[mjc123]

Can anybody explain the scoop-shaped negative feature at the beginning of the chromatogram?
(Reverse phase HPLC, acetonitrile 1 mL/min, sample dissolved in acetonitrile, DAD detection at 210 nm.)
The feature does not appear when running a solvent blank, but does not seem to be a component of the sample because (i) there is no corresponding feature in the RI signal, (ii) its magnitude does not vary with injection volume.

[Corribus]

No HPLC expert by any means, but what happens if you just inject a blank (some of the mobile phase)- do you still see the anomaly? Interestingly, it doesn't look like a peak, per se, just a broad negative region.

If you don't get any help here, you might try the chromatography forums.

http://www.chromforum.org/

[DrCMS]

It appears the solvent used to dissolve the sample has a lower absorbance at 210nm than the HPLC solvent.

Are the HPLC solvent and the solvent used for the sample prep exactly the same grade?  From the same bottle?
Is the HPLC solvent eluting from the end of the column recycled or discarded?

[mjc123]

The solvent and eluent are identical, in fact the solvent for samples was extracted from the eluent reservoir.
Eluting solvent is discarded.

[Arkcon]

These sorts of problems are little hard top pin down, because when someone is so sure they have systematically excluding everything, its not uncommon for them to rush through it, and not actually do something they think they did.  So that's a caveat thoroughout this whole question.

That said ...

There is something in the sample, with much less absorption than a solvent blank, even though the sample matrix is the same. (Right?  Conform this please.)  Not a refractive index change, because its not on the RI detector.  And it appears to be not retained at all.

Consider extending the re-equilibation time longer, long enough to keep this negative peak on the same run as the previous injection.  This might pin down the source better -- this negative peak may be from something very well retained from the previous injection.  This might not work, but see if it changes the negative peak in the next injection.  It might even disappear -- that sort of thing happens when we try to pin these sorts of things down.  We get more careful, as we try to isolate a problem, then it disappears, and we can't say for sure why.

[mjc123]

even though the sample matrix is the same. (Right?  Conform this please.) 

Yes
Extending the re-equilibration time had no effect; the negative feature didn't appear in the same run, but appeared in the next injection at exactly the same time as before.

[Arkcon]

If extending the re-equilibration doesn't move, or reduce or otherwise change this peak (may even make it bigger,) then we'll have to consider this is some sort of instrument defect -- somehow the injector is sticking in its rotation, or electronic noise or some detector electronics fault is causing a faulty baseline zeroing.

Thank you for keeping us up to date on your problem.  And let us know what eventually fixes it.