Poor precision only affects one of the two peaks of interest

[chris taylor]

Hello,

I've come across an interesting issue in which I am analyzing 2 peaks, which have a method specification of < 1.5% RSD by area and one of the peaks passes the specification and the other does not. Initially, the earlier of the 2 peaks (peak A) was failing the RSD spec. The method is a gradient which ramps immediately upon injection and uses a PDA from 200-400nm as well as monitoring 230 and 308 nm.

After looking thru the system and finding no obvious problems, I ran a calibration standard (which uses iscratic conditions) and the RSD result was 0.3%, well below our in-house calibration spec of < 1.0% RSD.

With this in mind, I ran the test in question using the method conditions and solvents/solutions and the specification was not met. I also ran the method isocratically since the calibration standard passed the specification under isocratic conditions. Only this time, Peak B (the later peak) did not meet the spec.

If it is an injector problem then logic tells me that both peaks would be affected.

My question: Why is precision poor for only one analyte in the chromatogram and not the other? Is this a system issue or do the analytes not 'like' this system? Furthermore, the analysis in question was moved to another LC and everything worked fine. I am stumped.

Any insight would be much appreciated.

Thanks

Chris

[tom jupille]

Chris, I agree with you that an injector problem would likely affect the precision of both peaks. Actually, a good check is to look at the precision of the *ratio* of the peak areas. If the ratio looks better than the individual areas, then it is an "upstream" problem (injector or sample prep). If the areas look better than the ratio, then it's a "downstream" problem (peak shape, integration, baseline noise, etc.).

What you're seeing is consistent with an integration problem. I could speculate that the baseline on one system might be a bit noisy, or than it might have some extra-column volume giving tailing.

[chris taylor]

Tom,

Thanks for the reply. I actually did calculate the 2 peaks 'internally' as you mentioned but failed to include that info in my first entry. When calculating the ratios, the precision is about the same (i.e. around 1.9 - 2.0 % RSD). So in this respect, I'd say that it is definitely not *better* than the externally-calculated areas, so downstream issues could just be it. Perhaps your theory regarding integration is the next item to look at...

Thanks for your input. There is a great sense of community on this forum!

Chris