Chromatography Forum

Hi All, I'm new here and have just started a new job.

I need to develop a Headspace GC method (using the Agilent HS) to analyze fermentation broth for solvents produced during process development using either an FID (preferably) or an MS detector.

I've done a lot of headspace work for residual solvents for small molecule drug delivery development in the past. These had very well defined relatively simple matrices and external standards always worked well.

But I have not had to deal with a messy matrix like broth before. Also the levels I'm looking for here are in a much broader range ( 250PPM to 25%) to follow the fermentation process.

For HPLC work, I centrifuge the broth and analyze the filtered supernate, which gives analyte concentrations in the supernate and not in the broth itself.

For HS-GC, in theory I should be able to just add 'raw' broth to the headspace vial and test that and get concentrations relative to the volume of added broth. Doing that, I realize it is highly unlikely I could get decent results using external standards.

To get the best results I would have to do some other method of quantitation.

The thing is they also want results as close to real time as possible. I think I can develop an isothermal method for the analytes, so how i do quantitation can affect turn around time significantly.

With the column I'm going to use my isothermal GC run time may be under 5 min and should almost certainly be under 10 min. I can always run standards early in the day before samples start coming in, so it's the per sample analysis time that matters most.

The options for quantitation that I can see are:

1) Standard addition:
Potentially most accurate, but slow (multiple injections per sample) and i don't know how well it would work for a broad concentration range.

2) Multiple-HeadSpace- Extraction:
Gets rid of matrix effects but slow (multiple injections per sample). Also I'm not sure how accurate it is in practice (I have never used it)

3) Full Vaporization:
Potentially fastest because vial equilibration time should not slow things down and could be a single injection per sample. It should also either eliminate or minimize matrix effects.
But I don't think I could accurately sample small volumes of raw broth because of the solids present.

4) Internal Standard
Potentially second fasted after Full vaporization as it would use a single injection per sample.
Theoretically not as accurate, but I MIGHT be able to get away with an internal standard (or two ISes) if the internal standards are similar enough chemically to the analytes to have partition coefficients affected to about the same degree as the analytes by the matrix.
Theoretically quantitation would not be as accurate as the above methods, but might be good enough practically.


If I work with filtered supernate then Full vaporization becomes an option I think,but I still worry about the precision and accuracy reproducibly of adding such small volumes for lots of samples.

With the supernate at least, maybe there is even one more option:


5) Internal and maybe even simple External standards with dilution.

Essentially try to dilute out the matrix effect. If I have enough sensitivity I could try taking say 50 or 100 ul of filtered supernate and adding 0.5 or 1 mL of Water or dilute internal standard in water to a 10 mL HS vial. That might give good enough results.

I'm looking for the best way to get started on this method development, which also needs to be done ASAP.

So am I wrong in any aspect of my analysis of the situation? Am I missing something? Any advice/thoughts on what to try first?

Thanks for any help.

- Karen

It would sure be helpful to the discussion if you could mention the necessary analytes you MUST measure. I can only assume that ethanol is one of them.

How about acetaldehyde? or methanol? butanol? any mercaptains or thiols? esters? aldehydes?

"For HS-GC, in theory I should be able to just add 'raw' broth to the headspace vial and test that and get concentrations relative to the volume of added broth. Doing that, I realize it is highly unlikely I could get decent results using external standards. "

I am unaware of any such theory for accurate measurements, only coarse general 'trends'.

If you really want accurate results you have to use std addition. This will be highly accurate if you can keep your broth sample sizes below .25 mL.

0.1mL would be desirable, with a 'std addition volume' of 0.1mL so every vial contains 0.2mL. This will give enough sample for low concentrations and will be close to the limit for high % levels of analyte. Use 3 or 6mL vials for trace work and 22mL vials for % work.

Use two levels of addition making three samples total to give you your line for x-axis intercept. Your linearity should be greater than 0.998 at worst and 0.9995 or better if you can make your samples that accurately. The final accuracy is determined by how well you can prepare samples. I found the technique was not the limiting factor in accuracy of linearity but how good the chemist was in preparing samples. (and to a degree how reproducible the vial volumes were from vial to vial. with 2% changes in vial volume your accuracy can never improve above that unless serendipity is involved)

I would even go to smaller samples if you can.
Smaller = faster and more accurate.

Yeah, counter intuitive but it is true. I spent years proving it over and over again in 1995-2000 in USP and FDA work.

I would not try any other path as you will find 'gotchas' in every direction.

You can determine the cycle time by determining the time of the final eluting peak.

With smaller samples you need only heat for 8-10 minutes for the alcohols, even quicker for other analytes. (see AC 1997 June for my paper separating 18 solvents in 6 minutes) I have done the old USP test for the 5 volatiles (DCM to TCE) in 2 minutes with complete separation using an 80°C oven and 20 cc/min flow rates on megabore columns. So if the number of your measured analytes are small you may be able to do the testing very quickly.

Varying the headspace volume as you change the sample volume affects the partition of analytes from complex matrices such as your broth. You must use std addition to compensate for the broth composition variances from sample to sample and from day to day IF YOU WANT ACCURATE data. If you want general trend data which may not be suitable to make solid scientific conclusions, then all bets are off.

Good luck.

Rodney George

Diluting the broth sample in water for instance is a viable alternative which will reduce the matrix effect on analyte partitioning. I would still suggest that, if you follow that path, that you keep your sample size at or below 0.25mL.

With a set of three std addition preparations at the beginning of the day you should be able to get linear data from 500 ppm to several percent. You can always run a second set of std addition preps at the end and/or middle of the day to confirm your measurements during the day's run.

best wishes,

Rod

It would sure be helpful to the discussion if you could mention the necessary analytes you MUST measure.

EtOH is one... as for the others I'm not sure i can say.

"For HS-GC, in theory I should be able to just add 'raw' broth to the headspace vial and test that and get concentrations relative to the volume of added broth. Doing that, I realize it is highly unlikely I could get decent results using external standards. "

I am unaware of any such theory for accurate measurements, only coarse general 'trends'.

All I was saying is that unlike HPLC using some calibration method I should be able get results without without first doing centrifugation and filtration as is required for HPLC.

I said "in theory" because the broth has a lot of solids in it and I don't know if one can truly get a representative sample as it's not a uniform suspension. I worry that difficulties involved in sampling and accurate dispensing of small volumes with undissolved solids might make it impractical. I have no experience with these types of samples but they are pretty nasty.

0.1mL would be desirable, with a 'std addition volume' of 0.1mL so every vial contains 0.2mL. This will give enough sample for low concentrations and will be close to the limit for high % levels of analyte. Use 3 or 6mL vials for trace work and 22mL vials for % work.

Can the Agillent G1888A use anything but the nominal 20 and 10 mL vials? And in any case i know for sure I can't mix vial sizes within a run with it.

Also I don't know for sure what some of the sample concentrations will be... things are very variable at this point.

BTW why would full evaporation be an issue on the supernate?

Use two levels of addition making three samples total to give you your line for x-axis intercept.

Depending on run time that may still be faster than HPLC or it may not be. The other issue is that individual samples from several different reactors are taken every 3 hours.... between the different reactors and the change in the reactors over time, i suspect i would need to do SA on every sample


We can do the most important analytes by HPLC now in a 35 minute isocratic run... this needs to be significantly faster. If I need to do SA on every sample at best it will only be slightly faster at best.

Maybe SA with dilution would eliminate that need... but at that point I have to wonder if using an IS might be almost as good.

I would even go to smaller samples if you can.
Smaller = faster and more accurate.

Yeah, counter intuitive but it is true.

Perhaps for accuracy, but not for speed. That is exactly what i suspected for speed.

For a different type of work I often found with headspace higher split vent flows can give better effective sensitivity because of sharper peaks ... Now THAT was counter intuitive!

You must use std addition to compensate for the broth composition variances from sample to sample and from day to day IF YOU WANT ACCURATE data. If you want general trend data which may not be suitable to make solid scientific conclusions, then all bets are off.

Accurate is relative. This is early process development and 2 sig figs should be enough to make valid scientific conclusions.

I really do wonder if using say n-propanol as an internal standard for Ethanol at least with some dilution would be good enough.

If I can get the GC run time to under 5 min that Std Addition may be viable. If Not I may just check it against HPLC and/or do Standard Addition for awhile and compare those values against IS results. IS would would be so much faster on a per sample basis if it's good enough.

Right now I'm thinking of using 10 mL HS vials, 100uL of centrifuged/filtered broth and 200uL of n-Propanol in water as the IS and diluent. And make up SA standards to add 200 uL fro each concentration

Good luck.

Thanks
- karen

" I said "in theory" because the broth has a lot of solids in it and I don't know if one can truly get a representative sample as it's not a uniform suspension. I worry that difficulties involved in sampling and accurate dispensing of small volumes with undissolved solids might make it impractical. I have no experience with these types of samples but they are pretty nasty. "

I agree with you that sampling the broth without solids is the best path to follow.

" BTW why would full evaporation be an issue on the supernate?

The solids issue with very small samples required for full evaporation technique would be a deal breaker, but not the supernate, s long as they sample size was very small ---- 5µL ?

If you could dilute 25 or 50 µL of supernate with 200 µL of water then you could probably ignore any matrix caused variation which hopefully would be minor. I suspect if would be. The water could contain std additions and/or an Istd.

That is where I would strike first, but I would run std addition preps first WITH a low level of your internal std of n-PrOH to cover both options at the same time. Stay with the smaller 10mL vials. I would heat them at 85°C for 8 min with shaking.

If you are only looking at ethanol then analysis time should be shorter than the heating equilibration time.

Conversely,

If you inject supernate you should be able to do direct injections via packed column every 120 seconds or so. With a precolumn and or a packed injection liner that you could swap out daily or so (I would prefer a short precolumn with a 8 inch or so empty at its head, you could really get accurate numbers and speed.

In either case, I think with these parameters you will find success quickly.

good luck,

Rod