3. THE DISCOVERY OF QUASARS

Minkowski's studies of radio galaxies culminated with identification of 3C 295 with a member of a cluster of galaxies at the unprecedented redshift of 0.46 (Minkowski 1960). Allan Sandage of the Mt. Wilson and Palomar Observatories and Maarten Schmidt of the California Institute of Technology (Caltech) then took up the quest for optical identifications and redshifts of radio galaxies. Both worked with Thomas A. Matthews, who obtained accurate radio positions with the new interferometer at the Owens Valley Radio Observatory operated by Caltech. In 1960, Sandage obtained a photograph of 3C 48 showing a 16^m stellar object with a faint nebulosity. The spectrum of the object showed broad emission lines at unfamiliar wavelengths, and photometry showed the object to be variable and to have an excess of ultraviolet emission compared with normal stars. Several other apparently star-like images coincident with radio sources were found to show strange, broad emission lines. Such objects came to be known as quasi-stellar radio sources (QSRS), quasi-stellar sources (QSS), or quasars. Sandage reported the work on 3C 48 in an unscheduled paper in the December, 1960, meeting of the AAS (summarized by the editors of Sky and Telescope [Matthews et al. 1961]). There was a "remote possibility that it may be a distant galaxy of stars" but "general agreement" that it was "a relatively nearby star with most peculiar properties."

The breakthrough came on February 5, 1963, as Schmidt was pondering the spectrum of the quasar 3C 273. An accurate position had been obtained in August, 1962 by Hazard, Mackey, and Shimmins (1963), who used the 210 foot antenna at the Parkes station in Australia to observe a lunar occultation of 3C 273. From the precise time and manner in which the source disappeared and reappeared, they determined that the source had two components. 3C 273A had a fairly typical class II radio spectrum, F ~ ^-0.9; and it was separated by 20 seconds of arc from component `B', which had a size less than 0.5 arcsec and a "most unusual" spectrum, f ~ ^0.0. Radio positions B and A, respectively, coincided with those of a 13^m star like object and with a faint wisp or jet pointing away from the star. At first suspecting the stellar object to be a foreground star, Schmidt obtained spectra of it at the 200-inch telescope in late December, 1962. The spectrum showed broad emission lines at unfamiliar wavelengths, different from those of 3C 48. Clearly, the object was no ordinary star. Schmidt noticed that four emission lines in the optical spectrum showed a pattern of decreasing strength and spacing toward the blue, reminiscent of the Balmer series of hydrogen. He found that the four lines agreed with the expected wavelengths of H, H, H, and H with a redshift of z = 0.16. This redshift in turn allowed him to identify a line in the ultraviolet part of the spectrum with Mg II 2798. Schmidt consulted with his colleagues, Jesse L. Greenstein and J. B. Oke. Oke had obtained photoelectric spectrophotometry of 3C 273 at the 100-inch telescope, which revealed an emission-line in the infrared at 7600. With the proposed redshift, this feature agreed with the expected wavelength of H. Greenstein's spectrum of 3C 48 with a redshift of z = 0.37, supported by the presence of Mg II in both objects. The riddle of the spectrum of quasars was solved.

These results were published in Nature six weeks later in adjoining papers by Hazard et al. (1963); Schmidt (1963); Oke (1963); and Greenstein and Matthews (1963). The objects might be galactic stars with a very high density, giving a large gravitational redshift. However, this explanation was difficult to reconcile with the widths of the emission lines and the presence of forbidden lines. The "most direct and least objectionable" explanation was that the objects were extragalactic, with redshifts reflecting the Hubble expansion. The redshifts were large but not unprecedented; that of 3C 48 was second only to that of 3C 295. The radio luminosities of the two quasars were comparable with those of Cyg A and 3C 295. However, the optical luminosities were staggering, "10 - 30 times brighter than the brightest giant ellipticals"; and the radio surface brightness was larger than for the radio galaxies. The redshift of 3C 273 implied a velocity of 47,400 km s^-1 and a distance of about 500 Mpc (for H₀ 100 km s^-1 Mpc^-1). The nuclear region would then be less than 1 kpc in diameter. The jet would be about 50 kpc away, implying a timescale greater than 10⁵ years and a total energy radiated of at least 10⁵⁹ ergs.

Before the redshift of 3C 273 was announced, Matthews and Sandage (1963) had submitted a paper identifying 3C 48, 3C 196 and 3C 286 with stellar optical objects. They explored the popular notion that these objects were some kind of Galactic star, arguing from their isotropic distribution on the sky and lack of observed proper motion that the most likely distance from the sun was about 100 pc. The objects had peculiar colors, and 3C 48 showed light variations of 0.4 mag. In a section added following the discovery of the redshifts of 3C 273 and 3C 48, they pointed out that the size limit of 0.15 pc implied by the optical light variations was important in the context of the huge distance and luminosity implied by taking the redshift to result from the Hubble expansion.

A detailed analysis of 3C 48 and 3C 273 was published by Greenstein and Schmidt (1964). They considered explanations of the redshift involving (1) rapid motion of objects in or near the Milky Way, (2) gravitational redshifts, and (3) cosmological redshifts. If 3C 273 had a transverse velocity comparable with the radial velocity implied by its redshift, the lack of an observed proper motion implied a distance of at least 10 Mpc (well beyond the nearest galaxies). The corresponding absolute magnitude was closer to the luminosity of galaxies than stars. The four quasars with known velocities were all receding; and accelerating a massive, luminous object to an appreciable fraction of the speed of light seemed difficult. Regarding gravitational redshifts, Greenstein and Schmidt argued that the widths of the emission lines required the line emitting gas to be confined to a small fractional radius around the massive object producing the redshift. The observed symmetry of the line profiles seemed unnatural in a gravitational redshift model. For a 1 M object, the observed H flux implied an electron density N_e 10¹⁹ cm^-3, incompatible with the observed presence of forbidden lines in the spectrum. The emission-line constraint, together with a requirement that the massive object not disturb stellar orbits in the Galaxy, required a mass 10⁹ M. The stability of such a "supermassive star" seemed doubtful in the light of theoretical work by Hoyle and Fowler (1963a), who had examined such objects as possible sources for the energy requirements of extragalactic radio sources. Adopting the cosmological explanation of the redshift, Greenstein and Schmidt derived radii for a uniform spherical emission-line region of 11 and 1.2 pc for 3C 48 and 3C 273, respectively. This was based on the H luminosities and electron densities estimated from the H, [O II], and [O III] line ratios. Invoking light travel time constraints based on the observed optical variability (Matthews and Sandage 1963; Smith and Hoffleit 1963), they proposed a model in which a central source of optical continuum was surrounded by the emission-line region, and a still larger radio emitting region. They suggested that a central mass of order 10⁹ M might provide adequate energy for the lifetime of 10⁶ yr implied by the jet of 3C 273 and the nebulosity of 3C 48. This mass was about right to confine the line emitting gas, which would disperse quickly if it expanded at the observed speeds of 1000 km s^-1 or more. Noting that such a mass would correspond to a Schwarzschild radius of ~ 10^-4 pc, they observed that "It would be important to know whether continued energy and mass input from such a `collapsed' region are possible". Finally, they noted that there could be galaxies around 3C 48 and 3C 273 hidden by the glare of the nucleus. Many features of this analysis are recognizable in current thinking about AGN.

The third and fourth quasar redshifts were published by Schmidt and Matthews (1964), who found z = 0.425 and 0.545 for 3C 47 and 3C 147, respectively. Schmidt (1965) published redshifts for 5 more quasars. For 3C 254, a redshift z = 0.734, based on several familiar lines, allowed the identification of C III] 1909 for the first time. This in turn allowed the determination of redshifts of 1.029 and 1.037 from 1909 and 2798 in 3C 245 and CTA 102, respectively. (CTA is a radio source list from the Caltech radio observatory.) For 3C 287, a redshift of 1.055 was found from 1909, 2798, and another first, C IV 1550. Finally, a dramatically higher redshift of 2.012 was determined for 3C 9 on the basis of 1550 and the first detection of the Lyman line of hydrogen at 1215. The redshifts were large enough that the absolute luminosities depended significantly on the cosmological model used.

Sandage (1965) reported the discovery of a large population of radio quiet objects that otherwise appeared to resemble quasars. Matthews and Sandage (1963) had found that quasars showed an "ultraviolet excess" when compared with normal stars on a color-color (U-B, B-V) diagram. This led to a search technique in which exposures in U and B were recorded on the same photographic plate, with a slight positional offset, allowing rapid identification of objects with strong ultraviolet continua. Sandage noticed a number of such objects that did not coincide with known radio sources. These he called "interlopers", "blue stellar objects" (BSO), or "quasi-stellar galaxies" (QSG). ¹ Sandage found that at magnitudes fainter than 15, the UV excess objects populated the region occupied by quasars on the color-color diagram, whereas brighter objects typically had the colors of main sequence stars. The number counts of the BSOs as a function of apparent magnitude also showed a change of slope at ~ 15^m, consistent with an extragalactic population of objects at large redshift. Spectra showed that many of these objects indeed had spectra with large redshifts, including z = 1.241 for BSO 1. Sandage estimated that the QSGs outnumbered the radio loud quasars by a factor ~ 500, but this was reduced by later work (e.g., Kinman 1965; Lynds and Villere 1965).

The large redshifts of QSOs immediately made them potential tools for the study of cosmological questions. The rough similarity of the emission-line strengths of QSOs to those observed, or theoretically predicted, for planetary nebulae suggested that the chemical abundances were roughly similar to those in our Galaxy (Sklovskii 1964; Osterbrock and Parker 1966). Thus these objects, suspected by many astronomers to lie in the nuclei of distant galaxies, had reached fairly "normal" chemical compositions when the Universe was considerably younger than today.

The cosmological importance of redshifts high enough to make L visible was quickly recognized. Hydrogen gas in intergalactic space would remove light from the quasar's spectrum at the local cosmological redshift, and continuously distributed gas would erase a wide band of continuum to the short wavelength side of the L emission line (Gunn and Peterson 1965; Scheuer 1965). Gunn and Peterson set a tight upper limit to the amount of neutral hydrogen in intergalactic space, far less than the amount that would significantly retard the expansion of the Universe.

The study of discrete absorption features in quasar spectra also began to develop. An unidentified sharp line was observed in the spectrum of 3C 48 by Greenstein and Schmidt (1964). Sandage (1965) found that the 1550 emission line of BSO 1 was "bisected by a sharp absorption feature". The first quasar found with a rich absorption spectrum was 3C 191 (Burbidge, Lynds, and Burbidge 1966; Stockton and Lynds 1966). More than a dozen sharp lines were identified, including L and lines of C II, III, and IV and Si II, III, and IV. A rich set of narrow absorption lines was also observed in the spectrum of PKS 0237-23, whose emission-line redshift, z = 2.223, set a record at the time. Arp, Bolton, and Kinman (1967) and Burbidge (1967a) respectively proposed absorption line redshifts of z = 2.20 and 1.95 for this object, but each value left many lines without satisfactory identifications. It turned out that both redshifts were present (Greenstein and Schmidt 1967).

All these absorption systems had z_abs < z_em. They could be interpreted as intervening clouds imposing absorption spectra at the appropriate cosmological redshift, as had been anticipated theoretically (Bahcall and Salpeter 1965). Alternatively, they might represent material expelled from the quasar, whose outflow velocity is subtracted from the cosmological velocity of the QSO. However, PKS 0119-04 was found to have z_abs > z_em, implying material that was in some sense falling into the QSO from the near side with a relative velocity of 10³ km s^-1 (Kinman and Burbidge 1967). Today, a large fraction of the narrow absorption lines with z_abs substantially less than z_em are believed to result from intervening material. This includes the so-called "Lyman alpha forest" of closely spaced, narrow L lines that punctuate the continuum to the short wavelength side of the L emission line, especially in high redshift QSOs. The study of intervening galaxies and gas clouds by means of absorption lines in the spectra of background QSOs is now a major branch of astrophysics.

A different kind of absorption was discovered in the spectrum of PHL 5200 by Lynds (1967). This object showed broad absorption bands on the short wavelength sides of the L, N V 1240, and C IV 1550 emission lines, with a sharp boundary between the emission and absorption. Lynds interpreted this in terms of an expanding shell of gas around the central object. Seen in about 10 percent of radio quiet QSOs (Weymann et al. 1991), these broad absorption lines (BALs) are among the many dramatic but poorly understood aspects of AGN.

The huge luminosity of QSOs, rapid variability, and implied small size caused some astronomers to question the cosmological nature of the redshifts. Terrell (1964) considered the possibility that the objects were ejected from the center of our galaxy. Upper limits on the proper motion of 3C 273, together with a Doppler interpretation of the redshift, then implied a distance of at least 0.3 Mpc and an age at least 5 million years. Arp (1966), pointing to close pairs of peculiar galaxies and QSOs on the sky, argued for noncosmological redshifts that might result from ejection from the peculiar galaxies at high speeds or an unknown cause. Setti and Woltjer (1966) noted that ejection from the Galactic center would imply for the QSO population an explosion with energy at least 10⁶⁰ ergs, and more if ejected from nearby radio galaxies such as Cen A as suggested by Hoyle and Burbidge (1966). Furthermore, Doppler boosting would cause us to see more blueshifts than redshifts if the objects were ejected from nearby galaxies (Faulkner, Gunn, and Peterson 1966). Further evidence for cosmological redshifts was provided by Gunn (1971), who showed that two clusters of galaxies containing QSOs had the same redshifts as the QSOs. Also, Kristian (1973) showed that the "fuzz" surrounding the quasistellar image of a sample of QSOs was consistent with the presence of a host galaxy.

¹ Here we adopt the now common practice of using the term "quasi-stellar object" (QSO) to refer to these objects regardless of radio luminosity (Burbidge and Burbidge 1967). Back.