- Split View
-
Views
-
Cite
Cite
Eric Emsellem, Is the black hole in NGC 1277 really overmassive?, Monthly Notices of the Royal Astronomical Society, Volume 433, Issue 3, 11 August 2013, Pages 1862–1870, https://doi.org/10.1093/mnras/stt840
- Share Icon Share
Abstract
A recent claim has been made by van den Bosch et al. that the fast-rotator galaxy NGC 1277 hosts an overmassive black hole with a mass (1.7 × 1010 M⊙) larger than half its (central) stellar spheroid mass. We revisit this claim here by examining the predictions from simple dynamical realizations based on new multi-Gaussian expansion (MGE) models of NGC 1277, using the same inclination i = 75°, and constant mass-to-light ratios. We present realizations which fit well the observed photometry taking into account an approximation for the extinction due to the central dust ring. The mass-to-light ratio M/L is fixed following scaling relations which predict a Salpeter-like initial mass function for such a luminous early-type fast rotator, 60 per cent higher than the one of the previously derived best-fitting model. A model without a black hole provides a surprisingly good fit of the observed kinematics outside the unresolved central region, but not, as expected, of the central dispersion and Gauss–Hermite h4 values. A model with a black hole mass of 5 × 109 M⊙ allows us to fit the central dispersion profile, consistently with models of the same mass and M/L in van den Bosch et al. It departs from the central h4 values by only about twice the given uncertainty. A slightly varying M/L or the addition of high-velocity stars in the central spatially unresolved region would further lower the need for a very massive black hole in the central region of NGC 1277. These results do not, by themselves, rule out the presence of a presumed overmassive black hole at the centre of NGC 1277. However, they lead us to advocate the use of 3σ (as opposed to 1σ) confidence intervals for derived MBH as better, more conservative, guidelines for such studies. We also caution for the use of ill-defined spheroidal components as an input for scaling relations, and emphasize the fact that a MBH in the range 2–5 × 109 M⊙ would represent less than 5 per cent of the spheroid bulge-like mass of our models and less than 2.5 per cent of its total stellar mass. This would make the black hole in NGC 1277 consistent or just twice as large as what a recent version of the MBH–σ predicts, well within the observed scatter. We examine the impact of the presence of an inner bar by running simulations from the same MGE model but with extreme anisotropies. An inner small (600 pc diameter) bar forms, and an end-on view does get closer to fitting the central dispersion profile (and fits the h3 amplitude) without the need for a central dark mass, while adding a black hole of 2.5 × 109 M⊙, in line with the prediction from scaling relations, allows us to fit the dispersion peak and h3 profiles. Both models, however, still fail to fit the central h4 value (overpredicting the mean velocity). The claimed large mass of the presumed black hole therefore mostly relies on the measured positive high central h4 (at high dispersion), which can be associated with broad wings in the line-of-sight velocity distribution (high-velocity stars). This emphasizes the need to go beyond medium-resolution long-slit kinematics, with e.g. high-resolution integral-field spectroscopic data. In the specific case of NGC 1277, molecular or ionized gas kinematics (if present) within the central arcsecond (or at large scale) may provide a strong discriminant between these various models. We finally briefly discuss the fact that NGC 1277 resembles a scaled-up version of e.g. NGC 4342, another nearly edge-on fast rotator with a potentially large (but not overmassive) black hole.
1 INTRODUCTION
A recent claim has been recently made by van den Bosch et al. (2012) that the fast-rotator galaxy NGC 1277 hosts an overmassive black hole with a mass (17 ± 3 × 109 M⊙) quoted as larger than half the stellar bulge mass. This departs very significantly from predictions of the so-called MBH–σ relation (see e.g. Tremaine et al. 2002), or the bulge–black hole mass relation. The best-fitting model including a relatively small size black hole (108 M⊙) with respect to its spheroidal component, and using the same mass-to-light ratio, fails to fit e.g. the stellar velocity dispersion profile within the central 10 arcsec (see their fig. 3). This emphasizes the fact that a large 17 ± 3 × 109 M⊙ black hole would influence the potential of the galaxy far out, a few hundreds of parsec from its central location. If confirmed, the existence of this overmassive black hole (relatively to the host galaxy) could either be a rare and exotic case, or would actually shed new light on the relation between the formation and evolution processes of early-type fast rotators and their central dark mass.
In this short paper, we revisit this claim by building simple dynamical models based on the multi-Gaussian expansion (MGE) technique (Emsellem, Monnet & Bacon 1994), also used in the modelling approach adopted by van den Bosch et al. (2012). The goal of the present paper is not to perform new fits to the observed data set. Here, we explore further what are the main constraints motivating the need for a very large black hole mass, and wish to examine whether other slight variations in the input assumptions may help reconcile the specific case of NGC 1277 with the above-mentioned scaling relations. We also examine the potential impact of an inner bar for the observed stellar kinematics of such a fast rotator (see e.g. Gerhard 1988), and more specifically its potential effect on the central stellar velocity dispersion. This could result in important diagnostics and in a near future help refute or confirm the claim of van den Bosch et al. (2012, hereafter vdB+12). In Section 2 we provide some basic information about the observational data and methods we have used in this paper. We then proceed in Section 3 to present the results coming out of this focused study, and wrap things up in Section 4.
2 OBSERVATIONAL DATA AND METHODS
2.1 Photometry and stellar kinematics
For the present paper, we have used of the same input photometry and stellar kinematics as in vdB+12.
Images obtained with the Hubble Space Telescope (HST)/Advanced Camera for Surveys (ACS) in the course of the GO:10546 programme (PI: Fabian) with the F550M and F625W filters. We gathered the ACS images from the ST-ECF Hubble Science Data Archive and used the point spread function (PSF) as derived from the TinyTim software.
Long-slit stellar kinematics as presented in vdB+12 which consists of a major-axis profile from the Low Resolution Spectrograph (LRS) mounted at the Hobby–Eberly Telescope (HET), covering up to about a radius of about 20 arcsec for the stellar velocity V, velocity dispersion σ, as well as for the next two Gauss–Hermite parameters h3 and h4 (within the central 7 arcsec only). The spatial resolution is about 1.6 arcsec full width at half-maximum (FWHM) and is parametrized with a double Gaussian model (vdB+12).
The reader is encouraged to consult at the Supplementary Information provided by vdB+12 for further details.
2.2 MGE photometric model
The dynamical modelling relies on the MGE technique developed by Emsellem et al. (1994), itself partly based on developments by Bendinelli (1991) and Monnet, Bacon & Emsellem (1992). The MGE method is a flexible way of building photometric and dynamical models of axisymmetric galaxies (see Monnet et al. 1992; Emsellem et al. 1994, for a specific treatment in the triaxial case, and deconvolution) with complex multicomponent structures. A set of two-dimensional Gaussians with various widths and axis ratios is used to represent the (observed) surface brightness distribution. This MGE representation is uniquely deprojected, given viewing angles and assuming that each Gaussian component is itself deprojected into a three-dimensional component constant on homothetic ellipsoids. The mass distribution can be specified by attributing a mass-to-light ratio value to each Gaussian, and including additional components (black hole, or other Gaussians which may serve as a representation for a dark halo or a gas disc).
vdB+12 have built an MGE model for NGC 1277 using the ACS F550M image and galfit (Peng et al. 2002), the parameters of which are available in their paper (see their supplemental table 1). We built a new MGE model this time using the HST/ACS F625W image: we do not expect that model to differ much from the one built by vdB+12 even considering the different implementation for the MGE fitting part itself. This new model should help to at least provide some partial assessment on the expected variation in the modelling. We made various fitting attempts using the pymge python package (Emsellem et al., in preparation; see also Cappellari 2002 for a robust idl implementation), which led to very similar outputs. During the fitting process, we masked a few bright stars and a nearby companion, and we accounted for the (apparently) roundish early-type galaxy NGC 1278 by first creating an MGE model for that galaxy which was then subtracted from the ACS image (see Figs 1 and 2). We also attempted to mask out the region which suffered too much from the dust ring extinction visible within the central arcsecond of NGC 1277 so that our MGE model does not underpredict the amount of light at the very centre. The overall approach is similar (though not identical) to the procedure followed by vdB+12. Since we will subsequently convolve the predicted kinematics with a broad PSF (see below), we took into account the overall effect of the ACS PSF by approximating the corresponding TinyTim prediction with a set of two centred Gaussians.
# . | Σ0 . | σG . | q . |
---|---|---|---|
. | (L⊙ pc−2) . | (arcsec) . | . |
1 | 116 230.070 | 0.0100 | 0.700 |
2 | 167 697.701 | 0.0300 | 0.300 |
3 | 28 108.207 | 0.0700 | 0.700 |
4 | 11 919.362 | 0.1488 | 0.700 |
5 | 6308.461 | 0.2889 | 0.305 |
6 | 9483.126 | 0.3208 | 0.700 |
7 | 6954.700 | 0.6208 | 0.700 |
8 | 1866.762 | 0.9320 | 0.305 |
9 | 3551.888 | 1.2155 | 0.695 |
10 | 1086.092 | 2.6801 | 0.553 |
11 | 237.533 | 4.4573 | 0.286 |
12 | 344.707 | 6.5533 | 0.437 |
13 | 55.810 | 8.9404 | 0.650 |
14 | 11.970 | 10.5439 | 0.655 |
15 | 8.385 | 16.9824 | 0.655 |
# . | Σ0 . | σG . | q . |
---|---|---|---|
. | (L⊙ pc−2) . | (arcsec) . | . |
1 | 116 230.070 | 0.0100 | 0.700 |
2 | 167 697.701 | 0.0300 | 0.300 |
3 | 28 108.207 | 0.0700 | 0.700 |
4 | 11 919.362 | 0.1488 | 0.700 |
5 | 6308.461 | 0.2889 | 0.305 |
6 | 9483.126 | 0.3208 | 0.700 |
7 | 6954.700 | 0.6208 | 0.700 |
8 | 1866.762 | 0.9320 | 0.305 |
9 | 3551.888 | 1.2155 | 0.695 |
10 | 1086.092 | 2.6801 | 0.553 |
11 | 237.533 | 4.4573 | 0.286 |
12 | 344.707 | 6.5533 | 0.437 |
13 | 55.810 | 8.9404 | 0.650 |
14 | 11.970 | 10.5439 | 0.655 |
15 | 8.385 | 16.9824 | 0.655 |
# . | Σ0 . | σG . | q . |
---|---|---|---|
. | (L⊙ pc−2) . | (arcsec) . | . |
1 | 116 230.070 | 0.0100 | 0.700 |
2 | 167 697.701 | 0.0300 | 0.300 |
3 | 28 108.207 | 0.0700 | 0.700 |
4 | 11 919.362 | 0.1488 | 0.700 |
5 | 6308.461 | 0.2889 | 0.305 |
6 | 9483.126 | 0.3208 | 0.700 |
7 | 6954.700 | 0.6208 | 0.700 |
8 | 1866.762 | 0.9320 | 0.305 |
9 | 3551.888 | 1.2155 | 0.695 |
10 | 1086.092 | 2.6801 | 0.553 |
11 | 237.533 | 4.4573 | 0.286 |
12 | 344.707 | 6.5533 | 0.437 |
13 | 55.810 | 8.9404 | 0.650 |
14 | 11.970 | 10.5439 | 0.655 |
15 | 8.385 | 16.9824 | 0.655 |
# . | Σ0 . | σG . | q . |
---|---|---|---|
. | (L⊙ pc−2) . | (arcsec) . | . |
1 | 116 230.070 | 0.0100 | 0.700 |
2 | 167 697.701 | 0.0300 | 0.300 |
3 | 28 108.207 | 0.0700 | 0.700 |
4 | 11 919.362 | 0.1488 | 0.700 |
5 | 6308.461 | 0.2889 | 0.305 |
6 | 9483.126 | 0.3208 | 0.700 |
7 | 6954.700 | 0.6208 | 0.700 |
8 | 1866.762 | 0.9320 | 0.305 |
9 | 3551.888 | 1.2155 | 0.695 |
10 | 1086.092 | 2.6801 | 0.553 |
11 | 237.533 | 4.4573 | 0.286 |
12 | 344.707 | 6.5533 | 0.437 |
13 | 55.810 | 8.9404 | 0.650 |
14 | 11.970 | 10.5439 | 0.655 |
15 | 8.385 | 16.9824 | 0.655 |
As mentioned, the dust ring strongly affects the very central region of the surface brightness distribution (see Fig. 1 , or fig. 1 in vdB+12). The outer luminosity profile is also not well constrained considering the limited depth of the ACS images. In order to make sure we do not extend our MGE model significantly beyond the one built by vdB+12, and to minimize its total flux, we further constrain the two outer Gaussians to be relatively flat with an axis ratio smaller or equal to 0.7, and to have the same major axis widths as the ones in vdB+12. The final model normalized in flux to the F550M (for comparison) is shown in Fig. 3 (see Table 1) and compared to the one of vdB+12. Both models are surprisingly similar although, as expected, our MGE model is slightly flatter in the inner and outer parts. The fact that we imposed a flattening constraint for the largest Gaussians also prevents our MGE model from showing an abrupt change in its ellipticity in the outer parts where the observed photometry does not constrain the model well. Our MGE model has a total V-band luminosity of about 1.8 × 1010 L⊙ as compared to about 2.0 × 1010 L⊙ for the model of vdB+12.
2.3 MGE dynamical models and comparison with observations
We built dynamical models from N-body realisations of the MGE axisymmetric models described above, using pymge. This is achieved by deprojecting the MGE models with an inclination of 75°, assuming a constant stellar mass-to-light ratio M/L for all Gaussians, and solving Jeans equations as in Emsellem et al. (1994) and Cappellari (2008), constraining the velocity ellipsoid associated with each Gaussian to have constant axis ratios, namely σR/σz and σθ/σR (where R, θ, z are the cylindrical coordinates). We further added a central point-like dark mass to the gravitational potential, playing thus the role of a black hole, but no outer dark component. We kept the freedom to reverse the orbital sense of rotation of stars, as this is a way to influence the projected odd velocity moments significantly (e.g. the projected mean velocity V and h3) without impacting the photometry (and mass) distribution. The realizations are compared with observations by first applying a simple projection, and then taking into account the seeing PSF, pixel size and nature of the observables (photometry, kinematics via its Gauss–Hermite moments). For the kinematic quantities, we reconstructed the line-of-sight velocity distributions (LOSVDs) with velocity bins of 25 km s−1 steps, which are subsequently fitted using Gauss–Hermite functions up to the fourth degree leading to a measure of V, σ, h3 and h4. We performed realizations using from 105 up to 4 × 106 particles, and we checked that the total number of particles does not significantly affect the output projected quantities.
For comparison with the HST/ACS images, we approximated the observed extinction due to the presence of a dust (and presumably gas) ring in the central arcsecond by adding a simple ‘screen’ model: the ring itself is modelled as a circular ring of dust in the equatorial plane (between 0.5 and 0.95 arcsec) and a maximum extinction factor of about 1.7 mag. This dust model is not meant as a realistic account of the dust extinction, but only as a way to get a first estimate of the effect of such a ring. This has obviously a significant effect on the reproduced HST photometry (see Fig. 1), but we find that it does not have a significant impact on the observed kinematics at the spatial resolution of the HET data.
As mentioned above, we used a constant mass-to-light ratio M/L as in vdB+12. We define a reference value, by simply following the known trend of more massive early-type galaxies having higher M/L (see Cappellari et al. 2013, and references therein, and e.g. Cappellari et al. 2006; Thomas et al. 2011). vdB+12 quote an aperture (within one effective radius but excluding the central arcsecond) stellar velocity dispersion of σe ∼ 333 km s−1, and the total V-band luminosity of NGC 1277 is around 2 × 1010 L⊙: this means that we expect a M/L consistent with a close to Salpeter-like initial mass function (IMF). The same is true if we refer to the mass of the galaxy which is between 1 and 2 × 1011 M⊙. We therefore adopted a reference value of M/LV = 10, about 60 per cent larger than the value of 6.3 for the best-fitting model in vdB+12. The circular velocity profiles in the equatorial plane for both this model (with and without a black hole) and the best-fitting model presented in vdB+12 (including an ‘overmassive’ black hole of 17 × 109 M⊙) are shown in Fig. 4. We should note that what drives the combined amplitudes of the stellar kinematics in the outer part (e.g. the second-order velocity moment V2 + σ2) is the total mass, inclusive of the stars and dark matter if present. Having M/L set to a Salpeter-like IMF without a dark halo, could be thus roughly equivalent to having a lower value of the stellar M/L but adding dark matter (which would require to additionally define its relative scale, mass normalization and shape): this is illustrated in the shift towards higher M/L of the chi-square contours obtained by vdB+12 (figs 1 and 3 of their Supplementary Material) when removing the dark halo in the models. For a galaxy of the mass of NGC 1277, we would expect a dark matter fraction of about 15 per cent (Cappellari et al. 2013) within one effective radius (5 arcsec), and probably dominating the mass budget beyond ∼15 arcsec. Since we focus here on simple models, and do not try to actually obtain best fits to the observed kinematics, we voluntarily exclude the use of an additional dark matter component in the outer part (a dark halo) or a varying M/L profile.
We used the gadget2N-body code developed by Springel (2005) to both test the stability of our MGE N-body realizations, as well as the impact of the formation of a bar. For the latter, we imposed a more extreme anisotropy with a smaller tangential dispersion σθ: this basically forces the model to start with dynamically cold initial conditions and is an efficient way to enforce the (artificial) triggering for the formation of a non-axisymmetric system (such as a bar).
3 RESULTS
For our reference (axisymmetric) models, we impose a simple anisotropy profile with σR/σz ∼ 1.15 and σR/σθ ∼ 1.4. The first ratio corresponds to |$\beta _{Rz} = 1 - \sigma _z^2 / \sigma _R^2 \sim 0.25$|, which is incidentally consistent with the general trend found by Cappellari et al. (2007) where βRz ∼ 0.7 ϵintr with ϵintr being the intrinsic ellipticity, assuming an ellipticity of about 0.3 for NGC 1277. The relatively high tangential anisotropy of the model leads to a high amplitude for the azimuthal rotation, making it a clear fast rotator (Fig. 5). We built realizations with various black hole masses using this reference model: in Fig. 6 we illustrate two models, one without a black hole (MBH = 0, Model A), and one with a black hole (MBH = 5 × 109 M⊙, Model B) twice as big as the one predicted by the MBH–σ relation recently provided by McConnell et al. (2011). Both models are stable when run with gadget2 (pure N-body run with 4 × 106 particles with various tested softenings between 10 and 50 pc). Model B corresponds to a model which departs from the best-fitting model of vdB+12 at more than the 3σ level (without a dark halo, see their Supplemental fig. 3), while we obviously expect Model A (without a black hole) to be a very poor fit of the observed stellar kinematics. In both cases we reversed the sense of rotation for about 12 per cent of the stars for the rounder inner Gaussian components, as to follow the observed skewed LOSVD (h3 being significantly non-zero): this basically increases the azimuthal dispersion within the central 10 arcsec or so (see Fig. 5).
These models were not designed as actual fits for the observed stellar kinematics, but can reproduce the observed kinematics profiles rather well over most of the available radial ranges, and down to about 1 arcsec. This may look surprising considering the very poor fit of the model with a low-mass (108 M⊙) black hole presented by vdB+12 in their fig. 3. However, that fit was constrained to have the same mass-to-light ratio than the best-fitting model, with a low value of 6.3, thus being very far from the best-fitting model without (or with a low-mass) black hole. We also compared the best-fitting model with MBH = 108 M⊙ (Remco van den Bosch, private communication) with our Model A at the same input M/L: our model is a significantly better fit to the kinematics, but this is mostly in the outer parts while in the inner region, the two models are similar (the Schwarzschild technique provides a slightly but not significantly better fit as expected). The reason for this is certainly associated with the rounder outer Gaussians included in vdB+12 MGE model, which may render more difficult the fit of the (high V/σ) kinematics at large radii. It, however, does not invalidate the analysis in the central part. As expected, Model A (model without a black hole) does not fit the central dispersion value and has a slightly too shallow central velocity gradient. Model A, however, represents already quite a remarkable fit of the outer region down to the central 2 arcsec. The addition of a 5 × 109 M⊙ black hole (Model B) solves these two above-mentioned discrepancies. The h3 profile is fitted rather well independently from the mass of the black hole which is used: this is mostly the consequence of the presence of about 10 per cent of counter-rotating stars which skew the observed LOSVD. The only significant discrepancy between Model B and the observed stellar kinematics lies in the h4 profile at the very centre, where the predicted values are slightly two low (0.0–0.02 instead of 0.04). That difference is, however, rather small, and only twice the given uncertainty on these parameters. The model of vdB+12 which corresponds to the M/L and MBH values of our Model B provides a slightly better fit of the central region, while it provides a too low predicted V/σ in the outer part, again very probably due to its rounder Gaussians (a situation similar to the one described for Model A).
A high h4 value is associated with high-velocity wings of the LOSVD in the central arcsecond. A spurious h4 value could for example originate in a so-called template mismatch (a mismatch between the galaxy spectra at these positions and the stellar template used to extract the kinematic parameters). However, if confirmed, such a high h4 value (associated with a high central dispersion) would, for the central LOSVDs of NGC 1277, actually correspond to a few 108 M⊙ stars at velocities between 900 and 1100 km s−1: these are missing in Model B. To remove that discrepancy, we could artificially add stars with high angular momentum (circular orbits) within 0.1 arcsec (∼35 pc), where the circular velocity is in the right range (see Fig. 4). This would, however, lead to a significant discrepancy in the photometry (assuming a constant M/L). A more massive black hole (e.g. 17 × 109 M⊙, as in vdB+12) naturally predicts the required high-velocity wings by increasing the volume within which such velocities are reached. This central h4 value is therefore a strong signature of large mass excess in the region unresolved by the HET spectroscopy, and is (as expected) a strong discriminant between models with various black hole masses.
We also examined the impact of the presence from a potentially small bar in these models by running gadget2N-body simulations with more extreme anisotropies, this time setting up the black hole mass to zero (Model C) or to a value of 2.5 × 109 M⊙, consistent with scaling relations (Model D). As mentioned in Section 2.3, we force a small tangential anisotropy dispersion by setting a high value of σR/σθ, namely 2.0 here. With such initial conditions a small bar quickly forms after a few dynamical times in the simulation, and stays as long the simulation was run (200 Myr). The extent of the bar is small, with a diameter of about 600 pc, as illustrated in the face-on view in Fig. 7. When the bar is edge-on (and the model inclined at 75°), the photometry is clearly not consistent with the observed photometry. With the bar seen end-on, the fit to the photometry is rather good as shown in Fig. 8, which is a bit surprising considering the significant mass redistribution triggered by the formation of the bar.
We then compare the end-on and edge-on views of these two models (Models C and D) with the observed stellar kinematics of vdB+12 in Fig. 9. Using the same mass-to-light ratio M/LV = 10, we get a reasonable fit of the global profiles even though this model is now, as expected, slightly underpredicting the velocity dispersion and overpredicting the mean velocity. An end-on view of the bar leads to an increase in the central dispersion peak which gets close to the observed values (385 instead of 415 km s−1) for Model C and fits the dispersion peak for Model D. It is only a relative surprise considering that an end-on bar has already been discussed more than 20 yr ago as a potential way to maximize the observed velocities of stars (Gerhard 1988). It is also interesting to note that, in both models, the end-on bar naturally produces a significant kink in the h3 profile which follows rather closely the observed radial profile without the need for counter-rotating stars as in Models A and B. Ironically, the fit to the h4 profile is reasonably good when the bar is near edge-on, but with the predicted central h4 becoming negative for an end-on view. Again, the central h4 seems to be the most constraining observed parameter for the presumed central dark mass. Obtaining a fit using such an evolved (barred) model is clearly beyond the scope of this paper. Results from Models C and D seem to indicate, however, that the presence of such a non-axisymmetric system may eventually lower the required central dark mass (even though the co-existence of a central dark mass and an inner 600 pc bar would have to be explained). It is worth noting that the bar which forms in these models would be located just inside the observed dust ring: this is very probably just linked with the fact that the circular velocity of Model A (axisymmetric model, and no black hole) peaks near a radius of 1 arcsec.
4 DISCUSSION AND CONCLUSIONS
In this short paper, we revisit the recent claim for the presence of an overmassive black hole in NGC 1277 by vdB+12. We build new MGE models which fit the photometry, using an approximation for the extinction of the observed dust ring, and assuming a fixed M/L = 10 consistent with a Salpeter IMF, and a nearly constant anisotropy.
We show that a model without a black hole does indeed not fit the central dispersion and h4 values, as expected, but still provides a rather good global fit of the observed long-slit kinematics. The corresponding models (with similar M/L and MBH values) of vdB+12 have some difficulty to follow the high V/σ values observed in the outer parts: this may be partly linked with the intrinsic flattening of the outer Gaussians, and the flatter external components of our MGE models seem to significantly improve the situation in this context. The model with a very low MBH value presented in fig. 3 of vdB+12 corresponds to the (low) mass-to-light ratio of their overall best-fitting model. It is clearly a very bad fit to the observed kinematics over nearly the entire radial range. However, this may be somewhat misleading as it naturally emphasizes the tension between models with and without a central dark mass, not following the ridge of minimum chi-squared in the M/L–MBH diagrams in vdB+12. We find it remarkable that a model with a constant M/L following observed scaling relations, simple anisotropy profiles and no black hole provides a rather reasonable fit to the large-scale (major axis) stellar kinematics of the galaxy except for the central arcsecond. The addition of a 5 × 109 M⊙ central dark mass removes most of the discrepancy with the central stellar velocity dispersion value, and has a predicted h4 value within two times the quoted uncertainty on that parameter.
We must emphasize again that the models presented in this paper are not actual fits to the data, but only MGE models which fit the photometry and include simple assumptions for its mass-to-light ratio and anisotropy: they can therefore not lead to a direct statement with respect to the ‘best-fiting’ value for MBH, and the value claimed by vdB+12. By construction, the Schwarzschild technique does find the best-fitting value given a set of assumptions (e.g. MGE model). Still, there are reasons to doubt the strict validity of the confidence interval for the MBH value derived by Schwarzschild modelling. Some doubt may be raised when we witness the difficulty of the M/L = 10 models of vdB+12 to fit the outer high V/σ values very probably due to the roundish outer Gaussians of their MGE models. In that context, the role of the black hole becomes very prominent as to allow the flexibility (with a lower M/L value) to fit the entire radial range. The addition of dark matter, if in a spherical halo, should not impact much on these issues, except by lowering the need for a high M/L to fit the overall amplitude of the kinematics (e.g. V2 + σ2) in the outer part. Looking at the simple parametrization of Model A (constant M/L, no black hole) and the overall quality of the fit it provides outside the central 2 arcsec, it is tempting to use an argument in the spirit of Occam's razor to discard a very high MBH for NGC 1277. Since chi-square contours should provide (as presented in vdB+12) an objective assessment of the quality of the fits, but do not include the systematic changes which could be triggered by a change in the MGE model itself (via e.g. different Gaussians, or a slightly varying M/L), we believe that it solely means that we should be more conservative and advocate to quote uncertainties corresponding to 3σ confidence levels instead (which would nearly include Model B), namely the best-fitting value of 17 × 109 M⊙ (vdB+12) but a range [5–25] × 109.
We now briefly comment on the mass ratio between the black hole and spheroidal component of the galaxy. vdB+12 find that a 17 × 109 M⊙ black hole would represent 59 per cent of the central bulge-like component. However, this directly relies on the definition of what should be the ‘bulge’ fraction of the galaxy (central or global component). By focusing on a specific scale, and thus a presumed central component, there is a risk that such a fraction is artificially set on the high side. This again depends on a given scale and the (difficult) identification of what we should call a bulge. Since there is a direct connection between a given M/L and the corresponding best-fitting MBH (see supplemental figs 1 and 3 of vdB+12), it obviously also depends on the M/L itself (and the fraction of dark matter in the outer region), which directly influences the bulge mass. To further quantify this using our MGE model (with M/L = 10), and assuming we now identify the spheroidal component to be the sum of all Gaussians with axis ratios larger than 0.3, we find a mass of ∼1.8 × 1011 M⊙. If we focus only on the more central spheroidal components (within e.g. 2 kpc) this would amount to about ∼1011 M⊙. The large value advocated by vdB+12 for the black hole would thus represent between 10 and 20 per cent of these masses, already significantly lower than the mentioned value of 59 per cent. Such an estimate would not be significantly impacted by the presence of about 15 per cent of dark matter within one effective radius. With a lower MBH of 5 × 109 M⊙ as for Model B of the present paper, this would correspond to either 3 or 5 per cent, putting it in a range similar to what was derived for e.g. NGC 4342 (Cretton & van den Bosch 1999). With a measured σ of 333 km s−1 (vdB+12), the latter value for MBH would make the black hole in NGC 1277 only twice bigger than what a recent version of the MBH–σ relation predicts (McConnell et al. 2011), well within the measured scatter. NGC 1277 would then again resemble a scaled-up version of e.g. NGC 4342, another close to edge-on fast rotator with a relatively large black hole: the fact that these are two, close to edge-on, fast rotators is an interesting fact to emphasize. In both cases, the spheroidal component contributes significantly to the overall rotational support of the galaxy, with a relatively small tangential anisotropy (at least near the equatorial plane).
We have also presented an illustration of the impact of an inner bar in such models (Models C and D) and showed that the predicted central dispersion value could get relatively close (within 8 per cent) to the observed value when such a minibar is viewed close to end-on, even without a central dark mass (the model fitting the dispersion peak when MBH = 2.5 × 109 M⊙). Such models naturally reproduce the relatively high h3 amplitude, but fail to reproduce the observed positive central h4 value. We cannot comment further at this stage about the possibility of fitting all observed kinematics simultaneously with a bar, as this would require more finely tuned models to understand whether or not the central h4 value is obtainable for such systems. It is, however, probable that a central mass (of a few 109 M⊙) would anyway be required if the high positive h4 value is confirmed, as it corresponds to broad wings in the central LOSVDs, thus stars moving at speeds significantly above the predicted circular velocity (see Fig. 4).
This work emphasizes the fact that the high value claimed for the mass of the central dark mass, as quoted in vdB+12, therefore mostly relies on the robustness of the central (significantly positive) h4 value. This seems quite natural considering that a high h4 can be the direct consequence of e.g. a high central mass concentration. The differences (in the high velocity wings of the LOSVDs) between our model with a 5 × 109 M⊙ black hole and the measured central h4 values cumulate to about 5 × 108 of stars rotating at about 1000 km s−1. According to the predicted circular velocity profiles shown in Fig. 4, this could correspond to a concentration of high angular momentum stars within the central 70 and 25 pc for MBH of 17 and 5 × 109 M⊙, respectively: in the latter case, such a structure would not be resolved but should be detectable with integral-field spectroscopy and adaptive optics or obtained in excellent seeing conditions. Predictions from Models B and D for integral-field kinematics of the central few arcseconds at high resolution (FWHM of 0.3 arcsec) are provided in Figs 10 and 11: the shape of the central stellar velocity dispersion peak (at that resolution), as well as the higher order moments should be strong discriminants between these various models as well as to further constrain the black hole mass. Black holes of 2.5 × 109 and 5 × 109 M⊙would imply a very peaked central dispersion reaching about 445 and 500 km s−1, respectively, at that resolution. We should expect a significantly more massive black hole as advocated by vdB+12 to produce an even more extreme dispersion peak, reaching well beyond 500 km s−1. A central bar may show up a distinct kink in the h3 map within 2 arcsec with local minima along the major axis (see Fig. 11).
It may be difficult to rule out a lower mass black hole using potential ionized or molecular gas associated with the dust ring, as this ring has a diameter close to 1 arcsec where the circular velocity profiles of the various models (see Fig. 4) overlap, so that any interpretation would depend on the detailed characteristics of the gaseous orbits (circular or not). A proper test would again require two-dimensional spectroscopy (see e.g. Krajnović et al. 2005) at high spatial resolution to probe the central few 10th of arcseconds.1 Another way to start discriminating between these models would be if ionized gas is detected at larger radii (beyond a few arcseconds) where the circular velocity profile is assumed to be dominated by the stellar component and where the actual mass-to-light ratio would have strong predictive power.
Finally, we would like to emphasize the resemblance between a galaxy like NGC 1277 and NGC 4342, which had been studied by Cretton & van den Bosch (1999) who claimed for the presence of supermassive black hole which represents about 3 per cent of the total spheroid mass, one of the largest black hole mass to bulge mass known at the time. NGC 4342 is a close to edge-on fast rotator with a prominent inner disc and stellar kinematics which show similar radial major-axis profiles (though with different amplitudes), specifically with an extended dispersion profile (van den Bosch, Jaffe & van der Marel 1998). The derived best-fitting model has anisotropy profiles (see fig. 13 of Cretton & van den Bosch 1999) which are very similar to the ones derived for NGC 1277 here (Fig. 5) with σθ being lower in the outer part than the two other components of the velocity ellipsoid, while going up in the central region above σR, and σR/σz being rather constant around 0.9. It is not surprising to note that, according to the model of NGC 4342 presented in Cretton & van den Bosch (1999), the rise of σθ in the inner part corresponds to the observed rise in the stellar velocity dispersion. Cretton & van den Bosch (1999) also noted that the mass-to-light ratio derived from their best-fitting model is large relatively to other early-type galaxies. NGC 1277 can thus roughly appear as a scaled-up version of NGC 4342 (factor of about 8 in mass). Interestingly, the potential presence of an end-on bar was discussed for NGC 4342 by Cretton & van den Bosch (1999), to acknowledge the similarities with another fast rotator, NGC 4570, where evidence for bar-driven secular evolution (van den Bosch & Emsellem 1998) was found. Cretton & van den Bosch (1999) considered the hypothesis of an end-on bar in NGC 4342 unlikely, reflecting on the fact that the end-on view would be rather constrained, and because the cuspy luminosity profile of NGC 4342 would probably prevent the formation of a bar in the first place (see the brief discussion in their section 7.2.2). These two arguments could also apply to NGC 1277 and NGC 4570, while it is worth emphasizing that the latter is also very cuspy and there are evidences for NGC 4570 to have been shaped by the presence of a bar. One common denominator of both NGC 4342 and NGC 4570 is the presence of a thin disc at a scale of about 100 pc, and NGC 4570 harbours a relatively young (∼2 Gyr old) and very thin stellar ring at a radius of 140 pc. In NGC 1277, there is a clear ring-like dust structure at a radius of about 0.9 arcsec (∼320 pc, at a distance of 73 Mpc), blocking part of the view of the very central structure. We cannot comment further to whether or not these structures are various stages of an evolution process (see also the beautiful and puzzling case of NGC 1332, presented in Rusli et al. 2011), but while searching for massive black holes in fast rotators with similar morphologies (and inclination), it would be worth keeping these facts in mind.
I would like to warmly thank Davor Krajnović for his helpful remarks on an early version of the manuscript, as well as Michele Cappellari, and Glenn van de Ven for their input. I would like to specifically acknowledge very valuable feedback from Remco van den Bosch, including on the details of the Schwarzschild models presented in vdM+12, and thank him for the open and interesting discussions we had. I wish to thank the anonymous referee for very constructive comments.
Gemini/Near-Infrared Integral Field Spectrometer (NIFS) data has now been obtained (PI: Richstone), and should help tremendously in this context.