poor acuracy and precision for methanol/water

[Ricardo]

I am analyzing a 50/50 mix of methanol/water on an agilent 6890, split mode to a TCD. The column is a HP-Innowax from agilent, 30mx0.32mmx0.25um film thickness. I have tried different temperatures but 90C works best for peak shape. Inlet is 150, TCD is 180. split is 100/1. Liner is a general purpose 4mm I.D. (also used a liner with cup and packing with no change in results). I have tried 0.1 and 0.5 ul injections. Have checked for leaks. The best I can achieve is an occasional 44/54 ratio, but usually get about 40% ethanol/60% water by normalized percent. the samples are from a distillation column, so samples will range from 1 to 90% methanol. I have also tried different linear velocity, makeup and reference flow. Have made up fresh stds as well, of course.

Any ideas? Thank you in advance for your help!

Ricardo

[Peter Apps]

Hi Ricardo

Can you clarify a few things:

is the occasional 44/54 ratio the known composition of what you are injecting - in other words did you make up a standard with this composition ?. Note 44 + 54 = 98.

What is "normalized percent" - is this peak areas, or is there a calibration step ? If it is just peak areas the discrepancy between the expected and measured ratios is almost certainly due to the detector responding more strongly to one substance than the other.

Injecting aqueous solutions into a hot vaporizing inlet never works very well. Do you have access to an automated headspace sampler, an on-column injector, or a programmed temperature vaporizing inlet ?

Are there substances other than water and methanol in the samples - in other words could you determine the methanol and then call the remainder water ?. If so you could use an FID to detect the methanol.

What repeatability are you getting now ?

Happy New Year, Peter

[chromatographer1]

There are three obvious points where you can be discriminating the sample.

The first is the splitter itself. While you do not seem to be overloading the injector at 0.1µL you may not be placing a representative portion on the column for some reason. But don't forget you could need a new septum and are losing your methanol portion there. Use a hollow point or a side hole needle rather than a tapered point needle to see if that helps.

The second is the sample in the needle. If you are not ejecting from the needle a representative portion of the sample you may be discriminating your mixture (more of one component may be leaving the needle than is in the sample).

The third is that you may be leaving water or methanol in the injector or on the column so it is not quanitatively reaching the detector. Not very likely, but possible, especially if the column has had a history.

I suspect that the most likely culprit is splitter discrimination.

A more suitable method might be a packed inlet liner. And a packed column would also be a better option, yes, sometimes older solutions are better.

You can change the position of the capillary column inlet in the injector to see if that changes or improves the discrimination of the solution. You can leave the needle in the injector to ensure that the entire sample is vaporized and leaves the needle. Or you could use a heated valve to inject your liquid samples quanitatively.

best wishes,

Rod

[Ricardo]

yes, that was an error. I sometimes get 44/56 at best. It is a pure mixture with no other components. The results are based on the area counts. With a correction factor of 0.55 for water and 0.58 for methanol. The area counts are divided by the factor, then normalized to 100%, which should give mole percent according to the correction factor document.

I do not have any other injection methods available such as headspace or direct injection. I do have an FID on the 6890, but we were using the TCD so we would not have to calibrate,(only use the area counts of water/etoh) as the FID doesn't see the water, and doesn't like it as well.

Thank you for your responses. I will try some of your suggestions and see if it helps.

Ricardo

[chromatographer1]

Where did you get this document? Is it trustworthy? Do different ratios of water and alcohol always give you the same correction factors?

You have used both methanol and ethanol in your postings. Which is it?

I would assume that 50 mL of water mixed with 50 mL of either EtOH or MeOH would give a volume of less than 100 mL. Would this be a part of your calculations? Do 10/90 or 90/10 give you the same resulting volume?

Have you confirmed these response factors are correct on your instrument's detector?

Are your responses linear? To confirm, inject variable fractions of water and alcohol in a different balance solvent: 10% water and then 10% alcohol followed by 3% (or 1%) water and 3.0% (or 1%) alcohol in another solvent (acetone? pyridine? ethylene glycol?) to determine linearity of response.

But is this research necessary? Have you used your own or another's research as a basis of your calculations and are confident of their validity?

I am assuming that you are seeing variable results for the same standard injected repeatedly. If this is so then I would still believe splitter discrimination is your problem.

Sorry if I am beating a dead horse. A few more details would be helpful.

best wishes,

Rod