Downward Drift as %B Increases???

[DR]

If the "cleanout" portion of a gradient is executed too quickly, it is possible to retain some undesirable material on a column and have more of it elute during the following injection's cleanout.

Not sure that this is what's happening here, but there is a mechanism for what you describe.

[mburleson]

Update / New Information:
We received some additional information and new data this morning that I thought I would share. I've also got a new hypothesis that I'd like feedback on.

1) The lab confirmed that all recent runs have been on Aquity H-Class UPLCs with Quaternary pumps and TUV detectors. They confirmed that they have NOT been using a reference wavelength (so that hypothesis is officially shot down...)

2) They did perform a run on H-Class system with a PDA recently (again, no reference wavelength applied), and the profile did look significantly better (almost no downward drift); however, they confirmed that one other factor was changed at the same time for this run: They didn't filter the buffer for this run (they've typically filtered the stock buffer using a 0.2µm nylon membrane). So, this provides us 2 potential contributing (and currently confounded) factors for further evaluation: Detector type and filtration. I'm interested in anyone's thoughts on whether or a nylon filter may be contributing a contaminant to the MP.

3) I'm realizing that I did not specify the column temperature in my original post (50C). One of my colleague brought up the idea that changes in refractive index of the mobile phase may be contributing to the baseline drift. More specifically, if the mobile phase is at a high and/or inconsistent temperature when it reaches the flow cell, this can cause a higher degree of sensitivity to RI changes in the MP. We're thinking that this may align with the recent observation that baseline drift was different when using a PDA detector, as the PDA design may be less susceptible to this effect (maybe the PDA flow-cell has more thermal mass, thus stabilizing the MP temp in the detector?). Does anyone have experience with RI and/or MP temperature variation affecting the baseline?

Changes in the RI of the MP is a good reason to use a reference wavelength:
https://www.chromatographyonline.com/vi ... ction-hplc
https://www.agilent.com/library/support ... faq182.pdf

Is the method validated or could the use of a reference wavelength be assessed?

[AnalyticalWisco]

::reread OP::

pH 3 with phosphate - not ideal, (you're not in a pH range where the PO4 provides any buffering capacity)... so even if your stocks are perfect and clean, it could still be the MP causing problems.

What MP glassware are you using? Schott glass can be problematic with ACN and very sensitive UV detectors...

Yeah, I agree that phosphate is not the best buffer at pH 3.0. Ironically, we were forced to move away from ammonium formate to potassium phosphate because their systems were so noisy when they ran with ammonium formate (we couldn't get adequate S/N at LOQ). They spent quite a bit of time trying (unsuccessfully) to get noise down to acceptable level before making that switch. In hindsight, the buffers switch appears to have been a band-aid for a bigger issue...

I'm not sure what glassware they're using. If other's think this is a possibility, I'll raise it on our next call.

[AnalyticalWisco]

Hi Wisco,

That is a very important distinction, and I'm sure many of us did not realize that was a blank/diluent injection. If this is a blank injection, there is certainly something on the column or in your injection matrix that is causing all of those peaks, and no, that amount of peaks is pretty irregular for blank/diluent injections, this should be a concern.

Thanks for the comment. I have to admit that my hands-on HPLC experience is pretty old at this point. While I still consider myself an analytical chemist, it's probably been at least 15 years since I've actually touched an HPLC. I did find the number/magnitude of peaks in the wash phase of the gradient surprising, but I suppose I didn't have reason to question it, as all of the peaks of interest come off well before the wash.

General consensus from the folks in this forum seems to be that we're seeing a pretty major contamination issue. We've been pushing for months to reduce potential sources of contamination (eg, stop putting the pH probe into the bulk buffer, start using HPLC-grade reagents, etc), but it seems we still have a ways to go on that front. Unfortunately, this lab is overseas, so I don't have the option of spending a day in the lab identifying/studying potential contamination sources, so we may be stuck with what we have for the time being.

[AnalyticalWisco]

Changes in the RI of the MP is a good reason to use a reference wavelength:
https://www.chromatographyonline.com/vi ... ction-hplc
https://www.agilent.com/library/support ... faq182.pdf

Is the method validated or could the use of a reference wavelength be assessed

Yes, the method is fully validated at this point. May be worth a trial using a reference wavelength just to rule out RI/Temp effects, even if that change can't be easily implemented. If we find that it is a temperature effect, we could ask them to run on systems that have active temp control on the flow cell (I believe Agilent 1260 PDAs have this feature).

[AnalyticalWisco]

If the "cleanout" portion of a gradient is executed too quickly, it is possible to retain some undesirable material on a column and have more of it elute during the following injection's cleanout.

Not sure that this is what's happening here, but there is a mechanism for what you describe.

Good point, but I think we're being pretty conservative with the cleanup portion of the gradient (6 min @ 100% MPB). If you look at the example in the original post, there's nothing eluting for the last two minutes of the cleanup phase (ie, from 24 - 26 min). My interpretation of that is that the wash is adequate, but I could be misreading that.

[Multidimensional]

mburleson wrote:
"Changes in the RI of the MP is a good reason to use a reference wavelength:
https://www.chromatographyonline.com/vi ... ction-hplc
https://www.agilent.com/library/support ... faq182.pdf"

"Is the method validated or could the use of a reference wavelength be assessed"

----
Those "articles" are worthless sales notes. They do not understand HOW "reference wavelengths" work as a software feature (The articles on "chromatographyonline" are written by sales people). The Agilent sales note leaves out 99% of the actual info. Please do not get your training from such sources *At least the very short Agilent noted started out with one true statement, which was that the wavelength signal you selected start off by subtracting out the reference signal, CONTINUOUSLY, to create the reported wavelength signal ( a newly created signal made with no traceability ) and the deletion of the original signal.

Anyone knowledgeable in chromatography would understand how this type of manipulation of the data is completely unscientific and immediately invalidates both your HPLC method AND all data collected. I*We have consulted on many cases where labs were audited to find out that they were using the software programmed "Reference Wavelength" resulting in recalls, action letters or suspensions. TURN IT OFF !

Here is an authoritative article on the topic:

REFERENCE WAVELENGTHS (as used in HPLC UV/VIS); https://hplctips.blogspot.com/2011/03/r ... -hplc.html

[AnalyticalWisco]

Quick Update

The lab has now run the same mobile phase and column on three different UPLC systems, one with a PDA, and two with TUV detectors. The downward drift is only observed on ONE of the TUV systems (the one they happen to be using most frequently). So, it seems that the downward drift problem is related to a 'system-specific' issue. They have confirmed the the two TUV systems are virtually identical in terms of components/tubing/etc. This leads me to believe that the most likely cause is either 1) system contamination, or 2) a system performance issue.

I'll provide another update if they are able to further pinpoint the root cause of the issue, but I'm feeling a lot better about the situation now having confirmed that this is not a fundamental issue with the test method.

Thanks again to everyone who commented!

[DR]

Time to passivate the TUV system...

[TylerSmith123]

This thread has made me think about the process of validation differently. In my opinion, a blank injection such as the one in the OP would certainly be flagged during a validation test, however it seems that they had no issue. Is this simply because validation can be any sort of test/tests you want as long as there is no outside auditor putting regulations on you? What is the point of validating a method when you may have these issues going by with no-one asking questions? Is it simply for posterity of the "chromatographer" running the analytes? A little bit of making themselves feel good? Certainly if you export this method to a naiive UPLC, with proper controls and standards, you would not get the same result as OP in his original picture. So then what is the point of validation?

[Hollow]

Thank you for keeping up with updates.

hm, now not sure anymore if it's still contamination.
A detector issue seems more likely.

I guess the TUV are of the Acquity family with its light-guiding tube design?

May your customer do a test where they interchange the flow-cells of the two detectors? Does the issue change the system too? -> then the flow cell seems causing the issue.
I guess there won't be anything to repair, just an exchange.

If the downdrift stays on the system, then the optic bench and/or an alignment issue with the flow-cell assembly may be guessed.

my thoughts:
a negative baseline/peak means that more light is getting to the detector cell (photomultiplier). So search for something that may cause this... and its correlated with the increase of MP B.
The Acquity flow-cells makes use of total refraction to provide enough pathlength.
So it seems that somehow with the change of the MP A/B and therefore the change of its refractive index, more light (energy) is getting to the detector cell.
Either because it's getting less reflected (less pathlength) inside the flow-cell or by getting better aligned with the optic bench of the detector.