Injection Volume Linearity using RI

[Lothar1984]

Hi everyone,

I'm new to writing on this forum, so I hope I'm doing this right! I’m looking for some help to understand a phenomenon I encountered recently.

I’m designing the Qualification / Maintenance tests for an Agilent RI HPLC. During a dry run of all the tests we usually perform with our DAD HPLCs, everything passed except for the injector linearity test. This test involves using the same concentration sample with different injection volumes.

The response I observe is non-linear, more like a logarithmic curve or an asymptote. To rule out detector overload, I lowered the concentration of the standard to a third, but the non-linear behavior persisted.

I get that this test is pretty much worthless since all our methods are ran at constant inj volume. I only did it because we do it for the DAD HPLCs, and my organization tends to stick with “we’ve always done it this way,” so any change will require some convincing.

I did a bit of online research, but I couldn’t find a solid explanation for why this happens. It seems to be tied to the fundamental operation of the RI detector, but I don’t quite get why.

Does anyone have a good explanation for this?

Thank you very much!

[vmu]

What is the column? What is the eluent? What is the analyte? What is the sample solvent? Did you purge the reference cell of the RID before the analysis? How does a pure solvent injection look like? ...

RID is sensitive to all the components of the injected solution. If the analyte is not separated from the sample solvent, the response linearity cannot be achieved.

[Lothar1984]

Hi vmu,

Thanks for your answer.

Since this is a performance test for the instrument and not a sample analysis, I am not using any columns. Instead, I am using a restrictor capillary (15m, 0.17 ID).

The eluent is HPLC-grade water.

The "sample" is Thermo's pre-vialed 3325.0010 OQ/PQ Standards for RI-Detector, which is basically glycerine in water. The specific vial I am using contains 5 mg/mL of glycerine.

I purged the reference cell with the mobile phase extensively and also equilibrated thoroughly before the test. All other tests (noise, detector linearity, injector reproducibility, etc.) performed very well, which attests to proper purging and equilibration.

I also repeated the injector linearity test with a lower concentration to rule out detector maxing out.

There is no matrix in this "sample" and no separation since it is only a performance test for the module.

I want to emphasize that the detector linearity test passed (different concentrations of the glycerine standard, all at the same injection volume). This leads me to think that the issue with the injector linearity might be related to the injection volume itself. MY gut feeling is band broadening might affect the RID to a greater extent?

I am just trying to fully understand this phenomenon so I can explain to QA and other non-technical folks why we are not running this test on an RID when all our other instruments (DAD) run it.

I hope this provides additional information and clarifies the request.

Thank you for your help!

[vmu]

Does your instrument include both a UV detector and a RID? Does the injector linearity test pass well with the UV detector and fail with the RID on this particular instrument?

Try to inject different volumes of pure water and see if there are interfering "system" peaks (positive or negative) on the chromatograms. RID is sensitive almost to everything, including the gases (air) dissolved in the sample and mobile phase. The concentration of gases in the water (and in the glycerine solution) you inject differs from the concentration of gasses in the ((in-line) degassed?) water you pump through the capillary. This can result in a "system" peak.

Try to perform the injection linearity test using a column that retains glycerine (or another analyte) sufficiently and separates it from any "system" peaks observed with a pure water injection to the same column.

MY gut feeling is band broadening might affect the RID to a greater extent?

Differences in band broadening may cause response linearity problems if you use peak heights. If you use peak areas, the band broadening is not important.

To rule out detector overload, I lowered the concentration of the standard to a third

Try to lower the concentration 10-fold and 100-fold.

[Lothar1984]

Hi vmu,

Thanks for your suggestions!

Unfortunately, this specific instrument does not include a UV detector, and we don't have any available to move around at this point. We were considering trying it to completely rule out injector malfunction, but it's not feasible at this time.

I performed injections with different volumes of water. There is an interfering negative peak, but it is several orders of magnitude lower than the RI signal, so I don't think that is causing the non-linearity.

I am using peak areas, not peak height, for the evaluation.

The 1/3 reduction put the signal in the same range as the detector linearity test, so I don't see the point in further lowering the concentration. Am I missing something?

At this point, it doesn't seem like there are any interfering peaks. Do you have any other theories?

Thanks again for your help!

[vmu]

Try to do the inj. volume test with any suitable column, mobile phase and retained analyte. For this, you can run any analysis conducted regularly with this RID. If this test fails, the autosampler does not work properly. No more theories now.

[Peter Apps]

What range of volumes are you testing?, and what kind of injector do you have - I presume a loop that is biggeer than your biggest volume, being loaded with different volumes from a syringe?

Peter

[DR]

I think the main things to remember here are 1) Dynamic range of RI is orders of magnitude lower than for UV, so you have to come up with a pretty limited range of concentrations of volumes to test. 2) By varying injection volume, are you really testing the linearity of RI response? It seems that picking a single injection volume and an appropriate set of serial dilutions from a standard stock, you would be better off.