We have a method with 0.1%H3PO4/ACN (A/B) gradient and a C8 3.9x50 mm column for the analyses of three components. The mobile phase starts with 80%A and 20%B for 3 min then 10%A and 90%B in 27min. The first peak eluts at ~ 2 min. Peak 2 appears at ~8 min, peak 3 at 14 min. The method has been used for several months and results were consistant (include the validation results). However, recently, the results of peak 2 appears lower than the other two peaks. In one of the system suitabilty results, %RSD of repeatability of peak area of Peak 1 was ~3%. %RSD for Peak 2 was ~22%, and Peak 3 was ~3%. Why three peaks have different %RSD? If it is a sysmatic problem involved why Peak 2 affacted the most. I tried swich to different column and instrument with the same samples and the problem seams corrected (Resultsre consistant among all three components). Please tell me what you think casue the problem. Thanks.

I would say it is time to replace your column. When the column gets older, it's stable phase is partually destroed and some silica become available to the molecules in the sample. Looks like your compound # 2 forms stronger bonds with silica than the other two. It is simply partually retained on the column.

%RSDs are on the high side to start with! If peak 2 is usually smaller and/or broader than peaks 1 & 3, then it should have a higher %RSD - but not 22%. If all three peaks were small enough to be near the limit of quantitation, maybe the s/n ratio of the system is not what it used to be. How many hours has your lamp been in service? What kind of reference energy levels are you seeing and how do they compare to those of a fresh lamp? If all of your peaks are broad, is it because the samples are in more than 20% ACN? Is your gradient formation correct (many pumps do not pump pure ACN very well)?

I agree with DR, the "good" RSD's of ~3% are not really that good to begin with. Area RSD's should better than 1% and really your should be able to be better than 0.5%. Of course, there are many variables that affect RSD, but you should be able to do better than 3%. 22% indicates that something very bad is going on!

Actually, your problem may be an instrument related or a combination of both (column and instrument). I do not think a pump is a problem since no change in retention time is pointed out. A detector could be the case. It is easy to locate the sourse. Just try to use a new column on the same instrument and compare the results back to those obtained on this column and a different instrument.

Also, as a general thought not related to your current problem... There are a few things in your method that I would change if allowed to do so. It may be impractical since the method is already validated, but consider the following ideas:

1. Retention time 2 min for a peak of interest is too early. See at least J.Dolan's LC Troubleshooting in the February 2000 issue of LC/GC on this topic.

2. Composition of the components A and B in the mobile phase. An HPLC usually works better when there is some organic component in your aqueous phase (the two components are mixed easier).

3. Do you give your system to equilibrate back to 80% A & 20% B after it is been up to 10% A and 90%? At least a few minutes depending on the system configuration is needed.

I should add more details about the method and experiment.
1) The gradient are: time 0-3 min: a 0% b 20%; 23 min: a 10% and b 90%; 27 min a 10% b 90%; 30 min: 80% b 20% 35 min a 80% b 20%.
2) The % RSD for all three peaks normally were <0.5%.
3) A new lamp was replaced after experecing ther problem and the poor %RSD and low %LS still shown, espacially for peak 2.
4)In one experiment, the Rt of peak 2 was shift from 7.0 to 6.4 min (one sample only). In the same experiment, the Rts of peak 2 of standard injections (total 7 throughout the run) were close (%RSD <1%).

You still did not give the complete information necessary. One thing that we need to know is the flow rate, and a few details about your instrument.

It is not impossible, that the troublesome peak elutes right at the breakthrough time of the end of the first part of your gradient, which is programmed to be at 3 minutes. You need to know the gradient delay volume of your instrument. We can make an estimate of the column dead volume and see if all this information agrees with the elution of your second peak. Then we can figure out why the rsd is so high.

Flow rate: 1 ml/min. Void volumn: 0.47 ml.

Ok, we are getting closer. What I had asked you about was the gradient delay volume of your instrument, which is the volume between the point where the gradient is mixed and the top of the column. This volume can be as low as somewhere around 0.7 mL to very high numbers, 5 mL not being unusual. This means that your gradient does not reach the column top until this gradient delay volume is flushed.
The reason for this question is the following: if the gradient delay volume is 5 mL or slightly more, then the problem peak elutes during the steep portion of your gradient that you programmed to be at the beginning of your chromatographic run. In a steep gradient section, one often gets coelution of junk that has accumulated on the column or may come from other sources. This means that the quantitation of the peak around 8 minutes is not reliable due to coelution. While in an isocratic section or in a flat part of a gradient coelution can be avoided, it is more tricky to do this in a steep gradient.
Anyway, I don't know if this is the answer, but it is worth checking.