Uneven retention times (HPSEC method)

[michbrrt]

Sorry for the small figure. The pressure variation here was between 20.5 and 21.3 bar over 72 runs. This corresponds to the erratic Tr jumps in fig. 2, so I see no correlation. My own experience from the days without problem (almost 15 years non-stop) is that, at this rather low pressure (not comparable with that obtained with RP columns), pressure variations do not translate into Tr variations. On the other hand, intentional changes in flow rate give a proportional change in pressure. What is the use of an HPLC pump if it can't deliver a steady flow at variable pressure? In our case the pressure build-up has almost always been linked to dirty samples. Changing the pre-column filter has always brought the pressure back to normal.

It sounds more and more like a valve issue... Trying with a few more experiments. Will keep in touch.

[AdrianF]

As I said earlier in this discussion an accurate flowmeter is a great (Accurate to 3 decimal palces) diagnostic tool and would be able to confirm that the problem is flow. You then have clear evidence to show the pump manufacturer or service engineer.

Sometimes flow consistency can be improved by using a back pressure regulator. Fluctuations in pressure are then a much smaller percentage of the total.

[unmgvar]

if you still cannot solve this after looking at the check valves, will you mind sending me all the runs/sequences that you showed in your trend plot as a .cmb file?

if the problem turns out to be a check valve then maybe moving to the ceramic kind is more appropriate for you.

[michbrrt]

Looks like you work at Dionex! I'll send the data later if my problem doesn't get solved soon.

Yesterday, I decided to bypass the column selection valve (placed in the oven) because I had noticed injection peaks in almost all of my runs. Since then, the injection peaks are gone (so the valve was leaking!) but my Tr haven't improved.

Earlier this year, there was a strange incident: brown deposit was found inside the PEEK tubing pieces connecting the valve to the columns (we use a system with a pre-filter cartridge, then a guard column, then 2 columns in series; the alternate valve position is used for a bypass).

Do we need to use another type of rotor seal with our buffer (contains 0.15 M NaCl)?

[unmgvar]

michbrrt,

i do not work for Dionex (it would be written otherwise) but i do use their instrument and software extensively. i know my way with an ultimate 3000.
simply your case seems application induced and if i get a look at the chromatograms i can get a better idea of your problem.
there are 2 types of rotor seals. for less then ph 10 or more then ph 10.
the standard one can old up to ph 10.
the system can stand ph as low as 1. check the solution ph, but i doubt that it is in either extremes

[HW Mueller]

What is an injection peak? How do you think a leaking (leaking what?) selection valve introduced these peaks? Are you using one of those "last drop" filters in the mobile phase bottle which sucks in air when the mobile phase level is low enough?

[unmgvar]

michbrrt,

i dealt with a matter this morning that reminded me of you for 2 reasons.
1. RT problems, which in ,my case where related to unstable flow due to a leak.
2. there was something else there as well.

i believe that when faced with a complex problem to troubleshoot it is best to follow a step by step methodology and to "divide and conquer" dealing with one issue at the time.

in my case we started by RT shifts that would occur randomly over a period of 45 injections. 10 injections gone bad, only standards runs, normal RT was 9.5 minutes. bad runs would have anything between 10 to 15 min RT. peak area also changed, being bigger, sometimes with correlation to the ration of change of the RT. this pointed out to a flow precision problem but not only.
first step we washed the system from all buffers with water.
the leak at the pump head then became clearly apparent (sometimes with buffers the leak is shut by the dryed buffer, reestablishing normal flow)
we dealt with the leak, checked that the system could hold the pressure.
we ran an RSD of RT and area check. RT this time passes, area does not.
now that we have solved the flow issue we can move on to the second issue, probably somewhere in the autosampler.
looking at it we saw that the syringe tip needed replacement.
this was done, the check done again. system is back for service.
the fact that only standard runs had gone bad was simply weird statistics. it can happen and it has nothing to do with the case.

[michbrrt]

This is a reply to the last three posts.
1.
I don't think my problem is application-induced, for several reasons. This isn't a new application (already 15 years) and it has worked perfectly on 3 different systems (Gynkotek, Waters, and Dionex for the first year). All the last experiments have been done only with standards (ethylene glycol, EtOH and acetone). We also switched to a new set of columns and made the eluent ourselves. Concerning the hardware, the pump has been changed: that's why we now focus on the rest (autosampler and valves).
2.
The injection peak occurred at ~0.2 min and showed with RI for all standard, and with UV only for acetone. It disappeared when the valve was bypassed. The valve (6-port design) has 2 positions: columns and bypass. If it is loose, it is very possible that a small part of the flow (and the sample) goes directly to the detectors We don't use last-drop filters and the eluent has never reached critically low levels.
3.
We have eliminated the pump as possible cause, simply by replacing it. The rest is the responsibility of the vendor, since we have a maintenance contract! So far, the maintenance visit hasn't been performed because the diagnostic tests showed no problem! In all the experiments, the peaks look very good (plates and asymmetry), only RT varies randomly.

[HW Mueller]

I am not sure of your setup anymore , and I still don´t know what is an injection peak, but if you have air getting into your pump you can change to all the pumps in the world and still have the possibility that air gets into it.

[unmgvar]

how does the area of your peaks behaves?
is the RSD as bad as that of your RT?

[michbrrt]

About the two previous posts:
1.
The so-called injection peak was much smaller (~3%) than the main peak and had also properties that indicated it was of the same nature as the sample (namely UV absorption in the case of acetone and not with eth. gly. and EtOH). We are pretty sure no air gets in the pump from the eluent bottle. The pump has also a built-in degasser. If the system had a faulty design, it would not have worked well at the beginning (1 year without problems). See setup description in the first post.
2.
We have seen area RSD up to 4%, but generally below 0.5-1% (1% for a quantitative determination is quite good). Of course, this depends on how the baseline is traced. Sometimes, especially with eth. gly. which has neighboring salt peaks, we need to do some manual corrections. With the peak maximum determination, there is no ambiguity: the signal/noise ratio is quite good and RT is always automatically calculated by the software. We are mostly interested in MW calculation, so a high RT variation has a huge effect on it. I don't see a clear correlation between our bad RT RSD and bad area RSD, but the high area RSD (4%) could also indicate a problem with the autosampler.

[unmgvar]

there are 2 general types of behavior that can affect the flow.
the first one is constant over time. for example a "steady" leak then the flow is 0.8 ml/min instead of 1 ml/min for example.
the second one is a sudden change in the flow rate that "corrects" itself overtime, but the average flow is 0.8 ml/min instead of 1 ml/min for example.
in the first case your coumpound will be in the flow cell at a lower flow rate therefore the peak area at 0.8 ml/min will be bigger then that of the compound at 1 ml/min in inverse proportion. slower flow rate bigger area.
in the second case the disturbance might or might not happen when the compound is in the flow cell because it is happening very fast and most probably randomly, so in general peak area is with a decent RSD.

the thing that has me confused here is that you say the pressure changes if there are, do not look linked to the change in RT. if this is really the case then i can't see where the flow is at fault for the RT shift because there is a direct connection between the 2. even a problem in the autosampler would show as an interference in the pressure.
if the pressure cannot be linked to the problem then you should move away from a flow rate problem to another interference.

[HW Mueller]

michbrrt, since I am not convinced that you have no correlation between rt and flow I will expound on the air bubble in the pump problem. One loose connection, or one bad ferrule which no longer tightens, a fault in tubing can cause leaks. Replacing one piece (the pump) in the system may not effect the position of the leak at all.

unmgvar, are you suggesting that flow variations are only important if they happen just when sample is going through the cell?

[michbrrt]

Reply to the two last posts:
1.
Isn't the area of a peak directly proportional to the quantity of compound, therefore independent of the flow? If there is no loss, I would not expect an area change but rather a peak height decrease (and width increase). Regarding the pressure variation, they seem random and don't make much sense, but there's a continuous pressure increase in experiment with an important RT drift. What puzzles me is that I had seen wide pressure variations in the past years, without this affecting my data. Another strange thing here is that each time I have measured the flow (25 mL calibrated volumetric flask totally filled, then flow rate calculated from time needed to fill), it was ~0.94-0.95 mL/min instead of 1 mL/min. Not having a flowmeter, I am unable to measure the flow in a continuous mode.
2.
So far, no bubbles or leaks have been found but your remarks are perfectly valid.

[michbrrt]

Maybe this is a happy conclusion to our ordeal: last week, Dionex came and did a complete preventive maintenance on the system. In addition, all tubing in contact with the samples (sample loop...), and all rotor seals, have been changed. The system passed all tests with flying colors. Previous tests didn't spot the problem, but they were mainly done on the pump. As recently suspected, the problem may have been with the autosampler: intermittently plugged tubing, and maybe defective valve(s), causing the various rises in pressure, and flow and RT variations. Indeed, one common characteristic of all the bad data is that the RT were higher (up to 3 min!) than normally expected with this system. The exact cause of the problem has not been found (no obvious leak or damaged seal), but the technician pointed to us the poor quality of the Alltech seals (PTFE/silicone/PTFE) that we had been using: after several injections, there was a distinct hole in the septum, therefore possibly rubber particles in the flow path. Since then, we use again our old Waters seals, and pray for the best!

I'll stop following this topic regularly, but you are welcome to keep the discussion alive.