Agilent headspace GC, non-linear response

[bckelvin]

I am new to GC and headspace, and using agilent headspace GC-FID to analyze Methanol in biodiesel by following a standard method (EN14110).

Sound easy and the method is simple, preparing 3 standard calibration standard solution (0.5%, 0.1 %, and 0.01% MeOH in biodiesel, w/w), and then put the analyte and the standard into the headspace.

However, this simple 3 points calibration curve have problem. The y-intercept is too large, the 0.5% standard always has a lower-than-prediction signal, and the slope is too small. It seem that the calibration curve is non-linear.

Are there any parameters in the headspace setting could seriously affect the linearity of the response / extraction efficiency ? e.g. so-called "sample loop-filling time", sample loop temperature, pressurization, etc ?

Thx

[Peter Apps]

One of the drawbacks of only having three points is that it is difficult to work out which one is off the line - maybe your 0.1% point is too high ?

Have you tried making a fresh set of standards, starting right from the beginning ?

How are the standards made up, do you use serial dilution ?, if you do what are the more concentrated stocks dissolved in ?

Peter

[Bigbear]

Incubation temp.? You should never "boil" your analytes, so for MeOH you should incubate at no more than 65C.
Tell us more of your instrument set up. Cap column? Split injection?

[chromatographer1]

Each sample and each standard must be of equal volume and of equal matrix composition except for the solvent(s) involved. In this case if you spiked the biodiesel with DMF containing differing concentrations of methanol, equal volumes of DMF must be added to the biodiesel.

The amount of biodiesel and its methanol contamination must not saturate the headspace or equilibrium will not be achieved and linearity will not be achieved either.

Temperature must be kept 20C below the boiling point of the solvent unless small amounts of sample and the std additions will be fully evaporated.

In your situation it is highly advised that you keep your sample at less than 0.25mL and your vial size be at least 6mL if not 20mL.

best wishes,

Rod

[bckelvin]

One of the drawbacks of only having three points is that it is difficult to work out which one is off the line - maybe your 0.1% point is too high ?

Have you tried making a fresh set of standards, starting right from the beginning ?

How are the standards made up, do you use serial dilution ?, if you do what are the more concentrated stocks dissolved in ?

Peter

Thank you Peter,

I prepare the standards just before the determination, however, it still take about an hour for sample preparation, I am a bit scared that the MeOH would evaporated during that hour, indeed.
To prepare the 0.5 % w/w standards, about 22 g biodiesel is put into volumetric flask and then about 0.11g MeOH is injected into the biodiesel. The other 2 are prepared by serial dilution.

I also think only 3 points calibration is really a bad practice, but I have to strictly follow the standard method. I just checked, the 0.1% is fine, because I tried to input a smaller value of the signal of 0.1% standard, it cannot give a better cal curve...

A set of data is showed below, and almost all of my measurement give a similar result..

0.5151 2898
0.1030 608
0.0103 62.53

[bckelvin]

Incubation temp.? You should never "boil" your analytes, so for MeOH you should incubate at no more than 65C.
Tell us more of your instrument set up. Cap column? Split injection?

Thx Bigbear,

and what a nice name

Thx for the temperature suggestion, I have no idea that analyzing MeOH shall be using 65 oC or below. Firstly I follow the standard method suggestion, to use 80 oC. But later I tried different incubation temp from 60 to 95, and it doesn't help to make the cal curve better.

Other parameters are listed below

Loop temp: 110 oC
transfer line temp: 120 oC
incubation time: 45 min
loop fill time:0.05 min
loop equilibrium time 0.05 min
Pressurization time: 0.08 min
vial pressure: 5.8 psi

and GC condition:
column: DB-wax
Split ratio 25:1
injector temp: 150
oven temp: 50 oC
flow: 2 ml/min

Do you think the so-called "loop fill time", "loop equilibrium time", and "pressurization time" could probably affect signal and the linearlty of the cal curve ? I have no idea about these parameters.

[bckelvin]

Each sample and each standard must be of equal volume and of equal matrix composition except for the solvent(s) involved. In this case if you spiked the biodiesel with DMF containing differing concentrations of methanol, equal volumes of DMF must be added to the biodiesel.

The amount of biodiesel and its methanol contamination must not saturate the headspace or equilibrium will not be achieved and linearity will not be achieved either.

Temperature must be kept 20C below the boiling point of the solvent unless small amounts of sample and the std additions will be fully evaporated.

In your situation it is highly advised that you keep your sample at less than 0.25mL and your vial size be at least 6mL if not 20mL.

best wishes,

Rod

Thx Rod for many suggestions.
The preparation step and instrument setting have been just added.
I directly inject MeOH into biodiesel inside a V-flask rather than using DMF solution.
I tried changing the equilibrium time, and thus I am a bit sure that 45 min is enough for the standards to achieve equilibrium. I am using 2 ml sample volume and 20 ml headspace vial...
however, I didn't try a smaller sample volume and I would try it later

Moreover, I also listed a data set of previous cal curve. Although the R-square is fine, the recovery is really bad, about 80% at the low range (0.01%).

[Peter Apps]

One of the drawbacks of only having three points is that it is difficult to work out which one is off the line - maybe your 0.1% point is too high ?

Have you tried making a fresh set of standards, starting right from the beginning ?

How are the standards made up, do you use serial dilution ?, if you do what are the more concentrated stocks dissolved in ?

Peter

Thank you Peter,

I prepare the standards just before the determination, however, it still take about an hour for sample preparation, I am a bit scared that the MeOH would evaporated during that hour, indeed.As long as all the containers are kept closed it should not be a major problem. I am concerned that it takes so long to do the preparation though - what is slowing you down ?
To prepare the 0.5 % w/w standards, about 22 g biodiesel is put into volumetric flask and then about 0.11g MeOH is injected into the biodiesel. The other 2 are prepared by serial dilution. This is good practise - the matrix is the same for everything

I also think only 3 points calibration is really a bad practice, but I have to strictly follow the standard method. I just checked, the 0.1% is fine, because I tried to input a smaller value of the signal of 0.1% standard, it cannot give a better cal curve...

A set of data is showed below, and almost all of my measurement give a similar result..

0.5151 2898
0.1030 608
0.0103 62.53

This could just be random variation. What repeatability do you get if you inject 5 replicates of the 0.1 % standard ?

Your loop fill time and pressurization time are too short, Increase the loop fill to at least 30s (it is good to check that flow out of the loop vent has stopped) and increase the pressurization time in increments of 10 s until you get good repeatability.

I'll echo Rod; it is the boiling point of the matrix, not the boiling points of the analytes that limit the maximum temperature at which you can incubate.

Peter

[chromatographer1]

You wrote:

" I directly inject MeOH into biodiesel inside a V-flask rather than using DMF solution.
I tried changing the equilibrium time, and thus I am a bit sure that 45 min is enough for the standards to achieve equilibrium. I am using 2 ml sample volume and 20 ml headspace vial..."

This preparation procedure does not appeal to me.

If you are shaking the 2mL sample ( and I hope you are) then 45 min is very much too long a required time. 20 min is probably adequate for methanol. A 0.25mL sample would require MAYBE AS MUCH AS 8-10 minutes.

I agree with Peter, your vial pressurization time, vial equilibration time, and loop fill time are poorly chosen.

Vial pressurization time maybe 30 sec. vial equilibration time: maybe 15 sec. loop fill time: when the sample quits exiting the vent line (no more bubbles), maybe 30 sec or more. You MUST wait until the bubbles stop, but not much longer.

These times depend upon the dead volume of the sample transfer lines and the sample loop volume.

Good luck,

Rod
You should perform trials of identical samples

[mburleson]

One of the drawbacks of only having three points is that it is difficult to work out which one is off the line - maybe your 0.1% point is too high ?

Have you tried making a fresh set of standards, starting right from the beginning ?

How are the standards made up, do you use serial dilution ?, if you do what are the more concentrated stocks dissolved in ?

Peter

Thank you Peter,

I prepare the standards just before the determination, however, it still take about an hour for sample preparation, I am a bit scared that the MeOH would evaporated during that hour, indeed.As long as all the containers are kept closed it should not be a major problem. I am concerned that it takes so long to do the preparation though - what is slowing you down ?
To prepare the 0.5 % w/w standards, about 22 g biodiesel is put into volumetric flask and then about 0.11g MeOH is injected into the biodiesel. The other 2 are prepared by serial dilution. This is good practise - the matrix is the same for everything

I also think only 3 points calibration is really a bad practice, but I have to strictly follow the standard method. I just checked, the 0.1% is fine, because I tried to input a smaller value of the signal of 0.1% standard, it cannot give a better cal curve...

A set of data is showed below, and almost all of my measurement give a similar result..

0.5151 2898
0.1030 608
0.0103 62.53

This could just be random variation. What repeatability do you get if you inject 5 replicates of the 0.1 % standard ?

Your loop fill time and pressurization time are too short, Increase the loop fill to at least 30s (it is good to check that flow out of the loop vent has stopped) and increase the pressurization time in increments of 10 s until you get good repeatability.

I'll echo Rod; it is the boiling point of the matrix, not the boiling points of the analytes that limit the maximum temperature at which you can incubate.

Peter

Peter, do you have any literature or know where I can get more background information on the topics of loop fill time and pressurization time?

[rb6banjo]

I plotted the calibration data and it appears to be quite linear - in both cases where you allow an intercept and you don't in the regression analysis. The response factors (slopes) are within 0.3% of each other. Pretty good in my opinion. For y=mx+b:

b/m = 7.387/5619.343 = 0.001

For y=mx, m = 5636.229

so you'll be off by 0.001% when you're analyzing samples close to zero (limit-of-detection) concentration. Farther from the origin, the intercept has much less of an effect on the final result. If you only have to report 0.XX and you are a ways from the LOD, then the intercept contribution is noise to you.

What is the goal of the analysis? How accurate do you have to be in your determination? Are you looking for a needle in a haystack or are you just wanting to make sure there's less than some critical amount of methanol in the biodiesel?

We tend to get too hung up on the details. Lots of times, if it's not perfect to the gnat's-hind-end doesn't really matter in the end. Just my $0.02. Try not to lose track of the goal. Most of the time, the magnitude of the intercept is determined by points far from the origin. I don't think it is fair to assess the LOD using the intercept from a regression calculation.

https://i.postimg.cc/TPJX12KD/Biodiesel ... -Study.jpg

[Peter Apps]

"Peter, do you have any literature or know where I can get more background information on the topics of loop fill time and pressurization time?:

No, sorry, I last worked on static headspace more than a decade ago. Google is your friend.