I'm having difficulty achieving %RSD's of replicate sample injections of better than about 2%. The analyte is a metal chelating agent which requires approx 10 injections to attain a stable area count due to the analyte passivating the inner surface of the column. All of the other tubing and fittings are PEEK with the exception of the sample loop. Below are the conditions in the method:
0.5 ml/min/pump column:Waters IC Pac Anion HR, 6 micron, 4.6mm, 75mm. Mobile phase B = Pierce Flouraldehyde rgt. mobile phase is 0.01 % HNO3(Aqueous). Column and about 10 feet of 0.3mm Teflon reaction delay tubing is at 35 deg C. Detector is EX=338nm and EM=455nm.
The method validation indicated approx 1.6% RSD attained for 2 analysts on different instruments. I'm using a Newer Waters 717 with dual 515 pumps and a Waters 474 Fluor. detector.

By how much is the area count changing? How long is "10 inejctions"? What do you do between injections? Do you wash the column somehow? Why do you think that the analyte interscts with the inner surface of the column? Maybe the proble is realted to something that does not depend on the surface?

Are you sure this is a chemistry problem. Maybe your 717 is not working at 100%. What is your injection volume, a standard 717 (with a 250 uL syringe) does a better job with larger volumes (>25 uL). Are you degassing your solvent? A 717 injection reproducability and accuracy will benefit from degassed solvent. Did you do a purge and compression check of the 717 before you began. This helps. And finally, check for an air bubble in the 717 syringe, this sometimes happens and people do not check as often as they should. An air bubble in the syringe will cause non-reproducible results.

Mike,
I would check the validation for two items.

The relative standard deviation depends upon the concentration for measurements with non concentration dependent variation, (the most common for absorbance and fluorescence detection.) For example, if the variation for a measurement is +/- 1 unit, the RSD% at 100 is 1% at 10 it is 10%.

Secondly, I would examine the robustness validation and the number of samples used to calculate the RSD% Short term 1% (at a concentration greater than 100 times the mean divided by the standard deviation) is about right for an HPLC assay, post column detection adds more variation, e.g., 1% so the number 1.6% sounds like the most preciss run ever for a method. Check to see if your goal is reasonable, before beating a dead horse.
Jim

I'm surprised nobody mentioned the injection volume. At 1 uL, this is pretty good! At 100 uL, well........

OOPS! Sorry, Anonymous 3/7 5:59. I just didn't read your post closely enough.