B.7.2. Binary Quasars

Weedman et al. ([1982]) reported the discovery of the third "gravitational lens", Q2345+007, a pair of z = 2.15 quasars separated by 7."3. The optical spectra of the two images are impressively similar (e.g. Small et al. [1997]), but repeated attempts to find a lens have failed in both the optical (e.g. Pello et al. [1996]) and with X-rays (Green et al. [2002]). Q2345+007 is the founding member of a class of objects seen in the optical as a pair of quasars with very similar spectra, small velocity differences and separations 3."0 15."0. The most recent compilation contained 15 examples (Mortlock, Webster & Francis [1999]). The incidence of these quasar pairs in surveys is roughly 2 per 1000 LBQS quasars (see Hewett et al. [1998]) and 1 per 14000 CLASS radio sources (Koopmans et al. [2000]). The separations of these objects correspond to either very massive galaxies or groups/clusters. Obvious lenses on these scales, in the sense that we see the lens, are rare but have an incidence consistent with theoretical expectations (see Fig. B.52). If, however, even a small fraction of the objects like Q2345-007 are actually gravitational lenses, then dark lenses outnumber normal groups and clusters and dominate the halo population on mass scales above M 10¹³ M.

If the criterion of possessing a visible lens is dropped, so as to allow for dark lenses, proving objects are lenses becomes difficult. There are two unambiguous tests - measuring a time delay between the images, which is very difficult given the the long time delays expected for lenses with such large separations, or using deep imaging to find that the host galaxies of the quasars show the characteristic arcs or Einstein rings of lensed hosts (Figs. B.3, B.4). The latter test is feasible with HST ⁷ and will be trivial with JWST. Spectral comparisons have been the main area of debate. In the optical, many of the pairs have alarmingly similar spectra if they are actually binary quasars (e.g. Q2345+007 or Q1634+267, see Small et al. [1997]) - indeed, some of these dark lens candidates have more similar spectra than genuinely lensed quasars (see Mortlock, Webster & Francis [1999]). The clearest examples of dark lens candidates that have to be binary quasars are the cases in which only one quasar is radio loud. These objects, such as PKS1145-071 (Djorgovski et al. [1987]) or MGC2214+3550 (Muñoz et al. [1998]), represent 4 of the 15 candidates. Similarly, the dramatic difference in the flux ratio between optical and X-ray wavelengths of Q2345+007 is the strongest direct argument for this object being a binary quasar (Green et al. [2002]).

Two statistical arguments provide the strongest evidence that these objects must be binary quasars independent of any weighting of spectral similarities. The first argument, due to Kochanek, Muñoz & Falco ([1999]), is that the existence of binary quasars like MMGC2214+3550 in which only one of the quasars is radio loud predicts the incidence of pairs in which both are radio quiet. We can label the quasar pairs as either O² R², where both quasars are seen in the optical (O) and the radio (R), O² R, where only one quasar is seen in the radio, or O² where neither quasar is seen in the radio. Lenses must be either O² R² or O² pairs. Surveys of quasars find that only P_R 10% of quasars are radio sources with 3.6 cm fluxes above 1 mJy (e.g. Bischof & Becker [2000]). If all the quasar pairs were binary quasars and the probability of being radio loud is independent of whether a quasar is in a binary, then the relative number of O², O² R and O² R² binaries should be 1 to 2P_R = 0.2 to P_R² = 0.01. Given that we observed 4 O² R binaries we should observe 20 O² binaries and 0.2 O² R² binaries. This statistical pattern matches the data, and Kochanek, Muñoz & Falco ([1999]) found that the most probable solution was that all quasar pairs were binary quasars with an upper limit of only 8% (68% confidence) on the fraction that could be dark lenses. With the subsequent expansion of the quasar pair sample and the discovery of the first O² R² binary (B0827+525, Koopmans et al. [2000]), these limits could be improved.

The second statistical argument is that the dark lens candidates do not have the statistical properties expected for lenses. Three aspects of the quasar pairs make them unlikely to be lenses simply given the properties of gravitational lensing. First, there are no four-image dark lens candidates even though a third of the normal lenses are quads. Second, many of the dark lens candidates have very high flux ratios between the images - 4 of the 9 ambiguous quasar pairs considered by Rusin ([2002]) have flux ratios of greater than 10:1. Magnification bias makes such large flux ratios very improbable for true gravitational lenses (Section B.6.6, Kochanek [1995b]). Third, the suppression of central/third/odd images in the lens population is a consequence of baryonic cooling and the resulting increase of the central surface density. Standard dark matter halos with their shallow central cusps, r^-1, generally produce detectable third images. Since it is probably a requirement for a lens to remain dark that the baryons in the halo cannot cool (or they would form stars), you would expect the typical dark lens to resemble APM08279+5255 and have an easily detectable third image (Rusin [2002]). Thus, in the context of CDM we would expect dark lenses to be standard cuspy density distributions like the NFW model (Eqn. B.60). Rusin ([2002]) evaluated the likelihood of the quasar pairs assuming that dark lenses have the structure of CDM halos and found that the observed flux ratios and the lack of three-image dark lenses were extremely unlikely. Only the real lens APM08279+5255 had a significant probability of being produced by a dark CDM halo, although for this case I think the exposed cusp/disk lens explanation for the morphology is more likely.

The evidence overwhelmingly favors interpreting the quasar pairs as binary quasars. However, as originally pointed out by Djorgovski ([1991]), the one problem with the binary hypothesis is that the incidence of the quasar pairs is two orders of magnitude above that expected from an extrapolation of the quasar-quasar correlation function on scales of Mpc. As discussed in Kochanek, Muñoz & Falco ([1999]) and Mortlock, Webster & Francis ([1999]) the incidence can be increased if the incipient merger of the two host galaxies is triggering the quasar activity. The separation distribution of the binary quasars is crudely compatible with tidally triggered activity when the merger starts followed by a coalescence of the host galaxies driven by tidal friction. Small separation binary quasars ( < 3."0) are rare because the decay of the host galaxy orbits accelerates as their separation diminishes. Well-measured angular distributions of binary quasars, potentially obtainable from SDSS, might allow detailed explorations of the triggering and merging physics.

⁷ We detected the host galaxies of the Q2345-007 quasars in the CASTLES H-band image. Their morphology is probably inconsistent with the lens hypothesis, but we viewed the data as too marginal to publish the result Back.