ARG0003ifv - What's happening here?!?
-
by
JeanTate
The E source looks like a bright core (or lobe) with a faint jet. No obvious host, and not overedge.
The central one? A plume/relic? very faint IR host; in SDSS, z_ph 0.372 ± 0.0612/0.238 ± 0.1253 SDSS J010122.19+044124.8 is host?
Or are they both unusual lobes (doublelobe), with the host z_ph 0.584 ± 0.1126/0.542 ± 0.0991 SDSS J010123.83+044135.6?
Posted
-
by
JeanTate
And an overlay doesn't seem to be much help:
The E source could be #compact with a small #jet, and host UNRELIABLE SDSS J010125.05+044140.4. But the central source?!?
And just for fun, there's a nice, slow-moving asteroid in the field (SDSS J010120.83+044150.1):
Boilerplate: SDSS image per
http://skyservice.pha.jhu.edu/DR10/ImgCutout/getjpeg.aspx, FIRST (red) contours derived from the FITS file produced using SkyView with Python code described in this RGZ Talk thread. Image center (J2000.0) is the center of the ARG image ARG0003ifv.Posted
-
by
ivywong
scientist, admin
This is a good example when higher resolution infrared image of this field would be very worthwhile. This is also a good example which shows why we chose to use the WISE infrared images over the SDSS optical images. We don't see much in the WISE images but we see even less in the optical images...
Thank you all for giving it your best shots!
Posted
-
by
42jkb
scientist, admin
Higher resolution radio images would help here too. This can get rather difficult!
Posted
-
by
JeanTate
... shows why we chose to use the WISE infrared images over the SDSS optical images. We don't see much in the WISE images but we see even less in the optical images...
Now if there were a way to produce WISE and FIRST overlays on SDSS images, deriving the WISE and FIRST ones from FITS files ... 😉
Posted
-
by
ivywong
scientist, admin
You can always overlay the WISE images as a separate set of contours but the images will get a whole heap busier. My suggestion is to have a 2 panel image. Have one panel with FIRST contours over SDSS and the second panel can have the same FIRST contours over WISE. What do you think?
Posted
-
by
JeanTate
in response to ivywong's comment.
Something like this?
The contrast could be turned up a bit, and perhaps the WISE contours done in 'lime green'; nonetheless it highlights the different colors of many objects well ... bright SDSS ones that are faint in the WISE band, and vice versa. It also shows that the angular resolution of the WISE data is comparable to that of the FIRST, and that the SDSS has much sharper eyes. The asteroid was long gone from the WISE sky, of course.
I wonder what a 3-band* WISE mapped to RGB image would look like, Luptonized color and all ...?
Boilerplate: SDSS image per
http://skyservice.pha.jhu.edu/DR10/ImgCutout/getjpeg.aspx, FIRST (red) and WISE 3.4μ (green) contours derived from FITS files produced using SkyView with Python code described in this RGZ Talk thread. Image center per the ARG image (left; J2000.0).*three shortest wavelengths
Posted
-
by
JeanTate
in response to JeanTate's comment.
Even with the a mild re-weighting , the 3.4μ emission dominates.
I wonder what a 3-band* WISE mapped to RGB image would look like, Luptonized color and all ...?
Quite an alien sky, if all you know is SDSS (same piece of sky; same center, same scale):
Posted
-
by
JeanTate
in response to JeanTate's comment.
To "Luptonize"* is to produce an RGB JPEG colorized using the method described in Lupton+ 2004, "Preparing Red-Green-Blue Images from CCD Data" (arXiv preprint here).
Calling the WISE data "CCD data" may be a bit of a stretch (pun unintended), especially for the 12μ and 22μ bands, but the approach described is fun to try, whatever scientific merits it may (or may not) have.
For the three WISE bands I have chosen - 3.4μ mapped to B, 4.6μ mapped to G, and 12μ mapped to R - there are at least two consistent methods for defining I ("intensity"?, some sort of measure of total flux, combining all three bands):
- take the flux values^ in the FITS files downloaded from SkyView
- normalize so that the max value in each band is the same (after removing the sky, and whatever 'noise floor' above the sky)
The Python code I wrote, to produce my Luptonized images, is similar to that described in this RGZ Talk thread. In particular, the 'sky' is the median, and the 'noise floor' sigma (calculated using the MAD statistic) multiplied by a small number (typically between 1.0 and 3.5).
Here are two Luptonized images of the above field, produced using these two methods (same f(I) - asinh - same m and M, same β, etc), respectively:
Why do they look so different?!? 😮
And can FIRST and/or NVSS contours be overlaid on Luptonized WISE images?
Stay tuned ... 😉
*I have no idea if anyone else uses this term; maybe I made it up ...
^or whatever they are (UPDATE (19 August, 2014): the SkyView FITS headers state "PixelUnits: DN", so definitely not flux! 😮 but likely related 😃).
(new) Boilerplate: "Luptonized" image produced from WISE FITS files (3.4μ, 4.6μ, and 12μ bands) obtained from SkyView with Python code described in {this very post!} Image center per the ARG image (left; J2000.0).
Posted
-
by
JeanTate
in response to JeanTate's comment.
Why do they look so different?!? 😮
Here's my amateur's understanding; I do hope a professional will jump in and correct any mistakes, and also add to my explanation.
The first image is dominated by 'blue' objects because most of the sources were observed to have greater flux in the 3.4μ WISE band than any other, though some faint sources have greater flux in the 12μ band than any other (they appear red).
In the second image, ~two objects are ~white (or, more accurately, have ~white cores). This is a consequence of the normalization; they are the brightest sources in all three bands, so end up with r ~= g ~= b. Most of the (other) 'blue' sources in the first image are 'turquoise' in the second; this too is a consequence of the normalization, which increases g relative to b. The faint red sources in the first image are bright red ones in the second; again, a consequence of the normalization, which greatly increases r relative to both g and b. Some of the 'new' red sources in the second image are real - they become 'visible' due to the normalization - but most are likely not; instead, they are random fluctuations in the sky, promoted to 'objects' by the relatively low 'noise floor' and normalization.
Why is there apparently no 'blue noise' in the first image? And what happened to the many faint (apparent) images in the SkyView black-n-white 3.4μ image?
I don't know, yet; I suspect it has something to do with how my code decides which pixels (locations) have a combined I of zero (such pixels end up black in the RGB JPEG) ... I think it may be too aggressive.
Posted
-
by
42jkb
scientist, admin
Just a quick note on colour WISE images before I digest this conversation.
If you click on the WISE link, it takes you to the WISE image server. There is a tab above the image that will create the colour image for you.
Blue = sourcces with the infrared emission dominated by star light (3.4 um)
Red = sources with the infrared emission dominated by hot dust resulting from star formation or from the AGN heating the dust or a combination of both (24 um).You can also see this on the black and white images on the WISE server as it shows the emission at the four WISE bands (3.4, 4.6, 12, 24um).
Determining the amount of star formation vs AGN heating has yet to be solved and is one of the things we are trying to tackle with RGZ.
Posted
-
by
WizardHowl
in response to JeanTate's comment.
Here's my explanation, as someone who has also spent some time classifying for Disk Detective: what you have done is to highlight objects with an IR excess.
Most objects with IR excess will be galaxies and AGN/QSOs but also include stars that may be throwing off material, especially red giants (potential Disk Detective false positives also include Be stars and Cepheids, whilst the debris disks being searched for are relatively rare).
If you look at the Spectral Energy Distribution (SED) for a typical galaxy, from the optical through to IR (as in Disk Detective, which includes WISE band 4) it will be a V-shape. The left half of the V is the light emitted by stars in the galaxy, whilst the right half is the result of dust in the galaxy absorbing that light and re-emitting it in the IR. AGN and QSOs have different SEDs (sometimes very odd, depending on redshift and other factors) that start faint in the optical but typically also get brighter towards the IR.
By choosing a colour scheme with redder images being the later WISE bands, you have therefore highlighted those objects that are brightest at those wavelengths, which are usually extragalactic objects.
It would also seem to me that by normalising such that the max value in each wavelength band is the same will highlight the fainter WISE bands, essentially giving them a multiplier to make them match the others, which will also bring up any noise in those bands as well as genuine signals. Looking at the WISE band 3 and 4 images from the link in RGZ often shows nearly featureless fields with maybe one or two brighter sources. RGZ being what it is, there is always a chance some of this 'noise' might actually be due to a barely-detected background galaxy cluster, though!
Posted
-
by
JeanTate
in response to JeanTate's comment.
Thanks 42jkb and WizardHowl! 😃
Why is there apparently no 'blue noise' in the first image? And what happened to the many faint (apparent) images in the SkyView black-n-white 3.4μ image?
Turns out an important reason is that the method I was using to determine the sky - a method which works well with almost all FIRST images (but fails for some NVSS ones) - was inappropriate ... it results in a sky which is too bright. So I need to go back and modify my code (no sad smilie; this is fun!)
In the meantime, a couple of images produced by setting the noise floor below the sky, and in a band-dependent way; the first leaves fluxes unchanged; the second normalizes them (as described above):
Boilerplate: Background "Luptonized" image produced from WISE FITS files (3.4μ, 4.6μ, and 12μ bands) obtained from SkyView with Python code described in this RGZ Talk post. Image center (J2000.0) is the center of the ARG image ARG0003ifv.
Posted
-
by
JeanTate
in response to JeanTate's comment.
Turns out an important reason is that the method I was using to determine the sky - a method which works well with almost all FIRST images (but fails for some NVSS ones) - was inappropriate ... it results in a sky which is too bright. So I need to go back and modify my code
At least for this field, the sky is black. In all three WISE bands. The 3.4μ band data seems to have ~no large, diffuse emission (lots of ~point sources; diffspikes excepted); the 4.6μ one some such emission; the 12μ one, lots. Deciding appropriate 'noise floors' seems to me is thus pretty arbitrary.
I like this combo because is has a modest amount of 'green noise' (though perhaps still a bit too much) and 'red noise', in different parts of the field. There are also some, relatively faint, point sources with colors other than blue or red; yay! 😃
'Normalizing' the pixel values, as described above, produces an image which is not particularly informative (to me, anyway):
Boilerplate: Background "Luptonized" image produced from WISE FITS files (3.4μ, 4.6μ, and 12μ bands) obtained from SkyView with Python code described in this RGZ Talk post. Image center (J2000.0) is the center of the ARG image ARG0003ifv.
Posted
-
by
JeanTate
in response to JeanTate's comment.
And can FIRST and/or NVSS contours be overlaid on Luptonized WISE images?
Of course .... 😃
Boilerplate: Background "Luptonized" image produced from WISE FITS files (3.4μ, 4.6μ, and 12μ bands) obtained from SkyView with Python code described in this RGZ Talk post. SDSS image per
http://skyservice.pha.jhu.edu/DR10/ImgCutout/getjpeg.aspx, FIRST (red) and NVSS (cyan) contours derived from FITS files produced using SkyView with Python code described in this RGZ Talk thread. Image center (J2000.0) is the center of the ARG image ARG0003ifv.Posted
-
by
JeanTate
in response to 42jkb's comment.
Exploring the WISE Help pages ...
The pixel values in the Atlas Images (from which, presumably, the SkyView FITS are derived) are in units of "DN" (Digital Numbers"). To within some ~constants (hopefully, mostly close to 1), these correspond to fluxes (although the Cautionary Notes pages should be always kept in mind). In bands 3 and 4 (12μ and 24μ, respectively), diffuse background emission is not unexpected, especially in/near the galactic plane/bulge.
Except, perhaps, for very late-type stars (spectral class M?, L, T, or Y), 'naked' stars* have monotonically decreasing SEDs; the longer the wavelength (band), the lower the flux. 'Dead and red' galaxies should also, thus, have similar monotonically decreasing SEDs (being 'dead' means essentially all the dust has been destroyed). Unless their redshifts are > x (1?); but such distant galaxies will be too faint to be detected by WISE (unless, of course, they have AGNs).
Dusty galaxies, especially those with warm/hot dust, should stand out like sore thumbs ... they will be (bright) green or red (depending on z and average dust temperature) in band 1-3 Luptonized WISE images: Eos, post-starburst, merging, ...
An AGN is a wild child; its SED may totally dominate that of the rest of the galaxy, in all WISE bands (or not); its band 1 (accretion disk/jet) and band 4 (dusty torus) emission may be ~comparable ('V' or 'U'-shaped SED), or not; and so on.
Luptonized/colorized WISE images can surely help to get a 'first pass' understanding of the nature of candidate hosts, especially when combined with SDSS images (which are already Luptonized).
*most stars are naked; they do not have associated dust disks/remnant natal ('Bok globule'?) clouds/'candle smoke'
Posted
-
by
JeanTate
in response to WizardHowl's comment.
Most objects with IR excess will be galaxies and AGN/QSOs but also include stars that may be throwing off material, especially red giants (potential Disk Detective false positives also include Be stars and Cepheids, whilst the debris disks being searched for are relatively rare).
And away from the galactic plane, these should be vanishingly rare, right?
If you look at the Spectral Energy Distribution (SED) for a typical galaxy, from the optical through to IR (as in Disk Detective, which includes WISE band 4) it will be a V-shape. The left half of the V is the light emitted by stars in the galaxy, whilst the right half is the result of dust in the galaxy absorbing that light and re-emitting it in the IR.
Indeed. Except for 'dead and red' galaxies, without AGNs (and as long as z < ~1?). This is very helpful for us, in investigating candidate hosts of 'spirals with associated doublelobes'.
AGN and QSOs have different SEDs (sometimes very odd, depending on redshift and other factors) that start faint in the optical but typically also get brighter towards the IR.
Such non-conformity; why can't they be well-behaved like the other children? 😉
It would also seem to me that by normalising such that the max value in each wavelength band is the same will highlight the fainter WISE bands, essentially giving them a multiplier to make them match the others, which will also bring up any noise in those bands as well as genuine signals.
Yes, that's exactly what seems to happen (at least for this field).
Posted
-
by
42jkb
scientist, admin
in response to JeanTate's comment.
Here is a good paper that describes what the WISE bands will tell us about the IR source. It should be free, if not let me know.
http://adsabs.harvard.edu/abs/2011ApJ...735..112J
Posted