Astronomy & Astrophysics manuscript no.

SabhaGC2˙Pub c ESO 2018

May 24, 2018

The S-Star Cluster at the Center of the Milky Way ⋆

On the nature of diffuse NIR emission in the inner tenth of a parsec

N. Sabha1 , A. Eckart1,2 , D. Merritt3 , M. Zamaninasab2 , G. Witzel1 , M. Garcı́a-Marı́n1 , B. Jalali1 ,
M. Valencia-S.1,2 , S. Yazici1 , R. Buchholz1 , B. Shahzamanian1,2 , C. Rauch1 , M. Horrobin1 , and
C. Straubmeier1

1
I.Physikalisches Institut, Universität zu Köln, Zülpicher Str.77, 50937 Köln, Germany
e-mail: [email protected]
arXiv:1203.2625v2 [astro-ph.GA] 4 Jan 2013

2
Max-Planck-Institut für Radioastronomie, Auf dem Hügel 69, 53121 Bonn, Germany
3
Department of Physics and Center for Computational Relativity and Gravitation, Rochester Institute of Technology,
Rochester, NY 14623, USA
Received: 9 March 2012/ Accepted: 4 June 2012

ABSTRACT

Sagittarius A*, the super-massive black hole at the center of the Milky Way, is surrounded by a small cluster of high
velocity stars, known as the S-stars. We aim to constrain the amount and nature of stellar and dark mass associated
with the cluster in the immediate vicinity of Sagittarius A*. We use near-infrared imaging to determine the Ks -band
luminosity function of the S-star cluster members, and the distribution of the diffuse background emission and the stellar
number density counts around the central black hole. This allows us to determine the stellar light and mass contribution
expected from the faint members of the cluster. We then use post-Newtonian N-body techniques to investigate the effect
of stellar perturbations on the motion of S2, as a means of detecting the number and masses of the perturbers. We find
that the stellar mass derived from the Ks -band luminosity extrapolation is much smaller than the amount of mass that
might be present considering the uncertainties in the orbital motion of the star S2. Also the amount of light from the
fainter S-cluster members is below the amount of residual light at the position of the S-star cluster after removing the
bright cluster members. If the distribution of stars and stellar remnants is strongly enough peaked near Sagittarius A*,
observed changes in the orbital elements of S2 can be used to constrain both their masses and numbers. Based on
simulations of the cluster of high velocity stars we find that at a wavelength of 2.2 µm close to the confusion level for
8 m class telescopes blend stars will occur (preferentially near the position of Sagittarius A*) that last for typically 3
years before they dissolve due to proper motions.
Key words. Galaxy: center - infrared: general - infrared: diffuse background - stars: luminosity function, mass function
- stars: kinematics and dynamics - methods: numerical

1. Introduction ∼15.9 years, and was the star used to precisely deter-
mine the enclosed dark mass, and infer the existence of
Using 8–10 m class telescopes, equipped with adaptive op- a ∼4 million solar mass SMBH, in our own Galactic cen-
tics (AO) systems, at near-infrared (NIR) wavelengths has ter (GC; Schödel et al. 2002; Ghez et al. 2003). The first
allowed us to identify and study the closest stars in the spectroscopic studies of S2, by Ghez et al. (2003) and later
vicinity of the super-massive black hole (SMBH) at the Eisenhauer et al. (2005), revealed its rotational velocity to
center of our Milky Way. These stars, referred to as the be that of an O8-B0 young dwarf, with a mass of 15 M⊙
S-star cluster, are located within the innermost arcsecond, and an age of less than 106 yrs. Later, Martins et al. (2008)
orbiting the SMBH, Sagittarius A*(Sgr A*), on highly ec- confined the spectral type of S2 to be a B0–2.5 V main-
centric and inclined orbits. Up till now, the trajectories sequence star with a zero-age main-sequence (ZAMS) mass
of about 20 stars have been precisely determined using of 19.5 M⊙ . The fact that S2, along with most of the S-
NIR imaging and spectroscopy (Gillessen et al. 2009a,b). stars, is classified as typical solar neighborhood B2–9 V
This orbital information is used to determine the mass of stars, indicates that they are young, with ages between 6–
the SMBH and can in principle be used to detect rela- 400 Myr (Eisenhauer et al. 2005). The combination of their
tivistic effects and/or the mass distribution of the central age and the proximity to Sgr A* presents a challenge to star
stellar cluster (Rubilar & Eckart 2001; Zucker et al. 2006; formation theories. It is still unclear how the S-stars were
Mouawad et al. 2005; Gillessen et al. 2009a). formed. Being generated locally requires that their forma-
One of the brightest members of that cluster is the tion must have occurred through non-standard processes,
star S2. It has the shortest observed orbital period of like formation in at least one gaseous disk (Löckmann et al.
2009) or via an eccentricity instability of stellar disks
⋆
Based on observations collected at the European around SMBHs (Madigan et al. 2009). Alternatively, if they
Organisation for Astronomical Research in the Southern formed outside the central star cluster, about 0.3 par-
Hemisphere, Chile (ProgId: 073.B-0085)

1
 Sabha, Eckart, Merritt, Zamaninasab et al.: The S-star Cluster

Fig. 1. Left: 17.5′′ × 17.5′′ NACO Ks -band mosaic of the central cluster zoomed in to the inner 1–2′′ region around Sgr A*. The
inner region is indicated by a dashed (white) circle. Right: a) Map of the diffuse background light within a circle of 0.69′′ radius
centered on the position of Sgr A*, shown here as a cross at the center. The projected orbit of the star S2 is over-plotted as an
ellipse. b) The same map smoothed by convolution with a Gaussian beam of FWHM = 6 pixels. The contours levels are at 95%,
90%, 80%, 70%, 60%, 50%, 40%, 30%, 20% and 10% of the maximum flux value for each image.

sec core radius (e.g. Buchholz et al. 2009; Schödel et al. follow a steeper profile. Similar results were obtained later
2007), there are several models that describe how they may by Do et al. (2009) and Bartko et al. (2010).
have been brought in (e.g. Hansen & Milosavljević 2003; These surprising findings required new models to ex-
Kim et al. 2004; Levin et al. 2005; Fujii et al. 2009, 2010; plain the depletion in the number of late-type giants in
Merritt et al. 2009; Gould & Quillen 2003; Perets et al. the central few arcseconds around the SMBH. Such at-
2007, 2009). For a detailed description of these processes tempts involved Smooth Particle Hydrodynamics (SPH)
see Perets & Gualandris (2010). and Monte Carlo simulations which tried to account for
Stellar dynamics predict the formation of a cusp of the under density of giants by means of collisions with other
stars at the center of a relaxed stellar cluster around stars and stellar remnants (Dale et al. 2009; Freitag 2008).
a SMBH. This is manifested by an increase in the Another explanation could be the disturbance of the cusp
three dimensional stellar density of old stars and rem- of stars after experiencing a minor merger event or an in-
nants towards the center with power-law slopes of spiraling of an intermediate-mass black hole, which then
1.5 to 1.75 (Bahcall & Wolf 1976; Murphy et al. 1991; would lead to deviations from equilibrium; hence causing a
Lightman & Shapiro 1977; Alexander & Hopman 2009). shallower power-law profile of the cusp (Baumgardt et al.
The steep power-law slope of 1.75 is reached in the case 2006). Merritt (2010) explains the observations by the evo-
of a spherically symmetric single mass stellar distribution lution of a parsec-scale initial core model.
in equilibrium. For a cluster with differing mass compo- Mouawad et al. (2005) presented the first efforts to de-
sition, mass segregation sets in, where the more massive termine the amount of extended mass in the vicinity of
stars sink towards the center, while the less massive ones the SMBH allowing for non-Keplerian orbits. Using posi-
remain less concentrated. This leads to the shallow den- tional and radial velocity data of the star S2, and leav-
sity distribution of 1.5 (Bahcall & Wolf 1977). Later nu- ing the position of Sgr A* as a free input parameter, they
merical simulations and analytical models confirmed these provide, for the first time, a rigid upper limit on the pres-
results (Freitag et al. 2006; Preto & Amaro-Seoane 2010; ence of a possible extended dark mass component around
Hopman & Alexander 2006b). These steep density distri- Sgr A*. Considering only the fraction of the cusp mass
butions were expected for the central cluster considering MS2apo that may be within the apo-center of the S2 orbit,
its age, which is comparable to the estimates of the two- Mouawad et al. (2005) find MS2apo /(MSMBH + MS2apo ) ≤
body relaxation-time of 1–20 Gyr for the central parsec 0.05 as an upper limit. This number is consistent with
(Alexander 2005; Merritt 2010; Kocsis & Tremaine 2011). more recent investigations of the problem (Gillessen et al.
However, observations of the projected stellar number den- 2009b). Due to mass segregation, a large extended mass
sity, which can be related to the three dimensional density in the immediate vicinity of Sgr A*, if present, is unlikely
distribution, revealed that the cluster’s radial profile can be to be dominated in mass of sub-solar mass constituents. It
fitted by two power-law slopes. The slope for the whole clus- could well be explained by a cluster of high M/L stellar
ter outside a radius of ∼ 6′′ (corresponding to 0.22 parsec) remnants, which may form a stable configuration.
was found to be as steep as 1.8 ± 0.1, while inside the break From the observational point of view, several at-
radius the slope was shallower than expected and reached tempts have been made recently to tackle the missing
an exponent of 1.3 ± 0.1 (Genzel et al. 2003; Schödel et al. cusp problem. Sazonov et al. (2011) proposed that the de-
2007). These findings motivated the need to derive the den- tected 1′′ sized thermal X-ray emission close to Sgr A*
sity profiles of the distinct stellar populations, given that re- (Baganoff et al. 2001, 2003) can be explained by the tidal
cent star formation (6 Myr, Paumard et al. 2006) at the GC spin-ups of several thousand late-type main-sequence stars
gave birth to a large number of high-mass young stars that (MS). They use the Chandra X-ray data to infer an up-
would be too young to reach an equilibrium state. Using per limit on the density of these low-mass main-sequence
adaptive optics and intermediate-band spectrophotometry stars. Furthermore, using Hubble Space Telescope (HST)
Buchholz et al. (2009) found the distribution of late-type data, Yusef-Zadeh et al. (2012) derived a stellar mass pro-
stars (K giants and later) to be very flat and even showing file, from the diffuse light profile in the region < 1′′ around
a decline towards the Center (for a radius of less that 6′′ ), Sgr A*, and by that they explained the diffuse light to be
while the early-type stars (B2 main-sequence and earlier) dominated by a cusp of faint K0 dwarfs.

2
 Sabha, Eckart, Merritt, Zamaninasab et al.: The S-star Cluster

Up to now, the true distribution of the Nuclear Star 3. The central few tenths of parsecs
Cluster, especially the S-stars, is yet to be determined. No
investigations have confirmed or ruled out the existence of In NS10 we gave a stringent upper limit on the emission
a cusp of relaxed stars and stellar remnants around Sgr A*, from the central black hole in the presence of the sur-
as predicted by theory. An excellent dataset to investigate rounding S-star cluster. For that purpose, three indepen-
the stellar content of the central arcsecond around Sgr A* is dent methods were used to remove or strongly suppress
the NIR Ks -band (2.2 µm) data (see Figure 1) we used in the flux density contributions of these stars, in the central
(Sabha et al. 2010, hereafter NS10). In that case we sub- ∼ 2′′ , in order to measure the flux density at the posi-
tracted the stellar light contribution to the flux density tion of Sgr A*. All three methods provided comparable re-
measured at the position of Sgr A*. The aim of this work sults, and allowed a clear determination of the stellar light
is then to analyze the resulting image of the diffuse NIR background at the center of the Milky Way, against which
background emission close to the SMBH. This emission Sgr A* has to be detected. The three methods, linear ex-
is believed to trace the accumulative light of unresolved traction of the extended flux density, automatic and itera-
stars (Schödel et al. 2007; Yusef-Zadeh et al. 2012). We ex- tive point spread function (PSF) subtraction were carried
plain the background light by extrapolating the Ks -band out assuming that the extracted PSF in the central few
luminosity function (KLF) of the innermost (1–2′′ , corre- arcseconds of the image is uniform. Investigations of larger
sponding to 0.05 parsecs for a distance of 8 kpc to the GC) images (e.g. Buchholz et al. 2009) show that on scales of a
members of the S-star cluster to fainter Ks -magnitudes. We few arcseconds the constant PSF assumption is valid, while
compare the cumulative light and mass of these fainter stars for fields ≥ 10′′ the PSF variations have to be taken into
to the limits imposed by observations. We then extend our account.
analysis to explore the possible nature of this background
light by testing its effect on the observed orbit of the star Figure 2 is a map of the 51 stars adopted from the list
S2. Furthermore, we simulate the distribution of the unre- in Table 3 of NS10. The stars are plotted relative to the
solved faint stars (Ks > 18) and their combined light to position of Sgr A*. The surface number density of these
produce line-of-sight clusterings that have a compact, close detected stars, within a radial distance of about 0.5′′ from
to stellar, appearance. Sgr A*, is 68±8 arcsec−2 , with the uncertainty correspond-
The paper is structured as follows: Section 2 deals with ing to the square-root of that value. This value agrees with
a brief description of the observation and data reduction. the central number density of 60 ± 10 arcsec−2 given by
We describe in Section 3 the method used (§ 3.1–§ 3.3) and Do et al. (2009). Extrapolating the KLF allows us to test if
discuss the different observational limits (§ 3.4) employed the observed diffuse light across the central S-star cluster,
to test our analysis. Exploring the possible contributors to or the amount of unaccounted dark mass, can be explained
the dark mass within the orbit of S2 is done in Section 4. In by stars.
Section 5 we give the results obtained by simulating the dis-
tribution of faint stars and the possibility of producing line
of sight clusterings that look like compact stellar objects.
We summarize and discuss the implications of our results
in Section 6. We adopt throughout this paper Σ(R) ∝ R−Γ
as the definition for the projected density distribution of
the background light , with R being the projected radius
and Γ the corresponding power-law index.

2. Observations and data reduction

The observations and data reduction have been described in
NS10. In summary: The near-infrared (NIR) observations
have been conducted at the Very Large Telescope (VLT)
of the European Southern Observatory (ESO) on Paranal,
Chile. The data were obtained with YEPUN, using the
adaptive optics (AO) module NAOS and the NIR cam-
era/spectrometer CONICA (briefly “NACO”). The data
were taken in the Ks -band (2.2 µm) on the night of 23
September 2004, and is one of the best available where
Sgr A* is in a quiet state. The flux densities were mea-
sured by aperture photometry with circular apertures of
66 mas radius. They were corrected for extinction, us-
ing AKs = 2.46 derived for the inner arcsecond from
Schödel et al. (2010). Possible uncertainties in the extinc-
tion of a few tenths of a magnitude do not influence the gen- Fig. 2. Map of the 51 stars listed in Table 3 from NS10. The
eral results obtained in this paper. The flux density calibra- color of each star indicates its Ks -magnitude. The size of each
tion was carried out using zero points for the correspond- symbol is proportional to the flux of the corresponding star. The
ing camera setup and a comparison to known Ks -band flux position of Sgr A* is indicated as a cross at the center.
densities of IRS16C, IRS16NE (from Schödel et al. 2010;
also Blum et al. 1996) and to a number of the S-stars
(Witzel et al. 2012).

3
 Sabha, Eckart, Merritt, Zamaninasab et al.: The S-star Cluster

of stars per bin. Their method is applicable to small samples

of stars. We apply their method to our KLF calculation and
get d(log(N )/δKs )/d(Ks ) = 0.12±0.09, consistent with our
fixed sized binning method.

3.2. The diffuse NIR background

The methods we used in NS10 to correct for the flux density
contribution of the stars in the central 2′′ have revealed a
faint extended emission around Sgr A* (NS10 Figures 3b,
4b and 5). We detected ∼ 1.3 mJy (obtained by correct-
ing the ∼ 2 mJy we quote in NS10 for the AKs = 2.46
we use here) at the center of the S-star cluster. With a ra-
dius of 1′′ (about twice the FWHM of the S-star cluster)
for the Point Spread Function (PSF) used for the subtrac-
tion, we showed that a misplacement of the PSF for about
only five stars, located within one FWHM of Sgr A*, would
contribute significantly to the measured flux at the center.
For a median brightness of about 1.3 mJy for these stars, a
1 pixel ∼ 13 mas positional shift of each of these stars to-
wards Sgr A* would be required to explain all the detected
∼ 1.3 mJy at the center i.e. 0.26 mJy from each star. In
Fig. 3. KLF histogram of the stars detected in the central field, Sabha et al. (2011) we showed that a displacement larger
derived from the 23 September 2004 data. The dashed line in- than a few tenths of a pixel would result in a clear and
dicates the linear fit of the KLF slope of α = 0.18 ± 0.07. The
vertical dotted line (red) represents the current detection limit
identifiable characteristic plus/minus pattern in the resid-
for faint Ks -magnitudes. ual flux distribution along the shift direction. For a maxi-
mum positional uncertainty of 1 pixel, we showed that the
independent shifts of the five stars can be approximated
3.1. KLF of the S-star cluster by a single star experiencing five shifts in a random walk
pattern. This resulted in calculating a total maximum con-
Figure 3 shows the KLF histogram derived for the stars tribution of 0.26 mJy from all the five stars to the center,
detected in the central field, (Figure 2). We improve the which translates to about 20–30% of the flux density. Thus,
KLF derivation by choosing a fixed number of bins that more than two thirds of the extended emission detected to-
allows for about 10 sources per bin while providing a suf- wards Sgr A* could be due to faint stars, at or beyond the
ficient number of points to allow for a clear linear fit. The completeness limit reached in the KLF, and associated with
Red Clump (RC)/Horizontal Branch (HB) stars, around the ∼ 0.5–1′′ diameter S-star cluster.
Ks ≈ 14.5, are in one bin, so the RC/HB bump is visi- The diffuse background emission we detected (see
ble there (Schödel et al. 2007). For estimating the uncer- Figure 1a) could be compared to the projected distribution
tainty, we randomized the start of the first bin in an in- of stars Σ(R) ∝ R−Γ , with R being the projected radius.
terval between Ks = 13.0 to 14.2 and repeated the his- We found that the distribution of the azimuthally averaged
togram calculation 105 times. The number of sources in residual diffuse background emission, centered on the po-
each bin was then determined by taking the average of all sition of Sgr A*, not to be uniform but in fact decreases
iterations and the uncertainties were subsequently derived gently as a function of radius (see Figure 7 in NS10) with
from the standard deviation. We derive a least-square linear a power-law index Γdiffuse = 0.20 ± 0.05. In this investiga-
slope of d log(N )/d(Ks ) = α = 0.18 ± 0.07, which compares tion we re-calculate the azimuthally averaged background
well with the KLF slope of 0.3 ± 0.1 derived in NS10 and light from the iterative PSF subtracted image alone. The
also with the KLF slope of 0.21 ± 0.02 found for the inner azimuthally averaged background light is plotted as a func-
field (R < 6′′ ) by Buchholz et al. (2009). For the magni- tion of projected radius from Sgr A* in Figure 4. In this new
tudes up to Ks = 17.50 within the central 0.69′′ radius, calculation we find the power-law index to have a value of
we detect no significant deviation from a straight power- Γdiffuse = 0.14 ± 0.07. Both results are consistent with re-
law. This implies that the completeness is high and can be cent investigations concerning the distribution of number
compared to the ∼70% value derived for magK = 17 by density counts of the stellar populations in the central arc-
Schödel et al. (2007) where the authors introduced artifi- seconds, derived from imaging VLT and Keck data. For
cial stars into their NIR image and attempted re-detecting the central few arcseconds Buchholz et al. (2009), Do et al.
them. However, for Ks = 17.50 to 18.25 the stellar counts (2009) and Bartko et al. (2010) find a Γ ∼ 1.5 ± 0.2 for
drop quickly to about 20% of the value expected from the the young stars, but an even shallower distribution for the
straight power-law line; hence the last Ks -bin is excluded late-type (old) stars with Γ ∼ 0.2 ± 0.1. A detailed discus-
from the linear fit. sion concerning the different populations and their distribu-
Maı́z Apellániz & Úbeda (2005) propose an alternative tion is given in Genzel et al. (2003); Schödel et al. (2007);
way of binning when dealing with stellar luminosity and ini- Buchholz et al. (2009); Do et al. (2009) and Bartko et al.
tial mass functions (IMF). Their method is based on choos- (2010).
ing variable sized bins with a constant number of stars in The small value we obtain for the projected diffuse
each bin. They find that variable sized binning introduces light exponent Γdiffuse and the high degree of completeness
bias-free estimations that are independent from the number reached around Ks = 17.5, makes this data set well suited

4
 Sabha, Eckart, Merritt, Zamaninasab et al.: The S-star Cluster

for analyzing the diffuse background light. Especially in in- between stars, the heavier objects sink towards the center
vestigating the role of much fainter stars, beyond the com- while the lighter objects move out. Their volume density
pleteness limit, in the observed power-law behavior of the will be significantly reduced and they may even be expelled
background. from the very center. Freitag et al. (2006) show that the
main-sequence stars begin to be expelled outward by the
cusp of stellar-mass black holes (SBH) after a few Gyrs,
just shorter than the presumed age of the stellar cluster
at about 10 Gyrs. While the reservoir of lower mass stars
may be replenished by the most recent - possibly still on-
going - star formation episode about 6 million years ago
(Paumard et al. 2006), we assume that stars well below our
low mass limit of ∼ 0.34 M⊙ with Ks -band brightnesses
around Ks = 25 are affected by depletion.
Figure 5 shows the KLF slope of α = 0.18 and the
upper limit imposed by the uncertainty in the fit (α = 0.25)
plotted as dashed and dash-dotted lines, respectively. The
extrapolated Ks -bins are shown as hollow circles. We adopt
a Monte Carlo approach for calculating the number of stars
N from the KLF, taking into account the uncertainty in the
slope. After 105 trials we find as a result for each bin, the
median number N and median deviation dN .

Fig. 4. Azimuthal average of the diffuse background emission as

derived from manual PSF subtracted 23 September 2004 image.
The squares (mean flux and 1σ uncertainty per pixel) have been
calculated in annuli of 39.8 mas (3 pixels) width. The black
dashed line marks a fit to the data points with an exponential
decrease of 0.14.

3.3. Extrapolating the KLF of the S-star cluster

Motivated by the power-law behavior of the diffuse back-
ground emission and assuming that the drop in the KLF
counts at magnitude ∼ 18 is caused only by the fact that we
have reached the detection limit, we extrapolate the KLF to
fainter magnitudes in order to investigate how these faint
stars contribute to the background light. The true shape
of the luminosity function for Ks -magnitudes below the
completeness limit of ∼ 17.5 has yet to be determined. Fig. 5. Extrapolation of the KLF power-law fit. The KLF slope
of α = 0.18 and the upper limit imposed by the uncertainty in
Investigations into the IMF of the S-cluster have shown the fit (α = 0.25) are plotted as dashed and dash-dotted lines,
that it can be fitted with a standard Salpeter/Kroupa IMF respectively. The black filled circles represent the data while the
of dN/dm ∝ m−2.3 and continuous star formation histories hollow circles represent new points based on the extrapolated
with moderate ages (below 60 Myr, Bartko et al. 2010). KLF slope. The approximate location of the detection limit is
Here, we estimate an upper limit on the stellar light by as- indicated by the vertical dotted/red line.
suming that the KLF exhibits the same behavior observed
for brighter magnitudes without suffering a break in the
slope toward the fainter end. Using the extrapolated Ks -magnitudes, the correspond-
We use the KLF slope we obtained for the innermost ing flux densities are calculated using the following relation
central region, 0.18 ± 0.07 (Figure 3) and extrapolate it
over five magnitudes bins to Ks ∼ 25. The Ks -magnitude (Knew star −KS2 )
bins between 18–25 (translating to stellar masses in the fnew star = fS2 × 10−0.4 , (1)
range of ∼ 1.68 to 0.34 M⊙ ) correspond to the bright-
ness of the expected main-sequence stars (luminosity class where fnew star and Knew star are the flux density and
V) which are likely to be present in the central cluster. Ks -magnitude for each new star in the extrapolation. The
However, we assume that due to mass segregation effects flux and magnitude for the star S2 were adopted from NS10,
in the Galactic nucleus (Bahcall & Wolf 1976; Alexander Table 3, and corrected for the extinction value we use here
2005), driven by dynamical friction (Chandrasekhar 1943) (see § 2). The new values are fS2 = 14.73 mJy and KS2 =

5
 Sabha, Eckart, Merritt, Zamaninasab et al.: The S-star Cluster

14.1. The accumulative flux density for each Ks -bin fbin is

obtained via

fbin = fnew star × Nnew star . (2)

The number of stars per bin Nnew star is randomly picked

from the interval between [Nnew star − dNnew star ] and
[Nnew star + dNnew star ]. In 105 trials the accumulative flux
per bin and its uncertainty are determined as the median
and median deviation of the randomly drawn fluxes fbin .
We then add up all the accumulative flux densities for all
the new Ks -band bins and obtain the integrated brightness
of the extrapolated part of the S-star cluster,
25
X
FExtra Stars = fbin = (25.72 ± 14.31) mJy. (3)
Ks ≃18

We assume that the faint, undetectable stars follow the

distribution of the azimuthally averaged background light,
as shown in Figure 4. Thus, the light from the faint stars
that we introduced in the 0.69′′ radius region can be com-
pared to the measured background light from our data Fig. 6. Relative azimuthally averaged light density, for the
for the same region. This is achieved by using the total background light taken from the observations and the extra stel-
flux density FExtra Stars to derive the peak light density lar light calculated from the extrapolation, plotted as a function
(IExtra Stars ) that would be measured inside one resolution of distance from Sgr A*. They are represented by a dotted line
with circles (black) and dashed lines with squares (blue), respec-
element of 0.033′′ radius centered on the position of Sgr A*,
tively. The stellar light density is normalized to the peak light
using the following relation: density of the background at the central resolution element. The
Z upper limit of the extrapolated extra stellar light is shown as the
FExtra Stars = f (r, φ)rdrdφ blue dashed line with no symbols.

′′
Z 0.690
= 2πIExtra Stars r1−Γ dr, (4) 3.4.1. Limits on the stellar light
0.033′′
Following the previous calculations and the result displayed
with Γ = Γdiffuse = 0.14 (see § 3.2). The peak light den- in Figure 6, we perform our analysis for a range of KLF
sity for the extra stars is then IExtra Stars = (15.24 ± slopes in order to test if the observed background light can
8.48) mJy arcsec−2 . To compare the light caused by the be solely obtained by the emission of faint stars. The range
extra stars with the measured background emission, we of KLF slopes we use is based on the values and uncer-
plot the stellar light density caused by our new stars with tainty estimates of the following published KLF slopes for
the azimuthally averaged measured light density of the the central 2′′ : 0.13 ± 0.02 (Buchholz et al. 2009, early-type
background (Figure 6). For illustration purposes we nor- stars), 0.27 ± 0.03 (Buchholz et al. 2009, late-type stars),
malize the observed peak stellar light to the measured 0.21 ± 0.02 (Buchholz et al. 2009, all stars) and 0.30 ± 0.1
background value within the central resolution element, (NS10), in addition to the improved newly fitted slope of
IBackground = (254.30 ± 58.45) mJy arcsec−2. It is clear the KLF in this work 0.18 ± 0.07.
that the peak light introduced by the new faint stars, as We extrapolate each KLF slope to a Ks -magnitude of
calculated from the extrapolation of the 0.18 ± 0.07 KLF ≃ 25. The peak light density (IExtra Stars ) is calculated us-
slope, is very small and below that of the background. The ing Equation (4). The peak light density of the extra stars is
dotted line (black circles) represents the background light plotted for the extrapolated KLF slopes in the range of 0.11
while the dashed line (blue squares) corresponds to the ex- to 0.40 in Figure 7 . The limit imposed by the peak light
tra stellar light. The upper limit of the extrapolated extra density of the measured background light (Figure 1) is plot-
stellar light contribution is presented as a dashed line with ted as a horizontal dashed line (blue). In addition, the KLF
no symbols. The figure shows that the upper limit of the slopes derived in this work and by NS10 and Buchholz et al.
extrapolated light contribution of the S-star cluster is lower (2009) are plotted as purple, yellow and green data points,
than 15% of the measured background light. respectively. Figure 7 clearly shows that almost all of the
KLF slopes result in a peak light density below the ob-
served limit, except for very high slopes > 0.37 which are
3.4. Observational limits on the stellar light and mass not in agreement with the observations.
Our analysis shows that if there was a population of very
faint stars, following the extrapolated Ks -band luminosity 3.4.2. Limits on the stellar mass
function and central cluster profile obtained for the brighter
stellar population (less than Ks = 18), the additional stellar Using the same range of KLF slopes, we estimate the mass
light and mass lie well below the limits given by observa- that would be introduced to the central region as a result
tional data. See following sections and Figures 7 and 8. of the KLF extrapolation. We obtain the stellar mass cor-

6
 Sabha, Eckart, Merritt, Zamaninasab et al.: The S-star Cluster

Fig. 7. Estimated peak light density from stars derived from Fig. 8. Estimated stellar mass from the added stars for dif-
for different KLF slopes. Slopes of Buchholz et al. (2009) (for ferent KLF slopes. Slopes of Buchholz et al. (2009) (for dif-
different stellar populations) are shown in green, NS10 in yellow ferent stellar populations) are shown in green, NS10 in yellow
and the KLF slope derived here, in § 3.1, in purple. A limit and our fitted slope in purple. A limit imposed by the enclosed
imposed by the measured peak light density from the measured mass within the S2 orbit is plotted as horizontal dotted (gray)
background light is plotted as a horizontal dashed line (blue). and dash-dotted (red) lines from Gillessen et al. (2009b) and
Mouawad et al. (2005), respectively.

responding to the extrapolated Ks -bins by calculating their

luminosity via
4. Dynamical probes of the distributed mass
LKs = 10−0.4(MKs −M⊙Ks ) L⊙Ks , (5) If the gravitational force near Sgr A* includes contributions
from bodies other than the SMBH, the orbits of test stars,
where, LKs and MKs are the luminosity of a star and its ab- including S2, will deviate from Keplerian ellipses. These de-
solute magnitude in Ks -band, respectively. L⊙Ks & M⊙Ks viations can be used to constrain the amount of distributed
are the Ks luminosity and absolute magnitude of the Sun. mass near Sgr A* (Mouawad et al. 2005; Gillessen et al.
Then, the mass for each Ks magnitude is calculated using 2009b). But they can also be used to constrain the “gran-
ularity” of the perturbing potential, since the nature and
magnitude of the orbital deviations depend both on the to-
m = (LKs )(1/4) (6)
tal mass of the perturbing stars, and on their individual
masses.
from Duric (2004); Salaris & Cassisi (2005). For example,
a Ks -magnitude around 20 corresponds to 1 M⊙ main- Investigations of a single scattering event were explored
sequence stars of F0V, G0V, K5V spectral types. by Gualandris et al. (2010) using high-accuracy N-body
simulations and orbital fitting techniques. They found that
In Figure 8 we show the estimated extra mass for all
an IMBH more massive than 103 M⊙ , with a distance com-
the KLF slopes in units of solar mass. The figure also
parable to that of the S-stars, will cause perturbations of
shows, dash-dotted/red line , the upper limit for an ex-
the orbit of S2 that can be observed after the next peri-
tended mass enclosed by the orbit of the star S2, cal-
bothron1 passage of S2. Here we examine the effect many
culated by Mouawad et al. (2005), where they use non-
scatterers (i.e. smaller masses for the scatterers but shorter
Keplerian fitting of the orbit to derive the upper limits,
impact parameters) will have on the trajectory of the star
assuming that the composition of the dark mass is sources
S2 as it orbits. Around Sgr A*, the stars and scatterers are
with M/L ∼ 2. The dotted/gray line represents the tighter
moving in a potential well that is dominated by the mass of
upper limit obtained later by Gillessen et al. (2009b) who
the central SMBH. In this case the encounters are of a cor-
derive the mass using recent orbital parameters of S2. They
related nature and hence cannot be considered as random
assume that the extended mass consists of stellar black
events.
holes (Freitag et al. 2006) with a mass of 10 M⊙ using esti-
mations from Timmes et al. (1996) and Alexander (2007). An important deviation from Keplerian motion occurs
It can be concluded from the figure that the introduced as a result of relativistic corrections to the equations of
stellar mass, within a radius of ∼ 0.69′′ , lies well below the
upper limits imposed by the S2 orbit with a semi-major axis 1
Peri- or apobothron is the term used for peri- or apoapsis for
of ∼ 0.123′′ (Gillessen et al. 2009b). See Figure 1 (right) for an elliptical orbit with a black hole present at the appropriate
a comparison of the sizes of the two regions. focus.

7
 Sabha, Eckart, Merritt, Zamaninasab et al.: The S-star Cluster

motion, which to lowest order predict an advance of the orbital plane. The latter can be described in a coordinate-
argument of peribothron, ω, each orbital period of independent way via the angle ∆θ, where
6πGM• L1 · L2
(∆ω)GR = . (7) cos(∆θ) = (13)
c2 a(1
− e2 ) L1 L2
Setting a = 5.0 mpc and e = 0.88 for the semi-major axis and {L1 , L2 } are the values of L at two times separated by
and eccentricity of S2, respectively, and assuming M• = ∆t. If we set ∆t equal to the orbital period of the test star,
4.0 × 106 M⊙ , the changes in its orbital elements due to RR are expected
to be
(∆ω)GR ≈ 10.8′ . (8) √ m
|∆e|RR ≈ Ke N , (14)
The relativistic precession is prograde, and leaves the ori- M•
entation of the orbital plane unchanged. √ m
(∆θ)RR ≈ 2πKt N , (15)
The argument of peribothron also experiences an ad- M•
vance each period due to the spherically-symmetric compo-
nent of the distributed mass. The amplitude of this “mass where N is the number of stars having a-values similar to,
precession” is or less than, that of the test star and {Ke , Kt } are constants
which may depend on the properties of the field-star orbits.
Because the changes in S2’s orbit due to RR scale differ-
 
p M⋆ (r < a)
(∆ω)M = −2πGM (e, γ) 1 − e2 . (9) ently with m and N than the changes due to the smoothly-
M•
distributed mass, both the number and mass of the per-
Here, M⋆ is the distributed mass within a radius r = a, and turbing objects within S2’s orbit can in principle be inde-
GM is a dimensionless factor of order unity that depends pendently constrained. For instance, one could determine
on e and on the power-law index of the density, ρ ∝ r−γ M⋆ = mN from Equations √ (7) and (11) and a measured
(Merritt 2012). In the special case γ = 2, ∆ω, then compute m N by measuring changes in e or θ
 p −1 and comparing with Equations (14) or (15).
GM = 1 + 1 − e2 ≈ 0.68 for S2 (10) We tested the feasibility of this idea using numerical in-
tegrations. The models and methods were similar to those
so that described in Merritt et al. (2010). The N field stars were
  selected from a density profile n(r) ∝ r−2 , with semi-major
M⋆ (r < a) axes extending to amax = 8 mpc. Initial conditions assumed
(∆ω)M ≈ −1.0′ . (11)
103 M⊙ isotropy in the velocity distribution. Two values for the field
star masses were considered: m = 10 M⊙ and m = 50 M⊙ .
Mass precession is retrograde, i.e., opposite in sense to the One of the N -body particles was assigned the observed
relativistic precession. mass and orbital elements of S2; this particle was begun at
Since the contribution of relativity to the periboth- apobothron, and the integrations extended for one complete
ron advance is determined uniquely by a and e, which period of S2’s orbit. Each of the N field-star orbits were in-
are known, a measured ∆ω can be used to constrain the tegrated as well, and the integrator included the mutual
mass enclosed within S2’s orbit, by subtracting (∆ω)GR forces between stars, as well as post-Newtonian corrections
and comparing the result with Equation (11). So far, this to the equations of motion. The quantities ∆ω, ∆e etc. for
technique has yielded only upper limits on M⋆ of ∼ 10−2 M• the S2 particle were computed by applying standard formu-
(Gillessen et al. 2009b). lae to (r, v) at the start and end of each integration. 100
The granularity of the distributed mass makes itself random realizations of each initial model were integrated,
felt via the phenomenon of “resonant relaxation” (RR) allowing both the mean values of the changes, and their
(Rauch & Tremaine 1996; Hopman & Alexander 2006a). variance, to be computed.
On the time scales of interest here, orbits near Sgr A* re- Figures 9 and 10a show changes in ω for S2. The me-
main nearly fixed in their orientations, and the perturbing dian change is well predicted by Equation (11). However
effect of each field star on the motion of a test star (e.g. S2) there is a substantial variance. We identify at least two
can be approximated as a torque that is fixed in time, and sources for this variance. (1) The number of
proportional to m, the mass of the field star. The net effect √ stars inside
of the torques from N field stars is to change the angular S2’s orbit differs from model to model by ∼ N , resulting
in corresponding changes to the enclosed mass, and hence
momentum, L, of S2’s orbit according to
to the precession rate as given by Equation (11). (2) When
|∆L| √ m ∆t N is finite, the same torques that drive resonant relaxation
≈K N (12) also imply a change in the field star’s rate of peribothron
Lc M• P
√ advance as compared with Equation (11), which assumes
where Lc = GM• a is the angular momentum of a circu- no
√ tangential forces. While the dispersion scales roughly as
lar orbit having the same semi-major axis as that of the N , as evident in Figure 9, the√fractional change in ∆ω
test star. (Equation 12 describes “coherent resonant relax- due to this effect scales as ∼ 1/ N (Merritt et al. 2010).
ation”; on time scales much longer than orbital periods, Additional variance might arise from close encounters be-
“incoherent”√resonant relaxation causes changes that in- tween field stars and S2, and from the fact that the mass
crease as ∼ ∆t.) The normalizing factor K is difficult to within S2’s orbit is changing over the course of the integra-
compute from first principles but should be of order unity tion due to the orbital motion of each field star.
(Eilon et al. 2009). Changes in L imply changes in both Whereas the (average) value of ∆ω depends only on the
the eccentricity, e, of S2’s orbit, as well as changes in its mass within S2’s orbit, the changes in e and θ depend also

8
 Sabha, Eckart, Merritt, Zamaninasab et al.: The S-star Cluster

Fig. 9. Histograms of the predicted change in S2’s argument

of peribothron, ω, over the course one orbital period (∼ 16 yr).
The shift due to relativity, (∆ω)GR ≈ 11′ , has been subtracted
from the total; what remains is due to Newtonian perturbations
from the field stars. Each histogram was constructed from inte-
grations of 100 random realizations of the same initial model,
with field-star mass m = 10 M⊙ , and four different values of the
total number: N = 200 (solid/black); N = 100 (dotted/red);
N = 50 (dashed/blue); and N = 25 (dot-dashed/green). The
average value of the peribothron shift increases with increasing Fig. 10. Average values of the changes in ω, e and θ for S2
N m, as predicted by Equation (11). The reasons for the spread over one orbital period (∼ 16 yr) in the N -body integrations.
in ∆ω values are discussed in the text. Filled circles are from integrations with m = 50 M⊙ and open
circles are for m = 10 M⊙ ; the number of field stars was
N = {25, 50, 100, 200} for both values of m. The abscissa is the
on m, as shown in Figures 10b and c. The lines in those distributed mass within S2’s apobothron, at r ≈ 9.4 mpc. In
figures are Equations (14) and (15), with each frame, the points are median values from the 100 N -body
integrations, and the error bars extend from the 20th to the 80th
Ke = 1.4, Kt = 1.0. (16) percentile of the distribution. a) Changes in the argument of
peribothron. The contribution from relativity, Equation (7), has
(We have defined N in Equations (14) and (15) as the num- been subtracted. The solid line is Equation (11). b) Changes in
ber of field stars inside a radius of 9.4 mpc, the apobothron the eccentricity. Solid and dashed lines are Equation (14), with
of S2.) For a given value of the enclosed mass, M⋆ = N m, m = 50 M⊙ and m = 10 M⊙ respectively and with Ke = 1.4.
Figure c) The angle between initial and final values of L for S2. Solid
√ 10 shows that
√
the changes in e and θ indeed scale as
and dashed lines are Equations (15) with Kt = 1.0.
∼ 1/ N or as ∼ m, as predicted by Equations (14) and
(15).
We can use these results to estimate the changes in
ω, e and θ expected for S2, based on theoretical models (Hopman & Alexander 2006b) we find
of the distribution of stars and stellar remnants at the
|∆e|RR ≈ 5.4 × 10−5 , (20)
GC. In dynamically evolved models (Freitag et al. 2006;
Hopman & Alexander 2006b), the total distributed mass (∆θ)RR ≈ 0′ .8, (21)
within S2’s apobothron, r ≈ 10 mpc, is predicted to be ∼ (∆ω)M ≈ −2.5′ . (22)
a few times 103 M⊙ . About half of this mass is in the form
of main-sequence stars and half in stellar-mass black holes, For obtaining the dispersion in the value of Equation (22),
with a total number N ≈ 103 . When there are two mass we scaled the dispersion given in Figure 10a for the sin-
groups, expressions like Equations (14) and (15) generalize gle population case, N = 50, m = 50 M⊙ of the same
to total extended mass, to the two populations case we are
 √ √  investigating here. The dispersion obtained from the √ sim-
m1 N 1 + m2 N 2 ulations is ∼ 4′ . We scale it using the relation ∆ω/ N in
|∆e|RR = Ke (17)
M• order to account for the SBH and MS populations, indepen-
 √ √ 
dently. The dispersion for the new configuration then be-
m1 N 1 + m2 N 2
(∆θ)RR = 2πKt (18) comes ∼ 1.43′ , lower than the single population case. This
M•
is attributed to the fact that the number of main-sequence
assuming stars is much larger than the stellar-mass black holes, hence
they lower the dispersion in the total Newtonian periboth-
m1 = 1 M⊙ , m2 = 10 M⊙ , N1 = 103 , N2 = 150 (19) ron shift (∆ω)M .

9
 Sabha, Eckart, Merritt, Zamaninasab et al.: The S-star Cluster

Considering a higher value for the enclosed mass M⋆ = much lower density near Sgr A* and an uncertain fraction
104 M⊙ while keeping the same mass scales and abundance of stellar-mass black holes (Merritt 2010; Antonini et al.
ratios of the scattering objects, 2012). The number of perturbers is so small in these mod-
els that their effect on the orbital elements of S2 would
m1 = 1 M⊙ , m2 = 10 M⊙ , N1 = 4000, N2 = 600 (23) be undetectable for the foreseeable future, barring a lucky
close encounter with S2.
one gets changes of
In addition to the small amplitude of the perturbations,
|∆e|RR ≈ 1.1 × 10−4 , (24) the potential difficulty in constraining N and m comes from
the nonzero variance of the predicted √ changes (Figure 10).
(∆θ)RR ≈ 1.7′ , (25)
The variance in ∆ω scales as ∼ ∆ω/ N and would be small
(∆ω)M ≈ −10′ . (26) in the dynamically-relaxed models with N ≈ 103 . Another
The dispersion in Equation (26) can be compared, as source of uncertainty comes from the dependence of the
we did before, to the case considered in the simulations amplitude of ∆ω on γ (Equation 9), which is unknown. We
(N = 200, m = 50 M⊙ ) by scaling the ∼ 8′ dispersion do not have a good model for predicting the variances in
(Figure 10a) to become ∼ 2.86′ for the two mass popula- |∆e| and ∆θ, but Figure 10 suggests that the fractional
tion. variance in these quantities is not a strong function of N
Repeating the same analysis as before to the M⋆ = or m, and that it is large enough to essentially obscure
105 M⊙ gives the following numbers for the stellar black changes due to a factor ∼ 5 change in m at fixed M⋆ . On
holes and low-mass stars the other hand, considerably more information might be
available than just ∆e and ∆θ for one star; for instance,
m1 = 1 M⊙ , m2 = 10 M⊙ , N1 = 40000, N2 = 6000 (27) the full time-dependence of (r, v) for a number of stars. We
leave a detailed investigation of how well such information
that result in could constrain the perturber m and N to a future work.
|∆e|RR ≈ 3.4 × 10−4 , (28)
(∆θ)RR ≈ 5.2′ , (29) 4.1. Fighting the limits on the power of stellar orbits
(∆ω)M ≈ −100′ . (30) The results from the previous sub-sections clearly show that
deriving the net-displacement for an ideal elliptical orbit for
Similar to the above cases, the dispersion in Equation (30) a single star will not be sufficient to put firm limits on both
can be compared to the single mass case by scaling the the total amount of extended mass and on the nature of
∼ 25.3′ dispersion to become ∼ 9.1′ for the two mass the corresponding population. However, the situation may
population.
√ The ∼ 25.3′ value is obtained by scaling with be improved if one studies the statistics of the time and
∆ω/ N from the value shown in Figure 10a for the 104 M⊙ position dependent deviations along a single star’s orbit or
extended mass. instead uses the orbits of several stars.
We would like to stress that making a definite prediction
about the N -dependence of the variance is beyond the scope
of the current paper. However, we have noted that in both 4.1.1. Improving the single orbit case
cases considered in Figure 10 the relative variance is of the
The actual uncertainty in projected right ascension or dec-
order of unity or larger i.e. the dispersion is of the order of 2
lination, σposition , can be thought of as a combination of
the Newtonian peribothron shift. 2
The positional uncertainty is currently of the order of several contributions. Here σapparent is the apparent posi-
1 mas. For the highly eccentric orbit of S2 this implies that tional variation due to the photo-center variations of the
the accuracy with which the peribothron shift can be de- star while it is moving across the sea of fore- and back-
tected is of the order of 24′ . As can be seen for the case of ground sources. The scattering process results in a variation
2
M⋆ = 105 M⊙ , the shifts are at the limit of the current of positions described by σscattering . Finally, systematic un-
instrumental capabilities if the total enclosed mass was en- certainties due to establishing and applying an astrometric
2
tirely composed of massive perturbers. The shifts given in reference frame give a contribution of σsystematic .
2
Equations (22) and (26) can be measured if the accuracy The value of σposition can be measured in comparison
is improved by at least one order of magnitude using larger 2
to the orbital fit. The value of σapparent can be obtained
telescopes or interferometric methods in the NIR. However, experimentally by placing an artificial star into the imag-
considering the variances in the calculated shifts one would ing frames at positions along the idealized orbit. A reliable
need to observe more than one stellar orbit in order to infer 2
estimate of σapparent is achieved by comparing the known
information on the population giving rise to the Newtonian positions at which the star has been placed and the posi-
peribothron shift. By comparison, the current uncertainty tions measured in the image frames. As for the case of the
in S2’s eccentricity is ∼ 0.003, and uncertainties in the systematic variations, they can be estimated by investigat-
Delaunay angles i and Ω describing its orbital plane are ing sources that are significantly brighter or slower than
∼ 50′ (Gillessen et al. 2009b). In both cases, an improve- the S-stars. Finally, the value that describes the scattering
ment of a factor ∼ 50 would be required in order to detect process, and therefore gives information on the masses of
the changes given in e and θ. the scattering sources, can be obtained via
Dynamically-relaxed models of the GC have been criti-
2 2 2 2
cized on the grounds that they predict a steeply-rising den- σscattering = σposition − σapparent − σsystematic . (31)
sity of old stars inside ∼ 1 pc, while the observations show
2
a parsec-scale core (Buchholz et al. 2009; Do et al. 2009; Alternatively, σscattering could be measured directly by
Bartko et al. 2010). Dynamically unrelaxed models imply a near-infrared interferometry with long baselines. Measuring

10
 Sabha, Eckart, Merritt, Zamaninasab et al.: The S-star Cluster

the position of S2 interferometrically as a function of time bin above ∼ 18, a number of stars N get their Ks -band
with respect to bright reference objects could allow observ- magnitudes according to the KLF. From this pool of stars
ing the effects of single scattering events. Here the assump- we then randomly pick objects to fill the cells of the grid
tion is that they happen infrequently enough such that one such that they obey the power-law radial number density
can build up sufficient signal to noise on the σscattering mea- profile. Then, the fluxes of the stars in each individual cell
surement provided that the uncertainties in the interfero- are added up and compared to the value of 0.76 mJy which
metric measuring process are sufficiently well known. is the flux density of the faintest stellar source in our S-star
cluster data, i.e. Ks = 17.31 (NS10). We ran the simulation
104 times in order to get reliable statistical estimates for the
4.1.2. Improving by using several stars brightnesses in each resolution cell. Hence we can estimate
If scattering events contribute significantly to the uncer- how likely it is to find strong apparent clusterings along
tainties in the determination of the orbits, a number of stars the line of sight that are brighter than the faintest star we
may help to derive the physical properties of the medium identified in the S-cluster (flux larger than 0.76 mJy).
through which the stars are moving. While the influence
of the extended mass imposes a systematic variation of the
orbits through the Newtonian peribothron shift, the varia- Table 1. Probabilities of detecting a false star (brighter
tions due to scattering events will be random. This implies than Ks = 17.31) in a 1.38′′ × 1.38′′ region.
that for individual stars the effects may partially compen-
sate or amplify each other. Averaging the results of N stars, Ks -band Power-law index
that will then essentially sample the shape of the distribu- magnitude
tions shown in Figure 9, may therefore result in an im- cutoff 0.19 0.30 0.35
provement proportional to N −1/2 in the determination of KLF slope = 0.11
the extended mass.
20.99 0.0000 0.1471 0.1500

5. Simulating the distribution of fainter stars 24.67 0.0285 0.0292 0.0224

KLF slope = 0.18
In NS10 we detected three stars that were either previ-
ously not identified at all (NS1 & NS2 stars, Figure 1 in 20.99 0.0848 0.1286 0.3016
NS10) or only allowed an unsatisfactory identification with
previously known members of the cluster (S62, as pointed 24.67 0.2058 0.2426 0.2927
out in Dodds-Eden et al. 2011). In addition we have the KLF slope = 0.25
case of the star S3 which was identified in the Ks -band
in the early epochs 1992 (Eckart & Genzel 1996), 1995 20.99 0.7776 0.7442 0.9085
(Ghez et al. 1998) and lost after about 3 years in 1996/7
(Ghez et al. 1998), 1998 (Genzel et al. 2000). We investi- 24.67 0.9462 0.9725 0.9802
gate this phenomenon in our modeling by extrapolating the
KLF in the inner 1–2 arcsec region, surrounding Sgr A*, to
stars fainter than the faintest source (Ks = 17.31) we de- Taking into account the uncertainties of the quantities
tected in our 30 August and 23 September 2004 dataset, in that describe the central S-star cluster we have repeated the
which Sgr A* shows very low activity (NS10). In this sec- simulation for a combination of three KLF slopes (0.11, 0.18
tion we describe the method we use to simulate the distri- and 0.25), three radial profile power-law indices (Γ = 0.19,
bution of these faint stars, and the possible false detections 0.30 and 0.35) and two Ks -magnitude cutoffs for the extrap-
that can be caused by the combined light of many stars olation, 21 and 25 (corresponding to 0.0258 and 0.0009 mJy,
appearing in projection to be very close to each other, such respectively). Here the brighter cutoff is very close to the
that they cannot be individually resolved with 8–10 m class brightness of the faintest stars that have been detected. The
telescopes. choice for the KLF slope satisfies the range of the power-law
The calculations were done by taking all the extra (ex- fit Γ = 0.18 ± 0.07. The power-law indices were taken from
trapolated) faint stars in the Ks -magnitude interval of 18 to Table 5 of Schödel et al. (2007) for the cusp radial profiles.
25. The stars were then distributed in a 23 × 23 grid that The results of the simulations are summarized in
corresponds to 529 cells. Each cell has the dimensions of Table 1. Three different realizations of a cluster simula-
0.06′′ × 0.06′′ , i.e. about one angular resolution element in tion as well as the average of 104 simulations are shown
Ks -band, this grid, therefore, simulates observations of the in Figure 11. We find that for the measured KLF slope of
inner 1.38′′ ×1.38′′ projected region surrounding Sgr A*. We 0.18, a measured power-law index of Γ = 0.3 and a faint Ks -
distributed the faint stars in the grid such that their radial magnitude cutoff we obtain a false star in about a quarter
profile centered on Sgr A* reproduces that of the stellar of all simulations. For steeper KLF and power-law slopes Γ
number density counts of the inner region of the central we get this result in more than 70% of all cases independent
stellar cluster with a power-law index of Γ = 0.30 ± 0.05 of the cutoff magnitude.
from Schödel et al. (2007). This way each cell has a specific In Table 2 we show the same statistics as in Table 1
number of stars that can be inserted into it, with the max- but for the central cell in the grid, at the projected posi-
imum number of stars being located in the central cell, i.e. tion of Sgr A*. Also given, in parentheses, is the number
the peak of the radial profile. Our algorithm fills each cell of stars in the central cell that gives rise to the detection
with its specified number of stars by choosing them ran- of a false star at a distance of less than one angular reso-
domly from a pool of stars created from the extrapolated lution element away from the line of sight to Sgr A*. We
KLF. The pool is created such that for each Ks -magnitude find that for a KLF slope of 0.25 we get a false star in

11
 Sabha, Eckart, Merritt, Zamaninasab et al.: The S-star Cluster

Fig. 11. Upper panel : Three different snapshots of the simulation for the 0.25 KLF slope, power-law index Γ = 0.3 and 24.67
Ks -magnitude cutoff. Lower panel : The same as upper but with only the detectable blend stars visible. Right: Average of all the
104 simulation snapshots for the same setup.

Table 2. Probabilities of detecting a false star (brighter in the simulated S-star cluster is a few 1000. This is consis-
than Ks = 17.31) at the position of Sgr A*. tent with the number of main-sequence stars assumed by
Freitag et al. (2006).
Ks -band Power-law index
magnitude
cutoff 0.19 0.30 0.35 6. Summary and conclusion
KLF slope = 0.11 By determining the KLF of the S-star cluster members from
infrared imaging, using the distribution of the diffuse back-
20.99 0.0000(2) 0.0752(4) 0.0757(4)
ground light and the stellar number density counts, we have
24.67 0.0094(6) 0.0103(6) 0.0099(6)
been able to shed some light on the amount and nature of
KLF slope = 0.18 the stellar and dark mass associated with the cluster of high
velocity S-stars in the immediate vicinity of Sgr A*.
20.99 0.0423(4) 0.0438(4) 0.1181(5) The amount of light from the fainter S-cluster members
is below the amount of residual light after removing the
24.67 0.0345(15) 0.0591(18) 0.0821(20) bright cluster members. One implication could be that both
KLF slope = 0.25 the diffuse light and dark mass are overestimated. However,
while NS10 estimate that only a maximum of one third of
20.99 0.3120(8) 0.3149(8) 0.5448(10) the diffuse light could be due to residuals from the PSF
subtraction, we find that faint stars at or beyond the com-
24.67 0.3223(59) 0.4756(70) 0.5291(74) pleteness limit reached in the KLF can account only for
Notes. The number of stars contributing to the detected flux
about 15% of the background light. Additional light may
of the false star is given in parentheses for each considered case. also originate from accretion processes onto a large number
of 10 M⊙ black holes that may reside in the central re-
gion, covered by the S-stars. We find that the stellar mass
derived from the KLF extrapolation is much smaller than
30% to 50% of all simulations, independent of the power- the amount of mass that may be present considering the
law index Γ and the cutoff magnitude. This is consistent uncertainties in the orbital motion of the star S2. Higher
with the offsets found in different observational epochs of angular resolution and sensitivity are needed to resolve the
Sgr A* light curves (Witzel et al. 2012; Dodds-Eden et al. background light and analyze its origin.
2011). In this case the blend consists of 8 to 74 stars below By investigating the effects of orbital torques due to res-
the unresolved background in the S-star cluster region. For onant relaxation, we find that if a significant population of
flatter KLF slopes (i.e. 0.11 and 0.18) we find that a blend 10 M⊙ black holes is present, with enclosed masses between
star only occurs in less than about 10% of all cases, which 103 M⊙ and 105 M⊙ (see e.g. Freitag et al. 2006), then for
appears to be well below the upper limit found from obser- trajectories of S2-like stars, contributions from scattering
vations. For a KLF slope of α = 0.25 and a number density will be important compared to the relativistic or Newtonian
power-law index of Γ around 0.3 the total number of stars peribothron shifts. This clearly shows that observing a sin-

12
 Sabha, Eckart, Merritt, Zamaninasab et al.: The S-star Cluster

gle stellar orbit will not be sufficient to put firm limits on References
the total amount of extended mass and on the importance Alexander, T. 2005, Phys. Rep., 419, 65
of relativistic peribothron shift. In this case only the ob- Alexander, T. 2007, arXiv e-prints:0708.0688
servation of a larger number of stars will allow to sample Alexander, T. & Hopman, C. 2009, ApJ, 697, 1861
the statistics of the effect, i.e. the distributions in Figure 9. Antonini, F., Capuzzo-Dolcetta, R., Mastrobuono-Battisti, A., &
However, if the distribution of 10 M⊙ black holes is cuspy Merritt, D. 2012, ApJ, 750, 111
Baganoff, F. K., Bautz, M. W., Brandt, W. N., et al. 2001, Nature,
then this may become even more difficult (and close en- 413, 45
counters should be frequent in this region). Baganoff, F. K., Maeda, Y., Morris, M., et al. 2003, ApJ, 591, 891
In general, the inclusion of star-star perturbations al- Bahcall, J. N. & Wolf, R. A. 1976, ApJ, 209, 214
Bahcall, J. N. & Wolf, R. A. 1977, ApJ, 216, 883
lows us to probe the distribution and composition of mass Bartko, H., Martins, F., Trippe, S., et al. 2010, ApJ, 708, 834
very close to the SMBH simultaneously, if the astromet- Baumgardt, H., Gualandris, A., & Portegies Zwart, S. 2006, MNRAS,
ric accuracy can be improved by an order of magnitude by 372, 174
using either larger telescopes or interferometers in the NIR. Blum, R. D., Sellgren, K., & Depoy, D. L. 1996, ApJ, 470, 864
Buchholz, R. M., Schödel, R., & Eckart, A. 2009, A&A, 499, 483
With measurements and extrapolations of the S-star Chandrasekhar, S. 1943, ApJ, 97, 255
cluster KLF slope, and number density counts with as- Dale, J. E., Davies, M. B., Church, R. P., & Freitag, M. 2009, MNRAS,
sumptions on the KLF cutoff magnitude, we can show that 393, 1016
the contamination for the members of the cluster, and es- Do, T., Ghez, A. M., Morris, M. R., et al. 2009, ApJ, 703, 1323
pecially at the position of Sgr A*, by blend stars is fully Dodds-Eden, K., Gillessen, S., Fritz, T. K., et al. 2011, ApJ, 728, 37
Duric, N. 2004, Advanced astrophysics
consistent with measurements. We show that for 8–10 m Eckart, A. & Genzel, R. 1996, Nature, 383, 415
class telescopes the presence and proper motion of faint Eilon, E., Kupi, G., & Alexander, T. 2009, ApJ, 698, 641
stars close to the confusion limit in the region of the S- Eisenhauer, F., Genzel, R., Alexander, T., et al. 2005, ApJ, 628, 246
star cluster is highly contaminated by blend stars. Due to Freitag, M. 2008, in Astronomical Society of the Pacific Conference
Series, Vol. 387, Massive Star Formation: Observations Confront
the 2-dimensional velocity dispersion of the stars within the Theory, ed. H. Beuther, H. Linz, & T. Henning, 247
S-star cluster of about 600 km/s the blend stars will last Freitag, M., Amaro-Seoane, P., & Kalogera, V. 2006, ApJ, 649, 91
for about 3–4 years before they fade and dissolve. Close to Fujii, M., Iwasawa, M., Funato, Y., & Makino, J. 2009, ApJ, 695, 1421
the center, we find the probability of detecting blend stars Fujii, M., Iwasawa, M., Funato, Y., & Makino, J. 2010, ApJ, 716, L80
at any time is about 30–50%. At the central position the Genzel, R., Pichon, C., Eckart, A., Gerhard, O. E., & Ott, T. 2000,
MNRAS, 317, 348
change from the appearance of a blend star to the appear- Genzel, R., Schödel, R., Ott, T., et al. 2003, ApJ, 594, 812
ance of another may also give the illusion of high proper Ghez, A. M., Duchêne, G., Matthews, K., et al. 2003, ApJ, 586, L127
motions for 8–10 m class telescopes. Such a prime exam- Ghez, A. M., Klein, B. L., Morris, M., & Becklin, E. E. 1998, ApJ,
ple would be S3, detected close to the position of Sgr A*, 509, 678
Gillessen, S., Eisenhauer, F., Fritz, T. K., et al. 2009a, ApJ, 707, L114
which had both a limited lifetime and high proper mo- Gillessen, S., Eisenhauer, F., Trippe, S., et al. 2009b, ApJ, 692, 1075
tion (Eckart & Genzel 1996; Ghez et al. 1998; Genzel et al. Gould, A. & Quillen, A. C. 2003, ApJ, 592, 935
2000). Blending of sources along the line of sight may also Gualandris, A., Gillessen, S., & Merritt, D. 2010, MNRAS, 409, 1146
severely contaminate the proper motion measurements of Hansen, B. M. S. & Milosavljević, M. 2003, ApJ, 593, L77
individual stars close to the confusion limit. Only with the Hopman, C. & Alexander, T. 2006a, ApJ, 645, 1152
Hopman, C. & Alexander, T. 2006b, ApJ, 645, L133
help of proper motion measurements over time significantly Kim, S. S., Figer, D. F., & Morris, M. 2004, ApJ, 607, L123
longer than 3 years one will be able to derive reliable or- Kocsis, B. & Tremaine, S. 2011, MNRAS, 412, 187
bital parameters for a single star. Also, spectroscopy may Levin, Y., Wu, A., & Thommes, E. 2005, ApJ, 635, 341
help to resolve blend stars, however, the objects are faint Lightman, A. P. & Shapiro, S. L. 1977, ApJ, 211, 244
Löckmann, U., Baumgardt, H., & Kroupa, P. 2009, MNRAS, 398, 429
and spectroscopy will be difficult. Madigan, A.-M., Levin, Y., & Hopman, C. 2009, ApJ, 697, L44
These findings clearly demonstrate the necessity of Maı́z Apellániz, J. & Úbeda, L. 2005, ApJ, 629, 873
higher angular resolution, astrometric accuracy and point Martins, F., Gillessen, S., Eisenhauer, F., et al. 2008, ApJ, 672, L119
source sensitivity for future investigations of the S-star clus- Merritt, D. 2010, ApJ, 718, 739
Merritt, D. 2012, Black Holes and the Dynamics of Galactic Nuclei
ter. They would also greatly improve the derivation of the (Princeton University Press)
amount and the compactness of the central mass as well Merritt, D., Alexander, T., Mikkola, S., & Will, C. M. 2010,
as the determination of relativistic effects in the vicinity of Phys. Rev. D, 81, 062002
Sagittarius A*. Merritt, D., Gualandris, A., & Mikkola, S. 2009, ApJ, 693, L35
Mouawad, N., Eckart, A., Pfalzner, S., et al. 2005, Astronomische
Nachrichten, 326, 83
Acknowledgements. We thank the anonymous referee for the help- Murphy, B. W., Cohn, H. N., & Durisen, R. H. 1991, ApJ, 370, 60
ful comments. N. Sabha is member of the Bonn Cologne Graduate Paumard, T., Genzel, R., Martins, F., et al. 2006, ApJ, 643, 1011
School (BCGS) for Physics and Astronomy supported by the Deutsche Perets, H. B. & Gualandris, A. 2010, ApJ, 719, 220
Forschungsgemeinschaft. D. Merritt acknowledges support from the Perets, H. B., Gualandris, A., Kupi, G., Merritt, D., & Alexander, T.
National Science Foundation under grants no. AST 08-07910, 08- 2009, ApJ, 702, 884
21141 and by the National Aeronautics and Space Administration un- Perets, H. B., Hopman, C., & Alexander, T. 2007, ApJ, 656, 709
der grant no. NNX-07AH15G. We thank Tal Alexander for useful dis- Preto, M. & Amaro-Seoane, P. 2010, ApJ, 708, L42
cussions. M. Garcı́a-Marı́n is supported by the German federal depart- Rauch, K. P. & Tremaine, S. 1996, New A, 1, 149
ment for education and research (BMBF) under the project number Rubilar, G. F. & Eckart, A. 2001, A&A, 374, 95
50OS1101. M. Valencia-S. and B. Shahzamanian are members of the Sabha, N., Witzel, G., Eckart, A., et al. 2010, A&A, 512, A2
International Max-Planck Research School (IMPRS) for Astronomy Sabha, N., Witzel, G., Eckart, A., et al. 2011, in Astronomical Society
and Astrophysics at the Universities of Bonn and Cologne supported of the Pacific Conference Series, Vol. 439, Astronomical Society of
by the Max Planck Society. Part of this work was supported by the the Pacific Conference Series, ed. M. R. Morris, Q. D. Wang, &
German Deutsche Forschungsgemeinschaft, DFG, via grant SFB 956 F. Yuan, 313
and fruitful discussions with members of the European Union funded Salaris, M. & Cassisi, S. 2005, Evolution of Stars and Stellar
COST Action MP0905: Black Holes in a violent Universe and PECS Populations
project No. 98040. Sazonov, S., Sunyaev, R., & Revnivtsev, M. 2011, MNRAS, 1987

13
 Sabha, Eckart, Merritt, Zamaninasab et al.: The S-star Cluster

Schödel, R., Eckart, A., Alexander, T., et al. 2007, A&A, 469, 125
Schödel, R., Najarro, F., Muzic, K., & Eckart, A. 2010, A&A, 511,
A18
Schödel, R., Ott, T., Genzel, R., et al. 2002, Nature, 419, 694
Timmes, F. X., Woosley, S. E., & Weaver, T. A. 1996, ApJ, 457, 834
Witzel, G., Eckart, A., Bremer, M., et al. 2012, ApJ, submitted
Yusef-Zadeh, F., Bushouse, H., & Wardle, M. 2012, ApJ, 744, 24
Zucker, S., Alexander, T., Gillessen, S., Eisenhauer, F., & Genzel, R.
2006, ApJ, 639, L21

14