Headspace Equilibrium: Choice of Diluent

[4meljones]

My run showed that the linearity (using one injection per level) for methanol (r2 = 0.996) and IPA (r2 = 0.998) are good, and methylene chloride (r2 = 0.992) is ok (could be better), but n-hexane is way off (r2 = 0.979). The plot for nhexane shows that athough I used a 25 µL syringe (one-shot spiking), my 100, 120, and 150% injections are quite a ways off the best fit line compared to the points representing lower concentration, and the 150% level is below the line with a lower response than expected. Surprisingly, when using the 100% level to calculate expected responses, the accuracy was ok if not a little variable, even at the 150% level. I am suspecting there is something wrong with my technique (something is escaping somewhere some of the time) or that there is "too much going on" in the vial to be consistent - perhaps some interaction between analytes. I am going to try using a stock standard with half the concentration (i.e. based on a 50 mg sample rather than a 100 mg sample per vial) to see if this helps even things out.

I notice that when an injection is not in line with the others, it is almost always that the methanol and IPA are a little low and the n-hexane and methylene chloride are a little high. Is it a weird coincidence that IPA and methanol (alcohols, polar) are always great or ok and the others (water insoluble) are problems? Is it a problem if the analyte favors the headspace too much?

Recap: The diluent is 50/50 DMI/H2O (2mL per vial) and the headspace temperatures are 80/120/150°C for the headspace oven, 1 mL loop, and transfer line. the GC inlet is 180°C (Agilent VI inlet - no liner, just small inert area to transfer headspace sample to the column) and the initial temperature is set to 40°C. I am using a DB-624 (ultra-inert) 0.53 mm ID column. My 10mL vial is equilibrated with "high" mixing for 20 minutes. I thought this was more than enough to reach an equilibrium, but I might also try longer, say 30 minutes.

Regards
Melissa

[Peter Apps]

Hi Melissa

When you add the volatiles to the matrix, do you add a mixture of volatiles in a single shot, or do you add them one at a time ? If one at a time, in which order do you add them ? You will get better performance from the hexane and CH2Cl2 if you add them last - otherwise they will be lost form the headspace when you lift the cap to add the methanol and IPA.

Peter

[4meljones]

I add all 4 analytes at once. When I add a stock standard to the vial, I use a gas-tight syringe. Normally, I would use a 10µL syringe and then, for the 150% level, I would spike 5µL, then 10µL with the syringe tip under the diluent before capping the vial. Up til now, I have not had any problems with this technique. For the most recent runs, I spiked using a 25µL.

The seemingly otherwise random variability that was worse when spiking at the higher levels made me wonder if it was my spiking technique causing some of the problems. So, for the last couple of runs (run on Friday), I tried spiking "slow", taking extra care to place the stock standard at the very bottom of the vial (under the diluent) and not allowing it to distribute in the diluent til the cap was on. This seemed like a desparate tiny detail, but my results turned out to be less variable using this techique. If I spiked by pushing the syringe plunger too quickly, I could see the stock standard spreading more quickly. This has never been a problem before and I previously hadn't considered this since I thought the analytes were soluble enough to stay in solubtion, but hexane and methylene chloride (in this diluent) may not be that soluble when the diluent is 50 % water, or maybe the difference in density between hexane (0.65) and DMI and/or water (~ 1.0 g/mL) causes the hexane to separate out of the diluent?

I also tried using a stock standard that is half as concentrated (based on a 50 mg sample weight). The responses are about half what they were using my other stock, as expected, but the repeatability was very good - ≤ 1 % for each analyte for 6 injections. Using ths stock and my slow spike technque, I am now trying the linearity and accuracy again. I'll let you know how it goes.

Regards
Melissa

[Peter Apps]

Hi Melissa

It's the desperate tiny details that are often critical. It will be interesting to see the results.

Peter

[4meljones]

Is there a way to post an attachment (excel or pdf)?

My linearity data was obtained using only one injection at each level (0, 10, 20, 50, 80, 100, 120, and 150%). The linearity (r squared) of methanol and IPA are both 0.9997 and for methylene chloride, 0.9982. (My requirement is r squared ≥ 0.990, but for other methods I often get at least 0.998 or better.) For hexane, r squared is 0.9924, which just meets my requirement. It would have been better but 2 of the 8 points were a bit more off the best fit line (for hexane only, but only slightly off for methylene chloride for the same points). My full validation will plot the average of at least 3 points at each level, which may help. I'm wondering, though, if n-hexane will always be a bit less well behaved, or is there something more I could do?

My accuracy (recovery) is rather good, calculated using only peak areas, where the detected response (corrected for the control that has not been spiked) is compared to the "expected response" and my requirement is 85 to 115 %. I calculate the expected response based on the average response of the standard prepared at the 100% level. (For my full validation, I will be calculating concentrations more precisely using the standard and sample weights and responses, but this will do for now.) At the 10% level, the recovery is a little higher (108 to 113 %) and a bit more variable, but that may be because it is harder to accurately measure 1.0 µL (from a 25 µL syringe). The recoveries at the 100 % and 150 % levels for methanol, IPA, and methylene chloride are mostly about 98 %. The recovery of n-hexane is about 98 to 102 %. That sounds good to me.

What do you think? should I be worried about linearity (or repeatability) for n-hexane? For the run, I analyzed 5 consecutive standards, and the RSD was 1 % for each analyte which I consider very good.

Regards
Melissa

[Peter Apps]

Hi Melissa

1% rsd for those volatiles is very good indeed !, the recoveries and linearities are good. I would say that you have cracked it !

It might be worth trying a smaller syringe for the smaller volumes and the 25 ul just for the higher part of the curve, but if you are gong to be be doing check weighing and calculating from actual quantity added, then the variability at the low end should largely disappear even with the 25 ul.

You should think about writing this up for a publication.

Peter