First things first: I just noticed that you are in Finland. That would explain why you are always posting so "late at night."

Just trying to be useful. It helps my understanding to try and explain it.

But before we go any further, I should ask how much you know about spectroscopy to begin with? Do you know what a thermal (blackbody) spectrum is, vs., say, a synchrotron (power law) spectrum? What causes absorption vs. emission lines? Things that can broaden emission lines?

Understanding those is necessary for really understanding what is going on with the spectra that I listed, and why they are different.

No, I think I wasn't clear. The light-profile of a source can tell you whether it is strictly a point source (that is, smaller than the resolution of the camera), and determining the light profile of a known point-source (say, a star) is the first step toward subtracting the PSF; the PSF of the camera is what an ideal point source with infinite signal to noise should look like.

If you have a well determined PSF, you can use the light profile to determine the size of things that are just beyond the PSF, as Mike Brown et al. did to determine the size of Eris (middle of the How Big is it? section, about HST). They got a very good constraint on Hubble's PSF, and used that knowledge to determine the angular size of Eris. They didn't subtract the PSF, just determined what it was for comparison.

They are actually probably better than you think: one of the two photo-z methods (the PhotoZ2 table in Sky Server) is described at this website. The absolute error plots (done on the validation set, which have spectra) are given for a few different brightness bins. Basically, for galaxies with r-band magnitude brighter than 20, the photo-z estimator has quite small errors.

I don't think Sky Server has photo-z's for quasar candidates listed yet (at least, computed based on knowledge of quasar spectra), but they are also quite accurate. I can't find a good paper about it online (the only ones are based on the SDSS early data release, which was years ago, and the methods are much better now), but there are some Bayesian statistical techniques that work quite well, with small scatter.

That depends. There are certain combinations of spectral lines that could potentially confuse an automatic classification system. Again, we're talking about the corner cases here, most of the time everything is fine. As you say, if there are many spectral lines, then there's generally nothing to worry about. But an invalid redshift can result from having few and/or weak lines. If they are weak (low signal-to-noise), or if the spectrum is otherwise odd, the spectral classifier can get it wrong (mis-interpreting lines, thinking that noise features are lines), and thus the redshift can be wrong.

Well, as you can see, there is definitely a galaxy there. There is also a very odd nuclear source. The spectrum of the nucleus in this case might include some of the galaxy's spectrum shining through, but I'm not really sure. Remember, NED just contains a list of objects that were described in a paper somewhere, it is not a telescopic survey itself. This object was probably first discovered as a galaxy (in DSS plates, I'd bet), and, from the NED reference list, the BL Lac was discovered in 1989 by Remillard et al. after follow-up from X-ray observations.

I'm rather curious what Arp et al. have to say about the immense X-ray and radio emission from these sources. We're talking outrageous fluxes here (note I said fluxes, not luminosities!). The standard model, where we are looking down the jet from the black hole, explains the features of these systems quite well. Some of the details are still a bit tricky...

That depends. Jet-induced blueshift could be one reason why the redshift might be wrong. Another reason is that there are very few lines in the spectrum, so it is hard to get a redshift in the first place (see my discussion above). In think the redshift of this particular system is ok, because there are enough absorption lines (possibly from the host galaxy) to get a good redshift determination. But for many BL Lacs, this is not the case, because the nuclear source is so much brighter than the host galaxy itself, and the optical spectrum of a BL Lac is nearly featureless.

Yup! That means it was targeted for spectroscopy because the pipeline thought it might be a quasar (though you may have already figured that out). The spectroscopy targeting flags are listed in this table and details on the spectroscopic target selection algorithm are available, if you'd like to peruse them. They aren't easy to digest, though.

Ok, here's a few questions for you (being the present potential supporter of Arp and company), though I'm sure they've all been asked before: how do you propose determining whether two objects are at the same distance, if you don't think their redshifts are relevant quantities for distance determination? I understand that Arp et al. claim that the redshifts of quasars are actually due to some, unknown, undetermined and undescribed physical process, but then how do you determine the distances to things? Remember, there are over a million objects in SDSS DR6 with spectroscopy, including more than a hundred thousand quasars, ~800,000 galaxies and ~300,000 stars... And if the redshifts to galaxies are trustworthy, why not quasars?

And what about galaxies where some other measure is used to determine the distance, and it agrees with the redshift (to a fairly good degree, if not exactly)? What about other measures of distance, say estimates based on normalized galaxy sizes, or rough luminosity distances? Those aren't precise, but they qualitatively match the theory that redshift is a measure of distance: further things are smaller and dimmer.

It does distinguish, but you have to understand the actual spectral shape. That's why I asked about Blackbody vs. power-law spectra above. Stars are generally nearly perfect blackbodies, with some absorption lines due to the photosphere. Quasars generally have a power-law spectrum (flux proportional to frequency^alpha), which you can see in the rate at which the flux increases toward blue wavelengths.

A good place to start about stellar spectra might be the Stellar Spectral Types Project from the Advanced Projects page (the one from Basic Projects is a simpler subset of that one, but the Advanced one has a lot more description).

And now we've run into the problem with using the quick-look spectra in SDSS! They are not designed to do science from, just for a quick glance.

In the star-forming galaxy's spectrum, the "quite broad line at ~7000Å" is, infact, the Hα, [NII] blended doublet. You can kind of make out the doublet in the image, but it is much more obvious in the complete FITS spectrum. The quasar spectrum, on the other hand, has little sign of the [NII] line, and very broad wings (look at the base of that line, as well as Hβ, in particular). This particular quasar has relatively narrow lines, compared to some!

If you want to examine the FITS spectra in detail, you can use specview a free Java application from the Space Telescope Science Institute.

Perhaps, but the broad lines are a give-away. Again, just how broad the lines are compared with the star-forming galaxy above is more apparent in the full FITS spectrum. Also, the continuum does not fall off towards short wavelengths, as happens with galaxy spectra. Again, flip back and forth between it and the one we both agree is a quasar (1): the similarities are striking.

I'm not sure I understand the question, but I'll try and answer what I think you were asking. A spectrum like this is the sum (linear combination, to be exact) of two very different blackbody spectra. It is hard to have two wildly different blackbodies together on the same object: they will thermodynamically equilibrate rather quickly (what's the temperature of things on the surface of the Earth? All roughly the same, and Earth isn't even a blackbody!)

However, what is a galaxy? Lots and lots of stars (and gas and dust, as well). So a galaxy's spectrum (ignoring the gas and dust, for now) should be the sum of a lot of blackbodies of different temperatures. The exact shape of a galaxy's spectrum depends on the stars in it: if they are old stars, it will be redder (cooler), if young stars, bluer (hotter). The gas and dust complicate things because dust preferentially absorbs blue light, and gas can produce absorption or emission lines, depending on how dense it is. So a starforming spiral galaxy might have a blue spectrum like the star earlier, with emission lines from hot gas and some dust absorption, while an old elliptical galaxy in a cluster might show just old, red stars and no signs of gas or dust, because it has all either turned into stars or been kicked out due to interactions.

You'll find that note on all the SDSS quick-look spectra (including stars). Almost everything data-related in SDSS is automatically generated. There are just too many sources for any kind of manual intervention, in general. The few cases where a person was in the loop are usually mentioned (e.g. the MANUAL_MAPPED specZWarning flag).

In this particular case, the spectroscopic pipeline found that the spectrum was best fit by a quasar template--hence the QSO identifier. But it doesn't have PNe templates to draw on when fitting, so this type of object would cause confusion. The redshift looks to be correct, though, since there are plenty of lines and it identified them correctly (again, compare with the owl nebula).

I may have to rescind that statement! Looks like this one has a supernova in the spectrum, as described in Madgwick et al. 2003! It may well be a galaxy spectrum (though at very high redshift) with the supernova providing some of the stranger features. Looks like I got caught out on that one... And the SN probably isn't visible in the image at all, because the spectroscopy happened a year or more later.

As I said, I'm still learning!

__________________
"What do you care what other people think?" -- Richard Feynman
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Feynman, at the conclusion of his Challenger report