Precision in Beverage EtOH by HS-GC

[EggsBenedict]

Hello,

I am currently using a Rxt-BAC2 column to analyze the EtOH content in wines by static headspace GC, but am struggling with both sample carryover and a lack of precision.

As for precision, I am seeing about ±1% difference between samples. Although most GC methods I have read (including BAC SOPs from crime labs) consider a ±5% difference to be acceptable, our lab’s goal is repeatability to a tenth of a percent EtOH at 14% v/v EtOH, which would mean results within ±.05% v/v EtOH, or a ±0.3% difference between samples. Does anyone know whether this level of precision is possible? If so, what column(s) and HS-GC method(s) are able to achieve this? If our current column and equipment is adequate, which method parameters should I try changing?

Below is our current method (I can provide more details if needed).

Vial composition: 1 mL sample volume (950 uL of 0.5% IsoP as ISTD + 50 uL wine or EtOH std) in a 22 mL vial
Column: Rtx-BAC2
Gases: N2 for pressurize, carrier, and makeup; Air and H2 for FID

Headspace Autosampler (Tekmar 7000):
Platen - 65 C
Sample Equil Time - 8 min
Vial Pressurization Time - 1 min
Loop Size - 100 uL (cut from 1 mL)
Loop/Valve/Needle temp - 150 C
Loop Fill Time - 0.25 min
Loop Equil Time - 0.25 min
Inject Time - 0.5 min
Transfer Line Temp - 165 C

GC (HP 5890 Series II):
Oven Temp - 40 C isothermal
Split - 40:1
Injector Temp - 200 C
Detector Temp - 280 C


Thank you for your help!

Lora

[Peter Apps]

Hi Lora

I very much doubt that you will get below 1% repeatability with off-the-shelf hardware. If you do hardware modifications you might get 0.3%. To get 0.1% you would need to design and build the sampler from scratch to get rid of the problems with current designs.

You best approach will be to do multiple replicates and work with the means.

Chromatographer 1 on this forum is the headspace guru, I hope that he will post something.

Peter

[EggsBenedict]

Thanks for responding, Peter!

Although I'm not where I want to be with this method (and may never completely get there), it is kind of a relief having a more experienced person say that I am pushing the limits already.

My equipment rep recommended that I replace any non-silcosteel parts in my autosampler with silcosteel parts to reduce the carryover I am experiencing. Do you think that might help?

Other than that, he and the technical service representatives at Restek and Sigma-Aldrich all ditto what you say--that it would be a frustrating exercise to try to attain that level of precision (it has).

In case you were wondering why I feel I need that much precision, it is because in the wine industry there is a very important ($$$) difference between 14.0% abv and 14.1% abv (just as there is between .07% and .08% in blood alcohol content), and I want to be confident that the results are on one side or the other.

Thanks again,

Lora

[Peter Apps]

Hi Lora

Silcosteel will probably help with the carryover, at least it can do no harm. What levels of carryover are you seeing at present, and how do you determine it ? How are you connecting the headspace transfer line to the inlet. If you are plumbing it into the carrier gas line, that area is a likely spot for adsorption and carryover.

What hardware do you have ?

If 14% is a regulatory cut-off, then how to the regulatory authorities determine alchohol content ?

Peter

[rb6banjo]

Agreed. What method does the TTB have that can provide that kind of precision? I would argue that they can't measure it that well either.

EtOH in beer/wine/spirits is a bit of a problem because it's at such high concentration. A 2% RSD on 5% (abv) in beer is +/- 0.1% in the abv - and that's only 1 standard deviation unit. For 95% confidence, the interval is somewhere close to double that. Most of us (even the methods used in the pharmaceutical industry) take 2% RSD as a good repeatability criterion. At 14% abv that same 2% RSD is more like +/- 0.3% in the abv (or +/- 0.6% @ 95% confidence).

2% RSD on a 10 ppb impurity gives you +/- 0.2 ppb. That's much more palatable than +/- 0.2% abv on 10% abv. I don't think you'll get to where you want to go with the GC/headspace equipment available on the market today.

Have you thought of trying to use an indirect method? There are analyzers out that based on NIR technology that can handle the high EtOH pretty well. As with all NIR-based technologies, they're only as good as the training set you give them during calibration.

[KM-USA]

EtOH in beer/wine/spirits is a bit of a problem because it's at such high concentration.

We use USP <611> to assay for ethanol in hand sanitizer products, which are about 64% ethanol by volume. We cGMP-validated the procedure to show applicability to our products. I'm not crazy about the column phase used by the USP (G43) because ethanol does tail on it, especially when injecting from water, but I didn't invent the procedure. But it does work, met cGMP precision and accuracy requirements.

First step of <611> is making an aqueous solution of about 2% ethanol, then procede from there. Capillary GC procedure, not headspace.

[Peter Apps]

EtOH in beer/wine/spirits is a bit of a problem because it's at such high concentration.

We use USP <611> to assay for ethanol in hand sanitizer products, which are about 64% ethanol by volume. We cGMP-validated the procedure to show applicability to our products. I'm not crazy about the column phase used by the USP (G43) because ethanol does tail on it, especially when injecting from water, but I didn't invent the procedure. But it does work, met cGMP precision and accuracy requirements.

First step of <611> is making an aqueous solution of about 2% ethanol, then procede from there. Capillary GC procedure, not headspace.

What injection repeatability and overall repeatability do you get with that method ? Depending how you do your dilutions the liquid handling is an additional source of variability.

When you are aiming for the 0.1% accuracy that Lora wants the uncertainty of the calibration starts to loom large as well.

Peter

[rb6banjo]

Here's a reference where they analyzed for EtOH in a fermentation liquor using the full evaporation technique for sampling.

J Chromatogr A. 2009 Jan 2;1216(1):169-72.

Basically, you use a very small amount of sample in a headspace vial. Heat the sample to a temperature that is well above the boiling point of all liquids (it seems they chose 105 °C) so that everything is gaseous. No partitioning phenomena to worry about. Then, analyze the vial by static headspace and GC.

[EggsBenedict]

(Sorry for the delay—harvest is in high gear.)

I am new to GC, so I am unaware of how chromatographers usually measure carryover. I first noticed it when I was running two identical 10-sample sets one after another--the second sample to run of each pair always showed greater peak area, amounting to an average of 0.2% v/v EtOH more than the first one. After increasing my dilution factor, I now see an average increase of 0.1% v/v EtOH in the second vial of a pair that is spaced 10 vials apart. Even if I cap some blank vials at home, or try to purge the vial with an inert gas, I can't avoid at least a very small EtOH ghost peak (smallest I have been able to get is ~2 or 3x noise), which leads me to believe that there are some good EtOH hiding places somewhere. But if I take that ghost peak on a freshly baked column to be "0%", and then run a blank after each sample in a run, I can see ghost peaks of between 0.03% and 0.1% v/v EtOH, perhaps depending on both the concentration of the preceding sample and the total number of samples that have been run so far during that run or since baking the column (I will do more extensive carryover testing after adjusting some settings).

We have a split/splitless inlet with a direct column connection from the transfer line, I believe. The autosampler is a Tekmar 7000, and the GC is an HP 5890 Series II.

The TTB (Alcohol and Tobacco Tax and Trade Bureau), uses AOAC Official Method 983.13 from the 17th ed. This is a very general (and antiquated--1988) direct-injection/packed column method, which of course, gives no information on quality of results whatsoever! I agree with rb6banjo’s suspicion that it is likely little/no better than what I am getting now.

rb6banjo, thank you so much for that numerical comparison! And yes, the high concentration has been the root cause of most of the problems in method development; I was wondering why I couldn't find any recommendations for working with high concentration samples besides “just dilute them enough”—I guess the reason is people generally choose a different method for high concentration samples!

The most common method in use today for determining alcohol in wine is ebulliometry (boiling point), which is sometimes said to have ±0.5% abv accuracy and fairly good precision (provided the atmospheric pressure doesn’t change between calibrating and running samples) but in reality, sugars and other solutes can be major interferents. Distillation/densitometry is ancient but not popular due to long analysis time and large sample requirement, but it can get ±0.1% abv accuracy (and excellent precision, too). GC-FID was the first modern approach developed for beverage alcohol determination, but, as you mention, NIR-based methods such as Anton-Paar’s Alcolyzer are growing rapidly in popularity due to their ease of use and purported quality of results (http://www.anton-paar.com/Alcolyzer-Win ... ifications).

I like the fact that a GC is a more fundamental piece of equipment (for when we develop methods for other analytes), but it might be tempting when some used Alcolyzers start coming up on the market…

I will take a look at that full evaporation reference--thanks!

Thank you everyone for all your help!

[gapeev]

EggsBenedict, you are likely beyond "mechanical limits" of this system.

I'd look into using isotopically labeled ethanol as IS with GC-MS. It's cheap
This way you can eliminate autosampler effects altogether by us ing just a good gas-tight syringe.

We will be happy to talk to you and explore possible ways we can help you.

[Peter Apps]

Hi Lora

To measure carryover run a standard, then a blank (i.e. the matric with no alcohol) the ratio between the peak areas gives the carryover.

You cannot be using both a split-splitless inlet and a direct connection of the transfer line to the column, The two are mutually impossible. Follow the transfer line towards the GC, if there is a connecting union from the end of the transfer line to a short piece of stainless tube that goes into the side of the inlet just underneath the septum nut you have the recommended (by Agilent) connection into the carrier gas line. Its only benefit is that you can do liquid injections through the septum without having the remove the transfer line. A direct connection will have a union connecting the transfer line straight to the column.

If the transfer line goes in through the septum nut you have a more inert sample path.

Cold spots are the major problem, with the recommended connection they are inevitable unless you add some auxillary heat. With the connection through the septum nut you need to have a short a distance as possible between the heater sheath on the transfer line and the nut, and the gap needs to be insulated.

High ethanol concentrations should not be a problem (although there is a slight kink in the calibration at some level) - if your peaks are too big reduce the equilibration temperature, or increase the split ratio. NB that the lower limit of repeatability for any given instrument set up depends on the signal:noise of the peaks, and bigger peaks have bigger signal:noise.

An isotope labelled standard will help with some problems, but an MS is MUCH more trouble to keep working than an FID.

Peter

[Bigbear]

I think your loop/valve temp is too hihh as well as the injection port. I was having the same lack of precision using an HT3 ( blood ETOH )and Tekmar helped.
Maybe you should contact Tekmar . We transfered the method to another lab last year so the unit is not here.
Using 100 ul loop I feel you should not be splitting. What sise is your column ?
As I remember we were not getting complete equil @ 15 min.

[RogerBardsley]

Hi Lora

I am Roger Bardsley, the application chemist for Teledyne Tekmar, specializing in the headspace instruments. All of the suggestions that have been posted to date are good suggestions. Plesae feel free to contact me at roger.bardsley@teledyne.com to discuss your method further. I am currently working on some ofthe methods from TTB concerning these issues.