Unexpected drop in plate count after flow increase (Kinetex)

[Mattias]

Hi,

We have an isocratic method that runs on a 4.6x150 mm Kinetex C18 column (2.6 µm). Flow 1 ml/min, using phosphate buffer/acetonitrile.

Plate count of main peak with this setup is about 20.000 plates. To save some time I turned up the flow to 2 ml/min, and then the plate count was only 5000?? I put the flow back to 1 ml/min and got back my 20.000 plates.

The Van-Deemter plot of the Kinetex material cannot explain this drop (flat line almost at higher velocities). Quite strange?

[lmh]

is your pump/mixer combination doing a good job at 2mL/min?

[Mattias]

is your pump/mixer combination doing a good job at 2mL/min?

It is an Agilent G1312B, rated up to 5 ml/min.

retention time at 1 ml/min: 9.2 min
retention time at 2 ml/min: 5.0 min

So the retention times do not fit perfectly with the flow change

[Hollow]

so in fact, your peak width stays +/- the same but the tR is halved?
this would give the factor of 4.
N = 16(tR/W)^2

So I'm guessing, your peak width is mostly due to extracolumn effects than to separation?
How does your plate count correlates to the one of the CoA from Phenomenex?

-> check for dead volume in your system and/or your capillary diameters, detector cell?
-> check the flow rate volumetrically
-> if online mixed, try premixed solvent and use only one channel, so you can exclude mixing effects in this comparison
-> what about injection conditions? (injection solvent vs. mobile phase and injection volume?)

[tom jupille]

So I'm guessing, your peak width is mostly due to extracolumn effects than to separation?

The catch is that the relative contributions of extra-column and intra-column volume should be the same regardless of the flow rate. Your point about extra-column effects is a good one, but I would look at it in terms of time, not volume. If it were my problem I'd check the data acquisition rate and/or time constant. If that's too slow (and on an older system, it may well be), it could be responsible for most of the observed peak width, and that *would* scale the plate numbers the way you suggest.

[M_Farooq]

That is an interesting question. When we were doing very fast separations on a core-shell media (> 3 to 4 mL/min), increasing the sampling rate increased increased the plates count by 2000!

The suggestion to reduce extra-column tubing is highly recommended along with lowering injection volumes (if possible). You might be surprised by the increase in efficiency gains.

[sga]

As long as you are exceeding a predefined NTP value, you should be fine.
With the halving of the retention time, your NTP has reduced by 4 time, which seems to fit into the formula quite well. I don't see a problem here.

To me it appears that your column, pump and detector is running just fine. If you have concerns then try to run precision and linearity, to be sure that there is no randomness in your system. Also try the same tests with a new column of the same make. I trust it will be just fine.

[tom jupille]

With the halving of the retention time, your NTP has reduced by 4 time, which seems to fit into the formula quite well. I don't see a problem here.

The problem is that the observed results *don't* fit the formula. At double the flow, both retention time and peak width should be (approximately) halved. It looks as if retention time did behave almost as expected but peak width remained constant.

So what could cause that? Extra-column volume is always a suspect with low plate counts and should be minimized, but that should change with flow just like retention time. Sampling rate (which might include a time constant and/or bunching) on the other hand will make the same contribution to peak width regardless of the flow rate. In the old days (back when HPLC was young and mammoths still roamed North America) we used the rule of thumb that you needed about 20 measurements to adequately characterize a peak.

A retention time of 5 minutes with 20,000 plates implies a peak width around 8 seconds which suggests a sampling rate of at least 3 Hz.

[Mattias]

Hi!

Just checked the instrument parameters and the sampling rate of the detector was 2.5 Hz in these runs. A bit low maybe, but not enough to explain the plate count observations?

I have one idea and that is that the method runs at a high temperature (60 C). I use the mobile phase preheater of the column thermostat (Agilent), but maybe the heating is not so good at 2 ml/min (compared to 1 ml/min)? That could explain why the retention time did not (completely) go down 50%...

Sample is dissolved in just water and only 10 µl is injected (optimal conditions for RP-chromatography). I use VIPER-capillaries (0.13 mm).

[lmh]

along with sampling rate, do also check if your system has a "time constant" or similar associated with the detector. Many detectors have built-in electronic smoothing, and if they're smoothing over several seconds, they will smudge the peak-width. Also, if you're using a pre-written processing method, make sure that any smoothing parameters in the processing method aren't widening the peak (e.g. boxcar smoothing over too wide a number of points).

[Mattias]

On the detector page it says:

Peak width: >0.10 min
Response time: 2.0 seconds
2.5 Hz

We run on Chromeleon and it is unclear how the signal is treated in the software. I have a feeling Chromeleon contains some kind of "smoothing" as well.

Next point is that I have ordered 3.0 mm ID columns. If the problem is due to poor heating of the mobile phase, it should be better with a narrower column (and lower flowrates)

[HPLCaddict]

Decrease the response time! 2 seconds is far too high. That's exactly what Imh was aiming at. Due to the response time your detector is "too slow" to catch narrow peaks.
Depending on the peak width, the "peak width" parameter as well as your sampling rate may also be too high.

[Hollow]

today I've attened a presentation also giving a short overview about some key parameters of stat.phases.
One that let me think on this topic here, was the 'real' accessability of the molecules to the pores.
Limited access would be more pronounced at high flow resulting in peak broadening.

I doubt that this is the source here and causing the plate count to drop by a factor of 4 by doubling the flow rate. Just as another input for discussion.

[Mattias]

The plate count did not improve by changing the detector settings, so this must be due to something else. My main peak weighs about 900 u, so maybe the pores are too small (in combination with the high flow rate).

It also did not improve by using a 3 mm column (and scaled flow). It is the velocity of the solvent that makes a difference.

Conclusion: The Van-Deemter plot is not valid for all molecules

[Mattias]

Or more correct: different molecules have different Van-Deemter plots on the same column material