Bill Keel posted this note to sci.astro on Nov 4, 1996. I've edited it slightly. MWR
Having just lost most of my carefully written post on this (blast it, I hate hitting the caps lock key at the wring time), I'll just throw in the relevant bit of TeX notes from last time I taught my grad class on galaxies - this summarizes the main arguments with references: Bill Keel Astronomy, University of Alabama
Much of our understanding of the physics of AGN depends on knowing their absolute properties (luminosities, size scales) and thus their distances. There is a small but vocal school which claims that much of the redshift of QSOs (at least) arises not in the Hubble flow but in exotic physical processes, and thus that redshift distances to (some?) QSOs are nonsense. This point of view has been defended in Arp's book ( Quasars, Redshifts, and Controversies, Interstellar Media, Berkeley), with some of his best cases.
The observational suspicion that some AGN might be at noncosmological distances seems to have first arisen when Arp (1967 ApJ 146, 321) noted an association between several low-redshift peculiar galaxies and quasars in a rough pairing sense, with pairs of QSOs on each side of the galaxy. Specific two-color searches led to identification of numerous QSOs in te fields of nearby galaxies (see, for example, Arp 1981 ApJ 250, 31 and references therein).
There has been much fruitless discussion of what might appear a straightforward statistical problem - are there or are there not excess QSOs in the directions of bright galaxies? The difficulties lie in the fact that QSO searches are still quite inhomogeneous over the sky, and thus a search may be deep enough to tell us something but cover too little solid angle, or cover the whole sky with too few QSOs. For example, there are four close galaxy-QSO pairs in the 3C catalog (Burbidge, Burbidge, Solomon, and Strittmatter 1971 ApJ 170, 233). But with only about 100 quasars over half the sky, the statistics are too sparse to do more. Perhaps large-scale automated surveys will be able to resolve this. The methodology Arp has frequently adopted doesn't help - starting from a galaxy and searching outward until a quasar shows up, then if it's ``interestingly" close keep on going outward. This is guaranteed to produce an apparent excess, on the ``seek and ye shall find" principle. A final problem with a statistical analysis is that it is not always clear what it is whose likelihood we want to assess. Some papers talk about QSO-galaxy pairs, some about QSO pairs with discordant redshift, lines of quasars... Statistics after the fact has a bad reputation.
If any of these claims hold up, extragalactic astronomy is in for a real shock. We will examine the direct issues individually, hoping to avoid the ``oh yes it is - oh no it's not" tone of many published papers.
There are additional cases sometime discussed; see the list on p. 86 of Arp's book. One may also add a discrepant-redshift galaxy in the chain or compact group VV172.
This is the crux of the debate on quasar redshifts. First, is there compelling evidence of quasars occurring in the direction of nearby galaxies? And second, can such an excss be explained by something like gravitational microlensing?
Arp conducted numerous searches using U-B colors for quasars near bright galaxies, and concluded that certain kinds of companion galaxies were especially likely to have associated QSOs. The radial distribution of these objects is not too different from that of a spheroid (de Vaucouleurs profile), as shown by Keel 1982 (ApJLett 259, L1) in checking for evidence of lensing.
However, some of this is an artifact of Arp's center-outward search procedure. Real progress here will require large, objectively selected QSO samples covering significant solid angle. Also, the ststistical analysis for associations has often been clouded by uncertainies as to what one is testing for. We need to know not just the probability of seeing what we do, but of seeing something ``sufficiently interesting", be that pairs, lines, associations in angle, or whatever. Various papers dispute probabilities by factors of $10^7$.
There are clearly some very striking single objects, such as the three quasars seen along the line of sight to NGC 1073 (Arp and Sulentic 1979 ApJ 229, 496) and another three near NGC 3842 (Arp and Gavazzi 1094 A\&A 139, 240). However, the answer to this must rest on quantifiable statistics for which it is clear that large areas of sky with and without bright galaxies have been searched. Arp and Hazard have examined a few ``blank fields" and report interesting structure in the quasar distribution even there. With recent evidence on large-scale structure in galaxies, perhaps we are falling victim to a facile assumption that the quasar distribution is much more uniform at moderate redshifts $z=1-2$ than it really is. Note also that there is a minor literature now on gravitational-lensing explanations, as to whether objects in galaxy halos can reasonably account for an excess of the claimed magnitude - see for example Canizares 1981 (Nature 291, 620). Archival variability studies show that in every case we can check the quasar was there at the turn of the century (Keel 1982) so that this is not due to lensing by low-mass objects; proper motions would destroy the necessary precise alignment in a few decades. If Arp's objects turn out to represent a strong excess, there does not appear to be enough mass density in typical galaxies to do all of it via microlensing.
There are certain peculiarities, claimed or accepted, that suggest either strange behavior of redshifts or that we don't know how to measure them as well as we think. These take the forms of an inescapabale asymmetry in redshifts of binary galaxies, and claims that such redshift differences are quantized and completely disallow a dynamical interpretation.
Redshift asymmetries are found in almost all samples of paired galaxies with precise redshifts, especially where spirals are involved. The tendency is for the fainter galaxy to have a slightly larger redshift, with a peak in the distribution at 50-80 km/s. The form of the distribution suggests that this is independent of background contamination. Conventional explanations have focussed on problem in measuring redshifts of dusty rotatig disks (for example, if dust is stronger on the inside or outside of arms, the nuclear velocity may be distorted) or, for small groups, on expansion and perspective effects in unbound groupings (Byrd and Valtonen 1985 APJ 289, 535; 1986 ApJ 303, 523). This problem is not directly related to AGN, but letting one camel's nose into the Hubble tent might weaken its defenses for other applications.
Tifft has claimed that redshift differences in galaxy pairs are quantized, with intervals from 12-72 km/s depending on sample size (Tiffy 1982 ApJ 257, 442; 257, 442, for example). Such an effect must nullify velocity Doppler shifts since, under conventional dynamics, they would smear out any other fine structure in the velocity distribution into invisibility. These distributions have naturally been the subject of vigorous (that's the polite word) debate. There have been claims that such periodicity could not be found from samples of the sizes used (Newman et al 1989 ApJ 344, 111), and possibly most damaging, the finding that though different data sets show similar $\Delta V$ distributions, a single pair may move from one peak to another depending on the particular measurement (Sharp, Trieste proceedings 1985). This last would imply that the periodicities exist in the data but not in the sky, a depressing though for people who want to do precise redshift surveys. Most recently, Tifft has concentrated on 21 cm redshifts, identifying several previously ignored sources of low-level error. However, there is a limit to how precisely one can consider some moment of an H I profile as representing ``the" galaxy redshift - just look at a 21 cm map of M101, for example. Sample periodicity diagrams are taken here from fig 1 of Tifft 1982 (ApJ 257, 442).
Any of the above phenomena would require explanation through some sort of new physics, the sort that people get Nobel Prizes for working out. Some of the original impetus for noncosmological redshifts arose, oddly enough, from conventional physics - the ``synchrotron catastrophe", in which quasar luminosities would be too high to sustain against their own synchrotron self-absorption. However, Seyfert galaxies know how to do this perfectly well at smaller and better-determined distances, so this seems to be our problem and not the universe's. Furthermore, people such as Hoyle who found a staedy-state universe appealing on philosophical grounds needed some other avenue to make objects that appear at first glance to show cosmological evolution. What do we require of any mechanism that can mimic Doppler shifts? It must
In his book, Arp sets out an evolutionary scheme that he finds acceptable from his interpretation. Objects are ejected from galactic nuclei, possible at very high velocities, with initially large density, high temperature, and large redshifts (quasars and BL Lac objects). As they age, stars appear starting with early-type ones and the redshift decreases. Finally, extended halos or spiral features appear, and the noncosmological redshift nearly vanishes. This gives sort of a fireworks-display view of galactic history. Most QSOs then are not very large or bright - more like the brightest supergiants than galaxy-hiding monstrosities.
So what are the arguments directly favoring conventional cosmological distances for quasars? We may examine associated and host galaxies, gravitational lenses, and absorption-line systems.