Headspace - problem with calibration

[Mr.Brown]

Hello

I am trying to analyse beerepellents
(Benzaldehyde, Phenol, p-Dichlorobenzene, Nitrobenzene, p-Bromochlorobenzene(ISTD), Naphthalene, Thymol)
in honey using simple Headspace
(5 g honey, 2 g NaCl, 2.5 mL H2O, 20 mL Headspace Vial with screwcap)
Incubation at 80°C for 2 h
Injection 1000 µL
Split 1:5
Injector: 250 °C
Carrier Gas: He
flow rate 1 mL/min

calibrations are perfectly linear
however depending on the honey type (forest, polyfloral, limetree,..)
the calibrations give totaly different slopes

Only for Phenol, p-DCB and naphthalene the calibration is independent on the honey type used.

Any Ideas how to solve this problem or what could be the cause for this?

[Analysor]

Hi
It seems that some matrix effects interfere in your analysis.
Then, you should use standard addition method for calibration.
It will eliminates some matrix effects. However, it is more difficult comparing with conventional calibration method.

[chromatographer1]

The 'problem' is that different chemicals have different partition coefficients.

Solution? Change the laws of nature or follow the advice of Analysor.

and BTW

Analysor, kudos.

Good answer.

Rod

[James_Ball]

If you do matrix matched calibrations for each sample and you are getting perfectly linear curves then you should already be compensating for any differences in the matrix effects.

Do you have any of the compounds detected in a honey sample that you are using for calibration when analyzed as a blank? If so, then the small amounts can interfere with the lowest standards in your curve and will skew the slope even though they produce an acceptable r^2 for the line. Example is I have a calibration for volatile compounds on purge and trap, low standard is 0.5ppb and highest standard is 200ppb with 8 points to the curve. The r^2 can be 0.995 while the lowest standard is near 1.0ppb because of a 0.5ppb contamination for that analyte. Do the same calibration using average response factor instead of linear fit and you get a %RSD of over 30% which is failing, because of the low level contamination skewing the average.

Have you tried purging the honey mixture prior to spiking in the calibration standards to see if that normalizes the slopes?

[Mr.Brown]

thanks for your answers

some of the compounds I try to analyse are naturally occurring in honey.
If they are in my sample I would expect a parallel shift of the calibration curve (different axis intercept)
but not a different slope.
This should also be the case if there is some contamination interfering with my analysis.

Matrix matched calibration for each sample is not feasible due to several different honey types from several different countries.

could this be influenced by the ratio Headspace vial volume (20mL) to total sample volume (~ 7mL)?

[chromatographer1]

"some of the compounds I try to analyse are naturally occurring in honey.
If they are in my sample I would expect a parallel shift of the calibration curve (different axis intercept) but not a different slope."

This is correct.

"could this be influenced by the ratio Headspace vial volume (20mL) to total sample volume (~ 7mL)?"

As long as the headspace does not become saturated, the different slopes should continue, but all will change as the volume changes, but not become the same slope.

Using three samples of each matrix, two of which are spiked at appropriate levels will compensate for the differing slopes (different matrices) and allow you to calculate an accurate answer for your analysis.

This is the minimum procedure for conducting a regression analysis noted by Analysor.

best wishes,

Rod

[RogerBardsley]

The effect that you are observing might be related to the pH of the honey which can cause a matrix effect.

[Mr.Brown]

Then, you should use standard addition method for calibration.

Using three samples of each matrix, two of which are spiked at appropriate levels will compensate for the differing slopes (different matrices) and allow you to calculate an accurate answer for your analysis.
This is the minimum procedure for conducting a regression analysis noted by Analysor.

So fare non of my other trials worked out so I will have to go for spiking.
I never had to validate a spiking method
any suggestions or hints what I have to consider?

[chromatographer1]

The standard addition method is self validating, as long as your recoveries are linear.
If the recoveries are not linear, then your method is not good enough to validate.

Plot the amounts of analyte added to each sample preparation against the area of the analyte found. You should add amounts that are reasonable to the expected amount of analyte you expect to find. This is important.

If you expect 10 ppm, then add 20 and 40 ppm.

If you expect 100 ppm, add 200 and 400 ppm to your samples, or you could add 50 and 200 ppm.

If you expect 1% then add 2 and 4 % , or you could add 0.5% and 2%.

You can always add more additions for more points in your regression line, but two additions as well as your sample unspiked are the minimum required.

If the line through the area points is linear ( linearity R squared = > 0.995 ) the absolute value of the x axis intercept will indicate the amount of analyte in the sample.

The linearity value is debatable. More strict value has greater confidence in the results. If your choice is less than 0.990 expect close examination from regulatory authorities.

Review "standard addition methodology" for details in you are unfamiliar with this type of analytical method.

For example, I have determined the amount of water in acetonitrile using this method: Found 0.27% with two additions of 0.25 and 0.50 %.

best wishes,

Rod

[Mr.Brown]

@rod

sound easy enough

linearity is not an issue,
I tested 3 different honey types with 5 point calibration for linearity
where I then recognized that with each honey type I get a different slope for my analytes but a perfect linearity.

thanks for your help
i will keep you apprised

[chromatographer1]

That sounds like you have an excellent method at hand.

Just remember that each sample, even of the same type of honey, should have a minimum of 3 samples, two of which are spiked.

With additional research documented, it may be that each sample of a single type of honey might give the same slope. If that is the case then only one of several samples of the same type of honey need be given the std addition treatment, with the others being calculated from the common slope factor of your analysis. But how do you know for sure that each sample of a type of honey will ALWAYS behave in the same manner? There's the rub.

best wishes,

Rod

[James_Ball]

thanks for your answers

some of the compounds I try to analyse are naturally occurring in honey.
If they are in my sample I would expect a parallel shift of the calibration curve (different axis intercept)
but not a different slope.
This should also be the case if there is some contamination interfering with my analysis.

I have had other results which changes the slope of my calibration curve when an interference is present.

Assume you have 5ppb acetone contamination in your solvent and your calibration is 2ppb, 10ppb, 20ppb, 50ppb and 100ppb. Once you plug in the numbers the standard you assume is 2ppb is really 7ppb, the 10ppb standard is really 15ppb, the 20ppb is actually 25ppb, the 50ppb is 55ppb and the 100ppb is actually 105ppb. If you are unaware of the contamination you make the calibration plot not corrected for the extra 5ppb. On a plot with the area counts the lowest standard is shifted more from its theoretical amount than the highest one would be. The lowest standard area would be 350% of its theoretical amount but the high standard would only be 105%. This would not only shift the intercept but also the slope and if the interference is great enough moves the fit from linear to a quadratic fit. That is if the interference is in the matrix you are spiking into and not in the stock standard itself.

[blueman_2]

Hello Mr.Brown,

I am also trying analysing bee repellents in honey. Which column did you use for GCMS? Is headspace method works?

I appreciate it if you can help.

Best regards.

[mckrause]

Just looking briefly at your method the first thing that jumps out at me is that your ratios are all out of proportion. I would probably use 2 grams of honey, 1 gram of salt and 5-10 mL of water. You need to have enough water present to truly be in an aqueous matrix and you simply do not have that. This is the likely cause of your changing slopes; you are not in a matrix that can be considered to be reproducible (each honey will have differing amounts of water present).