Access provided to University of North Texas by Libraries

Nature | Letter

日本語要約

The diversity of quasars unified by accretion and orientation

Journal name:
Nature
Volume:
513,
Pages:
210–213
Date published:
DOI:
doi:10.1038/nature13712
Received
Accepted
Published online

Quasars are rapidly accreting supermassive black holes at the centres of massive galaxies. They display a broad range of properties across all wavelengths, reflecting the diversity in the physical conditions of the regions close to the central engine. These properties, however, are not random, but form well-defined trends. The dominant trend is known as ‘Eigenvector 1’, in which many properties correlate with the strength of optical iron and [O iii] emission1, 2, 3. The main physical driver of Eigenvector 1 has long been suspected4 to be the quasar luminosity normalized by the mass of the hole (the ‘Eddington ratio’), which is an important parameter of the black hole accretion process. But a definitive proof has been missing. Here we report an analysis of archival data that reveals that the Eddington ratio indeed drives Eigenvector 1. We also find that orientation plays a significant role in determining the observed kinematics of the gas in the broad-line region, implying a flattened, disk-like geometry for the fast-moving clouds close to the black hole. Our results show that most of the diversity of quasar phenomenology can be unified using two simple quantities: Eddington ratio and orientation.

Subject terms:

At a glance

Figures

View all figures
  1. Distribution of quasars in the EV1 plane.
    Figure 1: Distribution of quasars in the EV1 plane.

    The horizontal axis is the relative Fe ii strength, RFe ii, and the vertical axis is the broad Hβ FWHM. The red contours show the distribution of our SDSS quasar sample (with quasar density increasing from outer to inner contours), and the points show individual objects. We colour-code the points by the [O iii] λ = 5,007 Å equivalent width, averaged over all nearby objects in a smoothing box of ΔRFe ii = 0.2 and ΔFWHMHβ = 1,000 km s−1. The EV1 sequence1 is the systematic trend of decreasing [O iii] strength with increasing RFe ii. The grey grid divides this plane into bins of FWHMHβ and RFe ii, in which we study the stacked spectral properties.

    Full size image

    View in article

  2. Average [lsqb]O iii[rsqb] profiles in the EV1 plane.
    Figure 2: Average [O iii] profiles in the EV1 plane.

    Each panel shows the stacked [O iii] λ = 5,007 Å line of quasars in the RFe ii−FWHMHβ bins defined by the grey grid in Fig. 1 (in the same layout). RFe ii increases from left to right, and FWHMHβ increases from bottom to top. In each bin we further divided the quasars into different luminosity bins using the measured L5,100 Å continuum luminosities. We have normalized the line fluxes by the (host-corrected) average quasar continuum luminosity L5,100 Å for each stacking subset; hence, these stacked lines reflect the relative [O iii] strength among different samples. In addition to the decrease of [O iii] strength when RFe ii increases (that is, Fig. 1), we also observe a decrease in [O iii] strength with increasing luminosity8, 9. The [O iii] profile is in general asymmetric, with a blueshifted wing, whose relative contribution to the total profile increases when RFe ii or luminosity increases.

    Full size image

    View in article

  3. Cross-correlation functions between different quasar subsamples and a galaxy sample.
    Figure 3: Cross-correlation functions between different quasar subsamples and a galaxy sample.

    rp is the transverse comoving separation and wp is the projected two-point correlation function. a, Difference in the clustering strength when the quasar sample is divided at the median RFe ii. A significant difference (3.48σ) is detected: quasars with stronger RFe ii are less strongly clustered, indicating they have on average smaller black hole masses. b, Difference in the clustering strength when the quasar sample is divided by the virial black hole mass estimates based on FWHMHβ. No significant difference (1.64σ) is detected, indicating there is substantial overlap in the actual black hole masses between the two subsamples owing to the uncertainties in these FWHM-based virial black hole masses. Orientation-induced FWHMHβ dispersion can naturally lead to such uncertainties. Error bars are 1σ measurement errors estimated with jackknife resampling (Supplementary Information).

    Full size image

    View in article

  4. The effect of orientation on FWHMH[bgr].
    Figure 4: The effect of orientation on FWHMHβ.

    The large symbols represent the 29 low-redshift AGNs that have both reverberation mapping data and host stellar velocity dispersion (σ*) measurements. The small symbols represent a low-redshift SDSS AGN sample26 with σ* and AGN spectral measurements based on spectral decomposition. We use the stellar velocity dispersion measurements and the local relation between black hole mass and σ* from inactive galaxies25 to estimate the black hole mass ( ) in these objects. We also estimate the average broad-line region size (RBLR = cτ, where c is the speed of light, and τ is the measured reverberation mapping lag) in these objects, either from direct reverberation mapping measurements, or by using the tight correlation between the broad-line region size and AGN luminosity27. The ratio of to (that is, the virial coefficient f) is plotted as a function of FWHMHβ, for different values. The strong trends of f with FWHMHβ at a given suggest that the dispersion in FWHMHβ does not reflect the underlying virial velocity of the broad-line region gas, and tends to bias the black hole mass estimates. This is in line with the fact that there is little vertical trend in the [O iii] strength in the EV1 plane (Fig. 1).

    Full size image

    View in article

  5. Decomposed [lsqb]O iii[rsqb] [lgr] = 5,007 A luminosity.
    Extended Data Fig. 1: Decomposed [O iii] λ = 5,007 Å luminosity.

    The core component (a) and the wing component (b) are shown for each composite spectrum shown in Fig. 2. Error bars are 1σ measurement errors estimated using Monte Carlo trials of mock spectra generated using the estimated flux error arrays of the co-added spectra. Both luminosities are normalized to the quasar continuum luminosity L5,100 Å, hence reflecting the strength of [O iii]. The core [O iii] shows a prominent anti-correlation with both L5,100 Å and RFe ii, while the wing [O iii] shows weaker anti-correlations with L5,100 Å and RFe ii. For both [O iii] components there is no correlation with FWHMHβ, as shown in Figs 1 and 2. The Baldwin effect and EV1 correlation for [O iii] shown in Fig. 1 and Fig. 2 are then primarily associated with the core [O iii] component. The difference between the core and wing [O iii] components may suggest different excitation mechanisms for both components.

    Full size image

    View in article

  6. Kinematic properties of the decomposed core and wing [lsqb]O iii[rsqb] components.
    Extended Data Fig. 2: Kinematic properties of the decomposed core and wing [O iii] components.

    a, FWHM against luminosity for core [O iii]. b, FWHM against luminosity for wing [O iii]. c, Velocity offset against luminosity for core [O iii]. d, Velocity offset against luminosity for wing [O iii]. Error bars are 1σ measurement errors estimated using Monte Carlo trials of mock spectra generated using the estimated flux error arrays of the co-added spectra. The most significant correlations are the correlation between luminosity and the core [O iii] FWHM, and the correlations between the wing [O iii] blueshift and L/RFe ii. The former correlation is consistent with the scenario that more luminous quasars are on average hosted by more massive galaxies with deeper potential wells, hence having larger core [O iii] widths. The latter correlations are consistent with the scenario that the wing [O iii] component is associated with outflows.

    Full size image

    View in article

  7. Composite SDSS quasar spectra for several other lines in the same RFe ii-FWHMH[bgr] bins as defined in Fig. 1.
    Extended Data Fig. 3: Composite SDSS quasar spectra for several other lines in the same RFe ii–FWHMHβ bins as defined in Fig. 1.

    a, Hβ and [O iii]. b, Mg ii. c, [O ii] 3,727 Å. d, [Ne v] 3,426 Å. As in Fig. 2, each composite spectrum has been normalized by the continuum such that the integrated line intensity reflects the strength of the line. The composite spectra for the Hβ region are generated using the pseudo-continuum-subtracted spectra, while for each of the other three lines (Mg ii, [O ii] and [Ne v]) the composite spectrum is the median spectrum created using the full SDSS spectra and normalized at a nearby continuum window.

    Full size image

    View in article

  8. Distribution in the EV1 plane in terms of C iv properties.
    Extended Data Fig. 4: Distribution in the EV1 plane in terms of C iv properties.

    A sample of low-redshift quasars with both Hβ and C iv measurements is shown, colour-coded by the C iv strength. A clear trend of decreasing C iv strength with RFe ii is seen, consistent with that seen for the other forbidden lines. The typical 1σ measurement uncertainty in C iv equivalent width is about 7% (relative to the measurement), and hence is negligible compared to the strong EV1 trend observed.

    Full size image

    View in article

  9. Distributions of SDSS quasars in the EV1 plane in terms of the optical-infrared (r - W1) colour.
    Extended Data Fig. 5: Distributions of SDSS quasars in the EV1 plane in terms of the optical–infrared (r − W1) colour.

    r is the SDSS r band (6,166 Å) and W1 is the WISE W1 band (3.4 μm). a, r − W1 for quasars with 0.4 < z < 0.8, for which the band-shifting effect is small. We see a trend of increasing mid-infrared emission relative to optical emission with increasing RFe ii. b, A similar result, using the excess colour, Δ(r − W1), which is the deviation of r − W1 colour from the mean colour at each redshift. Using Δ(r − W1) removes the redshift dependence of colours, and we can apply this to all quasars in our sample. This test suggests that the torus emission is enhanced in quasars with larger RFe ii. Given that we have argued that RFe ii is a good indicator for the Eddington ratio, this result suggests that quasars with higher Eddington ratios have stronger torus emission, which may have implications for the formation mechanism of the dusty torus.

    Full size image

    View in article

  10. A detailed look at the median excess optical-WISE colour [Dgr](r - W1) in the EV1 plane.
    Extended Data Fig. 6: A detailed look at the median excess optical-WISE colour Δ(r − W1) in the EV1 plane.

    The same bins as defined in Fig. 1 are used. Error bars are the 1σ uncertainty in the median, estimated by the standard deviation divided by the square root of the number of objects in the bin. At fixed RFe ii, we see increasing relative torus emission when FWHMHβ increases. This is consistent with the orientation scenario: larger FWHMs indicate more edge-on systems, which suffer more from geometric reduction (the cosI factor) and/or dust extinction in the optical than in the infrared parts of the spectrum.

    Full size image

    View in article

  11. Distribution in the EV1 plane in terms of X-ray properties.
    Extended Data Fig. 7: Distribution in the EV1 plane in terms of X-ray properties.

    The subset of our SDSS quasars with available measurements of their soft X-ray photon index ΓX are shown. ΓX increases (becomes softer) with increasing RFe ii, consistent with earlier findings3, 5. CSC refers to objects from the Chandra Source Catalog and XMM refers to objects from the XMM-Newton Serendipitous Catalog. The contours are the distribution of all SDSS quasars in our sample, as in Fig. 1.

    Full size image

    View in article

  12. The same EV1 plane as in Fig. 1 in logarithmic FWHMH[bgr].
    Extended Data Fig. 8: The same EV1 plane as in Fig. 1 in logarithmic FWHMHβ.

    The dashed lines show the running median value as a function of RFe ii and the dotted lines show the 16% and 84% percentiles, for objects in different luminosity bins. The distribution of FWHMHβ at fixed RFe ii roughly follows a log-normal distribution, with a dispersion of about 0.15−0.25 dex, which we argued comes mostly from orientation-induced variations. Lower-luminosity objects tend to have slightly larger dispersion in FWHMHβ, possibly caused by a broader Eddington ratio distribution at lower luminosities, which introduces additional dispersion in FWHMHβ. LEdd = 1.3 × 1038(MBH/1  ) erg s−1 is the Eddington luminosity of the black hole.

    Full size image

    View in article

  13. Distributions of radio-loud and radio-quiet quasars in EV1 plane.
    Extended Data Fig. 9: Distributions of radio-loud and radio-quiet quasars in EV1 plane.

    The radio-loud population shifts to lower RFe ii and larger FWHMHβ, compared with the radio-quiet population. We further divide the radio-loud quasars into core-dominant and lobe-dominant subsets, but we caution that our morphological classification is very crude, and there is potentially a large mixture of true morphological types between the two subsamples. The core-dominant (more pole-on) radio quasars have systematically smaller FWHMHβ compared with the lobe-dominant radio quasars, consistent with the hypothesis that orientation leads to variations in FWHMHβ. The points with error bars are the median and 1σ uncertainty for the median in each RFe ii bin.

    Full size image

    View in article

  14. Distribution in the EV1 plane colour-coded by the FWHM/[sgr] ratio.
    Extended Data Fig. 10: Distribution in the EV1 plane colour-coded by the FWHM/σ ratio.

    The distribution has been smoothed over a box of ΔRFe ii = 0.2 and ΔlogFWHMHβ = 0.2. We show only points for which there are more than 50 objects in the smoothing box to average. The black open circles show the median FWHMHβ at fixed RFe ii (using all objects in that bin), with the error bars indicating the 1σ uncertainty for the median. The transition in FWHM/σ reflects the change in orientation of the broad-line region disk relative to the line of sight.

    Full size image

    View in article

Main

The optical and ultraviolet spectra of quasars show emission lines with a wide variety of strengths (equivalent widths) and velocity widths. However, despite their great diversity in outward appearance, quasars possess surprising regularity in their physical properties. A seminal principal-component analysis1 of 87 low-redshift broad-line quasars discovered that the main variance (Eigenvector 1, or EV1) in their optical properties arises from an anti-correlation between the strength of the narrow [O iii] λ = 5,007 Å and broad Fe ii emission. Along with other properties that also correlate with Fe ii strength2, 3, 5, these observations establish EV1 as a physical sequence of broad-line quasar properties. In the two-dimensional plane of Fe ii strength (measured by the ratio of Fe ii equivalent width within 4,434–4,684 Å to broad Hβ equivalent width, RFe ii ≡ EWFe ii/EWHβ) and the full-width at half-maximum of broad Hβ (FWHMHβ), EV1 is defined as the horizontal trend with RFe ii, where the average [O iii] strength and FWHMHβ decrease1, 2. Figure 1 shows the EV1 sequence for about 20,000 broad-line quasars drawn from the Sloan Digital Sky Survey (SDSS)6, 7 (see Supplementary Information for details of the sample).

Figure 1: Distribution of quasars in the EV1 plane.
Distribution of quasars in the EV1 plane.

The horizontal axis is the relative Fe ii strength, RFe ii, and the vertical axis is the broad Hβ FWHM. The red contours show the distribution of our SDSS quasar sample (with quasar density increasing from outer to inner contours), and the points show individual objects. We colour-code the points by the [O iii] λ = 5,007 Å equivalent width, averaged over all nearby objects in a smoothing box of ΔRFe ii = 0.2 and ΔFWHMHβ = 1,000 km s−1. The EV1 sequence1 is the systematic trend of decreasing [O iii] strength with increasing RFe ii. The grey grid divides this plane into bins of FWHMHβ and RFe ii, in which we study the stacked spectral properties.

The statistics of the SDSS quasar sample allows us to divide the sample into bins of RFe ii and FWHMHβ (the grey grid in Fig. 1) and study the average [O iii] properties in each bin. Figure 2 shows the average [O iii] line profiles in each bin, as a function of the quasar continuum luminosity L measured at 5,100 Å. In addition to the EV1 sequence, the [O iii] strength also decreases with L5,100 Å, following the Baldwin effect8, 9 initially discovered for the broad C iv line10. The [O iii] profile can be decomposed into a core component, centred consistently at the systemic redshift, and a blueshifted wing component. The core component strongly follows the EV1 and Baldwin trends, while the wing component only shows a mild decrement with L and RFe ii (Supplementary Information and Extended Data Figs 1–2). This may suggest that the core component is mostly powered by photoionization from the quasar, while the wing component is excited by other mechanisms, such as shocks associated with outflows11.

Figure 2: Average [O iii] profiles in the EV1 plane.
Average [lsqb]O iii[rsqb] profiles in the EV1 plane.

Each panel shows the stacked [O iii] λ = 5,007 Å line of quasars in the RFe ii−FWHMHβ bins defined by the grey grid in Fig. 1 (in the same layout). RFe ii increases from left to right, and FWHMHβ increases from bottom to top. In each bin we further divided the quasars into different luminosity bins using the measured L5,100 Å continuum luminosities. We have normalized the line fluxes by the (host-corrected) average quasar continuum luminosity L5,100 Å for each stacking subset; hence, these stacked lines reflect the relative [O iii] strength among different samples. In addition to the decrease of [O iii] strength when RFe ii increases (that is, Fig. 1), we also observe a decrease in [O iii] strength with increasing luminosity8, 9. The [O iii] profile is in general asymmetric, with a blueshifted wing, whose relative contribution to the total profile increases when RFe ii or luminosity increases.

In addition to the strongest narrow [O iii] lines, all other optical narrow forbidden lines (such as [Ne v], [Ne iii], [O ii] and [S ii]) show similar EV1 trends and the Baldwin effect. Hot dust emission detected using WISE12, presumably coming from a dusty torus13, 14, also increases with RFe ii. In the Supplementary Information (and Extended Data Figs 3,4,5,6,7) we summarize all updated and new observations that firmly establish the EV1 sequence.

The [O iii]-emitting region is photoionized by the ionizing continuum from the accreting black hole. But the EV1 correlation of [O iii] strength with RFe ii holds even when optical luminosity is fixed, as demonstrated in Fig. 2. This suggests that another physical property of black hole accretion changes with RFe ii, one that, in turn, affects the relative contribution in the ionizing part of the quasar continuum as seen by the narrow-line region. The most likely possibility is the black hole mass MBH, or equivalently, the Eddington ratio L/MBH, given that L is fixed. The much less likely alternative would be that the [O iii] narrow-line region changes as a function of RFe ii. Reverberation mapping studies of nearby active galactic nuclei (AGN)15 have suggested that a virial estimate of MBH may be derived by combining the broad-line region size RBLR (measured from the time lag between continuum and emission-line variability) and the virial velocity of the line-emitting clouds estimated from the linewidth: , where G is the gravitational constant. The average FWHMHβ does decrease by about 0.2 dex when RFe ii increases from 0 to 2, and this fact underlies the earlier suggestion that EV1 is driven by the Eddington ratio4, 16.

A remarkable feature in Figs 1 and 2 is that the sequence is predominantly horizontal: there is little trend with FWHMHβ at fixed RFe ii. The standard virial mass estimators15, 17 would suggest that there is a strong vertical segregation in MBH, by a factor of a few in the vertical bins in Fig. 1. If lower MBH (or higher Eddington ratio) leads to weaker [O iii], as in the EV1 relation (that is, the horizontal trend), we should also see a vertical trend in Fig. 1. The absence of such a trend suggests that there is substantial scatter between FWHMHβ and the actual virial velocity, and the vertical spread in FWHMHβ in the EV1 plane largely does not track the spread in true black hole masses.

We propose, instead, that the sequence in RFe ii is driven by MBH; but the dispersion in FWHMHβ at fixed RFe ii is largely due to an orientation effect, as expected in a flattened broad-line region geometry. We first demonstrate that the average MBH indeed decreases with RFe ii for our quasar sample. We achieve this by measuring the clustering of SDSS quasars with low and high RFe ii values. In the hierarchical clustering Universe, more massive galaxies (which contain more massive black holes) form in rarer density peaks and are more strongly clustered18. We therefore expect quasars with larger RFe ii are less strongly clustered. This exercise, however, requires a very large sample size to achieve statistically significant results and has not been possible until now. Here we take advantage of the largest spectroscopic sample of galaxies from SDSS-III19, and use the much larger (by a factor of about 40) galaxy sample to cross-correlate20 with our quasar sample at redshift z ≈ 0.5 to substantially improve the clustering measurements. The resulting cross-correlation functions are shown in Fig. 3a, for the two quasar subsamples divided at the median RFe ii. A significant clustering difference is detected at 3.48σ: quasars with larger RFe ii are indeed less strongly clustered, confirming that they have on average lower MBH.

Figure 3: Cross-correlation functions between different quasar subsamples and a galaxy sample.
Cross-correlation functions between different quasar subsamples and a galaxy sample.

rp is the transverse comoving separation and wp is the projected two-point correlation function. a, Difference in the clustering strength when the quasar sample is divided at the median RFe ii. A significant difference (3.48σ) is detected: quasars with stronger RFe ii are less strongly clustered, indicating they have on average smaller black hole masses. b, Difference in the clustering strength when the quasar sample is divided by the virial black hole mass estimates based on FWHMHβ. No significant difference (1.64σ) is detected, indicating there is substantial overlap in the actual black hole masses between the two subsamples owing to the uncertainties in these FWHM-based virial black hole masses. Orientation-induced FWHMHβ dispersion can naturally lead to such uncertainties. Error bars are 1σ measurement errors estimated with jackknife resampling (Supplementary Information).

In the EV1 plane (Fig. 1), the distribution in FWHMHβ at fixed RFe ii is roughly log-normal, with mean value decreasing with RFe ii and a dispersion of about 0.2 dex (Extended Data Fig. 8). We argued above that this dispersion is largely due to orientation-induced FWHM variations in the case of a flattened broad-line region geometry. For a small subset of quasars that are radio-loud (around 10% of the population), it is possible to infer the orientation of the accretion disk, and by extension, the broad-line region, using resolved radio morphology to deduce the orientation of the jet. Such studies21, 22 show that high-inclination (more edge-on) broad-line radio quasars have on average larger FWHMHβ, in accordance with the orientation hypothesis. Below, we perform a different test for the more general radio-quiet quasar population, and we provide further evidence to support this argument in the Supplementary Information and Extended Data Figs 9 and 10.

We compile a sample of 29 low-redshift AGNs with literature broad-line region size measurements from reverberation mapping15, host stellar velocity dispersion (σ*) measurements23, and optical spectroscopy24. We use the well-established local MBH−σ* relation25 to independently estimate black hole masses for the 29 AGNs. We supplement the 29 local AGNs with a sample of about 600 SDSS AGNs26, where the host stellar velocity dispersion was estimated from the spectral decomposition of the SDSS spectrum into AGN and host galaxy components, and the broad-line region size RBLR was estimated using the tight correlation between RBLR and the AGN luminosity found in reverberation mapping studies27. We can then define a virial coefficient, . At a given MBH, f should not depend on FWHMHβ, if the latter is a faithful indicator of the broad-line region virial velocity. However, if FWHMHβ is orientation-dependent, as suggested above, f will be anti-correlated with FWHMHβ.

Indeed, there is a strong dependence of f on FWHMHβ at fixed MBH, shown in Fig. 4, consistent with the orientation hypothesis. A direct consequence is that the standard virial black hole mass estimates using FWHMHβ are subject to a significant uncertainty (about 0.4 dex), owing to this orientation dependence. To test this, we perform the same cross-correlation analysis as above, but for quasar subsamples divided by their virial black hole mass estimates based on FWHMHβ. The results are shown in Fig. 3b: there is no significant detection (1.64σ) in the clustering difference between the two quasar subsamples. This is in accordance with there being substantial overlap in the true black hole masses between the two subsamples, owing to the uncertainty in virial black hole mass estimates induced by using FWHMHβ. The division by RFe ii provides a cleaner separation of high-mass black holes from low-mass black holes in our sample.

Figure 4: The effect of orientation on FWHMHβ.
The effect of orientation on FWHMH[bgr].

The large symbols represent the 29 low-redshift AGNs that have both reverberation mapping data and host stellar velocity dispersion (σ*) measurements. The small symbols represent a low-redshift SDSS AGN sample26 with σ* and AGN spectral measurements based on spectral decomposition. We use the stellar velocity dispersion measurements and the local relation between black hole mass and σ* from inactive galaxies25 to estimate the black hole mass ( ) in these objects. We also estimate the average broad-line region size (RBLR = cτ, where c is the speed of light, and τ is the measured reverberation mapping lag) in these objects, either from direct reverberation mapping measurements, or by using the tight correlation between the broad-line region size and AGN luminosity27. The ratio of to (that is, the virial coefficient f) is plotted as a function of FWHMHβ, for different values. The strong trends of f with FWHMHβ at a given suggest that the dispersion in FWHMHβ does not reflect the underlying virial velocity of the broad-line region gas, and tends to bias the black hole mass estimates. This is in line with the fact that there is little vertical trend in the [O iii] strength in the EV1 plane (Fig. 1).

The collective evidence from this work leads to a simple interpretation of the observed main sequence of quasars (Fig. 1): the average Eddington ratio increases from left to right, and the dispersion in FWHMHβ at fixed RFe ii is largely an orientation effect. The many physical quasar properties correlated with EV1 are thus unified as being driven by changes in the average Eddington ratio of the black hole accretion. Although we do not discuss any physical model here, we suggest that the trends with the Eddington ratio are most probably caused by the systematic change in the shape of the accretion disk continuum and its interplay with the ambient emitting regions, which may in turn change the ionizing continuum (as seen by the emission-line regions) by modifying the structure of the accretion flow.

Online Content Methods, along with any additional Extended Data display items and Source Data, are available in the online version of the paper; references unique to these sections appear only in the online paper.

References

  1. Boroson, T. A. & Green, R. F. The emission-line properties of low-redshift quasi-stellar objects. Astrophys. J. 80 (suppl.). 109–135 (1992)
  2. Sulentic, J. W., Zwitter, T., Marziani, P. & Dultzin-Hacyan, D. Eigenvector 1: an optimal correlation space for active galactic nuclei. Astrophys. J. 536, L5–L9 (2000)
  3. Wang, T., Brinkmann, W. & Bergeron, J. X-ray properties of active galactic nuclei with optical FeII emission. Astron. Astrophys. 309, 81–96 (1996)
  4. Boroson, T. A. Black hole mass and Eddington ratio as drivers for the observable properties of radio-loud and radio-quiet QSOs. Astrophys. J. 565, 78–85 (2002)
  5. Laor, A. The soft X-ray properties of a complete sample of optically selected quasars. II. Final results. Astrophys. J. 477, 93–113 (1997)
  6. Schneider, D. P. et al. The Sloan Digital Sky Survey quasar catalog. V. Seventh data release. Astron. J. 139, 2360–2373 (2010)
  7. Shen, Y. et al. A catalog of quasar properties from Sloan Digital Sky Survey data release 7. Astrophys. J. 194 (suppl.). 45 (2011)
  8. Stern, J. & Laor, A. Type 1 AGN at low z—III. The optical narrow-line ratios. Mon. Not. R. Astron. Soc. 431, 836–857 (2013)
  9. Zhang, K., Wang, T.-G., Gaskell, C. M. & Dong, X.-B. The Baldwin effect in the narrow emission lines of active galactic nuclei. Astrophys. J. 762, 51 (2013)
  10. Baldwin, J. A. Luminosity indicators in the spectra of quasi-stellar objects. Astrophys. J. 214, 679–684 (1977)
  11. Dopita, M. A. & Sutherland, R. S. Spectral signatures of fast shocks. II. Optical diagnostic diagrams. Astrophys. J. 455, 468–479 (1995)
  12. Wright, E. L. et al. The Wide-field Infrared Survey Explorer (WISE): mission description and initial on-orbit performance. Astron. J. 140, 1868–1881 (2010)
  13. Antonucci, R. Unified models for active galactic nuclei and quasars. Annu. Rev. Astron. Astrophys. 31, 473–521 (1993)
  14. Urry, C. M. & Padovani, P. Unified schemes for radio-loud active galactic nuclei. Publ. Astron. Soc. Pacif. 107, 803–845 (1995)
  15. Peterson, B. M. et al. Central masses and broad-line region sizes of active galactic nuclei. II. A homogeneous analysis of a large reverberation-mapping database. Astrophys. J. 613, 682–699 (2004)
  16. Laor, A. On black hole masses and radio loudness in active galactic nuclei. Astrophys. J. 543, L111–L114 (2000)
  17. Vestergaard, M. & Peterson, B. M. Determining central black hole masses in distant active galaxies and quasars. II. Improved optical and UV scaling relationships. Astrophys. J. 641, 689–709 (2006)
  18. Bardeen, J. M., Bond, J. R., Kaiser, N. & Szalay, A. S. The statistics of peaks of Gaussian random fields. Astrophys. J. 304, 15–61 (1986)
  19. Ahn, C. et al. The tenth data release of the Sloan Digital Sky Survey: first spectroscopic data from the SDSS-III Apache Point Observatory Galactic Evolution Experiment. Astrophys. J. 211 (Supp.). 17 (2014)
  20. Shen, Y. et al. Cross-correlation of SDSS DR7 quasars and DR10 BOSS galaxies: the weak luminosity dependence of quasar clustering at z ~ 0.5. Astrophys. J. 778, 98 (2013)
  21. Wills, B. J. & Browne, I. W. A. Relativistic beaming and quasar emission lines. Astrophys. J. 302, 56–63 (1986)
  22. Runnoe, J. C., Brotherton, M. S., Shang, Z., Wills, B. J. & DiPompeo, M. A. The orientation dependence of quasar single-epoch black hole mass scaling relationships. Mon. Not. R. Astron. Soc. 429, 135–149 (2013)
  23. Park, D., Kelly, B. C., Woo, J.-H. & Treu, T. Recalibration of the virial factor and MBH-σ* relation for local active galaxies. Astrophys. J. 203 (suppl.). 6 (2012)
  24. Marziani, P. et al. An optical spectroscopic atlas of low-redshift active galactic nuclei. Astrophys. J. 145 (suppl.). 199–211 (2003)
  25. Kormendy, J. & Ho, L. C. Coevolution (or not) of supermassive black holes and host galaxies. Annu. Rev. Astron. Astrophys. 51, 511–653 (2013)
  26. Shen, J., Vanden Berk, D. E., Schneider, D. P. & Hall, P. B. The black hole-bulge relationship in luminous broad-line active galactic nuclei and host galaxies. Astron. J. 135, 928–946 (2008)
  27. Bentz, M. C. et al. The radius-luminosity relationship for active galactic nuclei: the effect of host-galaxy starlight on luminosity measurements. II. The full sample of reverberation-mapped AGNs. Astrophys. J. 697, 160–181 (2009)

Download references

Acknowledgements

Support for the work of Y.S. was provided by NASA through Hubble Fellowship grant number HST-HF-51314.01, awarded by the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy for NASA, under contract number NAS 5-26555. L.C.H. acknowledges support from the Kavli Foundation, Peking University, and the Chinese Academy of Science through grant number XDB09030102 (Emergence of Cosmological Structures) from the Strategic Priority Research Program. This work makes extensive use of SDSS-I/II and SDSS-III data (http://www.sdss.org/ and http://www.sdss3.org/).

Author information

Affiliations

  1. Carnegie Observatories, 813 Santa Barbara Street, Pasadena, California 91101, USA

    • Yue Shen

  2. Kavli Institute for Astronomy and Astrophysics, Peking University, Beijing 100871, China

    • Yue Shen &
    • Luis C. Ho

  3. Department of Astronomy, School of Physics, Peking University, Beijing 100871, China

    • Luis C. Ho

Contributions

Y.S. and L.C.H. co-developed the idea. Y.S. performed the measurements and analysis. Both authors contributed to the interpretation and manuscript writing.

Competing financial interests

The authors declare no competing financial interests.

Corresponding author

Correspondence to:

Author details

Extended data figures and tables

Extended Data Figures

  1. Extended Data Figure 1: Decomposed [O iii] λ = 5,007 Å luminosity. (137 KB)

    The core component (a) and the wing component (b) are shown for each composite spectrum shown in Fig. 2. Error bars are 1σ measurement errors estimated using Monte Carlo trials of mock spectra generated using the estimated flux error arrays of the co-added spectra. Both luminosities are normalized to the quasar continuum luminosity L5,100 Å, hence reflecting the strength of [O iii]. The core [O iii] shows a prominent anti-correlation with both L5,100 Å and RFe ii, while the wing [O iii] shows weaker anti-correlations with L5,100 Å and RFe ii. For both [O iii] components there is no correlation with FWHMHβ, as shown in Figs 1 and 2. The Baldwin effect and EV1 correlation for [O iii] shown in Fig. 1 and Fig. 2 are then primarily associated with the core [O iii] component. The difference between the core and wing [O iii] components may suggest different excitation mechanisms for both components.

  2. Extended Data Figure 2: Kinematic properties of the decomposed core and wing [O iii] components. (229 KB)

    a, FWHM against luminosity for core [O iii]. b, FWHM against luminosity for wing [O iii]. c, Velocity offset against luminosity for core [O iii]. d, Velocity offset against luminosity for wing [O iii]. Error bars are 1σ measurement errors estimated using Monte Carlo trials of mock spectra generated using the estimated flux error arrays of the co-added spectra. The most significant correlations are the correlation between luminosity and the core [O iii] FWHM, and the correlations between the wing [O iii] blueshift and L/RFe ii. The former correlation is consistent with the scenario that more luminous quasars are on average hosted by more massive galaxies with deeper potential wells, hence having larger core [O iii] widths. The latter correlations are consistent with the scenario that the wing [O iii] component is associated with outflows.

  3. Extended Data Figure 3: Composite SDSS quasar spectra for several other lines in the same RFe ii–FWHMHβ bins as defined in Fig. 1. (595 KB)

    a, Hβ and [O iii]. b, Mg ii. c, [O ii] 3,727 Å. d, [Ne v] 3,426 Å. As in Fig. 2, each composite spectrum has been normalized by the continuum such that the integrated line intensity reflects the strength of the line. The composite spectra for the Hβ region are generated using the pseudo-continuum-subtracted spectra, while for each of the other three lines (Mg ii, [O ii] and [Ne v]) the composite spectrum is the median spectrum created using the full SDSS spectra and normalized at a nearby continuum window.

  4. Extended Data Figure 4: Distribution in the EV1 plane in terms of C iv properties. (236 KB)

    A sample of low-redshift quasars with both Hβ and C iv measurements is shown, colour-coded by the C iv strength. A clear trend of decreasing C iv strength with RFe ii is seen, consistent with that seen for the other forbidden lines. The typical 1σ measurement uncertainty in C iv equivalent width is about 7% (relative to the measurement), and hence is negligible compared to the strong EV1 trend observed.

  5. Extended Data Figure 5: Distributions of SDSS quasars in the EV1 plane in terms of the optical–infrared (r − W1) colour. (309 KB)

    r is the SDSS r band (6,166 Å) and W1 is the WISE W1 band (3.4 μm). a, r − W1 for quasars with 0.4 < z < 0.8, for which the band-shifting effect is small. We see a trend of increasing mid-infrared emission relative to optical emission with increasing RFe ii. b, A similar result, using the excess colour, Δ(r − W1), which is the deviation of r − W1 colour from the mean colour at each redshift. Using Δ(r − W1) removes the redshift dependence of colours, and we can apply this to all quasars in our sample. This test suggests that the torus emission is enhanced in quasars with larger RFe ii. Given that we have argued that RFe ii is a good indicator for the Eddington ratio, this result suggests that quasars with higher Eddington ratios have stronger torus emission, which may have implications for the formation mechanism of the dusty torus.

  6. Extended Data Figure 6: A detailed look at the median excess optical-WISE colour Δ(r − W1) in the EV1 plane. (194 KB)

    The same bins as defined in Fig. 1 are used. Error bars are the 1σ uncertainty in the median, estimated by the standard deviation divided by the square root of the number of objects in the bin. At fixed RFe ii, we see increasing relative torus emission when FWHMHβ increases. This is consistent with the orientation scenario: larger FWHMs indicate more edge-on systems, which suffer more from geometric reduction (the cosI factor) and/or dust extinction in the optical than in the infrared parts of the spectrum.

  7. Extended Data Figure 7: Distribution in the EV1 plane in terms of X-ray properties. (254 KB)

    The subset of our SDSS quasars with available measurements of their soft X-ray photon index ΓX are shown. ΓX increases (becomes softer) with increasing RFe ii, consistent with earlier findings3, 5. CSC refers to objects from the Chandra Source Catalog and XMM refers to objects from the XMM-Newton Serendipitous Catalog. The contours are the distribution of all SDSS quasars in our sample, as in Fig. 1.

  8. Extended Data Figure 8: The same EV1 plane as in Fig. 1 in logarithmic FWHMHβ. (352 KB)

    The dashed lines show the running median value as a function of RFe ii and the dotted lines show the 16% and 84% percentiles, for objects in different luminosity bins. The distribution of FWHMHβ at fixed RFe ii roughly follows a log-normal distribution, with a dispersion of about 0.15−0.25 dex, which we argued comes mostly from orientation-induced variations. Lower-luminosity objects tend to have slightly larger dispersion in FWHMHβ, possibly caused by a broader Eddington ratio distribution at lower luminosities, which introduces additional dispersion in FWHMHβ. LEdd = 1.3 × 1038(MBH/1  ) erg s−1 is the Eddington luminosity of the black hole.

  9. Extended Data Figure 9: Distributions of radio-loud and radio-quiet quasars in EV1 plane. (215 KB)

    The radio-loud population shifts to lower RFe ii and larger FWHMHβ, compared with the radio-quiet population. We further divide the radio-loud quasars into core-dominant and lobe-dominant subsets, but we caution that our morphological classification is very crude, and there is potentially a large mixture of true morphological types between the two subsamples. The core-dominant (more pole-on) radio quasars have systematically smaller FWHMHβ compared with the lobe-dominant radio quasars, consistent with the hypothesis that orientation leads to variations in FWHMHβ. The points with error bars are the median and 1σ uncertainty for the median in each RFe ii bin.

  10. Extended Data Figure 10: Distribution in the EV1 plane colour-coded by the FWHM/σ ratio. (736 KB)

    The distribution has been smoothed over a box of ΔRFe ii = 0.2 and ΔlogFWHMHβ = 0.2. We show only points for which there are more than 50 objects in the smoothing box to average. The black open circles show the median FWHMHβ at fixed RFe ii (using all objects in that bin), with the error bars indicating the 1σ uncertainty for the median. The transition in FWHM/σ reflects the change in orientation of the broad-line region disk relative to the line of sight.

Supplementary information

PDF files

  1. Supplementary information (333 KB)

    This file contains Supplementary text and additional references.

Additional data