Uneven retention times (HPSEC method)

[michbrrt]

For many months, we have had trouble with unstable retention times (Tr). The Tr RSD can sometimes be over 1% and my peak tables become useless. The vendor has already tested, then replaced the whole pump (with built-in degasser and proportioning valve), but there was no improvement. New columns, freshly-made buffer: no improvement! We have tested 3 different standards (ethylene glycol, ethanol, and acetone) to make sure this isn't related to our normal standard (EtOH). The filters (in eluent bottle, and in pre-column cartridge) have been changed several times. In some experiments, the pressure has increased dramatically between injections (e.g. from 20 to 35 bar), but this has never caused this type of Tr variations before. We have also checked the flow rate with a volumetric flask: it was only within 95% of the required rate (0.95 vs. 1 mL/min), but the vendor insured us that this was normal. Is it possible that the autosampler is causing all this trouble, by e.g. delaying injection?
Hardware: Dionex UltiMate 3000 (quaternary pump with built-in degasser and proportioning valve; autosampler; oven with column selection valve; 4-wavelength UV detector; RI detector)
Columns (HPSEC): Shodex OHpak SB-805 and -804, in series, with guard column
Method: isocratic, 1 mL/min, 30 min
Eluent: phosphate buffer, 10 mM, pH 7.2; with NaCl 0.15 M

[AdrianF]

It is worth checking the flowrate with an accurate flowmeter
eg http://www.v-kit.com/vtools/HPLC_Flowmeter.pdf
It is likely that the change in pressure has a transient effect on flow rate and hence retention time however you should be able to be able to set your peak window to still detect the peak.

The autosampler should not be a problem as it should trigger the start of run in a consistent way.

[michbrrt]

Thank you for the first comment on my post!
In my last experiment, the pressure was perfect at ~20 bar (over 24 injections), but I still got bizarre Tr variations.
Flow: isn't a calibrated 10 mL volumetric flask accurate enough? A variation of 1% is easily detected (6 sec at 1 mL/min). In this last experiment, I got Tr variations of 1-2% between two runs. I am currently checking the flow over several runs to see if there is any variation.

[unmgvar]

was the pump replaced because it was faulty?
have the technicians done a flow reproducability check according to their SOP?
dionex procedure include a flow check accuracy with a flow meter, but this is less important then the reproducabilty check.
the 95% is actually a bad result for accuracy but again the reproducability is more the issue.

from the little you wrote is it safe to say that you are doing a SEC separation of a protein for mono, di and HMW?
can i conclude that the sample is dissolve in the same buffer or something close (buffer)?

are you monitoring the pressure in your runs? if not then do it and see if
~20 bars is not more then +- 1 bars.
is the retention time changing in a random manner or does it go forward or backward?

when doing SEC it is always good to add an RT marker to all your injections this way you can adjust the currently occuring shifts that you have.

i have had a case of RT shifts because of clogging in the autosampler. that was due to dirty samples, but it was in a gradient method. we knew that was the cause because we monitored the pressure and saw huge uneven pressure changes right at the first seconds of the runs.

[michbrrt]

Thanks for the reply!
We are separating both proteins (little) and polysaccharides (a lot). They are dissolved in compatible solvents (water or buffer) but rarely in the same PBS buffer used for the separation.
Our current problem is that we are unable to get reproducible Tr for our standards (ethylene glycol, ethanol and acetone). We use these only to check the column plates and asymmetry. The results are perfect but the Tr varies wildly.
In the last experiment the pressure stayed at ~21 bar during 3 days but the Tr variations were again enormous and random (up and down).
I am getting desperate because the vendor just refuses to believe that the instrument could be faulty. By the way, the 1st pump was not deffective (according to the tests) but was replaced just in case.

[tom jupille]

Our current problem is that we are unable to get reproducible Tr for our standards (ethylene glycol, ethanol and acetone).

As I read that post, you have all three peaks in the same injection. If that's the case, the diagnosis is relatively easy. If the relative retentions of the peaks are constant (i.e. all peaks show the same % change in retention), then you have a flow rate problem (well, to be technical, a change in column volume could cause the same effect, but somehow I doubt that your column is growing/shrinking).

[michbrrt]

The standards are injected separately. Our normal calibration routine uses ethylene glycol (plates and asymmetry), then a large MW marker, then a series of MW markers (2 injections with 4 standards each: total 8 standards), and finally 2 known samples. Because of the Tr problems, we started looking at EtOH and acetone to replace eth. gly. but all standards are separately injected (4 inj. eth. gly., then 4 inj. EtOH, then 4 inj. acetone, and the whole series is repeated once more). The reason for the repeated injections is that we want to track the reproducibility problem by measuring the RSD of the Tr. Again, we get horrible results, where one standard doesn't behave well in one series of injections, and even worse in the next series. Day to day variation is even worse. We are convinced this is a hardware problem but have been unable to spot it so far (the new pump and columns didn't solve anything).

[tom jupille]

How about trying a series with a "mixed" standard (high MW + low MW) markers and then looking at the retention time ratios?

The issue is trying to figure out whether the problem is a flow variation (most likely, except that swapping out the entire pump should have fixed it) or something else (temperature maybe?).

[michbrrt]

This is what the data looks like (I hope the quality is OK):

The same is observed for the two other standards but they were injected separately. Nevertheless, after the Tr of one standard has increased abnormally, the following injections show the same trend. Until the next event. Each series of 24 inj. takes about 12 h, then it is repeated twice. Each arrow on the graph indicates a new series of inj. or a rather abnormal event.

[michbrrt]

This is what the data really looks like (RI profiles)

The two first inj. gave much earlier Tr then what came next. The same shift can be observed for the negative salt peak at 24 min.

[tom jupille]

The same shift can be observed for the negative salt peak at 24 min.

Aha! That says you have a flow problem. If the first graph you showed was representative (a sudden increase in retention followed by a gradual drift back toward the original value), then I'd be inclined to suspect an air bubble in the pump or a check valve which is gradually redissolving.

[unmgvar]

the reason your pump was changed was to show that it is not a mechanical problem of the pump itself but something induced by your application.
we are of course also assumimg that you have a column oven and that you are not doing the separation at room temp.

do you have pressure chromatograms of those injections?
if not then do collect some.
if you get sharp changes in a very short time during the run then it is air bubles.
if you have one slowly occuring change over time or even one sharp cahnge but slowly the pressure restores itself then you probably have to look at the check valves. generally your RSD should be bad as well.
the thing is that you need to understand why you have a check valve problem.

if your sample are dissolved in a buffer then the autosampler might be doing the problem but i doubt it.
for that you need the injection port to leak. this means that you are maybe not doing external washes of the needle or you have the wrong settings or using the wrong wash solvent, then slowly

[HW Mueller]

The best explanation of all the info we have to this point is that of Tom. Now I learned to work with such a best fit instead of rambling around. Thus, michbrrt should be well adviced to check whether his mobile phase is properly degassed or whether his plumbing ahead of the pump is tight.

[michbrrt]

In previous experiments there was a gradual increase of pressure that magically disappeared between experiments (by the way, I always collect pressure data). The variations in Tr were random, so no direct correlation (no gradual increase or decrease).



In the last experiment, there is absolutely no significant change in pressure (see fig. 3), but the Tr variation is still enormous (RSD ~1%; see fig. 1 from the 1st red arrow). I also checked the oven temperature data: it never deviates much from 35°C.

Should I still suspect the pump, even if it has been recently replaced?
Could it be a problem with the column selection valve (6 ports, 2 positions; the 2nd position is taken by a short bypass loop) in the oven?

[tom jupille]

It's hard to tell much from the small figure, but if those are pressure traces, it looks like quite a bit of variation. Other things being equal, pressure and retention time are both linear functions of flow rate. A 1% RSD in average pressure run-to-run means a 1% RSD in retention time.

HW and unmgvar are right, the problem is not necessarily in the pump, but could be an intermittently leaking fitting, a valve issue, an air bubble, or . . .