Noel W Davies This is a mix of just 14 of the 17 they have present. Nothing is missing and peaks have great symmetry otherwise. The others analytes are not in most mixes and I run them separately.

This issue would rule out measuring CBNa & exo-THC and hurts LOQs of the THCVa and d9.

Apart from the dip in your baseline the column performance seems to be OK (peak shape and resolution) so I wonder if the dip could be related to the sample injection. What solvent did you use to dissolve the samples and standards? Have you tried doing a blank injection of the dissolution solvent and doing a zero volume injection? Does the "dip" still appear?

A few Comments (Updated; Apologies for the long messy answer):

The "Dip" may or may not be normal. Update: I looked up the Agilent app note (Not "literature", a note) and see the chromatogram shown may be based on MS response, not UV (?). No UV bandwidth settings were reported, but detailed TOF MS settings were. The scale is labeled "response" not AU. The dip may not be seen using TOF MS, as Agilent used, but in fact present by UV (?). It is a very complicated note with many problems too. I also do not know the exact configuration and plumbing of - your HPLC system - so changes between two analysis runs may certainly differ (I would expect them too; see below for more info).

• NOTE: The application note is also very unusual in that it shows a gradient ending in 8.5 minutes; with a 100% MeOH (w/TFA) wash step from 11 to 13 minutes, when peaks are still eluting, to wash everything else off the column.
• At 4.5 minutes, the methanol mobile phase is introduced on top of the ACN mobile phase, changing the background UV absorbance. When accounting for the delay volume, and the fact that METHANOL absorbs more than ACN in the UV range (esp at 230nm), a change in the UV baseline should occur when the solution passes through the column to the UV detector's flow cell (normal). NO SUCH CHANGES ARE SHOWN IN AGILENT's App note (It can not be from actual UV data).

(1) The chromatogram that you show with the "dip" appears to have the DAD set up incorrectly. The software "Reference Wavelength" feature is turned ON with the factory default values of 360nm (100nm bandwidth) selected. This means the raw data obtained from your sample is first processed to subtract all signal from 310nm to 410nm from your data, then output the result as YOUR signal. This tells me that you have not had any formal training in HPLC method development and/or in how to setup and use the DAD (as you are using a feature that should be turned OFF). *This may or may not contribute to the "dip" issue observed, but is important. As the odd software feature is designed to destroy collected data, there is never any way to tell what is real, which is why it should be turned off (*Please Read the article I reference to learn more. We teach this in our basic HPLC training classes). Please turn it OFF.

(2) The example chromatogram provided by you (Agilent Application Note?) was run on a different HPLC-MS system. It is unreasonable to assume that you would obtain the same result as someone else if you used any other system except for their exact system. Normal changes are common because the HPLC systems, flow path, configuration, modules, plumbing and even software settings are different between two systems. small changes in dwell volume or mixing alone could account for it. Knowledge of HPLC design and operation stresses this fundamental as it is key to understanding the results obtained from these instruments). This is a training issue and you would not be expected to know this unless you had several years of formal training in HPLC method development. Troubleshooting HPLC methods requires a great deal of knowledge to make sense of all of these possible changes and how they contribute to the analytical results obtained. Mistakes often occur and the process can be very frustrating at times. Proper training in the fundamentals of HPLC will help make sense of many of these observations plus allow for better troubleshooting and more confident results.

(3) A change in pH of the mobile phase may also contribute to the changes observed (?). If you are NOT preparing your solution of methanol and formic acid freshly, then the pH of the solution may be changing over time (please refer to separate post with free article). Again, this may or may not be related to the question, but is important to point out as a reason for "change". In the original app note, the authors mainly use formic acid in their mobile phase to enhance ion formation by MS.

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

• "A Case of Changing Solution pH. Formic Acid Stability in Solution (Methanol)"; https://hplctips.blogspot.com/2019/09/a-case-of-changing-solution-ph-formic.html

Thank you William for that very thorough going analysis of the problem. I absolutely agree with your comments, especially those on the requirement for freshly preparing solvents. Any "savings" made in using aged solvents are more than negated by the potential problems generated.

I should probably have mentioned why I focused on the sample solvent as being a possible problem and that was because of the relatively large disturbance observed at, what I took to be, the void volume. However, I confess it is still hard to imagine what sample solvent could cause such a broad dip. But never underestimate the human capacity for the unexpected!! Haha.

William Letter

In response to (1) & (3)

I have had no issues with these analytes using the reference subtract. Here is the exact same signal with no reference. I did not know exactly how Agilent implemented this feature and had no issues so I didn't think too much of this. I ran it side by side thinking it was pretty harmless. I will turn this off going forward.

I prepared the mobile phases the same day as these samples were run. While this isn't the best idea with ACN:water mix due to the endothermic mixing I wanted to see if this method was viable.

I explained the primary differences in flow path being related to the Quad pump they use as opposed to the K1312b binary pump I currently am running. I am running the Bin pump in the standard config with about ~600-800 uL of delay volume.

I think you are right with regard to the reference chromatogram in their method not being UV-based due to the expected changes from the gradient not happening. I may be chasing an impossible method, or at least impossible on a UV detector.

Do you think it is worthwhile to attempt this after switching to a low delay volume config? (~120 uL)

Charles H Hocart

Solvent sample in this case is acetonitrile. The dip is seen with blank injections of methanol and acetonitrile. Injection volume is small, 2 uL.

William Arce: Thank you for the additional info.

A few comments which address some of your questions.

Regarding the use of the software feature "Reference Wavelength". When the feature is turned 'ON', the original data is destroyed so you will never know what was collected. It invalidates the data & method which is unscientific and why we do not use it. Thank you for turning it off and re-checking. Now you can forever leave it 'OFF' and have peace of mind.

• **BTW: I did not expect the REF to show a change in your case as I found the real reason for the "dip" being the change in solvent composition. Just wanted to save you from trouble down the road by keeping it 'OFF'.

Regarding the mixing of ACN and Water, you wrote;"While this isn't the best idea with ACN:water mix due to the endothermic mixing". Please do not worry about this. A trival issue and in five minutes the solution stabilizes (temp) quickly. The degasser takes care of any gas bubbles. Exotherm with MeOH/Water, also no concern.

Flow Path of the HPLC systems: You wrote: "Do you think it is worthwhile to attempt this after switching to a low delay volume config? (~120 uL)"? ***Not really. Even worse mixing results with too low a volume. There is a LOT more to the flow path between two different HPLC systems than if one has Binary pump and the other a Quaternary. All of the interconnecting tubing (lengths and ID's used, which are not standardized), mixer dimensions, fittings, flow cell choice, Autoinjector options etc all apply here. The gradient "delay" volume is just one of many issues which contribute to the final results. The Flow path should be optimized.

Agilent's note probably shows a normalized or "modified" chromatogram result. By UV, it is not possible to obtain a flat baseline using their solvent composition program. The method had other issues too (bad example).