An over-massive black hole in a typical star-forming galaxy, 2 billion years after the Big Bang

See allHide authors and affiliations

Science  10 Jul 2015:
Vol. 349, Issue 6244, pp. 168-171
DOI: 10.1126/science.aaa4506

Black hole out of kilter with theory

It is believed that black holes and their host galaxies coevolve, with the feedback from the black hole inducing star formation. Such a scenario requires certain timing and mass constraints for the black hole and the star-forming gas. Trakhtenbrot et al. looked at high–red shift galaxies, when the universe was only about 2 billion years old. They found a black hole that developed to maturity much earlier than would be expected and was about 10% of the total galactic mass—much more than expected. Moreover, star formation continued after it would have been expected to stop.

Science, this issue p. 168


Supermassive black holes (SMBHs) and their host galaxies are generally thought to coevolve, so that the SMBH achieves up to about 0.2 to 0.5% of the host galaxy mass in the present day. The radiation emitted from the growing SMBH is expected to affect star formation throughout the host galaxy. The relevance of this scenario at early cosmic epochs is not yet established. We present spectroscopic observations of a galaxy at redshift z = 3.328, which hosts an actively accreting, extremely massive BH, in its final stages of growth. The SMBH mass is roughly one-tenth the mass of the entire host galaxy, suggesting that it has grown much more efficiently than the host, contrary to models of synchronized coevolution. The host galaxy is forming stars at an intense rate, despite the presence of a SMBH-driven gas outflow.

Several lines of observational evidence, spanning a wide range of cosmic epochs, have led to a commonly accepted picture wherein supermassive black holes (SMBHs, MBH > 106 M ; M is the solar mass) coevolve with their host galaxies (14). Moreover, energy- and/or momentum-driven “feedback” from accreting SMBHs (Active Galactic Nuclei; AGN) is thought to quench star formation in the host galaxy (5). To directly test the relevance of such scenarios at early cosmic epochs (high redshifts, z) requires the most basic properties of SMBHs and their hosts, including masses and growth rates, to be observed. Several observational studies found that at z ≤ 2 (more than 3.3 billion years after the Big Bang), the typical BH-to-stellar mass ratio, MBH/M, increases toward higher redshifts (68), suggesting that some SMBHs were able to gather mass more efficiently, or faster, than the stellar populations in their hosts. To date, measurements of MBH at earlier epochs (z > 2) have only been conducted for small samples of extremely luminous objects [LAGN > 1046 erg s−1; (912)] representing a rare subset of all accreting SMBHs, with number densities on the order of 1 to 10 per Gpc3 [i.e., ∼ 10−9 to 10−8 Mpc−3; (13)]. Moreover, the high AGN luminosities in such sources overwhelm the host galaxy emission and prohibit a reliable determination of M, and therefore of MBH/M. We initiated an observational campaign aimed at estimating MBH in x-ray–selected, unobscured z ∼ 3 to 4 AGN within the Cosmic Evolution Survey field [COSMOS; (14)]. Such sources have lower AGN luminosities and are more abundant than the aforementioned luminous sources by factors of 100 to 1000 [e.g., (13, 15)] and thus form a more representative subset of the general AGN population. Moreover, the fainter AGN luminosities and rich multiwavelength coverage of AGN within the COSMOS field enable reliable measurements of the mass and growth rate of the stellar populations in the host galaxies (M and star-formation rate, SFR).

CID–947 is an x-ray–selected, unobscured AGN at z = 3.328, detected in both XMM-Newton and Chandra x-ray imaging data of the COSMOS field [see fig. S4 and sections S2 and S4 in the supplementary materials (16)]. We obtained a near-infrared (IR) K-band spectrum of CID–947 using the MOSFIRE instrument at the W. M. Keck telescope, which at z = 3.328 covers the hydrogen Hβ broad emission line (see details in section S1 in the supplementary materials). The calibrated spectrum shows a very broad Hβ emission line, among other features (Fig. 1). Our spectral analysis indicates that the monochromatic AGN luminosity at rest-frame 5100 Å is Embedded Image. The typical line-of-sight velocity [i.e., the full-width at half-maximum of the line] is Embedded Image (see section S1.2 in the supplementary materials). By combining this line width with the observed L5100 and relying on an empirically calibrated estimator for MBH, based on the virial motion of ionized gas near the SMBH (17), we obtain Embedded Image. All the reported measurement-related uncertainties are derived by a series of simulations and represent the 16th and 84th quantiles of the resulting distributions. These simulations indicate a SMBH mass larger than 3.6 × 109 M at the 99% confidence level (see sections S1.2 and S3 in the supplementary materials for more details). Determinations of MBH from single-epoch spectra of the Hβ emission line are known to also be affected by significant systematic uncertainties, of up to ∼0.3 to 0.4 dex. For a detailed discussion of some of the systematics and related issues, see section S3 in the supplementary materials. This high MBH is comparable with some of the most massive BHs known to date in the local universe (18) or with the masses of the biggest BHs in the much rarer, more luminous AGN at z ∼ 2 to 4 [e.g., (9)]. The bolometric luminosity of CID–947 is in the range Lbol ≃ (1.1 to 2.2) × 1046 erg s−1, estimated either from the observed optical luminosity or the multiwavelength spectral energy distribution. Combined with the measured MBH, we derive a normalized accretion rate of L/LEdd ≃ 0.01 to 0.02. This value is lower, by at least an order of magnitude, than the accretion rates of known SMBHs at z ∼ 3.5 [e.g., (9, 10)]. Further assuming a standard radiative efficiency of 10%, we obtain an e-folding time scale for the SMBH mass of at least 2.1 × 109 years (Gy) (see section S3 in the supplementary materials), which is longer than the age of the universe at z = 3.328. By contrast, even the most extreme models for the emergence of “seed” BHs predict masses no larger than Mseed ∼ 106 M at z ∼ 10 to 20 [e.g., (19)]. Therefore, the SMBH powering CID–947 had to grow at much higher accretion rates and at a high duty cycle in the past, to account for the high observed MBH only 1.7 Gy after z ≃ 20. CID–947 could have evolved from a parent population similar to the fast-growing SMBHs observed in Embedded Image quasars, which have L/LEdd ∼ 0.5 to 1 and MBH ≃ 109 M [e.g., (11, 12)]. The requirement for a high accretion rate in the very recent past is supported by the clear presence of a high-velocity outflow of ionized gas, observed in the rest-frame ultraviolet spectrum of the source (fig. S4). The broad absorption features of C IV λ1549 and Si IV λ1400 have maximal velocities of vmax ≃ 12,000 km s−1. Assuming that this outflow is driven by radiation pressure, these velocities require accretion rates of Embedded Image, as recently as 105 to 106 years before the observed epoch (see section S4 in the supplementary materials). We conclude that the SMBH powering CID–947 is in the final stages of growth and that we are witnessing the shut-down of accretion onto one of the most massive BHs known to date.

Fig. 1 The observed Keck/MOSFIRE spectrum and best-fit model for the Hβ emission complex of CID–947.

The data are modeled with a linear continuum (dotted), a broadened iron template (dot-dashed), and a combination of broad and narrow Gaussians (dashed), which correspond to the Hβ and [O III] emission lines (see section S1.2 in the supplementary materials for details regarding the spectral modeling). The broad component of Hβ has a full width at half maximum of FWHM(Hβ) = 11330 km s−1, which results in MBH = 6.9 × 109 M and MBH/M = 1/8. The red dashed line illustrates an alternative scenario, in which the SMBH mass derived from the Hβ line width would result in MBH/M = 1/100 [i.e., FWHM(Hβ) = 3218 km s−1], clearly at odds with the data. The spike at λrest ≃ 4640 Å is due to a sky feature. The bottom panel shows the residuals of the best-fit model.

The rich collection of ancillary COSMOS multiwavelength data available for CID–947 enables us to study the basic properties of its host galaxy (see details in section S2 in the supplementary materials). A previously published analysis of the observed spectral energy distribution of the emission from the source reveals an appreciable stellar emission component, originating from Embedded Image in stars (20). Our own analysis provides a yet lower stellar mass, of Embedded Image. However, we focus on the previously determined, higher stellar mass, as a conservative estimate. The source is also detected at far-IR and (sub)millimeter wavelengths, which allows us to constrain the SFR in the host galaxy to about 400 M year−1. The stellar mass of the host galaxy is consistent with the typical value for star-forming galaxies at z ∼ 3 to 4 [i.e., the “break” in the mass function of galaxies; (21)]. Similarly, the combination of M and SFR is consistent with the typical values observed at z ∼ 3 to 4, which appear to follow the so-called main sequence of star-forming galaxies (22). Thus, the host galaxy of CID–947 is a typical star-forming galaxy for its redshift, representing a population with a number density of about 5 × 10−5 Mpc−3 [e.g., (21)]. This suggests that neither the intense, ionizing radiation that emerged during the fast SMBH growth, nor the AGN-driven outflow, have quenched star formation in the host galaxy. The relatively high stellar mass and SFR of the host galaxy further suggest that it is unlikely that the AGN affected the host in yet earlier epochs. That is, even in this case of extreme SMBH growth, there is no sign of AGN-driven suppression of star formation in the host.

Our analysis indicates that the BH-to-stellar mass ratio for CID–947 is MBH/M ≃ 1/8. In comparison, most local (dormant) high-mass BHs typically have MBH/M ∼ 1/700 to 1/500 [see Fig. 2 and, e.g., (4, 23)]. The MBH/M value that we find for CID–947 is thus far higher than typically observed in high-mass systems in the local universe, by at least an order of magnitude and more probably by a factor of about 50. The only local system with a comparably extreme mass ratio is the galaxy NGC 1277, which was reported to have MBH/M ≃ 1/7 [with MBH = 1.7 × 1010 M ≃ 2.5 × MBH(CID–947); see (24), but also (25)]. At earlier epochs (still z < 2), the general trend is for MBH/M to increase slightly with redshift, but typically not beyond MBH/M ∼ 1/100 (see Fig. 3). Only a few systems with reliable estimates of MBH show MBH/M reaching as high as 1/30 [e.g., (68)].

Fig. 2 A comparison of CID–947 with a compilation of observed MBH and M estimates in the local universe

[adapted from (4), assuming the tabulated bulge-to-total fractions]. CID–947 (red star) has a very high BH-to-stellar mass ratio of MBH/M ≃ 1/10. The asymmetric error bars shown on MBH and M represent measurement-related uncertainties, while the symmetric ones demonstrate systematic uncertainties of 0.3 dex (on MBH) and 0.1 dex (on M). The masses inferred for subsequent growth scenarios are highlighted as empty red stars. The CID–947 system is expected to evolve only mildly in MBH (perhaps to ∼1010 M), but M should grow to at least 2 × 1011 M, and possibly to as much as ∼7 × 1011 M, by z = 0. The local galaxies NGC 1277 and M87, which could be considered as descendants of systems like CID–947, are highlighted as filled symbols [(25) and (27), respectively]. Some studies suggest these galaxies to have somewhat higher MBH, and therefore relatively high mass ratios, of MBH/M = 1/7 and 1/127, respectively (24, 28).

Fig. 3 The observed cosmic evolution of the BH-to-stellar mass ratio, MBH/M, and its extrapolation beyond z ∼ 2.

CID–947 (red star) has MBH/M = 1/8 at z ≃ 3.3, which is higher by a factor of at least ∼50 than the typical value in local, inactive galaxies (at most, MBH/M ∼ 1/500; dotted line). The error bars shown for CID–947 represent only the measurement-related uncertainties, propagating the uncertainties on MBH and on M. The different data points at z < 2 represent typical (median) values for several samples with MBH/M estimates, with uncertainties representing the scatter within each sample [filled symbols, open circles, and open triangles represent samples from (7), (29), and (6), respectively; adapted from (7)]. Even compared to the extrapolation of the evolutionary trend supported by these lower-redshift data, MBH/M ∼ (z + 1)2 [dashed line, scaled as in (30)], CID–947 has a significantly higher MBH/M.

Given the high masses of both the SMBH and stellar population in CID–947, we expect this system to retain an extreme MBH/M throughout its evolution, from z = 3.328 to the present-day universe. Because the MBH that we find is already comparable to the most massive BHs known, it is unlikely that the SMBH will experience any further appreciable growth (i.e., beyond MBH ≃ 1010 M). Indeed, if the SMBH accretes at the observed rate through z = 2, it will reach the extreme value of ∼1010 M, and by z = 1 it will have a final mass of ∼2.5 × 1010 M. As for the host galaxy, we can constrain its subsequent growth following several different assumptions. First, if one simply assumes that the galaxy will become as massive as the most massive galaxies in the local universe [M ≃ 1012 M; (26)], then the implied final mass ratio is on the order of MBH/M ∼ 1/100. Alternatively, we consider more realistic scenarios for the future growth of the stellar population, relying on the observed mass (M) and growth rate (SFR). Our calculations involve different scenarios for the decay of star formation in the galaxy (see section S5 in the supplementary materials) and predict final stellar masses in the range M(z = 0) ≃ (2 to 7) × 1011 M, which is about an order of magnitude higher than the observed mass at z = 3.328. The inferred final mass ratio is MBH/M ∼ 1/50. This growth can only occur if star formation continues for a relatively long period (≥1 Gy) and at a high rate (>50 M year−1). This would require the presence of a substantial reservoir, or the accretion, of cold gas, which, however, could not increase the SMBH mass by much. Finally, in the most extreme scenario, the star formation shuts down almost immediately (i.e., due to the AGN-driven outflow), and the system remains “frozen” at MBH/M ∼ 1/10 throughout cosmic time. If the SMBH does indeed grow further (i.e., beyond 1010 M), this would imply yet higher MBH/M. Thus, the inferred final BH-to-stellar mass ratio for CID–947 is, in the most extreme scenarios, about MBH/M ∼ 1/100, and probably much higher (see Fig. 2).

CID–947 therefore represents a progenitor of the most extreme, high-mass systems in the local universe, like NGC 1277. Such systems are not detected in large numbers, perhaps due to observational selection biases. The above considerations indicate that the local relics of systems like CID–947 are galaxies with at least M ∼ 5 × 1011 M. Such systems are predominantly quiescent (i.e., with low star-formation rates, SFR ≪ 1 M year−1) and relatively rare in the local universe, with typical number densities on the order of ∼10−5 Mpc−3 (26). We conclude that CID–947 provides direct evidence that at least some of the most massive BHs, with Embedded Image, already in place just 2 Gy after the Big Bang, did not shut down star formation in their host galaxies. The host galaxies may experience appreciable mass growth in later epochs, without much further black hole growth, resulting in very high stellar masses but still relatively high MBH/M. Lower-mass systems may follow markedly different coevolutionary paths. However, systems with MBH/M as high as in CID–947 may be not as rare as previously thought, as they can be consistently observed among populations with number densities on the order of ∼10−5 Mpc−3, both at z > 3 and in the local universe, and not just among the rarest, most luminous quasars.

Supplementary Materials

Data, Methods, and Supplementary Text S1 to S4

Figs. S1 to S4

Table S1

References (3181)

References and Notes

  1. Data and methods, supplementary text, figures and tables are available on Science Online.
  2. Acknowledgments: The new MOSFIRE data presented herein were obtained at the W. M. Keck Observatory, which is operated as a scientific partnership among the California Institute of Technology, the University of California, and the National Aeronautics and Space Administration. The Observatory was made possible by the generous financial support of the W. M. Keck Foundation. We thank M. Kassis and the rest of the staff at the W. M. Keck observatories at Waimea, HI, for their support during the observing run. We recognize and acknowledge the very significant cultural role and reverence that the summit of Mauna Kea has always had within the indigenous Hawaiian community. We are most fortunate to have the opportunity to conduct observations from this mountain. Some of the analysis presented here is based on data products from observations made with European Southern Observatory (ESO) Telescopes at the La Silla Paranal Observatory under ESO program ID 179.A-2005 and on data products produced by TERAPIX and the Cambridge Astronomy Survey Unit on behalf of the UltraVISTA consortium. We are grateful to A. Faisst and M. Onodera for their assistance with the acquisition and reduction of the MOSFIRE data. We thank S. Tacchella, J. Woo, and W. Hartley for their assistance with some of the evolutionary calculations. K.S. gratefully acknowledges support from Swiss National Science Foundation Professorship grant PP00P2 138979/1. F.C. acknowledges financial support by the NASA grant GO3-14150C. M.E. acknowledges financial support by the NASA Chandra grant GO2-13127X. B.T. is a Zwicky Fellow at the ETH Zurich.
View Abstract

Stay Connected to Science

Navigate This Article