Baseline ripple problem

[sdegrace]

We are having a fairly intractable problem with a baseline "ripple" or "wave" effect at certain gradients at 220 nm which is effectively killing our LOQ and LOD (small peaks look indistinguishable from this "noise", causing our noise level to effectively appear higher for s/n calculations). Examples are below. I would appreciate any insight anyone can offer!!

We are using a Waters Alliance 2695 system with Waters 2487 dual wavelength detector. We also have a Waters 2996 PDA detector on the same system and have collected data with both; the "ripple" looks identical between the two, which says to us the detector is not the problem. We have to do related impurities analysis at a fairly low concentration and need the sensitivity of the 2487; the contortions of the baseline effectively erases the difference between the two for us in terms of their usefulness.

We have been back and forth with Waters on this and tried many experiments they have asked of us including changing the suppliers for the mobile phases used (we were already using HPLC grade solvents; we also tried substituting MilliQ for the water). The solvents are modified with TFA, we have been adding J. T. Baker HPLC grade TFA shipped in 1 mL glass ampoules. We have tried zero volume injections, injections with nio column, etc. The problem disappears in cases where we inject with no column, but using a different column doesn't make the problem go away.

The mobile phases in all cases:

• * Mobile A: Water + 0.05% Trifluoroacetic Acid
  * Mobile B: Acetonitrile + 0.05% Trifluoroacetic Acid

Mobiles are prepared fresh and are not filtered, to avoid any contact with outside sources of contamination.

Column (for the examples below): Phenomenex Luna C18(2), 100A, 250x4.6 mm, 3 um packing.

UV detection at 220 nm. 10 uL injection (although identical effect seen with 0 uL injection).

Method 1:

Code: Select all

Time (min)   Flow (mL/min)   %A       %B
0            0.8             75.0     25.0
30.0         0.8             60.0     40.0
50.0         0.8             30.0     70.0
51.0         0.8             5.0      95.0
55.0         0.8             5.0      95.0
56.0         0.8             75.0     25.0
65.0         0.8             75.0     25.0

Method 2:

Code: Select all

Time (min)   Flow (mL/min)    %A       %B
0            1.2              95.0     5.0
10.00        1.2              75.0     25.0
15.00        1.2              5.0      95.0
20.00        1.2              5.0      95.0
21.00        1.2              95.0     5.0
30.00        1.2              95.0     5.0

These are methods I am transferring from a different lab and I don't have any choice about any of these parameters.

[Bruce Hamilton]

My initial suggestions:--

Check the absorbance of your mobile phase components at 220nm, it should be low.
Check the lamp energy/hours of use, ensure it's still OK.

If the mobile phase is absorbing strongly, and the lamp energy is low, you may see a more noisy baseline.

I'd try measuring the noise with some premixed isocratic systems, eg 75:25, 50:50, 5:95, and compare their noise to that obtained using the pump mixing of equivalent phases after 10 minutes isocratic.

The pump mixing system may be struggling when you apply some back pressure. The pressure from that column would be significant.

If the above suggest mixing is a problem, you could change your mobile phases to 75:25 and 5:95, and adjust the gradient accordingly, see if that improves the situation.

Bruce Hamilton
Bruce Hamilton

[Mark Tracy]

This is a classic TFA baseline problem. You have done all the right steps to avoid contamination and use quality reagents. There is a fundamental problem that you need to minimize. Because TFA is somewhat retained on the column as an ion-pairing agent, its concentration in the mobile phase after passing through the column is sensitive to minute fluctuations in the water/MeCN ratio. Since it also has significant absorbance <230nm the baseline is showing these concentration changes. In effect TFA amplifies mixing noise from the pump. When I have had this problem, I usually increase the solvent mixing with an extra (or larger) static mixer. The problem never goes away entirely, but you should be able to reduce it.

[Uwe Neue]

I agree with Mark. It is an issue mobile phases containing TFA at lower wavelength. I also agree with the fix: an additional mixer. If your service rep can't find the part number, I can find it for you.

[sdegrace]

I tried an additional experiment yesterday where I took the same mobile phases and column and sample off of the instrument where I am having the problem and over to another Waters Alliance 2695 system with 2996 PDA we have in a different building (this one without a 2487 dual wavbelength detector, so I can't actually use it for my method transfers, unfortunately). On this instrument the problem was significantly reduced, essentially seemed to be absent (I can put up a chromatogram later, I can't ssh into my web server from behind the firewall here). I don't know to what extent this evidence impacts your answers, but I thought I'd throw it out there for comment.

We are going to get after the Waters people today and pursue the avenue you guys have suggested. Hopefully it fixes our problem. Thanks very much!!

[HW Mueller]

Lower sensitivity of the PDA?

[sdegrace]

No, because we have acquired chromatograms at the same time at 220 nm using both the 2996 PDA and 2487 dual wavelength detector connected in series on the other instrument and the ripples are seen on both and are virtually identical down to the last detail - this is how we have eliminated the detector per se as the cause of the problem.

[sdegrace]

I ran the Method 1 with water and ACN unmodified with TFA and the ripple completely disappeared. I figure that has to be it. Thanks so much to everyone for their help - you saved me!

[mardexis]

Glad you were able to take care of your baseline ripple problem. I still have an interest in the problem, albeit a purely academic one. The ripple is uniform and has a periodicity on the order of a minute. How can that be a mixing problem? I don't know the Waters system that well... What are rough piston size and dwell volumes?

could it be a degasser or the column heater?

Thanks,
Marc

[sdegrace]

I honestly don't know. I noted the periodicity too, and hoped that it might spark something with the Waters reps I spoke to but it didn't seem to raise any particular flags with them.

I guess we will soon have an empirical confirmation one way or another because we are pursuing the course of installing a larger mixer, ASAP. Unfortunately we are under intense time pressures and we have spent a lot of time on this problem already - we have judged that now that we have a lead we have to run with it.

Of course, if the mixer doesn't solve the problem then we will have to look at other possible factors.

I will follow up and let you know how it turns out.

[Bruce Hamilton]

Glad you were able to take care of your baseline ripple problem. I still have an interest in the problem, albeit a purely academic one. The ripple is uniform and has a periodicity on the order of a minute. How can that be a mixing problem? I don't know the Waters system that well... What are rough piston size and dwell volumes?
Marc

The following article from LC-GC discusses the issues with TFA-containing phases and low wavelength detection. Admittedly it's on an Agilent system, but it's probably still relevant.

I prefer to know the actual absorbance of mixed mobile phase components , as it's the high absorbance that exacerbates the effect.

I don't see it on my Agilent 1100 till detection around 205nm, unless the lamp is losing energy from old age, so maybe they have better mixing?. If you can adjust the pump stroke ( 1100s can, but I don't bother ), you may also gain some improvement.

The Physicochemical Causes of Baseline Disturbances in HPLC,
Part I – TFA-Containing Eluents
http://www.lcgceurope.com/lcgceurope/da ... rticle.pdf

Bruce Hamilton

[Mark Tracy]

At 0.05% TFA, the absorbance will be around 150-200 mAU depending on wavelength, purity, pathlength and % MeCN. I have observed that the size of the baseline ripple is directly proportional to the %TFA.

One useful exercise is to collect the pressure data and overlay it with the UV. You should expect to see the ripples line up in time.

[Uwe Neue]

Mardexis:
"The ripple is uniform and has a periodicity on the order of a minute. How can that be a mixing problem?"

For one thing, I do not see it being uniform. The ripples have a different frequency at 10 minutes than at 5 minutes. In addition, what you see at the detector at 5 minutes has been mixed at the pump significantly earlier. The delay is the sum of the gradient delay volume of the instrument and the column volume. Also, there is a difference between high-pressure and low-pressure mixing systems. In the latter case, the ripples are not related to the pump strokes.

One can see the effect best under isocratic conditions with high-pressure mixing, where one can make the ripples move around with the pump frequencies, i.e. the composition.

[sdegrace]

I tried an additional experiment yesterday where I took the same mobile phases and column and sample off of the instrument where I am having the problem and over to another Waters Alliance 2695 system with 2996 PDA we have in a different building (this one without a 2487 dual wavbelength detector, so I can't actually use it for my method transfers, unfortunately). On this instrument the problem was significantly reduced, essentially seemed to be absent (I can put up a chromatogram later, I can't ssh into my web server from behind the firewall here). I don't know to what extent this evidence impacts your answers, but I thought I'd throw it out there for comment.

For what it's worth, here is the chromatogram:

[HW Mueller]

If I understand your chroms correctly then you have more drift and more intense and broader "ripples" in the last chromatogram.