of sample diluent. Is it counter-productive to use diluents of 2 components, such as water in DMSO (i.e. 1:1 mixture)? My analyte responses are higher in DMSO mixed with water (1:1) than in pure DMSO (and I can't increase the heaspace temperature any further - 80°C), but there is variability with 2 of my 4 analytes. The methanol and IPA recovery in samples and standards is good in this diluent, but methylene chloride and especially n-hexane peaks which elute later are much more variable. I read that hexane is not particularly soluble in DMSO (or water). (I am trying DMI now.) Should I avoid using mixtures of water in DMSO (or DMI)? I envision that, at the head of the column during a split (headspace) injection where my sample "plug" has condensed, analytes that are less soluble in water and/or DMSO may be escaping out the split vent before they have a chance to enter the column. Or maybe in the vial, if the analyte gets to the headspace too easily, it is more likely to be lost out the headspace vent. That is, must the analytes be appreciably soluble in the diluent to obtain acceptable variability? Should I limit my water content (intended to drive water insoluble analytes into the headspace) of my diluent to a certain percentage?

Any comments would be appreciated.

Regards
Melissa

[Peter Apps]

The only resons that I can think of why working with a mixed solvent would increase variability per se is variability in the composition of the the solvent mixture. How are you preparing your standards ?, what procedures do you use to measure out and mix the solvents and how and at what stage do you add the analytes ?

Peter

[4meljones]

Dear Peter

Thanks for your quick reply.

To asnwer your question, I prepare my stock std in DMSO where the analytes (methanol, dichloromethane, IPA, and n-hexane) appear to be completely miscible. My working standard (100% level, in a 10mL vial) is 2 mL of diluent (50% DMSO in water, cooled before use) spiked with 10µL of stock std using a 10µL gas-tight syringe that will also be used to prepare the std for rountine testing. My main problems are at the 150% level when I need to spike 15 µL of stock standard, or 5 µL plus an additional 10 µL using the same syringe, mainly for dichloromethane and n-hexane. Could it be that these analytes are so water insoluble that they are escapiing my diluent before I can slap the cap on? I have never had this issue with spiking this way before, but maybe these analytes are touchy? I can either find a way to spike each vial only once, e.g. use a different syringe for all spiking or make a different stock for the 150% level (?), or use a diluent with more DMSO or DMI and less water? What do you think?

Regards
Melissa

[dlbenach]

Using 2 solvents should not be a problem. I also have problems with hexane in DMSO and believe it is due to hexane's lack of solubility in DMSO. I have found that swirling the flask just prior to taking an aliquot helps. Preparing your standards differently may also help. What column and GC conditions are you running? My guess is that it is either a miscibility or system issue and not due to the 2 solvents. What % recovery are you looking to get?

[4meljones]

My capillary column is an Agilent DB-624UI, 30m x 0.53 mm ID, 3.0 µm film thickness. The inlet temperature is 180°C, split ratio of 5:1, using helium as a carrier gas at 3.5 mL/min. My column temperature program is 40°C for 9 minutes, ramped 20°C to 120°, then 30°/min to 220°C and held so that the run time is 25 minutes. The FID is set to 250°C and is in constant makeup+col flow mode. My headspace temperatures for the oven, 1mL sample loop, and transfer line are 80° / 120°/ 150°C, where the vial pressurization is set to 10 psi.

I require recoveries within 85 - 115 %, but I expect closer to 95 - 105 %. I generally am getting within this ideal range for methanol and IPA, but methylene chloride and definitely n-hexane is low, more often between 80 and 90 %. This happens for samples and even standards prepared at the 150 % in particular. Today, I tried using a larger gas-tight syringe (25µL) to spike 150 % level samples (i.e. 15 µL all at once instead of 5+10µL with the 10µL syringe) and obtained recoveries of 96 to 98 % for n-hexane - much better, very interesting. My other analytes for that run was 91 to 94 % which is, I guess, ok, but like I said, I was expecting closer to 100 %. Maybe I am expecting too much.

Melissa

[Peter Apps]

Hi Melissa

Almost certainly the two-step spiking with the 10 ul syringe is the cause of the problem - it doubles the losses of hexane form the headspace, and doubles the variability in the losses.

How are you calculating recovery ?

Peter

[James_Ball]

If you are using a single point calibration at the 100% level then some variance could be from a lack of linearity up to the 150% level.

Also if you use the 25ul syringe to spike the 150% level, also use it to spike the 100% level. I have found that if you can possibly use the exact same syringe for all standards it helps with the calibration curve preparation. For my purge and trap standards I use a 50ul syringe to spike from 1ul up to 200ul and find it gives better results than using 10ul, 50ul and 250ul syringes for the same spikes.

We have just started working with the headspace method for Residual Solvents so I will be interested to see how things work out on this.

[ntruong]

Hi Melissa,
DMA could be a better diluent of choice for your analysis
As far as recovery is of concern, I would prepare the spiking standard solutions using the volumetric pipettes and flasks (you could start with 200% of your nominal concentration and dilute it down to the concentration level with respect to your sample)... I wouldn't recommend using gas-tight syringe for your spiking recovery study, though...as the reproducibility is LOW
Good luck!

[Peter Apps]

Hi Melissa,
D... I wouldn't recommend using gas-tight syringe for your spiking recovery study, though...as the reproducibility is LOW
Good luck!

Not if you actually take the trouble to measure it: APPS, P.J. and ARCHER, M. 2010. Evaluation of the source of bias caused by losses of solvent vapour during sample preparation. Journal of Accreditation and Quality Control 15: 171–180.

Peter

[ntruong]

Thanks Peter-
I should have noted that Melissa is having no issue with recovery using gas-tight syringe

Respectfully,

[4meljones]

Thanks for all the comments.

I will consider using a 25µL syringe for all spikes. For other analytes/methods, I have had no issues with using a 10µL gas-tight syringe in this way (Hamilton, SGE, etc), even for the 150% level. I have even tested for hexane in DMSO, but looking back, the recoveries were closer to 90-94% where I am hoping for 100ish %. (My accepted range is 85 - 115 %, but I usually get 97 - 105 % in other methdos. Anything wider and I start to suspect matrix effects. But the repeatability AND recovery is a problem here, but mainly only for hexane.)

I have started trying using DMI for the stock standard and find that hexane is much more miscible than in DMSO. So far, the recovery is still a bit shaky for this analyte. Meanwhile, methanol and IPA recovery/repeatbility is pretty good, so I wonder if this may not be a system problem.

I am calculating recovery by comparing the peak response of the spike to that of a standard spiked at the 100% level. I know that the presence of sample material could have an effect, but this is also happening with pure standards.

If it was a system issue, what should I be looking at ?

Regards
Melissa

[Peter Apps]

I am calculating recovery by comparing the peak response of the spike to that of a standard spiked at the 100% level.

If it was a system issue, what should I be looking at ?

Regards
Melissa

Hi Melissa

If I understand correctly you run a 100% spike, and the expected peak area for a 150% spike is 1.5 times the area from the 100% spike. If so the issue is linearity rather than recovery, but the solution remains the same; one-shot spiking.

You only need to worry about the GC system if the problem persists with one-shot spikes.

Peter

[James_Ball]

Thanks for all the comments.

I will consider using a 25µL syringe for all spikes. For other analytes/methods, I have had no issues with using a 10µL gas-tight syringe in this way (Hamilton, SGE, etc), even for the 150% level. I have even tested for hexane in DMSO, but looking back, the recoveries were closer to 90-94% where I am hoping for 100ish %. (My accepted range is 85 - 115 %, but I usually get 97 - 105 % in other methdos. Anything wider and I start to suspect matrix effects. But the repeatability AND recovery is a problem here, but mainly only for hexane.)

I have started trying using DMI for the stock standard and find that hexane is much more miscible than in DMSO. So far, the recovery is still a bit shaky for this analyte. Meanwhile, methanol and IPA recovery/repeatbility is pretty good, so I wonder if this may not be a system problem.

I am calculating recovery by comparing the peak response of the spike to that of a standard spiked at the 100% level. I know that the presence of sample material could have an effect, but this is also happening with pure standards.

If it was a system issue, what should I be looking at ?

Regards
Melissa

As Peter mentioned it could just be the linearity of the system. For the 150% to calculate correctly using the 100% as the standard, then the response curve would have to be exactly 1:1 concentration versus response from the origin, through the 100% point up to the 150% level. If you make a calibration using 3 points say, 50%, 100% and 150% or maybe 200%, then the 150% check should calculate correctly. Single point calibrations usually only work well for concentrations within +/-20% of the known point at maximum. A linear calibration using 3 or more points is usually much better and should at least slightly exceed the values you are expecting from samples on both the high and low extremes(i.e. being lower than the lowest expected and higher than the higher expected values).

There are quite a few published methods, especially in things like the USP, that use only one calibration point, but is it actually good science or just what someone decided to use for convenience?

[4meljones]

I'm doing a quick linearity study using DMI, a 25 µL syringe, and I am giving the stock standard a little shake before each sampling. I have been using one point calibration for other similar studies. I thought the peak had to be huge for the linearity to be affected and I didn't think I could saturate the detector at this concentration (290 mg hexane in 10 mL, dilute 1 into 10 mL, then 10 µL (or 15 µL for the 150 % level) into a vial containing 2mL of diluent and 100 mg of sample material). The peak area (using empower software, FID detector) is expected to be about 7500 pAs at the 150 % level. My linearity curve I'm running right now has 7 points and should show if the reponse is leveling off at the higher concentrations, so we'll see.

[Peter Apps]

Huge peaks that "saturate" the detector are only one possible cause of poor linearity. It's really just a question of semantics; "recovery" usually means that you spike a sample with a known quantity of analyte, and compare the peak size with the peak for the same quantity in a standard. This is not what you are doing. Linearity is investigated by comparing responses to different quantities of analyte - which is what you are doing.

It is possible to have very good linearity with very low recoveries as long as the recoveries are consistent.

NB that synthetic chemists use recovery in a completely different way - they mean how much of what they should have got they actually did get.

Peter