RT shifts Waters UPLC-PDA

[neitz]

Hello all,

I have an issue with a method that I am currently validating.
The method is the following:

Pre-mixed 60:40 ACN:0.1% TFA in Water, isocratic.
Flow 0.6 mL/min
Column: Cortecs UPLC Shield RP18, 90 Å, 1.6 μm, 2.1 × 100 mm
dual channel 220/255 nm, injeção volume 0.7 uL, column T: 35 degrees

I saw this on a Waters Acquity and also a ThermoVanquish UPLC-PDA as well.

Issue:

I started to develop this by mixing A and B in the instrument, but then the common ripple/wavy baseline with TFA started to appear and, to avoid this I have decided to pre mix it and this sorts the issue.

However, since I started to do this I see that there is a retention time shift in my system suitability standards (all affected in the same way and in the same direction).

Example of a peak:

1.6582
1.6523
1.6480
1.6430
1.6393
1.6351
1.6326

This is enough for me to be unable to properly identify components in my botanical extracts (complex extracts in MeOH). On my system suitability stats I do not see loss of resolution between peaks.

After each sample set I always flush it with ACN as per column recommendations. Once resuming analyses, I give the system enough time to equilibrate (close to 60m).

I am aware that some ACN evaporation may be possible in the bottle, but I see retention time shifts between subsequent runs (i.e., if I re-inject right after the first injection, I see a shift). Also, the retention time shift if to the left, so earlier retention times. I would expect that ACN evaporation would lead to increasing retention.

I was wondering if this could be related to some stationary phase degradation (the columns are relatively new). The method was based on an application note that used the same mobile phases so I am surprised if this is due to degradation - however I don't see dramatic loss of resolution between the standards.

I am a bit lost on this one, any help would be greatly appreciated.

Thank you!

[Multidimensional]

Lots of possible reasons besides the obvious (too short equilibration before starting resulting in drift backwards as the system sloooooowly stabilizes)... but maybe no problem at all.

Start with the basics. For example:

What does the peak shape look like? is it at or near Gaussian? Peak shape needs to be close to ideal and sampling rate needs to be appropriate (20-30 points across width) to get reproducible data and show a trend.

The estimated linear flow rate to use with a 2.1 mm ID column and 1.6u particle would be close to 0.400 mL/min, not 0.600 mL/min. The higher flow rate may result in turbulence, frictional heating or both. These can both change the chromatography observed (shorter retention times). Frictional heating would slowly reduce the Rt values. Try optimizing the flow rate, then re-evaluate the method for RSD to make sure these issues are not causing problems. *We have seen a lot of clients using the sub 2u particles have this problem when they push the flow too high.

Your mobile phase is not buffered. Acids in water solutions tend to degrade over short times (pH goes up), even in just a day, changing the pH of the solution. Again, the chromatography can change so we make up mobile phase solutions fresh each day so we start with the same pH.

Additionally, there may be no problem of any kind here. In your example with the retention times changing from 1.6582 to 1.6326, you did not indicate how long this change took? Are these Rt values taken 1 hour apart, 20 minutes, one day ??? A Rt time difference of only 0.0256 minutes may or may not be a big deal. Looks like your column volume is estimated at 200 ul and at 0.6mL/min that suggests a Tzero of around 0.3 min, so your K prime may be OK. What is the injector RSD for the same sample (5x inj)? What is the resolution between peaks? Is it enough (what is 'R' ?)? Perhaps the method needs to be optimized and improved to show better resolution and larger K primes? We do not have enough info to know.

[vmu]

RSD of RT in the above example is 0.57 %. This is not perfect, but not too huge. The difference in RT between the first and the last injection is 0.0256 min (1.6 %). Assuming that the plate height is as low as 2*dp = 3.2 μm (N = 31000), the FWHM of the peak is 0.022 min, i.e. similar to the maximum RT difference. This is not an obstacle for the peak identification. If you use the standard chromatogram nearest to the sample chromatogram for the qualitative analysis and inject the standard, e.g., after each 4 samples, there is no problem with identification at all. What is the actual peak width? What was the RSD of RT when you mixed the phases from the two channels? Does RT decrease continuously from day to day even with a freshly prepared eluent?

I estimate the column hold-up volume as 190 μL for this column (55 % of the column volume). Then the hold-up time is 0.32 min and k' = 4.