The Astrophysical Journal, 528:L13–L16, 2000 January 1

q 2000. The American Astronomical Society. All rights reserved. Printed in U.S.A.

VIEWING THE SHADOW OF THE BLACK HOLE AT THE GALACTIC CENTER

Heino Falcke,1 Fulvio Melia,2,3 and Eric Agol4
Received 1999 September 14; accepted 1999 October 26; published 1999 December 7

ABSTRACT
In recent years, evidence for the existence of an ultracompact concentration of dark mass associated with the
radio source Sagittarius A* in the Galactic center has become very strong. However, unambiguous proof that
this object is indeed a black hole is still lacking. A defining characteristic of a black hole is the event horizon.
To a distant observer, the event horizon casts a relatively large “shadow” with an apparent diameter of ∼10
gravitational radii that is due to the bending of light by the black hole, and this shadow is nearly independent
of the black hole spin or orientation. The predicted size (∼30 mas) of this shadow for Sgr A* approaches the
resolution of current radio interferometers. If the black hole is maximally spinning and viewed edge-on, then the
shadow will be offset by ∼8 mas from the center of mass and will be slightly flattened on one side. Taking into
account the scatter broadening of the image in the interstellar medium and the finite achievable telescope resolution,
we show that the shadow of Sgr A* may be observable with very long baseline interferometry at submillimeter
wavelengths, assuming that the accretion flow is optically thin in this region of the spectrum. Hence, there exists
a realistic expectation of imaging the event horizon of a black hole within the next few years.
Subject headings: black hole physics — galaxies: active — Galaxy: center — relativity — submillimeter —
techniques: interferometric

1. INTRODUCTION The key spectral features of Sgr A* are a slightly inverted

centimeter-wavelength spectrum, an apparent excess (or bump)
High-resolution spectroscopy (especially with the Hubble at submillimeter wavelengths, and a steep cutoff toward the
Space Telescope) of galactic nuclei has produced an abundance
infrared (Falcke et al. 1998; Serabyn et al. 1997). The radio
of evidence for compact dark mass concentrations of up to 108
emission is circularly polarized but undetected in linear polar-
M, pc23, whose nature is strongly suspected to be indicative
ization (Bower et al. 1999a, 1999b). Proposed models for the
of supermassive black holes (Kormendy & Richstone 1995).
radio emission range from quasi-spherical inflows (Melia 1992,
Even better evidence exists for the galaxy NGC 4258 and the
1994; Narayan, Yi, & Mahadevan 1995) to a jetlike outflow
Milky Way, for which spectroscopic and proper-motion studies
(Falcke, Mannheim, & Biermann 1993; Falcke & Biermann
have provided an unprecedented three-dimensional view of the
1999).
kinematics of gas and stars around a central point mass, point-
The submillimeter bump is particularly interesting since this
ing to dark mass concentrations of greater than 1012 M, pc23
should be the signature of a very compact synchrotron-emitting
with very high significance (Miyoshi et al. 1995; Eckart &
region with a size of a few Schwarzschild radii (Falcke 1996;
Genzel 1996; Lo et al. 1998).
Complementary observations of galactic nuclei with very Falcke et al. 1998). The presence of compact radio emission
long baseline interferometry (VLBI) reveal the presence of in Sgr A* at a wavelength as short as 1.4 mm has been con-
compact radio cores (Zensus 1997) that appear to be coincident firmed recently by the first VLBI detection at this wavelength
with the central black hole candidates. An intriguing case is (Krichbaum et al. 1998). This detection is exciting for several
that of the Galactic center where the bright, compact radio reasons. First, it lies in a region of the spectrum where the
source Sagittarius A* lies at the dynamical origin (Menten et intrinsic source size should become apparent over scatter broad-
al. 1997; Ghez et al. 1998). The nature of Sgr A* is still unclear ening by the intervening screen (Melia, Jokipii, & Narayanan
since its structure is completely washed out by strong inter- 1992). Second, this component is sufficiently bright to be de-
stellar scattering at centimeter wavelengths (Lo et al. 1998). It tected with VLBI techniques at even shorter wavelengths, and
is only at millimeter wavelengths that we may begin to see third, Sgr A* is sufficiently close that the size scale where
some internal structure (Bower & Backer 1998; Krichbaum et general relativistic effects are significant could be resolved with
al. 1998; Lo et al. 1998). Although the dark mass concentration VLBI at submillimeter wavelengths. In addition, at submilli-
could in principle be distributed in the form of exotic objects meter wavelengths, the various models predict that the syn-
on a scale slightly larger than the size of Sgr A*—but with chrotron emission is not self-absorbed, allowing a view into
difficulties accounting for its radiation characteristics (Melia & the region near the horizon. The horizon has a size of [1 1
Coker 1999)—it is expected to be associated with Sgr A* itself (1 2 a∗2 )1/2 ] Rg, where Rg { GM/c 2, M is the mass of the black
since the latter, unlike the surrounding stars, has a tightly re- hole, G is Newton’s constant, c is the speed of light, a∗ {
stricted proper motion indicating that it is very heavy (Reid et Jc/(GM 2 ) is the dimensionless spin of the black hole in the
al. 1999; Genzel et al. 1997). range of 0–1, and J is the angular momentum of the black
hole.
1
Max-Planck-Institut für Radioastronomie, Auf dem Hügel 69, Bonn, Bardeen (1973) described the idealized appearance of a black
D-53121, Germany.
2
hole in front of a planar-emitting source, showing that it literally
Department of Physics and Steward Observatory, 933 North Cherry would appear as a “black hole.” At that time, such a calculation
Avenue, University of Arizona, Tucson, AZ 85721-0065.
3
Presidential Young Investigator and Sir Thomas Lyle Fellow.
was of mere theoretical interest and limited to calculating just
4
Henry A. Rowland Department Physics and Astronomy, Johns Hopkins the envelope of the apparent black hole. To test whether there
University, Baltimore, MD 21218. is a realistic chance of seeing this black hole in Sgr A* (Falcke
L13
L14 BLACK HOLE IMAGING Vol. 528

et al. 1998), we here report the first calculations obtained with Lindquist radial velocity, Q is the orbital frequency, D {
our general relativistic (GR) ray-tracing code that allows us to r 2 2 2r 1 a 2, and A { (r 2 1 a 2 ) 2 2 a 2D sin2 v. (We have set
simulate observed images of Sgr A* for various combinations G = M = c = 1 in this paragraph.) The second has the plasma
of the black hole spin, inclination angle, and morphology of orbiting in rigidly rotating shells with the equatorial Keplerian
the emission region directly surrounding the black hole and not frequency Q = 1/(r 3/2 1 a) for r 1 rms with vr = 0 and infalling
just for a background source. A more detailed description with constant angular momentum inside r ! rms (Cunningham
of our calculations is in preparation (E. Agol, H. Falcke, & 1975) with vv = 0 for all r.
M. Melia 1999, in preparation). In order to display concrete examples of how realistic our
proposed measurements of these effects with VLBI will be, we
2. THE APPEARANCE OF A BLACK HOLE have simulated the expected images for the massive black hole
candidate Sgr A* at the Galactic center. For its measured mass
We determine the appearance of the emitting region around (Eckart & Genzel 1996; Ghez et al. 1998) M = 2.6 #
a black hole under the condition that it is optically thin. For 10 6 M,, the scale size for this object is the gravitational radius
Sgr A*, this might be the case for the submillimeter bump Rg = 3.9 # 10 11 cm, which is half of the Schwarzschild radius
(Falcke et al. 1998) indicated by the turnover in the spectrum R S { 2GM/c 2.
and can always be achieved by going to a suitably high fre- To simulate an observed image, we have to take two addi-
quency. Here we simply assume that the overall specific in- tional effects into account: interstellar scattering and the finite
tensity In observed at infinity is an integration of the emissivity telescope resolution achievable from the ground. Scatter broad-
jn times the differential path length along geodesics (Jaroszynski ening at the Galactic center is incorporated by smoothing
& Kurpiewski 1997). In line with the qualitative discussion of the image with an elliptical Gaussian with a FWHM of
this Letter, we assume that jn is independent of frequency and 24.2 mas # (l/1.3 mm)2 along the major axis and 12.8 mas #
that it is either spatially uniform or scales as r22. These two (l/1.3 mm)2 along the minor axis (Lo et al. 1998). The position
cases cover a large range of conditions expected under several angle of this ellipse is arbitrary since we do not know yet the
reasonable scenarios, be it a quasi-spherical infall, a rotating spin axis of the black hole on the sky, and we have assumed
thick disk, or the base of an outflow. a position angle of 907 for the major axis. The telescope res-
The calculation of the photon trajectories and the intensity olution—in an idealized form—is then added by convolving
integrated along the line of sight is based on the standard for- the smoothed image with a spherical Gaussian point-spread
malism (Thorne 1981; Viergutz 1993; Rauch & Blandford function of FWHM = 33.5 mas # (l/1.3 mm)21(l/8000 km)21,
1994). Our calculations take into account all the well-known which is the possible resolution of a global interferometer with
relativistic effects, e.g., frame dragging, gravitational redshift, 8000 km baselines (Krichbaum 1996). In reality, the exact
light bending, and Doppler boosting. The code is valid for all point-spread function will of course depend on the number and
possible spins of the black hole and for any arbitrary velocity placement of the participating telescopes.
field of the emission region. In Figure 1, we show the resulting image of Sgr A* for a
For a planar-emitting source behind a black hole, a closed maximally rotating black hole viewed at an angle of i = 457,
curve on the sky plane divides a region where geodesics in- for a compact region in free fall, with an emissivity of jn =
tersect the horizon from a region where geodesics miss the n 0 r22. We first show the original, unsmoothed image of the
horizon (Bardeen 1973). This curve, which we refer to as the emission region as calculated with the GR code in Figure 1a
“apparent boundary” of the black hole, is a circle of radius and then present the simulated “observed” images at 0.6 and
(27)1/2 Rg in the Schwarzschild case (a∗ = 0 ), but it has a more 1.3 mm wavelengths in Figures 1b and 1c, respectively. The
flattened shape of similar size to a Kerr black hole, slightly two distinct features that are evident in Figure 1a are (1) the
dependent on inclination. The size of the apparent boundary clear depression in In —the shadow—produced near the black
is much larger than the event horizon because of the strong hole, which in this particular example represents a modulation
bending of light by the black hole. When the emission occurs of up to 90% in intensity from peak to trough, and (2) the size
in an optically thin region surrounding the black hole, the case of the shadow, which here is 9.2Rg in diameter. This represents
of interest here, the apparent boundary has the same exact shape a projected size of 27 mas, which is already within a factor of
since the properties of the geodesics are independent of where 2 of the current VLBI resolution (Krichbaum et al. 1995). The
the sources are located. However, photons on geodesics located shadow is a generic feature of various other models that we
within the apparent boundary that can still escape to the ob- have looked at, including those with outflows, cylindrical
server experience strong gravitational redshift and a shorter emissivity, and various inclinations or spins.
total path length, leading to a smaller integrated emissivity, To illustrate the expected image for another extreme case,
while photons just outside the apparent boundary can orbit the we show in Figure 1d the analog to Figure 1a for the case with
black hole near the circular photon radius several times, adding a∗ = 0 (i.e., no rotation), an emitting plasma orbiting in Kep-
to the observed intensity (Jaroszynski & Kurpiewski 1997). lerian shells (as described above), and a uniform jn for r !
This produces a marked deficit of the observed intensity inside 25Rg. Even though these conditions are distinctly different com-
the apparent boundary, which we refer to as the “shadow” of pared with those of Figure 1a, the black hole shadow is still
the black hole. clearly evident; here it represents a modulation in In in the
Here we consider a compact, optically thin emitting region range of 50%–75% from peak to trough (Fig. 1d) with a di-
surrounding a black hole with spin parameter a∗ = 0 (i.e., a ameter of roughly 10.4Rg. In this case, the emission is asym-
Schwarzschild black hole) and a maximally spinning Kerr hole metric due to the strong Doppler shifts associated with the
with a∗ = 0.998. In the set of simulations shown here, we take emission by a rapidly moving plasma along the line of sight
the viewing angle i to be 457 with respect to the spin axis (with velocity vf).
(when it is present), and we consider two distributions of gas The important conclusion is that the diameter of the
velocity v. The first has the plasma in free fall, i.e., vr = shadow—in marked contrast to the event horizon—is fairly
2[2r(a 2 1 r 2 )]1/2 D/A and Q = 2ar/A, where vr is the Boyer- independent of the black hole spin and is always of order 10Rg.
No. 1, 2000 FALCKE, MELIA, & AGOL L15

Fig. 1.—Image of an optically thin emission region surrounding a black hole with the characteristics of Sgr A* at the Galactic center. The black hole is here
either (a–c) maximally rotating (a∗ = 0.998 ) or (d–f) nonrotating (a∗ = 0 ). The emitting gas is assumed to be in free fall with an emissivity proportional to r22
(top panels) or on Keplerian shells (bottom panels) with a uniform emissivity (viewing angle i = 457). (a, d) GR ray-tracing calculations; (b, e) images seen by
an idealized VLBI array at 0.6 mm wavelength, taking interstellar scattering into account; and (c, f) images seen for a wavelength of 1.3 mm. The intensity
variations along the x-axis (solid green curve) and the y-axis (dashed purple curve) are overlayed. The vertical axes show the intensity of the curves in arbitrary
units, and the horizontal axes shows the distance from the black hole in units of Rg, which, for Sgr A*, is 3.9 # 1011 cm ∼ 3 mas.

Indeed, this is consistent with the observed 0.8 mm–size limit function of wavelength and emissivity; however, the size of
being greater than 4Rg for Sgr A* owing to a lack of scintillation the shadow will remain of similar order, and under no circum-
(Gwinn et al. 1991). The presence of a rotating hole viewed stances can it become smaller.
edge-on will lead to a shifting of the apparent boundary (by The technical methods to achieve such a resolution at wave-
as much as 2.5Rg or 8 mas) with respect to the center of mass lengths shortward of 1.3 mm are currently being developed,
or the centroid of the outer emission region. and a first detection of Sgr A* at 1.4 mm with VLBI has already
Interestingly, the scattering size of Sgr A* and the resolution been reported. The challenge will be to push this technology
of global VLBI arrays become comparable to the size of the even further toward 0.8 or even 0.6 mm VLBI. Over the next
shadow at a wavelength of about 1.3 mm. As one can see from decade, many more telescopes are expected to operate at these
Figures 1c and 1f, the shadow is still almost completely washed wavelengths. Depending on how short a wavelength is required,
out for VLBI observations at 1.3 mm, while it is very apparent the projected timescale for developing the necessary VLBI
at a factor of 2 shorter wavelength (Figs. 1b and 1e). In fact, techniques may be about 10 yr. A fundamental problem pre-
already at 0.8 mm (not shown here), the shadow can be easily venting such an experiment is not now apparent, but in light
seen. Under certain conditions, i.e., a very homogeneous emis- of our results, planning of the new submillimeter telescopes
sion region, the shadow would be visible even at 1.3 mm should include sufficient provisions for VLBI experiments.
(Fig. 1f). A potential problem with our model may occur if jn has an
inner cutoff that is larger than that of the horizon, making the
3. HOW REALISTIC IS SUCH AN EXPERIMENT? shadow larger than predicted due to a decrease in emissivity
The arguments for the feasibility of such an experiment are rather than to GR effects. However, first of all, the truncation
rather compelling. First of all, the mass of Sgr A* is very well of accretion disk emission at the marginal stable orbit rms is
known within 20%, the main uncertainty being the exact dis- somewhat arbitrary (Cunningham 1975), and secondly, if it
tance to the Galactic center. Since, as we have shown, the exists, such a cutoff would likely be frequency dependent, while
unknown spin of the suspected black hole contributes only there will be a frequency-independent minimum radius due to
another 10% uncertainty, we can conservatively predict the the GR effects we have described. Another problem could be
angular diameter of the shadow in Sgr A* from the GR cal- the unknown morphology of the emission region. Anisotropy,
culations alone to be ∼30 5 7 mas, independent of wavelength. strong velocity fields, and density inhomogeneities would make
As seen in Figure 1, the finite telescope resolution and the an identification of the shadow in an observed image more
scatter broadening will make the detectability of the shadow a difficult. However, inhomogeneities are unlikely to be a major
L16 BLACK HOLE IMAGING Vol. 528

issue since the timescale for rotation around the black hole in 4. SUMMARY
the Galactic center is only a few hundred seconds and hence
much less than the typical duration of a VLBI observation. The importance of the proposed imaging of Sgr A* at sub-
The strong shear near the black hole would tend to smooth out millimeter wavelengths with VLBI cannot be overemphasized.
any inhomogeneities very quickly. Indeed, submillimeter var- The bump in the spectrum of Sgr A* strongly suggests the
iability studies on such short timescales (Gwinn et al. 1991) presence of a compact component whose proximity to the event
have yielded negative results. The same argument applies to horizon is predicted to result in a shadow of measurable di-
emission models that are offset from the black hole (e.g., one- mensions in the intensity map. To our knowledge, such a feature
sided). Since the shadow of the black hole has a very well is unique, and Sgr A* seems to have all the right parameters
defined shape, it would under any conditions appear as a distinct to make it observable. The observation of this shadow would
feature, given that the dynamic range of the map is large enough confirm the widely held belief that most of the dark mass
(i.e., *100 : 1, considering a range of emission models; E. Agol concentration in the nuclei of galaxies such as ours is contained
et al. 1999, in preparation). within a black hole, and it would be the first direct evidence
Finally, synchrotron self-absorption could pose a problem. of the existence of an event horizon. A nondetection with suf-
So far, the available submillimeter spectra show a flattening of ficiently developed techniques, on the other hand, might pose
the spectrum around 1.3–0.6 mm, indicating a turnover toward a major problem for the standard black hole paradigm. Because
an optically thin spectrum. Given the current observational of this fundamental importance, the experiment that we propose
uncertainties, one could in principle construct simple models here should be a major motivation for intensifying the current
where the flow does not become optically thin until 0.2 mm. development of submillimeter astronomy in general and mil-
Improved simultaneous measurements at submillimeter wave- limeter and submillimeter VLBI in particular.
lengths are therefore highly desirable in order to measure ex-
actly the spectral turnover since the experiment we propose
here will only work for an optically thin flow. At hundreds of We thank P. L. Biermann. T. Krichbaum, A. Zensus,
microns, the atmosphere becomes optically thick, making much O. Blaes, R. Antonucci, and M. Reid for useful discussions.
more expensive space-based observations necessary. At X-ray This work was supported in part by a Sir Thomas Lyle Fel-
wavelengths, the accretion flow will be optically thin to electron lowship (F. M.), NASA grant NAG5-8239 (F.M.), DFG grants
scattering, so there may be a better chance of detecting the Fa 358/1-1 and 2 (H. F.), and NSF grant AST 96-16922 (E. A.).
shadow with future space-based X-ray interferometry as pro- E. A. would like to thank the ITP at the University of California
posed in the MAXIM experiment. at Santa Barbara for their hospitality.
REFERENCES

Bardeen, J. M. 1973, in Black Holes, ed. C. DeWitt & B. S. DeWitt (New Krichbaum, T., Britzen, S., Standke, K., Witzel, A., Schalinski, C., & Zensus,
York: Gordon & Breach), 215 J. 1995, Proc. Natl. Acad. Sci., 92, 11,377
Bower, G. C., & Backer, D. C. 1998, ApJ, 496, L97 Krichbaum, T. P., et al. 1998, A&A, 335, L106
Bower, G. C., Backer, D. C., Zhao, J. H., Goss, M., & Falcke, H. 1999a, ApJ, Lo, K. Y., Shen, Z. Q., Zhao, J. H., & Ho, P. T. P. 1998, ApJ, 508, L61
521, 582 Melia, F. 1992, ApJ, 387, L25
Bower, G. C., Wright, M. C. H., Backer, D. C., & Falcke, H. 1999b, ApJ, in
———. 1994, ApJ, 426, 577
press
Cunningham, C. T. 1975, ApJ, 202, 788 Melia, F., & Coker, R. 1999, ApJ, 511, 750
Eckart, A., & Genzel, R. 1996, Nature, 383, 415 Melia, F., Jokipii, J. R., & Narayanan, A. 1992, ApJ, 395, L87
Falcke, H. 1996, in IAU Symp. 169, Unsolved Problems of the Milky Way, Menten, K. M., Reid, M. J., Eckart, A., & Genzel, R. 1997, ApJ, 475, L111
ed. L. Blitz & P. J. Teuben (Dordrecht: Kluwer), 169 Miyoshi, M., Moran, J., Herrnstein, J., Greenhill, L., Nakai, N., Diamond, P.,
Falcke, H., & Biermann, P. L. 1999, A&A, 342, 49 & Inoue, M. 1995, Nature, 373, 127
Falcke, H., Goss, W. M., Matsuo, H., Teuben, P., Zhao, J. H., & Zylka, R. Narayan, R., Yi, I., & Mahadevan, R. 1995, Nature, 374, 623
1998, ApJ, 499, 731 Rauch, K. P., & Blandford, R. D. 1994, ApJ, 421, 46
Falcke, H., Mannheim, K., & Biermann, P. L. 1993, A&A, 278, L1 Reid, M. J., Readhead, A. C. S., Vermeulen, R. C., & Treuhaft, R. N. 1999,
Genzel, R., Eckart, A., Ott, T., & Eisenhauer, F. 1997, MNRAS, 291, 219 in ASP Conf. Ser. 186, The Central Parsecs of the Galaxy, ed. H. Falcke,
Ghez, A. M., Klein, B. L., Morris, M., & Becklin, E. E. 1998, ApJ, 509, 678
A. Cotera, W. J. Duschl, F. Melia, & M. J. Rieke (San Francisco: ASP), 24
Gwinn, C. R., Danen, R. M., Tran, T. K., Middleditch, J., & Ozernoy, L. M.
1991, ApJ, 381, L43 Serabyn, E., Carlstrom, J., Lay, O., Lis, D. C., Hunter, T. R., & Lacy, J. H.
Jaroszynski, M., & Kurpiewski, A. 1997, A&A, 326, 419 1997, ApJ, 490, L77
Kormendy, J., & Richstone, D. 1995, ARA&A, 33, 581 Thorne, K. S. 1981, MNRAS, 194, 439
Krichbaum, T. 1996, in Science with Large Millimetre Arrays, ed. P. A. Shaver Viergutz, S. U. 1993, A&A, 272, 355
(Berlin: Springer), 96 Zensus, J. A. 1997, ARA&A, 35, 607