I have been working on an HPLC assay (C18 100 x 4.6 mm, 3 micron - mobile phase 40%ACN : 60% 25mM pH7 phosphate) with UV detection (270 nm).

When I run linearity standards from 50-150% of the assay concentration I get a significant y-intercept from my linearity plot (2.5% of the 100% response).

What are some common causes of this type of linear curve? My standards are prepared in 40:40:20 MeOH:ACN:H20 and injection volume is 10 mcL. Peak shape is good and blanks show no peaks which interfere with the analyte of interest.

Also of interest is that if I prepare standards of the same concentrations mentioned above by spiking into placebo samples, I get a very small intercept (0.3% of the 100% response).

Any suggestions?

Thanks in advance

I would run your calibration curve down as low as you can go, maybe around 5 or 10 % of your target conc. with a few points between there and 50%. Ideally you will be able to show that the lower you bring the calibration curve the more the y-intercept approaches zero. If you can show that this is the case then you can force your line through zero. Not sure about the placebo, I would assume it is being separated away from your std on column. The main in thing is that the slope is consistent more than the y-intercept.

I think that you may be outside the upper linearity range of your detector. At the upper end, the signal goes down. If your upper calibration signal is anywhere close to 1 AU, this is the problem. If you are not there, it is still possible that the whole thing is caused by a dirty cell window, or a low lamp intensity.

If Uwe's first suggestion is correct, you should be able to see it by comparing the shape of the tops of peaks - the high concentration samples (if non-linear) will appear to be different from peaks that are in the linear range of the LC - specifically, check for plateaus. Excessive fronting as concentrations rise is indicative of too little column capacity for the amounts of analyte being injected...

The signal tops out around 200 mAU, so I don't think that I am exceeding the linear range of the detector. The peak shapes are consistent through the entire range I studied.

I checked the lamp hours on Uwe's suggestion, and indeed, this lamp is pretty old. I have set up a run on a second instrument where the lamp was recently changed. Perhaps this was the problem all along.

Thanks for your input.

Non-linearity can occur long before the detector is saturated and produces a plateau.

Any of the "normal" deviations from Beer's Law may be present as analyte concentration increases. Then, regardless of the analyte, the detector will have a linear range due to its design.

Indeed the run repeated on a different instrument with a recently installed lamp gave a "zero" intercept.

Thanks for your suggestions.