How do I know the LC autosampler injecting 5ul injects 5ul; not less or not more?

Are you talking about a autosampler with a variable loop, otherwise your accuracy isn't very important, only repeatability.

If it's variable, you can inject 5µl, 10µl, 20µl and 50µl for example, you can make a lineair curve, and your fault from you injector is b in the equation:
y = ax + b ( equation from a lineair curve that does not go tru 0 )

You can try this if you have 5 point balance;

fill your vial with water, put it on tray, let the water temperature reach the ambient. Use reference thermometer to measure the water temperature. Weigh your vial, inject 5ul, and weigh again. Calculate injected volume in ul using water density at current temperature. Repeat at least six time to be sure it is working.

I am having precision problem and I just heard this from a collague. My method uses 20 ul and I got mean 19,98 (n=6), rsd%=0,22. This turnes my problem away from the autosampler to the method.

Here is a plan if you have a spectrophotometer. It can be modified for ICP and any alternative trace analytical method, or even GC or LC. Basically it is to use the autosampler as a pipet and inject into a vol flask, dilute to volume and analyze.

So for example...

Make stock of chromate (or other chromophore). Inject into LC system without column and pump sample into a flask. Dilute to volume and measure absorbance. Compare with pipet delivery of 5 uL.

I'll come back to the suggestion made by A.nonymous (second post) and ask the original poster: "why do you want to know? ".

If you are doing quantitative HPLC, you will be comparing peak areas or heights obtained from your sample with areas or heights obtained from calibrator solutions of known concentration run *under identical conditions*. This means that you do not have to know accurately how much was injected, but you must know that precisely the same amount was injected every time.

If you are using the autosampler for some other purpose, then you may need to determine the accuracy.

By the way, even fixed-loop injectors can be subject to considerable inaccuracy. Because loops are not made from precision-bore tubing, they can be off by quite a bit (I've heard +/- 10%, but don't have the numbers first-hand) from the "nominal" volume.

-- Tom Jupille / LC Resources Inc.

Tom- We check autosampler accuracy and precision over its entire range to give us the flexibility of calibrating using mass on column. Also, can program the autosampler to do some basic sample/standard prep dilution, spiking, etc.

A.nonymous - there is no connection between injector accuracy and its intercept. Install a 900uL syringe in a HP1100 autosampler and configure it for a 100uL syringe. Your accuracy will be way off but your response could still be linear with no intercept.

Water's uses the weighing method in their instrument OQ. I use a method similar to bill's.

Tom (second one)- Your right, its just when the difference is always the same (eg 21µl, 51µl, ...) and not 21µl, 42µl, ... but there is some test for this, I'm gonna check this, but I'm 3 weeks at home, so this could cost some time, maybe there is someone here who is fammiliar with Waters AQT, that protocol use such test.