HEEEELP! Gradient bad shape!

[hplc.chem]

Dear all,
I am struggling since two weeks with my HPLC. I am running a gradient with the following conditions:
0-5 min: Isocratic at 40%ACN/Water
5-25 min: Linear Gradient 40% ACN/Water to 100% ACN
25-35 min: Isocratic at 100%ACN
35-45 min: Isocratic at 40%ACN (Re-Equilibration)

Low-pressure mixing pump, Flow = 1.5 ml/min, UV Detector at 254nm, Temp.=30 oC, C18 column 250x4.6mm 5um with guard column, online degasser, autoinjector set at 20 ul injection. Sample solvent is ACN and samples have low levels of analytes.

The PROBLEM is:

I used to get a good (I think!) linear gradient shape with the two isocratic plateaus at the start and end of the gradient with a maximum height around 0.0005Volts. However, after running a number of samples the shape started to deteriorate with a peak (or a sudden jump) at about 17 min, and absorbance levels increased.

Below is a chromatogram with the BAD shape:



In addition to the BAD shape, the absorbance levels are increased gradually from an injection to the following one!

My investigations: ( ALL INJECTIONS WERE ACETONITRILE)

1st : I washed the column with 100%ACN for 30min, then repeated 3 injections and got the chromatogram below:



Better shape and lower absorbance! But there was still some deterioration in the shape and the absorbance levels are still increased gradually from replicate 1 to 3 and deteriorated more with the following injections.

2nd : I consulted this forum and watched the Mini-tutorial on gradient artifact peaks posted thankfully by Tom Jupille. I then ran the "three blank gradients" test. The results are shown below:




I think there was no clear evidence of contamination of the A solvent (Water), so I continued having fun! I suspected that something is happening from the injections.

3rd : I injected ACN 10ul, 20ul, 60ul, 100ul and then 10ul again to check if the sample solvent is causing the contamination. However, there was no clear trend of the absorbance with the injected volume.
4th : I suspected the Injector so I replaced the Injector wash and washed the Injector with 25 injections of 100ul ACN. Then ran 4 replicate injections of ACN. The results are shown below:




I think things are becoming worse! The absorbance level increased more and the shape deteriorated more!

5th : now it is the column time! I disconnected the column and replaced it with a tubing to get some backpressure and then ran 2 injections. I got good shape and absorbance: linear gradient shape with the two isocratic plateaus, and a maximum absorbance level at 0.0007V. So, I suspected the column has accumulated some garbage from somewhere.

6th : I disconnected the guard column and made 2 injections to get the below:



The absorbance level dropped significantly BUT the shape is still not OK and the absorbance increased from the first to the second injection!

7th : I tried a another column (a used C18, 150x4.6mm, 5u) BUT I got almost the same shape and level of absorbance.

8th: I back to my column (the 250mm) and washed the column and the system with 80%ACN at 0.5ml/min for about 20 hours!!! I got the below:




Better shape and absorbance levels but the increasing levels of absorbance between injections is still there!

9th: I stopped having fun and started to cry because I wasted most of the ACN purchased for the current project!!!

So, I decided to consult you in this valuable forum through this post!

Any suggestion or comment will be appreciated.

Many thanks in advance.

[Hollow]

Hi

first, thanks for the detailed informations and chrom.
if possible, tell us also the modell of your HPLC and software.

my first thoughts:
- re-equilibration volume (15ml) seems to be on the low end for a 4.6x250 column (V0 ca. 2.5ml). Especially when I guess your dwell volume (gradient delay) to be about 7 ml (ca. 5 min delay until your baseline starts to rise). Therefore your re-equilibration would be around 3 column volumes only. Rule of thumb 5-10 V-Col.

- are you sure your gradient proportioning valve is working properly? Your chroms look a bit weird.
-> perform a linear gradient test as it is used to determine the dwell volume
find the procedure on the troubleshooting wizard from LC Resources (diagnostic section).
(personally, I use to introduce an additional isocratic step of about 1-2 min at the beginning, which I then need to correct during calculation).
If you have more than only two lines, perform it also with other combinations than only A/B e.g C/D
-> if you don't get a nice straight line for the gradient or C/D is different than A/B, perform also the gradient step test to check for proportioning accuracy (use again differnt combinations of solvent lines).
-> check also the flow rate accuracy (just to be sure. (7 ml of dwell volume is quite a lot..))
-> if possible collect also the pressure channel for these experiments as well as for some your real method

- if gradient proportioning is ok:
-- test the purity of your ACN. Do you have another Brand/Quality/Lot at hands?
-- if needed (e.g. Agilent/EZChrom): are you using the correct compressibility settings (can have a hughe influence on pressure ripple as well as baseline)
-- increase your re-equilibration time until the pressure is same as at start
-- if possible decrease the ACN content of your sample or inject less

[hplc.chem]

Thanks a lot Hollow for the suggestions.

The model is Shimadzu LC-10A vp series and it is not perfectly plumbed! It has a bit long tubing between some of the units (eg from Inj to Oven) which may contribute to the high dwell volume. Software is Class vp.

I have done a linear gradient without the column as I mentioned in 5th in the last post and I got much better shape.

I also checked the flow rate at the detector outlet and it was accurate enough.

I did all the gradient tests few months ago, just before starting this method and all the results were OK.

I use "gradient purity" ACN and I tried two different batches but they both gave similar results.

I collected the pressure readings for all the previous runs and checked the pressure pattern and it has a smooth linear gradient shape.

What do you suggest to examine in the pressure output?

Regarding sample volume, I mentioned in 3rd the injection of different volumes and they all have very similar shape.

Many thanks again for the valuable suggestions and I will wait for more generous thoughts!

[hplc.chem]

This is a FRESH chromatogram for a run without the column.



I hope it can help.

[tom jupille]

If that's intended to be a linear gradient with no column, you have *horrible* mixing and proportioning problems. I'd follow Hollow's suggestion and at least run the stair-step test.

[hplc.chem]

Thanks tom for the reply. It is an injection of ACN without the column, following the above described method. However, the pressure was between 900psi for 40%ACN and only 250psi for the 100%ACN.
Can I ask why you think that I have a HORRIBLE mixing and proportioning problems, based on the above chrom.?

[tom jupille]

Can I ask why you think that I have a HORRIBLE mixing and proportioning problems, based on the above chrom.?

What you should see is two horizontal regions (at the beginning and end), a linear ramp in between, and *slight* curvature where they join. What you actually got is this:

Waaay too much curvature at the end and significant deviations from linearity in the ramp.

The "usual suspects" are the mixer and the GPV.

[hplc.chem]

Thanks Tom and Hollow. I will do the Gradient Tests and get back to post the results soon.

[Hollow]

Thanks a lot Hollow for the suggestions.

you're welcome

The model is Shimadzu LC-10A vp series and it is not perfectly plumbed! It has a bit long tubing between some of the units (eg from Inj to Oven) which may contribute to the high dwell volume. Software is Class vp.

- thanks for update. since I'm not familiar with this model nor with the SW, I can only give generic assistance
- Do you know anything about the diameter of the tubing?
long but small can be still better than short but big (remember volume of a tube; same holds for peak broadening)
- checked for any dead volumes?

I also checked the flow rate at the detector outlet and it was accurate enough.

what's "accurate enough" for you? Is it also accurate enough for Shimadzu (or Tom) (specifications)?

What do you suggest to examine in the pressure output?

a) if the pressure ripple differs between channel A and B (during gradient) this could point you where to look at (or if it is too much at all, recheck the compressibility settings, check valves, pump problems etc.)
b) if pressure isn't fully back to its initial value at the end of the run, your equilibration step is too short
c) gives you an idea of your mixinig characteristic (e.g. when you transfer a method from one system to another. Even if you translate the reequilibration step according to dwell volume etc. for the new system, if it has a very differnt mixing behaviour, you probably have to extend this step, see Pt. a)
d) compare inj-to-inj or sequence-to-seq perfomance
(we normaly record the pressure but rarely do anything with it. But in cases of troubleshooting you have valuable info at hands (e.g. late retention time + pressure lower than ususal -> probably flow problem; leak; maybe too low %organic; etc.). And as of today, hard-disk space (usually) isn't an issue any more in workstation environments.

Regarding sample volume, I mentioned in 3rd the injection of different volumes and they all have very similar shape.

yep, noticed and it wasn't my primary focus. just to keep in mind.
Especially when moving to another system with different injector geometry. Could be working on one system but is too much on another.
If you've tested it and it works, it's fine.

For next, follow our guidance for proportioning and mixing tests.
Use same solvent on both channels (e.g. MeOH or Water), one with an UV absorbant like Acetone 0.1% for good and meaningfull results (no influence of viscosity mixing, real UV signal etc.)

[hplc.chem]

Hi All,

I ran Linear Gradient and Step Tests as described in the "troubleshooting wizard". The flow rate was 3 ml/min, and I checked its accuracy during the run by measuring the volume of the solvents collected from the detector outlet over a known time period. I used a measuring cylinder and a volumetric flask for confirmation. I hope this method can be acceptable! (and met Tom's specifications ).

For the Linear Gradient, the Gradient time was 20 min and I added 5 min at the end, and nothing at the start.

I got the below trace.



I used the data to calculate the dwell time and volume. Dwell time was found to be 1.32 min and therefore, volume was 3.96 ml.

The step test was conducted by making 10% steps and two extra steps at 45 and 55% as suggested by Dolan in LC/GC troubleshooting.

The trace is below:




Shimadzu specification for the step test (for this pump) is +/-1% and all steps passed this level.

I hope the two tests are acceptable enough to exclude the problems in the mixing performance, and back again to the main problem!

Many thanks for any suggestion or comment.

[Hollow]

ok, that looks nice.
So, GPV and pump is working fine under these conditions.
3.9 ml VDwell is high, but not uncommon for older systems, especially when the tubing from the GPV to the pump has a large diameter.

(Keep the information about the Dwell volume for this system. If you don't change the tubing, it won't change either. Maybe label it on the system for everyone.
And of course it's actually part of your gradient method (which should also be included in the documentation )

But back to topic:
With regard to the VDwell your netto re-equilibration volume is about 11 ml (about 4 VColumn). As said, it's on the low end but could be ok if proven.
What makes me sceptic is the mixing chromatogram from your last post.
I guess with that mixing characteristic, your column isn't well reequilibrated.

(I have to admit that I don't have any clue why this should lead to an increasing trend of baseline, rather than to uncorrelated fluctuations. But I've learned not to exclude anything until checked.)

-> Could you please show us some of the pressure channel records of consecutive injections as well? Maybe some of "good" ones and "bad" ones? Or collect them during the next proposal.

-> record some runs with prolonged re-equib time (incl. pressure)
(eg. inj1+2 = Runtime 45 min; inj3+4+5 = Runtime e.g 60 min
(Notice: Inj1 is used as equib for inj2
Inj2+3 can be evaluated after 45 min just as usual, but inj3 is used as equib for inj4
Inj4+5 have are those with the prolonged re-equib time)

How would chroms 4+5 look like compared to the normal ones (2+3)?
Any improvements? Increased garbage peaks? Does the UV absorbance stabilize somewhen with better reequilibration?

Another question: is your water buffered or acidified somehow? (e.g. 0.05% H3PO4).
Could there anything be absorbed from air within reservoir A (like CO2 -> change in pH etc.)?
If your analytes are non ionizable at all they shouldn't change their retention behavior. So maybe you could give it a try if anything else is already tested. But this will introduce another source of potential contamination, so if not needed/improvements, leave it...

Good luck

[hplc.chem]

I attached below some of the pump pressure traces.

The first one below belongs to the first chromatogram I attached in my first post (the BAD shape). This was one of the worst chormatograms and it also has high pressure at max=3700psi. It was the first one diagnosed as BAD.



The second one below belongs to the 4th chromatogram I attached in my first post which was done after I washed the column and injector.



There was a clear drop in the pressure (max) from about 3500 to about 2500!

The third one below belongs to the LAST chromatogram I attached in my first post (point 8th) which was done after I washed the column for a longer time and the run was also done without the guard column in place.



there is also some decrease in the pressure but I thought it might be due to the removal of the guard column.

For any consecutive injections, the pressure pattern and level is almost identical.

Regarding the mobile phase, I am using only Water and ACN with no buffers. My analytes are PAHs.

Many thanks for all contributers (and readers as well ) and I hope these may add something to go further!

[tom jupille]

Your test profiles look fine, which tells us that there are no problems with your proportioning and mixing system (so much for my diagnosis! ). The pressure profile is about what you would expect (remember that water and acetonitrile have different viscosities and that viscosity is a non-linear function of composition. You have checked out the autosampler/injector.

That leaves the column and the detector. What happens if you swap in a new column? And, is there the possibility of swapping in a different detector (misaligned detector flow cells can cause funny baseline issues because they exacerbate refractive index differences.

[hplc.chem]

Thanks Tom and Happy to hear that half of my instrument is working fine!

I can try a new column but I was afraid to damage more columns if something is building up in the column inlet!

I will try at least one or few injections and see what I can get.

Regarding the detector, unfortunately I have only one UV detector.

But why do you think that the detector might be contributing in the problem although it seems to give acceptable results in the absence of the column?

Also, how can the column cause such an effect?

[tom jupille]

But why do you think that the detector might be contributing in the problem although it seems to give acceptable results in the absence of the column?

I'm thinking of refractive index sensitivity. Because it's impossible to perfectly collimate a light beam, there will always be some light that enters the flow cell at an angle. When the RI of the cell contents changes, that light will "bend" differently. All UV detectors show this effect to some extent (that's why we typically see a blip at t0). If the flow cell is out of alignment, then more light is coming in at an angle, and the RI response will be greater. Refractive index of ACN/water mixtures is a non-linear function of composition, so the net effect is a curved offset in the baseline.

Also, how can the column cause such an effect?

I'm grasping at straws. To quote from Sherlock Holmes: "when you have eliminated the impossible, whatever remains, however improbable, must be the truth"