Peaks creeping forward on Shimadzu HPLC

[osp001]

Running a C18 column, 25/75 (0.1% formic acid/water, 0.1% formic acid/acetonitrile), do my morning warm-up and everything looks great. Over the course of the day, normally the peaks don't shift at all. But some days, after 5-10-20 runs, some of the peaks towards the end of the run will speed up 0.05 minutes or more, and it "caterpillars" forward to where peaks that elute earlier start to come off earlier, too.

Some days it does this, some days it's bang on for the entire day. I'm not an HPLC guy, so I don't know if this is normal or not. The oven holds temperature just fine, the room is climate-controlled, everything seems to be operating consistently from run to run.

Anything in particular I should look for?

[bunnahabhain]

Your description implies that the next morning the retention times are back to normal. So what is changed before the next warm-up?

[osp001]

That's a great question and I shouldn't have glossed over it.

Normally I run samples first thing in the morning, and I idle the machine at very low flow rates (0.1 mL/minute, ~160 psi) once samples are done. Sometimes I run them in the afternoon or the evening, and even if I "re-start" the instrument, the shift is still there.

The instrument gets powered down overnight, and then started up the next morning with the same routine as always: turn on heat (to 50C), turn on pump at low flow rate (0.1 mL/minute) until at operating temperature, and then run a blank and a standard. Sometimes the solvent blank comes up with a few tiny glitches, but nothing outstanding. And the standard always gives me nice peaks at the usual retention times.

About 1/3 the time, they shift within half an hour to an hour of operating at 1.5 mL/minute. Sometimes it's abrupt, sometimes it's gradual.

[itspip]

When you say that you "powered down" the machine, do you actually turn the power off, or are you just setting the flow to 0mL/min and using the software to turn off any lamps?

[osp001]

The instrument is shut down for the evening using a shutdown method from a file; it allows 5 minutes for the column to cool down (inadequate, I know, for the column to go from 50C to ambient, but I cannot see any harm in the column at 40C by the time it powers off), and "Power Off after shutdown" option is selected for this instrument. By the time it is done, the fans have kicked off, the display is blank, and the instrument is quiet.

Interestingly, today I powered it up, ran the solvent blank and standards, started running- and got beautiful results, up until about 2 PM when the shift suddenly appeared. Once I plugged in fresh values for the retention times on the various compounds (cut and pasting the same set that was generated yesterday), the peaks were all identified correctly and quantitation was where it should be.

[ccharles]

you say that the peaks creep forward for 0.05 minutes.
0.05 minutes is 0.05 *60 seconds = 3 seconds.
that is very little time. Your HPlc software should recognize them anyhow. They generally have a 5% time window.
For example if your peak elutes at 5 min the time window would be: 5*60 *.05 = 15 seconds.
If this is the case I wouldn`t worry to much. If the peaks are not recognize by the software, then there could be a problem.
Time shift can ocur if you do not have an oven in your system. When the ambient temperature heats up the the peaks will elute earlier.

[osp001]

But the bandwidth is normally set for 0.05 minutes, and if I open it up (0.08, 0.10, whatever), then it starts to identify peaks incorrectly, saying that "X" is actually an adjacent compound. It's defaulting to the wrong identification when I do that.

It is more convenient to simply change the retention times, deducting a few hundredths in order to get the correct ID.

Why this happens in mid-day (if it happens at all) is what is concerning.

[Perreman]

Could be worth taking a look on the pressure plot during the run and see if you can notice any changes when this effect is first noticed.
Are you using a quartenary pump or a binary pump? Could possibly be an issue with a bubble stuck in the inlet valve of a quartenary pump.

[DR]

Just make your retention windows (what you called bandwidth) a little wider.

[Peter Apps]

But the bandwidth is normally set for 0.05 minutes, and if I open it up (0.08, 0.10, whatever), then it starts to identify peaks incorrectly, saying that "X" is actually an adjacent compound. It's defaulting to the wrong identification when I do that.

It is more convenient to simply change the retention times, deducting a few hundredths in order to get the correct ID.

Why this happens in mid-day (if it happens at all) is what is concerning.

If a 3s shift in retention causes the software to misallocate peaks your retention windows must be very close together, which implies a very small separation between peak apices and unexpectedly narrow peaks - more like GC than HPLC.

The consistency of the shift is a puzzle in itself - if the column was deteriorating or the pump was playing up it would be more erratic. What is the resolution of the time scale readout - how many decimals of minutes ?

What are your samples, and what are they dissolved in ?, and does running a particular type of sample precede the retention shift ?

Peter

[itspip]

Is your solvent fresh every day? I'm thinking that there may be something going on the the pH due to dissolved air that the online degasser isn't taking care of. Or that the degasser may be on its way out.