Heliocentric corrections variation with the field

[segasai]

Hi,

Currently DESI applies heliocentric corrections based on the ra,dec of the center of the field, as opposed to star by star. The difference between those can be computed easily: It can reach .8 km/s and has a stddev of 0.35 km/s

This effect is detectable, by looking at the delta(barycentric correction) (i.e. barycentric correction computed for the center of the field minus barycentric correction for a given star) vs delta RV in external surveys (SoS in this case). The plot is made using z-band only RVSpecfit measurements in iron single exposures.

Given that I was able for the first time detect this, one can try to start thinking about how to actually fix that.

And the easiest way of doing that I think is through trace_shifts.py where one can simply add a term for barycentric correction of the fiber minus barycentric correction of the center of the frame.

Thoughts, comments ?

[sazabi4]

I do want to point out that the shift is significant for a good fraction of stars and therefore will affect our science results using these stars. So it would be good to fix this in Year 3 data release.

[djschlegel]

Adding my old notes on this: Thinking about the barycentric corrections again, the worst-case scenario for errors by using the tile center instead of the fiber position is (cos(45+1.6 deg) - cos(45 deg)) * 30 km/s = 0.6 km/s Across a tile, that’s 1.2 km/s peak-to-peak error. (I don't see how we get the larger 0.8 km/s shown above.)

Sky-subtraction needs to be done on the as-observed wavelength scale, meaning either before the barycentric correction or with the same done per fiber.

But a way to fix this would be to make small offsets to the final resolution matrix to account for the residual per-fiber errors. The central fibers have no change, nor do fibers offset in the perpendicular direction to the earth’s motion vector. Fibers parallel to that motion would have their resolution matrices shifted by ~0.01 - 0.02 pix.

This approach would give us data products that are unchanged, with the same wavelength scale for all fibers. But for the detailed analysis, like projecting stellar models to the as-observed spectra, this would formally have the correct velocity information.

[segasai]

@djschlegel

The largest possible difference is sin(np.deg2rad(1.6))*30 = 0.83 km/s This is the case where the correction is zero in the center of the field. That matches my calculation

[segasai]

Regarding the implementation of this through the resolution matrix, personally I would be against it, if it can be avoided. In my understanding the wavelength corrections machinery in trace_shifts would allow us to correct for this, but maybe I am mistaken there.

[julienguy]

• I suggest we simply write in the spectra fibermap the correction term to bring to the wavelengths of each target to the correct frame.
• For coadds, we would report the weighted average of the correction term. The extra blur due to the average of spectra with tiny shifts of wavelength corresponding to <0.03% of a spectral pixel is probably negligible, but this assumption can be challenged.
• Radial velocity and redshift codes would apply this correction.