Velocity Calibrations

We want to check the accuracy and consistency of the velocity (wavelength) measurements of PNe with the PN.S.

Galactic PNe

We first look at PNe in our own Galaxy. These should be much brighter than their central star at the [O III] wavelengths, and should be unresolved/pointlike so that centroiding is not ambiguous. A pair of images of one Galactic PN (PNG 019.2-02.2) can be seen below (120-sec exposure; m5007~15.1).

Left Right

We first centroid the position of the PN images (left and right) using IRAF's digiphot.apphot.phot. Then we perform spectral calibrations of the appropriate arc-lamp images (taken immediately before and/or after the PN image, at the same position on the sky).
Below left is a schematic of the calibration mask's layout: central longslit surrounded by a grid of 1-arcsec round holes.
Below right is a dispersed calibration image.


The spectrum from one of the holes can be seen below in close-up. There are two lines used in the calibration: 5017.1Å and 5037.8Å (left and right marked spots, respectively?).

Using an automated script (see description), the spectral lines are identified and centroided over the entire field, and a polynomial wavelength solution is fit with 11 terms (3rd-order in x and y, 1st-order in lambda). The rms residuals of this fit are ~0.1 pixel in each of the x- and y- directions, which nominally corresponds to a total velocity uncertainty of 11 km/s. Applying this solution to the measured dispersed positions of the PN, the wavelength and velocity are calculated (heliocentric correction made using IRAF's astutil.rvcorrect).

The results on five Galactic PNe are given in the table below. The literature values come from the Strasbourg-ESO Catalogue of Galactic Planetary Nebulae (Acker et al. 1992): see V/84 catalogues in VizieR.

Name PN.S measured heliocentric velocity (km/s) Literature heliocentric velocity (km/s) Comments
PNG 018.6-02.2 (M3-54) 178.2 157.0±5.0
PNG 019.2-02.2 (M4-10) 55.4 50.0±5.0
PNG 019.4-05.3 (M1-61) 17.2 40.8±10.1 possible blended image
PNG 027.6-09.6 (IC 4846) 106.9 151.0±3.0 image distortion -- resolved source?
PNG 038.7-03.3 (M1-69) 32.0 7.0±25.0 image distortion

We calculate the chi-squared fit of the PN.S velocity measurements to the literature values, assuming a constant velocity uncertainty for the PN.S. This yields a measurement uncertainty of 25 km/s for the PN.S for high-S/N objects -- see the plot below, where the band shows values in agreement within ±25 km/s.

How much flexure is there in the instrument? (I.e., what error would be introduced if we used the "wrong" arc-lamp calibration image for a given PN?) Testing on the same science image (PNG 019.2-02.2) with different calibration images (at different telescope elevations), it appears that flexure is a minor concern: the inferred velocity varies by ~8 km/s. (But if a refocussing takes place, this can introduce a change of ~40 km/s.) So the current error budget must be dominated by the uncertain wavelength solution of a given image: either not high enough order terms used, or the two spectral lines aren't sufficient.

Does the wavelength solution maintain its integrity over the whole field-of-view? So far, these calibrations have all been done at the center of the field, where distortions are minimal. One check is to observe a Galactic PN in different parts of the field. This has been done for PNG 018.6-02.2, which has been observed in the center and the four corners of the field. However, the results are puzzling. The inferred velocity varies by 20-40 km/s, but this appears to be a monotonic function of time, which doesn't make sense. So we need to re-do these observations with more arc-lamp images, and more field positions, to figure out what's going on. One idea was that the field-varying passband might cause the centroids of the lines to vary, since at 6°, the variation is strong, but these shouldn't be, because the lines are virtually delta functions in wavelength space as they go through the filter, so the passband should be irrelevant; what is still possible is that the PSF of the holes could be distorted in the corners, producing centroiding variations -- check. Also make sure these PNe are unresolved.

Local Group PNe


We have observed a few fields in M31, and can compare our PN velocities with those of Halliday, Carter, Bridges, & Jackson (unpublished).
Below is a section of a field in the left arm (top) and right arm (bottom). On the left are the original images with three PNe identified (m5007 ~ 22.0). One PN is hidden beneath a stellar trail, but it pops out quite clearly when a 21x1 median filter is applied and then subtracted (right).

In field NE2, we have 12 PNe in common with Halliday et al. Our velocities are systematically lower than theirs by 43 km/s. If this difference is subtracted off, then our results are consistent with theirs if our velocity uncertainty is a constant 13 km/s (chi-squared=11).

In field NE1, we have 8 PNe in common with Halliday et al. In this case, we find our systematic offset is -17 km/s and our uncertainty is 18 km/s. However, for this field, we didn't get an arclamp exposure at the same position angle -- the one used for the calibration was 30 arcmin away, which shouldn't cause a large effect, but another one 50 arcmin away gives some significantly different velocities, implying that this field's calibration can't be trusted.

In fields NE4, SW1, (centre), we should have some PNe in common with Halliday et al.

In field NE3, we may have some repeated measurements in common with field NE2.


Compare with Ford velocities...

More distant extragalactic PNe

And IV

This is a dwarf galaxy at a distance of ~5-8 Mpc. Ferguson, Gallagher & Wyse (2000) have measured velocities of 3 emission line regions. Our (still rough) velocities of these objects are fully consistent with theirs, even if we assume our uncertainties are zero (note these are all near the centre of our field):

Name PN.S vhelio (km/s) Literature vhelio (km/s)
And IV 3 242.1 244±15
And IV 4 263.3 250±13
And IV 6 287.5 273±19


Compare with Feldmeier et al. velocities...


Compare with Puerari et al. velocities.

NGC 5866

Compare with ISIS velocities...

Below is a summary of all the calibration data. The white dotted line is a perfect fit to the literature data. The green points are the Galactic PNe, with a ±25 km/s band shown by dashed lines. The blue points are the And IV objects. The red points are the PNe from M31 field NE2, with a dotted line showing a -42 km/s offset. The magenta points are the PNe from M31 field NE1.

NGC 3379

The PNe velocities we calculated for NGC 3379 show an offset of ~20 km/sec if compared to the velocities published by Sluis et al.

I run 2 tests in order to figure out the cause of this offset. In the first test I compared the wavelength solution obtained with the reference mask (r484614un / r484615un) with the wavelength solution obtained using two different masks. The new masks I used are the ones taken immediately after (r484618un / r484619un) and immediately before (r484608un / r484609un) the reference mask.
In the following plot I show:
red diamonds:The velocities calculated with the r484608/09 mask minus the velocities calculated with the reference mask r484614/15
black diamonds:The velocities calculated with the r484618/19 mask minus the velocities calculated with the reference mask r484614/15

The differences are quite small, so we can exclude an error associated to the mask itself.

In the second test I tried to figure out if the shift is generated somewhere in the pipeline. In particular in the alignspots, undistort and calctrans procedures.
I took the L/R reference masks, I considered its emission lines spots as Planetaries and I run pnsvel on them to calculate their wavelengths. Then I calculated the velocities which belong to every measured wavelengths (W_PNS) comparing it with its nominal value W_lab:
V = c*(W_PNS - W_lab)/W_lab
If the error is somewhere in the script it will be present also in this calculation.So I expect to measure a systematic shift of about ~20 km/sec. But, as the following plots show, the "velocity" of the spots is consistent with zero, as we would expect if no shift is present.

The velocities associated with the 5009AA emission lines. The median value is 4.6 km/sec , with a sigma of 6.9 km/sec

The velocities associated with the 5017AA emission lines. The median value is -6.9 km/sec , with a sigma of 3.2 km/sec

The velocities associated with the 5023AA emission lines. The median value is -1.8 km/sec , with a sigma of 15.8 km/sec

The velocities associated with the 5031AA emission lines. The median value is 3.6 km/sec , with a sigma of 5.3 km/sec

The velocities associated with the 5038AA emission lines. The median value is -1.2 km/sec , with a sigma of 4.4 km/sec

It appears that the shift is not related to calctrans, alignspot or undistort. Moreover, since the shift seems to be related only to the target images and not to the masks, it must be originated in some steps which involve only the galaxy images. Thus,there are 2 hypothesis left for the shift's origin:
1) The masks are generated "in" the telescope, while the galaxy images come from infinity. Maybe the shift is generated in the optical path of the instrument itself.
2) The galaxy images are processed by xstartrails, while the masks are not. Maybe the xstartrail procedure is not able to find the correct center of the stellar trails, because of the shape of the spectrum. The same alignment error is repeated for every image on the stuck, because the stellar trails are identical. This may generate the shift.

last modified by AJR, 4 February 2002
Additional comments on NGC 3379 by LC, March 2006