Last ditch hail mary effort on Residual Solvents

[bor9455]

Alright, I come to all of you hoping someone has an idea or two that can kickstart this problem I'm having. First off, a little background, which probably creates more problems for me than anything else.

I'm running a method looking for the residual solvent chloroform in an Active Pharmaceutical Ingredient (API). No big deal, right? Wrong. This method has been around for many years, and it was either developed by or adopted by the previous company I worked for. I know that at the time we had only 5890s (the old ones without the EPC modules) and no headspace sampler. The method was always a bit of trouble, but we had one guy we worked with (developer I believe) who could always get the RSD to pass our limit of 15%. Fast forward a few years, my previous company is no longer making that product, and they transferred it out to the company I currently work for.

I come to the new company, and they only have 6890s and a 7890 that is a raging pile of crap. When they "transferred" this method in, all they did was directly copy/paste the method from my previous company. They never went through the steps to verify it, and alas I'm here today because I have spent about two weeks pounding my head against a brick wall. I can make modifications to the method that are allowed within USP <621> but I cannot change the column and other key things.

Some of the details. Sample and standard solvent is water (I believe this is the major problem) and the final concentration of my standard that I'm having RSD issues with is the following: The final concentration per mL is : 12.0 µg of methylene chloride, 1.6 µg of trichloroethylene, 7.6 µg of 1,4-dioxane, and 1.2 µg of chloroform. The trichloroethylene and the chloroform are the two that I can't get to an RSD of less than 20 no matter what I have tried.

Column: Restek Rtx-5, 30 m x 0.53 mm ID , 5 µ df (with a 5m guard column (phenyl-methyl deactivated, 5 m x 0.53 mm ID, or equivalent)). This is an Integragard column.

FID at 250
Inlet at 75
Splitless with a purge on at 1.5 and a purge off at 10 min (unable to set purge off time)
Purge flow of 25 mL/min
Method calls for a 2mm splitless liner, and a column install of 20-25 mm above ferrule. I have tried the method with these parameters to no avail and have changed the liner and reverted to a typical column install between 4-6 mm above ferrule. I have tried a cup splitter, a 4mm splitless liner and a 4 mm liner with glass wool. The liner with the glass wool pledget got me closest to RSD than I have been. The temperature profile is as follows:

35˚C for 1.5 min
240˚C
4˚C/min to 65˚C
Isothermal at 65˚C for 0 min
25˚C/min at 240˚C
Isothermal at 240˚C for 5 min

The method calls for a 1uL injection volume. During a series of test injections, I was seeing very little response, area counts in the teens for the methylene chloride and the dioxane peaks. I began injecting 1 uL, 2 uL and 5 uL and found that with 5 uL I was finally seeing the chloroform and trichloro peaks with area counts of between 5-7, which I have been able to reproduce with other runs, but not this particular run.

At this point, I come to some of you experts hoping that you can think of something or a direction to try here. I don't have a lot of wiggle room with the temperature profile on the run, but I can change liners, injection volumes, and even run this split if that is what someone suggests.

Thanks in advance and if you have further questions, I will try to answer them as I'm at a complete loss as to my next move here.

[Peter Apps]

Hydrophobic substances with water as a solvent - bad.

Water injections to inlet at 240C - bad.

"Sample and standard solvent is water (I believe this is the major problem)" + "I cannot change the column and other key things." = insoluble problem.

Sorry, but this method needs reworking from scratch.

Peter

[bor9455]

Hydrophobic substances with water as a solvent - bad.

Water injections to inlet at 240C - bad.

"Sample and standard solvent is water (I believe this is the major problem)" + "I cannot change the column and other key things." = insoluble problem.

Sorry, but this method needs reworking from scratch.

Peter

Tell me something that I didn't already know sir! Thanks for your input on this. I'm in the QC dept, so the time for this method to be fixed would have been when my group was the customer end of things not when we are trying to execute it.

The inlet I'm injecting to is actually at 75, the final temperature in my ramp is 240.

I know this method is bad. The expansion volume of 1 uL of water is bigger than any liner volume out there, which already means backflash issues and you are already behind the ball on it. Like I had mentioned previously, this method was created in lieu of using a headspace, since my old company didn't have a HS sampler when they first started making that product. Since we perform residual solvent testing for a ton of different APIs, creating a simple HS method for this analysis would be a snap.

[Peter Apps]

Sorry , misread the temperatures.

Another bad thing - an FID has a really low response to chloroform.

OK, strategy. If you get this method limping along, just good enough on a good day then you are stuck with it forever, and every other method your company has that is not really fit for purpose will land on your bench for similar treatment. If you fail dismally at your current task (making a useless method just barely usable) then they will have to let you set up and validate a headspace method with a detector (ECD) that is selective for the target analytes.

Having said that - With the inlet at 75 C your are doing purge and trap onto the column, actually not such a bad idea if you can optimise the right parameters. Try a range of inlet temperatures (all below 100C, and a range of splitless times - you want maximum volatiles and minimum water on the column.

With water as a solvent you will always have issues with losses into the headspace of containers, in fact you could cheat by changing the height of the needle in the vial to take a HS sample - nobody but a chromatographer would see needle height as a major change would they ?

Peter

[bor9455]

Sorry , misread the temperatures.

Another bad thing - an FID has a really low response to chloroform.

OK, strategy. If you get this method limping along, just good enough on a good day then you are stuck with it forever, and every other method your company has that is not really fit for purpose will land on your bench for similar treatment. If you fail dismally at your current task (making a useless method just barely usable) then they will have to let you set up and validate a headspace method with a detector (ECD) that is selective for the target analytes.

Having said that - With the inlet at 75 C your are doing purge and trap onto the column, actually not such a bad idea if you can optimise the right parameters. Try a range of inlet temperatures (all below 100C, and a range of splitless times - you want maximum volatiles and minimum water on the column.

With water as a solvent you will always have issues with losses into the headspace of containers, in fact you could cheat by changing the height of the needle in the vial to take a HS sample - nobody but a chromatographer would see needle height as a major change would they ?

Peter

At this point I would settle for anything other than a direct inject method of water. We currently have a lot of headspace methods that detect for chloroform using FID (unfortunately all I have is FID available) and they work fine. RSDs of less than 1 for multiple different solvents.

[gcguy]

If you have a system that allows for a pulsed splitless injection this might help with the vapour volume problem. I have looked at several options for VOCs in samples that contain water. Generally if the compounds in question form azeotropes with water it is likely that there will be an issue with obtaining reproducible results using a split or splitless injector. I have used on-column injection systems with some success but this means of injection is not without it's own problems. Might be worth a try though.

GCguy

[DR]

Can you program the Agilents for a very slow injection rate? This may also alleviate the volume issue somewhat. We used to do this sort of thing on an old Varian...