arXiv:0907.3786v2 [astro-ph.GA] 17 Sep 2009
Simultaneous Multi-Wavelength Observations of Sgr A*
during 2007 April 1-11
F. Yusef-Zadeh1, H. Bushouse2, M. Wardle3, C. Heinke4, D. A. Roberts5, C.D. Dowell6, A.
Brunthaler7, M. J. Reid8, C. L. Martin9, D. P. Marrone10, D. Porquet11, N. Grosso11, K.
Dodds-Eden12, G. C. Bower13, H. Wiesemeyer14, A. Miyazaki15, S. Pal16, S. Gillessen12, A.
Goldwurm17, G. Trap18, and H. Maness13
We report the detection of variable emission from Sgr A* in almost all wave-
length bands (i.e. centimeter, millimeter, submillimeter, near-IR and X-rays) during
a multi-wavelength observing campaign. Three new moderate flares are detected
simultaneously in both near-IR and X-ray bands. The ratio of X-ray to near-IR flux
in the flares is consistent with inverse Compton scattering of near-IR photons by
submillimeter emitting relativistic particles which follow scaling relations obtained
from size measurements of Sgr A*. We also find that the flare statistics in near-IR
wavelengths is consistent with the probability of flare emission being inversely pro-
portional to the flux. At millimeter wavelengths, the presence of flare emission at 43
GHz (7mm) using VLBA with milli-arcsecond spatial resolution indicates the first
direct evidence that hourly time scale flares are localized within the inner 30×70
Schwarzschild radii of Sgr A*. We also show several cross correlation plots between
near-IR, millimeter and submillimeter light curves that collectively demonstrate the
presence of time delays between the peaks of emission up to three hours. The ev-
idence for time delays at millimeter and submillimeter wavelengths are consistent
with the source of emission being optically thick initially followed by a transition to
an optically thin regime. In particular, there is an intriguing correlation between the
optically thin near-IR and X-ray flare and optically thick radio flare at 43 GHz that
occurred on 2007 April 4. This would be the first evidence of a radio flare emission
at 43 GHz delayed with respect to the near-IR and X-ray flare emission. The time
delay measurements support the expansion of hot self-absorbed synchrotron plasma
blob and weaken the hot spot model of flare emission. In addition, a simultaneous
fit to 43 and 84 GHz light curves, using an adiabatic expansion model of hot plasma,
appears to support a power law rather than a relativistic Maxwellian distribution of
Subject headings: accretion, accretion disks — black hole physics — Galaxy: center
– 2 –
The black hole at the center of our own galaxy was first detected as the radio source Sgr A*
over 30 years ago (Balick & Brown 1974). It was found to lie at the center of a cluster of young
massive stars. Submillimeter and far-infrared observations showed that Sgr A* is encircled by a
torus of gas approximately 10 light-years across, which orbits with a speed of 100 km s−1(e.g.
Genzel & Townes 1987). The gravity required to hold onto this material implies a mass of several
million solar masses, although a portion of this is contributed by the stars in the stellar cluster.
These measurements suggested that Sgr A* could be a black hole.
through studies of the light distribution of stars in the cluster, as well as the motions of ionized and
molecular gas clouds orbiting Sgr A*. These measurements implied a mass of approximately 3–4
million times that of the sun (Genzel 2000; Genzel & Townes 1987) lying within a third of a light
year of the radio source. Recently, more precise measurements of fast moving stars in close orbits
around Sgr A* have conclusively demonstrated that it has a mass of ∼ 4×106M⊙(Ghez et al. 2005;
Eisenhauer et al. 2005; Ghez et al. 2008; Sch¨ odel et al. 2002; Gillessen et al. 2009) and that the size
of the radio source is about ∼4 times its Schwarzschild radius at 230 GHz (Rs) (Doeleman et al.
2008). This dark, massive object has also been uniquely identified through the proper motion of
the radio source, which show that Sgr A* must contain > 4 × 105M⊙(Reid & Brunthaler 2004).
Taken together, these measurements provide strong evidence that Sgr A* is a black hole with mass
∼ 4×106M⊙. No other known category of astrophysical object can easily fit so much mass into a
More detail was provided
1Dept. of Physics and Astronomy, Northwestern University, Evanston, Il. 60208
2STScI, 3700 San Martin Drive, Baltimore, MD 21218
3Department of Physics and Engineering, Macquarie University, Sydney NSW 2109, Australia
4Dept. of Physics, University of Alberta, Room #238 CEB, 11322-89 Avenue, Edmonton AB T6G 2G7, Canada
5Adler Planetarium and Astronomy Museum, 1300 South Lake Shore Drive, Chicago, IL 60605
6Cal Tech, Jet Propulsion Laboratory, Pasadena, CA 91109
7Max-Planck-Institut f¨ ur Radioastronomie, Auf dem Huegel 69, 53121 Bonn, Germany
8Harvard-Smithsonian CfA, 60 Garden Street, Cambridge, MA 02138
9Oberlin College, Dept. of Physics and Astronomy, Professor 110 N. St.,Oberlin, OH 44074
10National Radio Astronomy Observatory; University of Chicago, 5640 South Ellis Avenue, Chicago IL 60637
11Observatoire astronomique de Strasbourg, Universit´ e de Strasbourg, NRS, INSU, 11 rue de l’Universit´ e, 67000
12Max-Plank-Institut f¨ ur Extraterrestrische Physik 1312, D-85471, Garching, Germany
13Radio Astronomy Lab, 601 Campbell Hall, University of California, Berkeley, CA 94720
14Institut de RadioAstronomie Millimetrique, 300 rue de la Piscine, Domaine Universitaire 38406 Saint Martin
d’Heres, France, on leave to IRAM Granada, Spain
15Mizusawa VLBI Observatory, National Astronomical Observatory of Japan, Mizusawa, Oshu, Iwate 023-0861,
16School of Physics, University of Western Australia, 35 Stirling Highway, Crawley, WA, 6009, Australia
17Service d’Astrophysique / IRFU / DSM, CEA Saclay, Bat.
AstroParticule & Cosmologie (APC) / Universit´ e Paris VII / CNRS / CEA / Observatoire de Paris Bat. Condorcet,
10, rue Alice Domon et L´ eonie Duquet, 75205 Paris Cedex 13, France
709, 91191, Gif-sur-Yvette Cedex, France and
– 3 –
This massive black hole is a hundred times closer to us than the next nearest example, presenting an
unparalleled opportunity to closely study the process by which gas is captured by black holes. It is
therefore the subject of intense scrutiny. The energy radiated by Sgr A* is thought to be liberated
from gas that is falling into the black hole after being captured from the powerful winds of members
of its neighboring cluster of massive stars (e.g., Melia 1992). The broad band spectrum of Sgr A*
peaks at submillimeter wavelengths (Zylka et al. 1992; Falcke et al. 1998); this is thought to be
the dividing line between optically thick and optically thin emission at low and high frequencies,
respectively. The bolometric luminosity of Sgr A* ∼ 100
below that predicted given its expected rate of capture of material from stellar winds, prompting
a number of theoretical models to explain its very low efficiency (Narayan et al. 1995; Liu & Melia
2001; Yuan et al. 2003; Goldston et al. 2005; Liu et al. 2004; Falcke et al. 2009).
L⊙ is several orders of magnitudes
Now that the quiescent spectrum of emission from Sgr A* has been characterized from radio to X-
rays, attention has turned to variability of emission in multiple wavelengths. These measurements
probe the structure and the physical parameters of the hot plasma in the vicinity of the black hole
by measuring the time variations of its flux in different wavelength bands as well as their cross-
correlation with each other. Flaring activity on < 1 − 4 hour time scale is seen in all wavelength
bands in which quiescent emission has been detected.
Flaring X-ray emission from Sgr A* has been detected and has been argued to originate within
a few Schwarzschild radii of the ∼ 4 × 106M⊙black hole (Baganoff et al. 2003; Goldwurm et al.
2003; Porquet et al. 2003; B´ elanger et al. 2005). At near-IR (NIR) wavelengths (Genzel et al. 2003;
Yuan et al. 2003; Ghez et al. 2004; Hornstein et al. 2007), flare emission from Sgr A* is shown to
be due to optically thin synchrotron emission, whereas the long-wavelength flaring activity in
submillimeter, millimeter and centimeter bands is due to optically thick synchrotron emission.
The exact frequency at which the transition from optically thick to thin flare emission occurs is
A variety of mechanisms have been proposed to explain the origin of the variability of Sgr A*. Many
of these models have considered different energy distributions for the relativistic particles to explain
the origin of submillimeter emission (Markoff et al. 2001; Yuan et al. 2002; Melia 2002; Liu & Melia
2002; Yuan et al. 2003; Nayakshin & Sunyaev 2003; Eckart et al. 2004, 2006a,b; Yusef-Zadeh et al.
2006a; Gillessen et al. 2006; Goldston et al. 2005; Liu et al. 2006; Falcke et al. 2009); Melia and
Falcke (2001 and references therein). The direction that has been taken in the past in interpreting
the flaring activity of Sgr A* is within one of the established paradigms for the accretion flow
that have been developed based on the time-averaged emission – for example a thin accretion
disk, a disk and jet, outflow, an advection-dominated accretion flow, radiatively inefficient accre-
tion flow, accretion disk inflow/outflow solutions (Melia 1992; Yuan et al. 2003; Falcke & Markoff
2000; Falcke et al. 2009; Narayan et al. 1998; Blandford & Begelman 1999) and then the predicted
spectrum is compared with the observed spectrum.
We have recently analyzed the NIR flaring of Sgr A*, which is produced by synchrotron emission
from a transient population of particles produced within ∼ 10 Schwarzschild radii of the massive
black hole (Genzel et al. 2003; Eckart et al. 2006a; Gillessen et al. 2006).
∼ 2−3 hour duration of submillimeter flares could not be due to synchrotron cooling when observed
simultaneously with a NIR flare (estimated to be ∼ 20 minutes and ∼ 12 hours at 1.6µm [188 THz]
and 850µm [350 GHz], respectively). The decline in submillimeter light curves was interpreted to
be due to adiabatic cooling associated with expansion of the emitting plasma (Yusef-Zadeh et al.
2006a,b) under the assumption that the same accelerated population of particles is responsible
We argued that the
– 4 –
for NIR and submillimeter emission. Time delays detected between peaks of flare emission at
radio, submillimeter and NIR/X-rays wavelengths are consistent with this picture (e.g., Yusef-
Zadeh et al. 2006b; Marrone et al. 2008; Yusef-Zadeh et al. 2008; Meyer et al. 2008; Eckart
et al. 2008). However, the lack of long simultaneous coverage have not placed strong constraints
in time delay measurements, especially between radio and NIR wavelengths. Simple modeling of
the total and polarized intensity of the hot expanding plasma provide predictions that can be
tested observationally by carrying out observational campaigns such as the one we coordinated
during April 2007 to examine the mechanisms for the variability, with implications on the nature
of the accretion flow. The results presented here are the third in a series of papers that came
from the multi-wavelength observing campaign that took place on 2009 April (Porquet et al. 2008;
Dodds-Eden et al. 2009). The results of soft γ-ray observations are given separately (Trap et al.
The structure of this paper is as follows. §3 presents light curves of all the useful data that were taken
in this campaign, following the observational details described in §2. In §4 we analyze the statistical
properties of flare emission and the corresponding spectral and power spectrum distributions at
NIR wavelengths, as well as cross-correlation analysis of light curves. We then discuss in §5 the
origin of X-ray emitting flares and provide observational support for the expanding hot plasma
model of flare emission. The polarization results will be given elsewhere.
2. Observations and Data Reduction
The primary purpose of observations made during 2007 April 1-11 was the coordination of several
telescopes operating at many wavelengths to monitor the emission from Sgr A* and measure the
time evolution of its spectrum. There were a total of 13 observatories that participated in this cam-
paign, including XMM-Newton, the Hubble Space Telescope (HST), the International Gamma-Ray
Astrophysics Laboratory INTEGRAL, the Very Large Array (VLA) of the National Radio Astron-
omy Observatory18(NRAO), the Very Long Baseline Array (VLBA18), the Caltech Submillimeter
Observatory (CSO), the Very Large Telescope (VLT), the Submillimeter Array (Blundell 2004),
the 30m Pico Veleta Telescope of the Institute for Millimeter Radioastronomy (IRAM), the Sub-
millimeter Telescope (SMT), the Nobeyama Millimeter Array (NMA), the Combined Array for
Research in Millimeter-wave Astronomy (CARMA), and the Giant Meterwave Radio Telescope
(GMRT). The campaign was organized by first obtaining observing time with XMM-Newton (PI:
D. Porquet) and HST (PI: F. Yusef-Zadeh), and then coordinating the ground-based facilities to
the allotted space-based schedules.
Figure 1 shows the schedule of all observations and their rough durations.
Newton and VLT/NACO observations have already been reported in Porquet et al. (2008), and
(Dodds-Eden et al. 2009), respectively. Summary of the results from VLT/VISIR and INTEGRAL
observations has also been given by Trap et al. (2009). Briefly, XMM observations were carried out
using three observations for a total of 230 ks blocks of time during 2007 March 30 to April 4, and
the VLT/NACO observations took place between 2007 April 1–6 using H (1.66 µm), K (2.12µm),
and L’ (3.8µm) bands. INTEGRAL observations took place in parallel to XMM-Newton on April
1 and 4 for a total effective exposure time of 212 ks for IBIS/ISGRI (20–100 keV) and 46 ks for
Details of XMM-
18The National Radio Astronomy Observatory is a facility of the National Science Foundation, operated under a
cooperative agreement by Associated Universities, Inc.
– 5 –
JEM-X 1 (3–20 keV).
2.1. HST NICMOS: 1.45µm and 1.70µm
We obtained 40 orbits of HST observations using the NICMOS camera 1, with the orbits distributed
over seven consecutive days between 2007 April 1–7. The observations used the NICMOS F145M
and F170M filters, with exposure times of 144 sec in each filter and readout samplings of ∼16 sec
within each exposure. Use of this pair of filters has several advantages. First, they have no overlap
in wavelength space and are therefore suitable for spectral index measurements. Second, they are
well matched to one another in terms of throughput, so that we can use identical exposure and
readout times, thus producing time series data that are evenly sampled in both filters. This has
great advantages for making periodicity measurements. Third, their relatively high throughput
also allows us to use relatively short exposure times, so that we can cycle back and forth between
the two filters fairly rapidly. The near-simultaneous observations then allow us to make meaningful
spectral index measurements of flare events. This is especially true for the peaks of flare events,
where the flux from Sgr A* does not change rapidly over the course of several minutes or more.
The spectral index distribution can not be measured accurately, however, during the rise and fall of
flare events because the overall flux is changing more rapidly than the cadence of our filter cycling
The IRAF “apphot” routines were used to perform aperture photometry of sources in the NICMOS
Sgr A* field, including Sgr A* itself. For stellar sources the measurement aperture was positioned on
each source using an automatic centroiding routine. This approach could not be used for measuring
Sgr A* because its signal is spatially overlapped by that of the orbiting star S0-2 and S17. Therefore
the photometry aperture for Sgr A* was positioned by using a constant offset from the measured
location of S0-2 in each image. The offset between S0-2 and Sgr A* was derived from the orbital
parameters given by Ghez et al. (2003). The position of Sgr A* was estimated to be 0.16′′south
and 0.01′′west of S0-2 at the time of the HST observations. To confirm the accuracy of the position
of Sgr A*, two images of Sgr A* taken before and during a flare event were aligned and subtracted,
which resulted in an image showing the location of the flare emission.
We used a measurement aperture with a diameter of 3 detector pixels, which corresponds to ∼ 0.13′′.
This size was chosen as a suitable compromise between wanting to maximize the fraction of the
Sgr A* PSF included in the aperture, while at the same time limiting the amount of signal coming
from the wings of the PSF from the adjacent star S0-2. We derived aperture correction factors by
making measurements of a reasonably well isolated star in the field through a series of apertures of
increasing size. The aperture corrections, which convert the fluxes measured through our 3-pixel
diameter aperture to a semi-infinite aperture, are 2.91 and 3.40 for the 1.45 and 1.70 µm bands,
respectively. Absolute calibration of the photometric measurements was accomplished using the
latest calibrations for the NICMOS F145M and F170M obtained from STScI.
De-reddened fluxes were computed using the appropriate extinction law for the Galactic center
(Moneti et al. 2001) and their extinction value of A(K)=3.3 mag. These translate to extinction
values for the NICMOS filter bands of A(F170M)=5.03 mag and A(F145M)=6.52 mag, which then
correspond to correction factors of 103 and 406, respectively. Because there is some contribution
from the neighboring star S0-2 within our measurement aperture, we have determined the flux of
Sgr A* when it is flaring by subtracting the mean flux level measured during “quiescent” episodes.
The resulting net flux can therefore be safely attributed to Sgr A* flares.
– 6 –
2.2.VLA: 43 and 22 GHz
We used a fast-switching technique to observe Sgr A* simultaneously using the VLA D configuration
at 43.3 GHz (7mm) and 22.4 GHz (13mm) GHz. These observations took place on 2007 April 1–4,
each lasting for ∼7 hours, using 8 VLA and 18 eVLA antennas. The two IFs were separated by
50MHz in each observation, except for those on 2007 April 4, when the two IFs were centered at
non-standard frequencies of 43.1851 and 43.5351 GHz, which corresponds to a frequency separation
of 350 MHz. The separation was used to carry out polarization measurements, the result of which
will be given elsewhere.
We cycled between Sgr A* and the fast-switching calibrator 17444-31166 (2.3 degrees away from
Sgr A*) for 90 sec and 30 sec, respectively, throughout the observation. On 2007 April 4, we also
used the fast-switching phase calibrator 17459-28204, which is weaker, but closer (∼ 21′) to Sgr A*.
3C286 was used as the flux calibrator and NRAO530 was observed as a polarization and additional
phase calibrator. The light curves at 43 GHz restricted data to a uv range greater than 90 kλ with
full width at half point of 2.45′′×1.3′′(PA=−40). We used NRAO530 for pointing every 30 minutes;
the bootstrapped flux of NRAO530 at 43 GHz is 2.43±0.05 Jy. At 22 GHz, the strong continuum
emission from ionized gas associated with extended features surrounding Sgr A* overwhelmed the
flux of Sgr A* itself, making the variability analysis uncertain. Therefore the 22 GHz data are not
useful and are not presented here. In all the measurements presented here, we used only antennas
that had constant gain curves with similar values, thus many of the eVLA antennas were not used.
In all cases, at least two and sometimes three phase calibrators were used in order to ensure that
amplitude variability or calibration errors of one of the calibrators would not be introduced into the
light curve of Sgr A*. In the case of multiple phase calibrators, the same calibrator used to calibrate
the gains of Sgr A* was used to cross-calibrate the other calibrators. In cases where calibrator light
curves are shown as alongside those of Sgr A*, they were obtained from cross-calibration using
one of the other phase calibrators and not from self-calibration. Additionally, as a check, all light
curves of Sgr A* made using phase calibrations from the principle phase calibrator (usually 17444-
31166) were compared against light curves of Sgr A* made using the other phase calibrators. These
comparisons were used to identify bad data in the calibrators and after editing and recalibration
light curves using different calibrators were consistent.
For the final light curves of Sgr A*, the data were calibrated using the principle phase calibrator
(17444-31166). A phase self-calibration was applied to Sgr A* before the determination of a light
curve. No amplitude self-calibration was done to Sgr A* or any backup phase calibrators, whose
light curves are shown as a reference, since amplitude calibration would remove time variation from
the light curves. After phase self-calibration, large images were made and found no confusing point
sources above the rms noise (typically 2.5 mJy beam−1in a full run at 43 GHz) when the selected
uv data were greater than 90kλ.
In order to derive the light curve in the visibility plane, the Astronomical Image Processing System
(AIPS) task DFTPL was used. DFTPL plots the direct Fourier transform of a vector averaged
set of measured visibilities as a function of time. Since we use this on data that has been phase
self-calibrated using the point source Sgr A*, the vector average gives the flux of Sgr A*. Visibilities
are averaged in bins with defined time widths, however, since the number of visibilities in each bin
varies, the error associated with the average will not be constant and are derived at each time
– 7 –
2.3.VLBA: 43, 22, and 15 GHz
We observed Sgr A* with the VLBA in two different experiments. One took place on 2007 April 1,
5 and 11 at 43 GHz under program BR124. All observations employed four 8 MHz bands in dual
circular polarization each. These observations were made at 43 GHz and involved rapid switching
between Sgr A* and the two background continuum sources J1748-291 and J1745-283. Sources
were changed every 15 seconds in the sequence Sgr A* – J1748-291 – Sgr A* – J1745-283 – Sgr A*,
yielding an on-source time of ∼ 10 seconds. Before, in the middle, and after each observation
16 quasars were observed within ∼ 40 minutes. NRAO530 was also observed as fringe-finder.
The total observing time including the quasars was 8 hours for each observation. The data were
correlated with 16 spectral channels per frequency band and an integration time of 0.131 seconds.
The ∼ 1 hour gaps in the light curves of Sgr A* are due to geodetic measurements.
In the second experiment (proposal code BB230), the observations on 2007 April 2 and 10 involved
rapid switching between two frequencies on Sgr A* . We observed 43 and 22 GHz on 2007 April 2
and 43 and 14 GHz on April 10. We changed the receiver every 20 seconds, yielding an on-source
time of ∼10 seconds for each frequency. The observations were interrupted three times by 20 minute
observations of four different quasars, including the fringe-finder 3C345. All data were correlated
with 16 spectral channels per band and an integration time of one second. The total observing
time including the quasars was 6 hours for each observation.
The VLBI data were edited and calibrated using standard techniques in AIPS. First, we applied
the latest values of the Earth’s orientation parameters. A-priori amplitude calibration was applied
using system temperature measurements and standard gain curves. We performed a “manual phase-
calibration” using the data from NRAO530 or 3C345 to remove instrumental phase offsets among
the three frequency bands. Then, we fringe fitted the data from Sgr A* using only the five inner
VLBA antennas (PT, KP, FD, OV, and LA). Then, we discarded all data with elevations below 15◦
and performed one round of phase self-calibration on Sgr A*. Finally, we divided the calibrated
uv-data by a model described in Bower et al. (2004), i.e. an elliptical Gaussian component of
0.71×0.41 mas with a position angle of 78◦. Lightcurves were extracted form the uv-data with the
AIPS task DFTPL.
2.4. CSO: 350µm, 450µm & 850µm
Nightly observations of Sgr A* were made at three wavelengths over the period 2007 April 1–6 UT,
using the SHARC-II camera. The observations on April 1–5 were made with the SHARP imaging
polarimeter module installed (Li et al. 2008), using the 350 µm half-wave plate. The observing
bands (selected by a cryogenic filter) were 450 µm on April 1–3 and 350 µm on April 4–5. The
observations on April 6 were made at 850 µm with SHARP removed from the optical path.
In this paper, we report only the total intensity measurements.
polarization instrument, and because the observations were made over cycles of half-wave plate
angles that fully modulate the polarization, our total intensity results are insensitive to the polar-
ization of the source.
Because SHARP is a dual-
Except for the stepping of the polarimeter half-wave plate between integrations, the observing and
analysis method was similar to past CSO observations (Yusef-Zadeh et al. 2006a, 2008). We used
Lissajous scans with typical full amplitudes of 100′′in both azimuth and elevation. The instanta-
neous field of view is 57′′× 57′′in polarimeter mode and is 154′′× 58′′without the polarimeter.
– 8 –
The measured beam sizes were 8.4′′at 350 µm, 10.1′′at 450 µm, and 18.8′′at 850 µm. We used
the Dish Surface Optimization System (DSOS) on April 1–5. Three quadrants were working fully
during the run, but the fourth quadrant of the system was available for only part of the run. For
observations at the elevation of Sgr A*, we expect no significant effect on the results from the non-
operational quadrant. Any change in the beam FWHM due to the status of the DSOS quadrant
was less than 3%.
Mauna Kea weather conditions were good overall during the April 1–6 campaign. Occasional thin
cirrus was observed visually and on satellite photos on April 3, 4, and 6; otherwise skies were clear.
Local humidity was < 30% during the observations. The wind speed for April 1 was noticeably
high (roughly 30 mph), but less than 20 mph on the other nights. The zenith atmospheric opacity
at 225 GHz was marginal for observations at 450 µm on April 1 (τ225≈ 0.07), as well as at 850 µm
on April 6 (τ225≈ 0.15), and rising in both cases. Atmospheric opacity on April 2–5 was excellent
and relatively steady, ranging from 0.03 to 0.06 at 225 GHz.
In producing the light curves for this paper, we reconsidered the image registration and absolute
calibration for all of our CSO observations of Sgr A* from 2004 September through 2008 May. The
absolute pointing of the images is based on hourly measurements of point-like calibration sources,
such as planets, and the pointing model for the telescope. This procedure appears to average
down in a reasonable manner. The 350 µm and 450 µm position that we measure for the variable
component of Sgr A* is within 0.3′′of the nominal position of α2000= 17h: 45m: 40s.03,δ2000=
−290: 00′: 28′′.1. The agreement at 850 µm, at which the telescope beam size is larger, is somewhat
worse at 0.9′′.
At any particular point in time, the telescope pointing model has only ∼ 2′′accuracy. Therefore,
we shifted the individual observations to align with the average of all the observations, using the
bright dust emission in the images as the reference. Subsequent photometry of Sgr A* assumes a
fixed position and beam size.
Minor changes have been made to the absolute calibration scale factor and Sgr A* “zero point”, in-
cluding data which have been published in the past (Yusef-Zadeh et al. 2006a, 2008; Marrone et al.
2008). For the scale factor, we have adopted the following brightness temperatures for calibration
at 350, 450, and 850 µm, respectively: Callisto (128, 122, 120 K), Neptune (61, 66, 81 K), and
Uranus (64, 70, 86 K), arranged in decreasing order of usage and with an estimated 10% uncertainty.
These brightness temperatures are not significantly different from our past assumptions. The zero
point relates to the difficulty of measuring the total flux of Sgr A* with ∼ 10′′resolution because
of confusion from surrounding dust emission. We estimate an additive uncertainty of 1 Jy in our
measurements of the absolute flux at 350 µm and 850 µm, and an additive uncertainty of 0.5 Jy
at 450 µm. To be consistent with the results published in this paper, the August–September 2004
measurements at 850µm reported by Yusef-Zadeh et al. (2006a) should be shifted upwards by ∼0.2
Jy; the 450 µm measurements for the same period are essentially unchanged. The 850 µm mea-
surements for July 2006, reported by Yusef-Zadeh et al. (2008) and (Marrone et al. 2008), should
be shifted upwards by ∼0.5 Jy. The 350 µm results for July 2005, reported by (Marrone et al.
2008), should be shifted upwards by ∼0.4 Jy; the 450 µm and 850 µm results for the same period
are essentially unchanged.
– 9 –
2.5. SMA: 230 GHz
The SMA observed Sgr A* on the nights of 1, 3, 4, and 5 April 2007, typically covering the interval
1200–1830 UT. On the first three nights the array was tuned to observe 231.9 (221.9) GHz in the
upper (lower) sideband, while on the last night the frequency was tuned to 246.0 (241.0) GHz. The
array was in its “compact-north” configuration, resulting in angular resolution of approximately
3′′. All eight antennas were used except on April 4, when one was lost to an instrument problem.
The SMA polarimetry system (Marrone & Rao 2008) was used in these observations to convert the
linearly polarized SMA feeds to circular polarization sensitivity, which prevents confusion between
linear polarization and total intensity variations.
The data were gain calibrated using the quasar J1733−130, which was observed approximately
every 10 minutes. The absolute flux density scale was derived from observations of Callisto and
has an uncertainty of 10%. To remove the effects of the extended emission that surrounds Sgr A*,
only projected baselines longer than 20 kλ were used in the light curve determination. Flux density
measurements were made by applying the quasar gains to the Sgr A* data, removing the average
phase on Sgr A* in each light curve interval via phase self-calibration, which reduces the effect
of baseline errors and phase drifts on the measurement, and fitting a central point source to the
calibrated visibilities. Errors in the flux density account for thermal noise, as well as the time-
variable uncertainty in the gain, which is estimated from the data themselves.
2.6. IRAM-30m Telescope: 240 GHz
Observations with the IRAM-30m telescope at Pico Veleta, Spain, were carried out on 2007 April 1-
4. Because of the low elevation of Sgr A* at Pico Veleta, gain drifts due to atmospheric fluctuations
are the most severe limitation to accurate flux monitoring. For the same reason, accurate peak-up
is important if flux variations are to be measured that are small with respect to the quiescent
flux. To account for both requirements, we alternated between Sgr B2 and Sgr A* with the
following procedure, applying a wobbling secondary mirror to remove the 240 GHz emission from
the atmospheric and from extended (> 70”) source structure. First, we pointed at Sgr B2 and
measured the position of its point-source component by fitting simultaneously a Gaussian and a
linear baseline (for a refined removal of extended emission) to the pointing subscans taken in on-
the-fly mode (two in azimuth direction, two in elevation). The positional correction was entered
and the procedure repeated to recover the correct flux. We used either the azimuth or elevation
subscan, depending on where the flux was larger (and thus a better peak-up was provided). Then
the antenna was moved to Sgr A*, where the same procedure was repeated. Thus, for each time
sample, there are two data points representing the flux of Sgr A*, one from the pointing, and
another one from the peaked up pointing. Both results were used if the pointing correction was
sufficiently small. Error estimates were made by comparing the results of subscans in the same
direction. To avoid effects due to instrumental and atmospheric gains drifts, only scaled fluxes
of Sgr A* were retained for further analysis, using Sgr B2 as a non-variable flux reference. Data
reduction was done with the MOPSIC software package19The average Sgr B2 flux density is
estimated to be 38.0±1.2 Jy and was derived using the HII region G10.62-0.38 as absolute flux
reference. The beam FWHM is 11′′2.
– 10 –
2.7.SMT: 250 GHz
Observations were undertaken at the SMT located at 3200m altitude on Mount Graham in eastern
Arizona Baars et al. (1999) using the 250 GHz channel of the facility’s four color bolometer Kreysa
(1990). This bolometer was used to observe at 250 GHz with a broad band, ranging between 200
and 290 GHz on 2007 April 1–4. We made use of the telescope’s beam switching mode, chopping
horizontally ±2′with the subreflector at a rate of 2 Hz along with an “off-on-on-off” observing
mode that shifted the position of the telescope every 10 seconds to remove any asymmetries in
the observations due to the chopping. Jupiter, Saturn, and Mars were used for focus and pointing
references confirming the telescope’s typical half power beam width at these frequencies of 30′′and
pointing accuracy of 2′′. While Jupiter was used to set the gain of the bolometer and skydips to
find the atmospheric opacity, NRAO 530, 1757-240 and G34.3 were also observed throughout the
observations as secondary calibrators to check the stability and repeatability of the measurements.
Finally, after splitting the data into 80 second increments (consisting of two iterations of the 40
second long “off-on-on-off” observing mode), the raw data were reduced using a version of the
standard GILDAS NIC reduction program customized for the four color bolometer. Because a
single calibrator was not used continuously during the first two days of observations, the flux
variation of Sgr A* was uncertain and thus the data are not presented here.
2.8. NMA: 150 & 230 GHz
Interferometric NMA observations were carried out simultaneously at 90 and 102 GHz in the 3-mm
band, and simultaneously at 134 and 146 GHz in the 2-mm band with bandwidth of 1024 MHz
on 2007 April 1–4. The 2 and 3 mm flux densities are measured to be 1.8±0.4 and 2.0±0.3 Jy,
respectively. The light curves of data from April 3 and 4 are presented using five and six antennas,
respectively. Th weather was bad on April 1 and 2 so we discarded the data on the first days of
observations. 1744-31 (J2000) was used as the phase calibrator and the data was binned every 3-4
minutes. The flux measurements of Sgr A* were estimated by fitting a point source model in the
uv plane restricted to distances > 20 kλ, in order to suppress the contamination from extended
components surrounding Sgr A*. The FWHM of the synthesized beam in the 2mm observation on
2007, April 4 is 6′′× 1.4′′. We used 3C279 as a passband calibrator and Neptune as the primary
2.9. CARMA: 94 GHz
Interferometric CARMA observations were done to observe Sgr A* on 2007, April 2-5. Observations
were made at 94 GHz using nine 6m diameter BIMA and six 10m diameter OVRO antennas with
the exception of observations on 2007, April 3 which did not include any OVRO antennas. In all
days, Uranus was used as the primary flux calibrator, 1744-312 as the complex gain calibrator and
1751+096 was used as a passband calibrator. The weather was poor for observing at 94 GHz on
the first half of 2007, April 3 and the second half of 2007 April 4. We did not include the data
during these times. Five frequency windows, each 469 MHz wide, were used at frequencies from 94
to 100 GHz. We used NRAO530 (1730-130) to cross calibrate 1744-312, in order to independently
track the amplitude stability of 1744-312. All calibration was done using MIRIAD package and
calibrated visibility data for each day were read into AIPS and the DFTPL task was used to extract
– 11 –
light curves for the source.
2.10. GMRT: 1.28 GHz
We observed Sgr A* using Giant Meterwave Radio Telescope (GMRT) in 1280, 610 and 325 MHz
frequencies with central observation time on MJD 54195.0, 54191.1 and 54190.1 (5.0 April, 1.1
April and 31.1 March 2007 UT) respectively. GMRT20consists of thirty fully steerable parabolic
antenna array, where fourteen antennas are randomly distributed in 1 km area and rest of the
sixteen antennas are placed in three arms, spread over 25 km area, forming nearly a shape like ‘Y’.
The diameter of each antenna is forty five meter. Observation band-width in each frequency was
32 MHz and integration time was 16.9 second. The source was observed for 6.1, 4.2 and 7.0 hr in
1280, 610 and 325 MHz respectively. We have done flux calibration using 3C286 and 3C48 and used
Baars et al. (1977) for setting flux density scale. J1830-360 was used as phase calibrator. The bad
data and radio frequency interferences (RFIs) are eliminated from the data set and the source is
self-calibrated. The original data has channel width of 125 KHz in the spectral line mode. To take
care of effect of the band width smearing in low frequency, we did not averaged all the channels
after calibration but averaged 32, 16 and 8 channels in 1280, 610 and 325 MHz respectively (forming
effective channel width of 4, 2 and 1 MHz in 1280, 610 and 325 MHz). The images are corrected
for the beam-shape. Because of the strong emission from the nonthermal emission surrounding Sgr
A*, the light curve of Sgr A* could have not been obtained reliably at these low frequencies 330,
630 and 1280 MHz. This is mainly due to the instantaneous elongated beam shape which contains
extended structures surrounding Sgr A*.
3. Light Curves: Individual Telescopes
The results of XMM and VLT observations in X-ray and NIR have already been presented elsewhere
(Porquet et al. 2008; Dodds-Eden et al. 2009). A detailed account of INTEGRAL observations are
given elsewhere (Trap et al. 2009). To present all the data that were taken during this campaign,
we include the XMM and VLT light curves again here and briefly review the results of these
observations that have already been published elsewhere.
3.1. NICMOS Photometric Measurements
Figure 2a shows the observed variability of Sgr A* in the NICMOS 1.45 and 1.70µm bands, where
there is good agreement between the two bands. The observed “quiescent” emission levels of Sgr
A* in the 1.45 and 1.70µm bands are ∼ 32 and ∼ 38.5 mJy, respectively, but some fraction of
this total signal is due to the neighboring star S0-2. During flare events, the emission is seen to
increase by anywhere from a few percent to 25% above these levels. In spite of the somewhat lower
signal-to-noise ratio for the 1.45µm data, due to the somewhat lower sensitivity of the NICMOS
detector and increased effects of extinction, the flare activity is still easily detected in this band.
In order to confirm that the observed variability of Sgr A* is not due to either instrumental or
data reduction effects, we have compared the Sgr A* light curves to that of the star S0-2 and to a
– 12 –
region of background emission with the NICMOS images, as shown in Figure 2b. The photometric
measurements for Sgr A* show obvious signs of variability in six of the seven windows of HST
observations, while the corresponding light curves of S0-2 and the background remain quite stable.
The panels of Figure 3a-e present detailed light curves of Sgr A* and, for comparison, S0-2 for
each of the seven HST observing windows. These plots show the time-ordered measurements in the
1.45 and 1.70µm bands, where we have now subtracted the mean “quiescent” flux level, leaving the
net variations in emission for both Sgr A* and S0-2. All light curves are aperture and extinction
corrected. Each observing window consists of 5 to 7 HST orbits, with each orbit covering ∼ 46
minutes. We have identified flaring activity in at least one orbit in each of the seven observing
windows. These activities are identified in the light curves with labels designating the day (1–7)
and the flare even within the day (A–C). A typical flare event lasts between 10 and 40 minutes.
The amplitudes and durations of the events are similar to what was found in our earlier HST
observations (Yusef-Zadeh et al. 2006a).
To examine the short time scale variability in more detail, the 1.70µm light curves are shown in
Figure 4 with a sampling of 64 seconds. The Sgr A* and S0-2 light curves are qualitatively similar
to those in Figure 3, except for the finer sampling and we show only the 1.70µm band because
the 1.45µm data do not have sufficient signal-to-noise in this shorter integration period. There are
16 identified flaring events, all of which are shown in 45min periods in two panels of eight flares.
One type of fast fluctuation that we have detected is generally associated with the rise or fall of
bright flares, or at the peaks of bright flare emission, as seen for the flares 1A, 2A, and 5A. Similar
minute time-scale variability has also been detected by Dodds-Eden et al. (2009). Another type of
fluctuation is the point-to-point variability seen during some of the quiescent phases of low-level of
activity, such as flares 1B, 1C, 2B, 3A, 4B, 6A, and 7A.
3.2. VLT NIR and Mid-IR Observations
The VLT observations used multiple bands to observe Sgr A* on 2007 April 1–7, using the two
instruments NACO (NIR) and VISIR (mid-IR). The results of these observations, which included
the identification of seven flaring events are discussed in detail by Dodds-Eden et al. (2009). The
brightest flare detected at 3.80µm coincides with the brightest X-ray flare on April 4. Figure 5
shows a composite light curve of VLT observations with labeled flares using 3.8µm, 2.12µm and
1.66µm NIR bands. No NIR spectral index measurements are available for the detected flares.
However, a 3σ upper limit of 57 mJy is placed at 11.88µm for the bright 3.8µm flare on April 4
with a peak flux density of ∼30 mJy (see also Trap et al. 2009). The brightest NIR flare detected in
this campaign consists of a cluster of overlapping flares that last for about two hours. The second
brightest flare detected by the VLT is identified as #6 in Figure 5. This flare precedes the bright
NICMOS flare 5A (April 5), as shown in Figure 3e. These flares are components of another period
of flaring activity lasting for about two hours.
3.3. X-ray Flaring Activity
The X-ray light curves between 2 and 10 keV with a time binning of 144s are shown in Figure 6.
A total of five flares were observed: one in 2007 April 2 (labeled #1) with a peak X-ray luminosity
L2−10keV = 3.3 × 1034erg s−1and four on 2007 April 4 (labeled #2, #3, #4, #5) with peak
– 13 –
L2−10keV=24.6, 6.1, 6.3, and 8.9×1034erg s−1, respectively (Porquet et al. 2008). For the first
time, within a time interval of roughly half a day, an enhanced incidence rate of X-ray flaring
was observed, with a bright flare (#2, with a duration of 2900 s) followed by three flares of more
moderate amplitude (#3, #4, #5, with durations of 300, 1300, and 800s respectively). An enhanced
rate of X-ray flares, although with lower amplitudes, was also reported in Belanger et al. (2005)
when one moderate and two weak flares were detected within a period of eight hours. These rates
of X-ray activity (Porquet et al. 2008; B´ elanger et al. 2005) are clearly higher than the typical duty
cycle of one X-ray flare a day (Baganoff 2003). The brightest event on 2007 April 4 represents the
second-brightest X-ray flare from Sgr A* after the X-ray flare with Γ =2.2 ±0.3 on 2002, October 3,
on record with a peak amplitude of about 100 times above the 2–10keV quiescent luminosity21This
bright X-ray flare exhibits similar light-curve shape (i.e.,nearly symmetrical), duration (∼3 ks) and
spectral characteristics to the very bright flare observed on 2002, October 3 with XMM-Newton
(Porquet et al. 2003). Its measured spectral parameters, assuming an absorbed power law model
including the effects dust scattering, are NH= 12.3+2.1
quoted errors are at the 90% confidence level. Therefore, the two brightest X-ray flares observed
so far from Sgr A* exhibited similar soft spectra Γ ∼ 2.2 − 2.3. The spectral parameter fits of the
sum of the three following moderate flares, while lower (NH= 8.8+4.4
are compatible within the error bars with those of the bright flares. However, fixing the column
density at the value found for the brightest flare (i.e. NH= 12.3 × 1022cm−2) leads to a larger
photon index value for the sum of these moderate flares, i.e. Γ = 2.1±0.4.
−1.8× 1022cm−2and Γ =2.3 ±0.3 where the
−3.2×1022cm−2and Γ =1.7+0.7
3.4. 43 GHz Time Variability: VLA
Figure 7a,b shows light curves measured during April 1–4 at 43 GHz with the VLA, using 87sec and
300 sec sampling, respectively. The light curve of the phase calibrator 17444-31165, which itself is
cross calibrated by NRAO530, is flat and is shown at the bottom of each panel in Figure 7a. Since
NRAO530 is not the primary calibrator, it provides a second check on instrumental stability and
that its light curve was flat also.
The light curves of Sgr A* show variations on a variety of time scales from as short as 30 min to
longer than five hours at 43 GHz. The fluctuations on time scales of several hours ∼ 5−6 hours can
be seen in Figure 7a,b. The slow flux variation over 5-6 hours could, in principle, result from the
contamination of the emission by an asymmetric distribution of extended structures surrounding
Sgr A* especially when a compact configuration of the VLA is used. However, the contamination
of flux by extended emission is minimal for uv data > 90kλ (or 2.3′′) and the variability on several
hour time scale is intrinsic to Sgr A*. Previous high resolution data taken with a wide configuration
of the VLA have also shown the presence of flux variation of Sgr A* on such time scales (Yusef-
Zadeh et al. 2006a,b and 2008). The contamination of extended emission was clearly seen at low
elevations in the uv data < 90kλ at 43 GHz, and our 22GHz data taken simultaneously with 43
GHz data on 2007 April 1-4 were useless for time variability analysis because of the limited uv
range (i.e., < 70kλ).
Most of the power of the 43 GHz fluctuations in four consecutive days of observations appears to
fall in a range between 30 minutes and few hours, as best shown all light curves of Figure 7b. For
21No detection was made using INTEGRAL in the 20–40 keV and 40–100 keV energy bands, leading to 3σ upper
limits of 2.63 and 2.60×1035ergs s−1, respectively (Trap et al. 2009).
– 14 –
example, fluctuations with ∼1h time scale are detected at a level of 200 mJy in the April 1 and
April 2 light curves centered near 13h and 11:15h UT, respectively. The light curve of April 4
shows largest flux variations at a level of ∼40% are seen to increase flux density from 1.1 Jy at 9h
UT to 1.6 Jy near 15h UT. Another interesting feature of the April 4 light curve is the presence of
multiple weak fluctuations at a level of 50 mJy on a time scale of ∼20-30 minutes. Figure 7c shows
the light curves of April 4 for simultaneous observations at frequencies of 43.1851 GHz and 43.5351
GHz with a 30sec sampling time. The frequency separation between these light curves is 345 MHz.
We note at least five 20–30 minute fluctuations that are seen in both light curves. A more detailed
account of the power spectrum analysis of the time variability of Sgr A* in radio wavelengths will
be given elsewhere.
3.5. 14, 22 and 43 GHz Time Variability: VLBA
Figure 8a shows 43 GHz light curves based on VLBA observations on April 1, 5 and 11, with a 60
sec sampling time. Figure 8b shows the light curves at 22 GHz and 43 GHz on April 2 whereas
Figure 8c shows the light curves at 15 GHz and 43 GHz on April 10. The flux density of Sgr A* show
variations on several hour time scales in these VLBA observations at multiple frequencies. These
light curves show the first measurements of the flux variation of Sgr A* on a VLBA (milli-arcsecond)
scale at several frequencies.
Fluctuations in phase coherence and amplitude errors could produce significant changes in flux on
short timescales. However, it is unlikely that calibration errors are similar at two frequencies, thus
the flux variation on ∼5-hour time scale (Fig. b,c) is intrinsic to Sgr A*.
126.96.36.199 GHz Light Curve: VLBA and VLA Comparison
Because VLBA and VLA measurements on April 1 and 2 are taken simultaneously at 43 GHz, we
compared the two light curves, as shown in Figure 9a,b with a 300 sec sampling time, respectively.
The comparison of the light curves examines directly the localization of flaring events at radio
wavelengths. The largest fluctuations in both VLA and VLBA light curves appear to agree with
each other. Peaks with hourly time scale durations occur in both light curves near 13h UT on
April 1, as seen in Figure 9a. Similarly, the slow decreasing trend in the flux of Sgr A* over few
hours is seen in the light curves of 2007 April 2 at 43 GHz using both the VLA and VLBA, as
shown in Figure 9b. The behavior of the light curves on hourly time scales measured with VLBA
provides the first direct evidence that flaring activity arises from the innermost region of Sgr A* on
milliarcsecond (mas) scales. The size of the flare emission is dominated by interstellar scattering.
The general agreement between the VLA and VLBA light curves imply that flaring region that has
been detected is unresolved with the VLA.
There are also discrepancies between the two light curves. One is the different values of “average
levels” of flux taken in the light curves measured with the VLA and VLBA. In all the measurements
shown in Figures 7a and 8a,b the average-level of VLBA flux appears to be lower than than that of
the VLA by ∼ 200 mJy. The second discrepancy is the flux variations do not agree with each other
on small time scales in VLA and VLBA light curves. The uncertainty in the absolute flux density
calibration of Sgr A* at 43 GHz using VLBA and VLA could easily explain the first discrepancy.
It is possible that the emission from Sgr A* could be contaminated by extended emission from the
– 15 –
surrounding medium, as measured with the VLA, even though we have selected data with uv >
100kλ. This could explain why the VLA and VLBA light curves do not agree with each other on
10-15 minute time scales. Lastly, it is possible that these discrepancies could be explained by a
core-halo structure of emission from Sgr A* in which the halo component is resolved out in VLBA
observations. Future simultaneous VLA observations using its most extended array configuration
and VLBA should be able to examine closely the reason for these discrepancies.
3.6. CSO 350µm, 450µm, 850µm Light Curves
Figure 10 shows the light curves at three submillimeter wavelengths. The data have been smoothed
to increase the signal-to-noise ratio with a sampling time of ∼6.5 minutes. As at radio wavelengths,
the flux of Sgr A* appears to be varying on hourly time scales. The largest Sgr A* increase is
detected at the beginning of the observation near 13:30 UT on 2007 April 1. These light curves
show evidence for hourly and intraday variability at 450µm, at a level of 14%. The mean daily flux
of Sgr A* at 450µm is ∼ 3 ± 0.25 Jy.
Figure 10b shows some of the first variability of Sgr A* at 350µm. The flux increase on 2007,
April 4 over 5 hours is about 50% of the initial flux of Sgr A*. This steady increase of flux density
over several hours is seen to continue at 90 GHz (see section 3.10). Figure 10c shows the light
curve at 850µm on 2007, April 6. unlike the other submillimeter light curves shown here, this
light curve appears to show time variability on a time scale of ∼ 10 minutes as seen near 13h:20m
UT. Such sharp variations, though with low signal-to-noise values, at 850µm at such a short time
scale resembles the recent light curve obtained with a different instrument (LABOCA of APEX)
(Eckart et al. 2008). The reality of such a short time scale variation needs to be confirmed.
3.7. SMA: 230 GHz Light Curves
Figure 11 shows the light curves taken from four days of observations with the SMA at 230 GHz.
The 2007 April 1 data shows an asymmetric profile indicating a duration of possibly ∼ 4 hours
considering that there is a gap between 16h and 17h:30m UT. Similar submillimeter characteristics
have been seen recently at 850µm (Yusef-Zadeh et al. 2008). The 2007 April 3 light curve shows
an emission peak near 14h UT, before a slow decay that lasts for about 4 hours. The light curve
obtained with SMT on the same day and at the same wavelength showed the rising part of the
light curve suggesting that the duration of the flare on this day could be as long as 8 hours. The
2007 April 4 light curve shows a typical profile of submillimeter flare emission, except for a dip in
the flux at a level of 100 mJy near 14h UT. The April 5 data shows a light curve with multiple
peaks as the light curve decays. The typical time scale for this variation is between ∼1-2 hours.
The overall percentage of flux variation during 6 hours of observations is between 10% and 30%.
3.8. SMT 250 GHz Light Curves
The SMT light curves of Sgr A* and calibrators (in blue) for 2007, April 1-4 are shown in Figure 12.
Because SMT and SMA observed Sgr A* at the same wavelength considering the broad bandwidth
of the SMT, we compared the SMT light curves with those of SMA on April 1, 3 and 4. An increase
in flux of ∼1 Jy in the rising part of the light curve is seen between 10h UT and 14h UT on 2007,
– 16 –
April 3. This increase is similar to the decrease in the flux of Sgr A* found in the decaying part of
the SMA light curve, as seen in Figure 11.
The low spatial resolution of the SMT results in a higher background level for the Sgr A* light
curve. The discrepancy in the zero level flux of Sgr A* using SMT and SMA on April 3 and 4 is
due to the fact that the emission from Sgr A* using SMT is contaminated by 3.5±0.2 Jy of flux
from extended features. If this flux is subtracted from the April 3 data and combined with the
SMA light curve, the duration of the variability is estimated to be ∼6 hours.
The 2007, April 1 shows the most dramatic flux variation of ∼2 Jy The reality of this feature can
not be confirmed as different calibrators were used at the beginning of the observation. However,
a 3.6±0.2 Jy subtraction from the SMT data matches well with the SMA data, thus suggests that
the sudden rise 11h UT is likely to be real.
3.9.IRAM 240 GHz Light Curves
Figure 13 shows the light curves of the two days of IRAM observations on 2007 April 3 and 4.
There is no evidence for any flux variations in these observations. The April 4 light curve overlaps
with the biggest NIR/X-ray flare detected during this campaign. However, there is no indication
that the 240 GHz flux density changed by more than 1.5±0.5 Jy between 5h and 7h UT, during
which the bright NIR/X-ray flare took place.
3.10.NMA 140 GHz and 230 GHz Light Curves
Figure 14 shows the NMA light curves of Sgr A* and 1744-312 based on two days of observations
under excellent weather conditions. The flux of the calibrator remains flat during these observations,
whereas the flux of Sgr A* increases by ∼0.5 Jy at 90 GHz on April 3. The April 4 light curve
shows a slight increase at 146 GHz before decaying strongly by more than 1 Jy. The duration of
the flare is roughly 2 hours.
3.11. CARMA Light Curves
Fig 15 shows the light curves of Sgr A* and the calibrator 1733-130 at 94 GHz taken for four days
of observations on 2007, April 2-5. There is flux variation on short and long time scales in all days
of observations. There is concern on the variation of the calibrator evident in almost all days of
observations. Due to this uncertainty, we compared the light curves with other 94 GHz and 43 GHz
measurements and we believe the large scale variation may reflect the intrinsic variable emission
from Sgr A*. However, the flux variation of Sgr A* on short time scale may not be valid.
3.12. GMRT Light Curves
As we observe Sgr A* at long wavelengths, the light curve of Sgr A* may be contaminated by
the extended nonthermal emission surrounding Sgr A* as well as by interstellar scattering which
becomes more important at long wavelengths. In order to avoid the contamination by extended
emission, we used the uv data at the highest elevation as well as restricted the uv distribution
– 17 –
be greater than 80kλ. Also, interstellar scintillation is expected to operate on longer time scales
than hourly time scales that we are sensitive to. A flux variation of ∼80 mJy over four hours was
detected. Given the limited resolution of the GMRT data at this frequency, it was not clear if this
variation reflects the flux of Sgr A* at 1.28 GHz or an artifact of the contamination of extended flux.
It is clear that higher resolution data are needed to separate Sgr A* from the extended features in
4.1. NIR Flare Statistics
Given the ability of HST to produce continuous observations over many 45 min orbital visibility
periods, along with its long-term photometric stability, the NIR NICMOS data provide an excellent
way to investigate the flare strength distribution over many flare episodes. Figure 16 shows a his-
togram of the NICMOS 1.70µm net flare emission for the 7 days of data obtained in this campaign.
The net flare emission is measured by first subtracting the background emission for each day before
the excess flux above the background is selected. Thus, the selected data points do not sample the
peak flare emission but rather the flux associated with flaring activity. The peak of values centered
at a net flux of zero represents the emission from Sgr A* during “quiescent” periods. The positive
half of the histogram, on the other hand, shows a tail of flare emission events extending out to
∼10 mJy. The “quiescent” distribution is best fitted with a Gaussian, which is expected from the
level of random noise in the observations. The tail of flare emission can be fitted with a power-law
distribution having an index of -1.19±0.27 and a low-energy cutoff at Sν= 1 mJy. The dotted line
in the figure shows the result of simultaneous Gaussian and power-law fits to these two components.
Yusef-Zadeh et al. (2006a) reported that distribution of flare activity seen in our more limited 2004
NICMOS observations could be fitted by two simultaneous Gaussians profiles. A reanalysis of
those data, however, now show that a power-law distribution with a low-energy cutoff yields a good
fit to the 2004 epoch data as well. Figure 17 shows a histogram of the 2004 1.60µm data, with
Gaussian and power-law fits to the two components shown by the broken lines. The best power-law
fit to these data has an index of -1.11±0.13, with a low-energy cutoff of Sν= 0.25 mJy. This is
remarkably consistent with the best-fit power-law index of the 2007 data. We note that the fraction
of observing time that flare activity has been detected in the 2004 and 2007 campaigns is ≥ 32%
and ≥ 37%, respectively.
The NIR flare histograms for the two epochs show that the probability of measuring flux Sνat any
instant is approximately proportional to 1/Sν. Presumably this reflects the statistics of the flaring
behavior of Sgr A* at NIR wavelengths. To explore this we construct a simple phenomenological
model for the flaring by simulating a light curve and then sample it to construct a simulated
histogram. This model shows that the observed 1/Sν behavior arises quite naturally, but does
constrain the statistics of the flaring.
Our phenomenological model represents the flaring as a sequence of 100 Gaussian profiles occurring
over 100 arbitrary time units, with flare i characterized by peak flux Si, timing of the peak ti, and
standard deviation of the flare σi, so that the net light curve may be written as
−(t − ti)2
– 18 –
The parameters Si, ti, and σiare drawn randomly and uniformly from the ranges [0,1], [0,100] and
[0,σmax], respectively, and the resulting light curve is evenly sampled every 0.2 time units to create
a flare histogram. Note that σmaxis the only independent parameter of this model, as increasing
the number of flares and changing the maximum flare amplitude can be accommodated by rescaling
the flux and time units. In addition, changing the sampling rate or the number of flares does not
affect the statistics, provided that the light curve has already been adequately sampled (which is
the case for our adopted sampling rate of 50 per time unit). We find that σmax<
∼0.5 yields the
A typical simulated light curve and the corresponding histogram for σmax= 0.5 are given in Figures
18a and b, respectively. The slope of S−1
is drawn on Figure 18b. Larger values of σmaxlead to
significant overlap between flares, tending to give a flatter dependence of the flux probability on
Sν. This does not, of course, prove definitively that the flares behave as given by equation 1. Other
choices of functional form or different statistics for Si may also yield the 1/Sν behavior of the
histogram. It does, however, seem to require that the flare events do not significantly overlap each
4.2. Spectral Index Distribution Between 1.45µm and 1.70µm
We have constructed a log-log distribution of spectral index based on the NICMOS 1.45µm and
1.70µm data. Figure 19 shows the ”color” distribution of all the data selected with signal-to-noise
S/N=3. The diagonal line (in red) shows the spectral index of β=0.6, where Fν ∝ ν−β. For
comparison, β of -4, -2, 2, and 4 are also plotted. This figure shows a tendency for the spectral
index of low flux values to be steeper than 1, whereas the high flux values are represented by a
flatter distribution of spectral index. Because the data points used in making Figure 19 are not
taken simultaneously at the two different wavelengths, we attempted to estimate spectral index
values of adjacent data points, where the flux of Sgr A* is not varying rapidly, such as during the
fast rise or fall of individual flares. The 1.45 and 1.7µm NICMOS images were acquired back-to-
back in long sequences, in which the exposures within each pair are separated in time by about
2.5 minutes. The points shown in Fig. 19 represent all adjacent pairs of measurements (adjacent
meaning ∼2.5 minute separation) for which the S/N in the individual measurements is greater than
3 (hence not all available pairs from Fig. 3 are included). The fact that the overall Sgr A* flux
could be changing within that 2.5 minute time scale is a concern and is why the spectral index
values listed in Table 1 where taken from only those measurements where we could see from the
light curves that the overall Sgr A* flux was not changing much on that ∼2.5 minute time scale.
The full light curves also indicate that the overall flux of Sgr A* does not often change on such short
time scales and therefore the number of suspect measurements in Fig. 19 should be a relatively
small fraction of all measurements. Hence we only deduce general trends from that diagram.
We identified five sets of data points associated with five different flares during which the overall
Sgr A* flux is not varying rapidly. Table 1 shows the corresponding flux and spectral index values
using data sampled at 144 sec intervals. The two brightest flares, 5A and 2A, have spectral indexes
0.73±0.16 and 0.97±0.27, whereas the weaker flares have indexes steeper than β=1.5. These indi-
vidual measurements are consistent with the spectral index trend shown in Figure 19. We also find
that the spectral index of the brightest flares are consistent with recent Keck measurements, which
yield a spectral index of 0.6 (Hornstein et al. 2007). The spectral index of low flux values is also
consistent with VLT measurements, which show a steep spectrum for weak flares (Eisenhauer et al.
– 19 –
2005; Gillessen et al. 2006). These measurements suggest that the spectral index of flares varies
with the NIR flare strength, support earlier measurements by Gillessen et al. (2006) and disagree
with measurements by Hornstein et al. (2007) who claim a constant spectral index in NIR wave-
lengths. The variation of spectral index with flare emission at NIR wavelengths has important
implications on the inverse Compton scattering mechanism of X-ray and soft γ-ray emission from
Sgr A* (Yusef-Zadeh et al. 2006a) as well as on the hypothesis that X-ray emission is due to syn-
chrotron mechanism (Dodds-Eden et al. 2009). It is possible that weak flares with a steep energy
index of particles are associated with low-level activity of the accretion disk of Sgr A*, whereas the
bright flares represent the hot magnetically-dominated events that are launched from the disk. Po-
larization characteristics of the weak and strong flares may constrain models of the flare emission.
The correlation of the spectral index and flux has been discussed in the context of electron heating
and cooling by a turbulent magnetic field (Bittner et al. 2007). In the synchrotron scenario, the
higher value of the spectral index at low NIR fluxes could be an indication of the cooling break.
4.3. NIR Power Spectrum Analysis
Genzel et al. (2003) had reported a possible 17 min NIR periodicity with implications for the
spin of the black hole. Our previous 2004 HST data (Yusef-Zadeh et al. 2006a) showed a marginal
detection of power at 33±2 minutes. We investigated the power spectrum of flare data taken with
the new NICMOS measurements. We created Lomb-Scargle periodograms (Scargle 1982) to search
for periodicities in our unevenly-spaced NIR measurements. We performed 1000 simulations of each
light curve, with the same sampling and variance as the data, and with simulated noise following
a power-law (P(f) ∝ f−δ) chosen to match the periodogram of the data as closely as possible
(following Timmer & Konig 1995, Mauerhan et al. 2005), typically with an index δ of 1.5 or 2.
Artificial signals are seen at the 90 minute orbital period of HST, and the 144 second filter switching
cycle, and discounted. For each point in a lightcurve, we identify the periodogram simulation at
the nth (where n=99, 99.9) percentile of the distribution, and thus create lines below which n% of
the simulations fall. Figure 20 shows the power spectrum as a solid line and the dotted lines show
the the spectrum of the noise using power-law distributions. Only one HST observation shows any
power above the 99.9 percentile line, on 2007 April 4, near 2 hours.
The 99.9 percentile refers to the local distribution; however, the chance of getting a point above
the 99.9 percentile line must be computed considering all trials (Benlloch et al. 2001). We sam-
ple 158 frequencies above 10.8 minutes, the lower limit of our simulation software, and perform
seven observations, so our total is 1106 observations, suggesting ∼1 peak above the 99.9 percentile
line. We have two adjacent points above the 99.9 percentile line, but these points are probably
not independent. Altering the index δ within a range consistent with the data does not change
the strength of the signal. We conclude that the significance of this possible periodicity is not
The lack of any significant power between 17 and 20 minutes supports the results from an earlier
analysis of HST data in 2004 (Yusef-Zadeh et al. 2006a). Recent analysis of data taken with the
combined VLT and the Keck observations shows no significant power on short time scales (Do et al.
2008; Meyer et al. 2008).
– 20 –
5.1. X-ray Flare Emission Mechanism
As described in §3.3, five X-ray flares were detected in the present observing campaign.
strongest X-ray flare (#2) coincided with a strong NIR flare observed with the VLT (Dodds-Eden
et al. 2003). HST observations detected three of the remaining four X-ray flares (#1, #4 and
#5) corresponding to flares 2A, 4A and 4B, respectively. Figure 21a,b show the X-ray light curves
of these newly detected NIR flares at 1.70µm. Table 1 presents the flux and spectral index of
the NIR flare 4A. The remaining X-ray flare detected in this campaign (#3) was not observed
contemperaneously in the IR. The simultaneous monitoring of Sgr A* in X-rays and the IR have
shown that X-ray flares are always accompanied by flaring in the IR but that the reverse is not
necessarily true (Porquet et al. 2009).
The NIR flare emission from Sgr A* have been shown to be highly polarized (Eckart et al. 2006a;
Meyer et al. 2006) and is therefore likely to be produced by synchrotron emission from GeV elec-
trons in the ∼ 10G magnetic field strengths thought to be present in the vicinity of Sgr A*. The
30 minute duration of the flares is broadly in line with the synchrotron cooling time scale of these
electrons. Substructure in the NIR flare lightcurves has been attributed to doppler beaming and
lensing of an orbiting hotspot (Meyer et al. 2006), although this remains to be confirmed.
The X-ray flares are always seen in concert with flaring in the IR (when the IR has been simultane-
ously observed). Thus scenarios for the X-ray emission are directly associated with the acceleration
of the GeV electrons responsibly for the IR synchrotron emission, either through upscattering of
submillimeter seed photons (Markoff et al. 2001, Yusef-Zadeh at al. 2006, 2008), synchrotron self-
Compton (Eckart et al. 2006a) simply as synchrotron emission from the high energy tail of the
accelerated electrons (Yuan et al. 2003, Dodds-Eden et al. (2009) or from the NIR emitting elec-
trons (Eckart et al. 2006a). Dodds-Eden et al. (2009) have recently shown that the ICS scenario
requires an uncomfortably small submillimeter source source size (<
X-ray and NIR fluxes for flare #2. In addition, although the other flares can be modeled with
source sizes of a few Schwarzschild radii and field strengths in the 10G range (e,g., Yusef-Zadeh
et al. 2006), the energy density of the GeV electrons in the NIR-emitting region exceeds the mag-
netic energy density by more than an order of magnitude, and their acceleration and confinement
becomes problematic (this is also the case for flare #2). Finally, the duration of the X-ray flare is
shorter than the IR flare whereas one might expect them to be identical. Dodds-Eden et al. 2009
therefore strongly preferred the synchrotron scenario. This implies that the acceleration mechanism
must continuously resupply the 100GeV electrons for the 30 minute duration of the observed flares
as the synchrotron loss time of the ∼ 100GeV electrons responsible for the synchrotron emission
is ∼ 30seconds.
Here we consider an alternative ICS scenario: the upscattering of NIR seed photons emitted dur-
ing the flare by the mildly relativistic ∼ 10MeV electrons responsible for the quiescent radio-
submillimeter emission. If the submillimeter emission region were optically thin this would pro-
duce a similar X-ray luminosity as the upscattering of submillimeter seed photons. However, as
the submillimeter source region is optically thick below 1000GHz, the observed submillimeter flux
is produced by a fraction of the underlying electrons. The emission region is optically thin to NIR
photons, and so all of these electrons are available to upscatter NIR seed photons to X-ray energies.
As a result, the ICS luminosity produced through this scenario will dominate that produced by the
original ICS scenario.
∼Rs) to match the observed
– 21 –
To estimate the resulting X-ray flux we characterize the electrons responsible for the submillimeter
emission by electron number density ne, a relativistic Maxwellian energy distribution at temperature
T, and a quasi-spherical region of size R. These electrons upscatter the NIR seed photons arising
from synchrotron emission from the relativistic electrons producing the NIR flare with observed
flux Sνat the earth. The ICS flux depends on the direction-averaged intensity, Jν, of seed photons
which in turn depends on the location and size of the flare emission region; we estimate this to
order of magnitude by simply assuming that the flare region is of similar size to the submillimeter
emitting region, such that Jν= (d2/πr2)Sν, where d = 8kpc is the distance to the Galactic center.
The energy of the upscattered photons is small compared to the MeV-range of the electron energies,
so we can use the Thomson scattering cross-section. To a good approximation, scattering by an
electron with energy E ≫ mec2boosts the seed photon energies by a factor (E/mec2)2irrespective
of the scattering angle. The differential ICS luminosity per unit energy interval is then simply
where N(E)dE is the total number of electrons in the energy interval [E,E + dE], and Jνis the
direction-averaged intensity of the seed photons at frequency ν = Eγ/(E/mec2)2. For temperatures
in excess of a few MeV, the vast majority of the electrons have v ≈ c, so we approximate the
relativistic Maxwellian distribution by f(E) ≈ (2kT)−1(E/kT)2exp(−E/kT) and then N(E) =
3πR3nef(E). If the spectrum of the seed photons is a simple power-law, ie. Sν= S0(ν/ν0)−β, the
differential luminosity is
Γ(3 + 2β)S0neR
where Eγ0= (kT/mec2)2hν0.
By way of illustration we adopt reasonable choices R = 10Rs, ne= 107cm−3, and compute the
ratio of 2–10keV luminosity to 2.2µm flux as a function of spectral index β. The results for various
adopted electron temperatures are plotted as solid curves in Figure 22. The X-ray to NIR flux ratio
declines with increasing spectral index β (where Sν∝ ν−βcreated by upscattering of optical/NIR
photons from an electron population with radius R = 10Rs, and uniform density ne= 107cm−3,
and temperatures of 3, 5, 7, and 10MeV). Because X-rays in the 2-10keV band are produced by
upscattering of photons that are shortward of 2.2µm, and for fixed flux at 2.2µm there are less of
these as β is increased. Also shown for comparison are the measured ratios and spectral indices of
the 7 coincident IR and X-ray flares seen to date (Yusef-Zadeh et al. 2006; Belanger et al. 2005;
Eckart et al. 2006; Hornstein et al. 2007; Marrone et al. 2008; Porquet et al. 2008; Dodds-Eden et
al. 2008; this paper). The high value of β at high temperature correlates with a high ratio of X-ray
to NIR flux, as shown in Figure 22. This correlation is consistent with the low flux value of NIR
flare emission for high value of the spectral index, as described in section 4.2. We conclude that
the fluxes of the observed X-ray flares are broadly consistent with this ICS scenario.
Our simple model predicts that the spectral indices of the X-ray and IR flares should be identical,
but this not need be the case for a broken power-law electron energy distribution. In the case of
the strongest X-ray flare (#2) which coincides with a strong NIR flare, the NIR spectral index and
X-ray spectral index are different with βX−ray= 1.3 ± 0.3 (90% confidence) whereas βNIR< 1.0
(3σ) (Dodds-Eden et al. 2009). This could be explained by a broken power law of NIR emitting
electrons with a steeper spectral index shortward of 3.8µm, perhaps resulting from a shorter time
scale for synchrotron cooling of electrons at high energies. In this scenario, the flux in the 2-10keV
– 22 –
band is produced predominantly by upscattering of photons with wavelengths shortward of 1µm.
This may explain why the width of the bright 2007 April 04 flare is less in X-rays than in the
infrared: the infrared flare may have decayed more rapidly shortward of 1µm than at 3.8µm and
so the X-ray flux declines without a corresponding decrease at NIR wavelengths.
The extent of the submillimeter-emitting electron population may on occasion give rise to significant
time delay between infrared flaring and their X-ray counterparts. A sufficiently hard IR flare
would lead to X-ray production by inverse Compton scattering on the extended ∼ 1000Rsouter
envelope of low-temperature electrons, producing weak post-main-flare X-ray emission lasting for
tens of minutes after the main flare has subsided. Theoretically, the electron temperature is set
by a balance between heating by Coulomb interactions with protons and by plasma effects and
synchrotron cooling. kT for the protons is a reasonable fraction of their virial energy because of
inefficient cooling in the accretion flow. This implies that the protons should be non-relativistic at
∼ 1000Rs, with MeV-range energies. The electrons are likely to have similar energies because of
formalization and are then mildly relativistic. Empirically, Loeb & Waxman (2007) estimate from
the radio/submillimeter spectrum that the electron temperature is a few MeV all the way out to
1000Rs. Detection of these “echoes” would confirm the scenario proposed here and help determine
the size of the outer region which is rendered inaccessible to direct by the effects of interstellar
scattering in radio wavelengths.
5.2. Cross-Correlation of Light Curves
As pointed out in the previous section the adiabatic expansion picture of the flare emission from
Sgr A* makes the predictions that i) the NIR and X-ray emission are expected to be simultaneous
and therefore optically thin whereas optical depth effects become important at lower frequencies,
thus a time delay is expected between their peak emission. To examine these issues, a great deal
of data have been obtained simultaneously in this campaign, which allows us to cross-correlate the
multi-wavelength data for each day of observation. The light curves that we have presented thus far
indicate that the flux of Sgr A* is constantly changing as there is low-level flare activity in almost
all wavelength bands. There are very few measurements that were taken simultaneously with the
same time coverage with few exceptions, as described below.
The cross-correlation analyses in this paper use the Z-transformed discrete correlation function
algorithm (Alexander 1997); see also (Edelson & Krolik 1988). This algorithm is particularly useful
for analyzing sparse, unevenly sampled light curves. We identify the peak likelihood value, and
a 1-σ confidence interval around that value, using a maximum likelihood calculation (Alexander
5.2.1. NIR, X-ray and Radio flare Emission
Altogether there are four detected NIR flares that have shown X-ray counterparts. There is no
evidence that there is time delay between the peaks of any of the detected flares, thus supporting
the fact that both NIR and X-ray emission are optically thin. The lack of time delay places a
strong constraint on the ICS picture.
The strong flare and simultaneous coverage in both X-ray and NIR wavelengths observed on 2007
April 4 is one example in which a cross correlation peak with small error bars can be obtained. The
– 23 –
cross-correlation of the light curves at X-ray and NIR wavelengths is shown in Figure 23. The peak
of the cross-correlation shows that X-ray emission is delayed by 29 seconds with a one-sigma error
bar of -6.5 and +7.0 minutes. In the ICS picture, as described in section 4.4, the region from which
NIR photons are upscattered should be less than 0.8 AU. Dodds-Eden et al. (2009) presented first
the simultaneity of X-ray and L′bands to within one sigma error bar of three minutes.
The relationship between radio and NIR/X-ray flare emission has remained unexplored due the very
limited simultaneous time coverage between radio, infrared and X-ray telescopes. The continued
variations of the radio flux on hourly time scale also makes the identification of radio counterparts
to infrared flares difficult. In spite of this, the strong flaring in NIR/X-ray wavelengths on 2007,
April 4 has given us an opportunity to examine whether there is a correlation with variability
at radio frequencies. One of the key motivation of our observing campaign was to examine the
adiabatic expansion picture of flaring activity of Sgr A*. One of the prediction of this model is
a time delay between the peaks of optically thin NIR emission and optically thick radio emission.
This implies a NIR flare with its short duration is expected to have a radio counterpart shifted
in time with a longer duration. Given that there is zero time delay between NIR and X-ray light
curves, as shown in Figure 23, we argue below for a radio counterpart to a strong X-ray flare by
shifting and stretching the time axis of the X-ray light curve.
Figure 24a shows composite light curves of Sgr A* obtained with VLA, VLT, HST and XMM on
2007, April 4. The flux increase at 43 GHz is ∼40% which is higher than those from the first three
days of VLA observations which is ∼ 20%. We also know that there was no significant variation
at 240 GHz during the period in which the strong NIR/X-ray flare took place. The IRAM-30m
observation shows an average flux of 3.42±0.26 Jy between 5 and 6h UT when the powerful NIR
flare took place. The flux is mainly arising from the quiescent component of Sgr A*. Comparing the
light curves of the 43 and 240 GHz data, there is no evidence for a simultaneous radio counterpart
to the NIR/X-ray flare with no time delays.
We now argue that the radio flare detected between 10h and 15h UT is a time-delayed counterpart
to the NIR/X-ray flare for the following reasons: i) the highest percentage of the flux increase at 43
GHz on 2007, April 4 compared to other three days of radio observations, ii) Similar morphology
between radio and X-ray light curves as well the presence of three peaks in NIR and radio light
curves and iii) the lack of significant flux variation above the quiescent flux of Sgr A* at 240 GHz
during the NIR/X-ray flaring events. We suggest that flare emission at 43 GHz is time delayed
with respect the NIR/X-ray flare emission. To explore this further, we have empirically shifted and
stretched the time axis of the X-ray light curves by 5.25 hours and a factor of 3.5, respectively.
The shift and stretch operation to the time axis is carried out by eye and then examined by cross
correlating the time-shifted and time-stretched X-ray data and the 43 GHz light curves. The top
panel of Figure 24b presents the light curve of the time-shifted and time-stretched X-ray data. The
middle plot shows a baseline subtracted radio light curve. The subtraction is used to remove the
contribution by the quiescent flux. We find that the best fit shows a peak in the cross correlation
plot of 4.6+9.4
−7.6minutes which is consistent with zero. The 1-σ error to the cross correlation peak
of the shift is given in Table 2.
Given that NIR/X-ray and radio emission are expected to be optically thin and thick, respectively,
the similarity in the substructures in radio and X-ray and NIR light curves and the way that they
trace each other, as shown in Figure 24, are remarkable. For example, two main peaks before and
after 12h UT are detected in both NIR, X-ray and radio light curves. The dips near 13:30h UT
and 13h UT between radio and NIR are also seen in Figure 24a, To make a stronger case that
– 24 –
the X-ray/NIR and radio flares are related to each other, the morphological agreements between
radio and NIR could have been improved had we used a a varying time shift to different subflares or
components in the NIR light curve. In fact, the shift and stretch values measured here is not unique
as it is possible to decrease and increase the time-shifted and time-stretched values, respectively,
and yet obtain a reasonable zero time delay between radio and X-ray data. A detailed account of
these light curves in the context of adiabatic cooling plasma model will be given elsewhere.
Given that there is continuous coverage for about ten hours between 5h-15h UT in X-ray, NIR and
radio wavelengths using XMM, VLT, HST, IRAM and VLA with an exception of a two-hour gap
between 7 and 9h UT in radio wavelengths and a 2.5-hour gap between 10.5h and 13h UT in NIR
wavelengths, we believe the lack of association between flaring activity in NIR and radio wavelengths
is highly contrived. Obviously, we can not prove conclusively that radio flare seen on April 4, 2007
is associated with the NIR/X-ray flare because of two gaps in our coverage. Nevertheless, the
comparison of the 43 GHz light curve with the NIR data suggest that these variations are tied
closely with each other.
5.2.2. Other Cross Correlations
There were no simultaneous observations that were taken with the same instrument except with
the VLA and NMA but with a small frequency separation. In spite of the small separation between
the observed frequencies at 134 GHz (2.23 mm) and 146 GHz (2.05 mm), the data are taken
simultaneously on 2007, April 4 with the NMA. The cross correlation of the light curves at these
frequencies, as shown in Figure 25, peaks with 6+6.6
−4.8minutes time delay.
Due to the limited UT coverage with individual telescopes as well as the lack of strong flaring event
in this campaign (with the exception of the strong NIR/X-ray flare on 2007, April 4), the cross
correlation of the light curves had difficulty following accurately the time evolution of a flare as
a function of frequency. In spite of these difficulties, we have obtained cross-correlation plots of
low-level fluctuations evident in four light curves. Although most of the individual cross correlation
peaks have low signal-to-noise, the peaks of optically thick emission all show a tendency to lag rather
lead, thus, consistent with the adiabatic expansion picture of flare emission. We have selected the
best light curves to show the time lag but in fact almost all light curves systematically showed a
time lag rather than a lead in their cross correlation peaks, though low signal to noise ratios. We
believe the data presented here supports the plausibility of the time delay, as has also been shown
in earlier cross-correlation measurements. We give four examples that indicate higher probability
that the peak flare emission at high frequencies leads those at low frequencies.
Figure 26a presents the cross-correlation plot of the light curves taken at 450µm using the CSO
and 230 GHz using the combined data taken from the SMA and SMT on 2007 April 3. The cross-
correlation plot at the bottom of the panel shows a maximum likelihood time delay at 1.32+1.66
hours. Another example shows the evidence for a time delay between 1.2mm (230 GHz) and 450µm
wavelength bands on 2007, April 1 and the cross correlation plot is displayed in Figure 26b. The
cross-correlation peak between these wavelength bands is 0.24+1.48
−0.04hours time delay.
The cross-correlation peak between 1.70µm and 1.3mm wavelengths on 2007, April 5 is shown in
Figure 26c and gives a time delay 2.64 hours with a a 2σ uncertainty range of -1.66 to 3.3 hours.
The NIR data for this plot combined the 1.70µm data of HST (flare 5A of Fig. 3) and the 3.8µm
VLT data (flare 6 in Dodds-Eden et al. 2009). Because the spectral index of the HST data is
– 25 –
determined, we assumed that the preceding flare detected by the VLT at 3.8µm has the same
spectral index and its duration is continuous with the brightest HST flare emission seen in this
campaign. Lastly, Figure 26d shows the light curves taken with CARMA and VLA on 2007 April
02 at frequencies of 94GHz and 43GHz respectively. The cross correlation plot shows the strongest
peak with a time delay of 1.02+0.16
−0.31hours. Given that there are three peaks shown in the cross
correlation plot, there is ambiguity in the determination of the true time delay from the comparison
of the peaks alone. However, when the cross correlation is considered in the context of many other
cross correlations plots showing similar delays of peak emission at long wavelengths following those
at short wavelengths, it is clear that the primary peak signifies most likely the real time delay.
As discussed before, the strongest NIR/X-ray flare was detected on 2007, April 4 showing a peak
at 5.9h UT with a full duration of about two hours in NIR wavelengths. The 2.1mm (140 GHz)
light curve taken with IRAM during this period of flaring activity placed a constraint by showing
a lack of flux variation with a one-sigma error of 0.26 Jy at millimetre wavelength during a strong
5.3. Adiabatic Expansion of Hot Plasma vs. Hot Spot Model
We discuss two models that attempt to explain the nature of the flare emission. One is an ex-
panding hot plasma model in which the peak frequency of emission (e.g., the initial optically thin
NIR flare) shifts toward lower frequencies (submillimeter, millimeter and then radio) as a self-
absorbed synchrotron source cools adiabatically away from the acceleration site (Shklovskii 1960;
van der Laan 1966; Yusef-Zadeh et al. 2006b, 2008). A variation of this model is a jet model in
which the expansion speed of the plasma is relativistic and is collimated in the form of outflow.
In the expanding blob model, polarized flare emission does not follow classical Faraday rotation
and a frequency-dependent rotation measure (RM) is predicted (Yusef-Zadeh et al. 2007). The
expanding blob picture considers hot plasma being launched from the disk. The cooling plasma is
dominated by the magnetic pressure as the plasma escapes or remains bound to the system. The
second model assumes that flares are hotspots that are orbiting within few Rsof the black hole
(Broderick & Loeb 2006) where Doppler boosting and GR effects become important. The hot spot
picture requires the hot plasma be embedded within the disk where the emission is dominated by
the gas pressure in the disk before the hot spot plunges into the hole. Quasi-periodic flaring events
are expected under the assumption that hot spots survive longer than the period of the last stable
orbit. There are several issues that the hot spot model appears to be inconsistent with observations.
One is the time delay between the peaks of flare emission which is not expected in this picture. It
is possible that the optically thin and thick flare emission is not related to each other and that the
hot spot model is applicable only to the NIR flare emission. However, recent measurements as well
as measurements presented in this campaign indicate time delay between the peaks of NIR/X-ray
and submillimeter flare emission (Yusef-Zadeh et al. 2008; Marrone et al. 2008; Eckart et al. 2009).
Another difficulty with the hot spot model is the lack of evidence for power on the quasi-periodicity
in NIR light curve (Do et al. 2008, Meyer et al. 2008). Claims of quasi-periodic variations in the
NIR lightcurves were the motivating observation for the hotspot model, but the most thorough
analysis of NIR lightcurve variability have not shown evidence for significant quasi-periodic power
(Do et al. 2008; Meyer et al. 2008). Furthermore, MHD simulations of accreting gas indicate that
hot spots can last less than an orbital time scale before they disperse (Hawley & Balbus 2002).
A major motivation for the observational study of Sgr A* in the 2007 observing campaign was to
– 26 –
test predictions of the plasmon model of flare emission such as the time delay between the peak
emission at different wavelengths. Previous observing campaigns to monitor Sgr A* have found
evidence for time delay between the peak emission at 43 and 22 GHz. (Yusef-Zadeh et al. 2006b;
Yusef-Zadeh et al. 2007a). These measurements were consistent with the predicted time delays in
the plasmon model. As described above, we carried out the cross correlation of the peak of radio
flare emission and the emission at other wavelength bands using simultaneous space- and ground-
based observatories. These measurements would allow us to determine if radio flares are correlated
with flaring in the near-IR, X-ray and sub-millimeter. The data presented here show an increasing
chance of high frequency flare emission leading the low frequency emission when simultaneous data
between 43 GHz, 94 GHz, 230GHz, 0.45mm and 1.70µm are examined. It is only the collection of
the cross correlation plots that make the time delay between the peaks of flare emission compelling.
Another intriguing example that could be used against the orbiting hot spot model is the relation-
ship between radio and NIR/X-ray flare emission. The light curve at 43 GHz using the VLA began
covering the flare 4.5 hours after the start of the NIR flare on 2007 April 4. The morphology of
the peaks in the two light curves appears to show a rise of flux followed by flattening of the emis-
sion. Figure 24b shows the comparison of the X-ray and radio light curves by shifting the X-ray
light curve by about 5.25 hours and stretching by a factor of 3.5. We suggested that shifting and
stretching of the light curves serve as the time delay and duration of flare as it evolves in time. The
stretching of the NIR light curve by a factor of 3.5 can be viewed in the context of the expanding
blob model of an initial flare or a compact blob observed in NIR followed by the expansion of a
blob of hot plasma emitting in radio wavelengths. Individual NIR and radio flares show typical
durations of ∼20 min and 1-2 hours, respectively. The ratio of the observed durations is similar to
the stretching factor that was applied to the time axis of the NIR light curve of flare emission. A
more detailed account of this analysis in the context of the plasmon model will be given elsewhere.
Recently, Marrone et al. (2008) criticized the expanding blob model on the grounds that their
measured ratio of 1.3mm to 850µm flux during a flare on 2006 July 17 was higher than expected
from the blob model, both in the optically thick precursors (where one would expect a spectral
index of 5/2 rather than the measured value β = 0.1 ± 0.5), and in the ratio of the amplitudes
of the flares at the two wavelengths. However, the continual variability at radio and millimetre
means that there is a large uncertainty in determining the underlying background flux level for a
particular flare, and we have determined that there are reasonable choices of the levels that renders
flare profiles that are consistent with the plasmon framework. Fig. 27 demonstrates simultaneous
fit to both the 1.3mm and 850µm flux of the 2006 July 17 flares (see Figure 3 of Marrone et al.
2008) using the plasmon model. The parameters of the successful fits to the first flare peaking near
5.7hr UT are p = 1, R0= 0.52Rs, v = 0.011c, and B = 73G whereas the parameters of the strong
flare peaking near 7.5hr UT are p = 0.5, R0= 0.42Rs, v = 0.003c, and B = 75G. These fits show
clear evidence that a simple picture of plasmon model can easily be applied to previously published
light curves in submillimeter wavelengths (see additional fits to light curves in Yusef-Zadeh et al.
The other criticism of Marrone et al. (2008) is that the model requires adiabatic expansion of the
blobs at speeds ∼ 0.03c much below than the canonical sound speed c/√3 approaching the black
hole that would be expected were the blobs filled with plasma and sitting in vacuum. Given that
the blob diameters are in the range of several Rsthey may well be located at 10 or more Rs, where
the local sound speed would be ∼ 0.1c. In any case, they may be embedded in the outer layers of
an accretion flow where they would be only mildly overpressured with respect to their surroundings
– 27 –
and the expansion time scale would be comparable to the buoyancy or orbital time scale.
5.4.Distribution of Electrons in the Flares
Previously we have modeled the time delays at submillimeter to radio frequencies in the expanding
hot plasma model assuming that the accelerated particles have a power-law energy distribution.
This is motivated by the long time scale of the flares compared to the synchrotron loss time for
the expected magnetic field strengths of 10–30G. These models assume a homogeneous sphere
threaded by a uniform magnetic field. As the region expands, the relativistic particles cool by
adiabatic expansion with E ∝ 1/R and the magnetic field is diluted as B ∝ R−2because of flux
freezing. Assuming that the relativistic electron energies run between 1MeV and 100MeV and
that they are in equipartition with the magnetic field, the models are characterized by the particle
spectral index, p (with n(E) ∝ E−p), the expansion speed v (assumed constant), and the timing
and amplitude of the flaring at a single frequency. From this we can infer a magnetic field strength
B and the size of the emitting region at t0, R0. Figure 28a shows an approximate fit to the CARMA
and VLA data obtained on 2007 April 02 using two flares. The derived parameters of the first flare
at 10.9h UT are p = 0.5, R0= 3.6Rs, v = 0.070c, and B = 15G, while the later flare at 14.7hr
UT has p = 1.5, R0= 9.8Rs, v = 0.065c, and B = 13G. These numbers should be regarded as
illustrative given the rough fitting, the simplicity of the model, and the freedom in choosing the
baselines at each frequency.
While a power-law electron spectrum is plausible, the inferred spectra are significantly harder than
the E−2expected on the basis of the simplest version of diffusive shock acceleration. This suggests
that the derived power law may instead be the effective power-law of the particle energy spectrum
over the small (5%) range of initial energies relevant to our observing frequencies at 96GHz to
43GHz. Other spectral forms are easily introduced within the context of this model. By way of
example, in Fig. 28b we show the “best” relativistic Maxwellian model, with the particle spectrum
characterized by the electron temperature Teat time t0 instead of p. The parameters in this case
(kTe= 0.25keV, R0= 4.4Rs, v = 0.077c, and B = 10G; kTe= 4.1MeV, R0= 10.4Rs, v = 0.091c,
B = 12G) yields similar emission region characteristics but is worse in matching the 43GHz data.
This is because the synchrotron spectrum is more strongly peaked than for a power-law electron
population. As a result, the flare amplitude declines more rapidly at successively lower frequencies
than is the case for power-law models (except for large choice of p). Future simultaneous light
curves with better time coverage are needed to confirm this results.
The main results of extensive observing campaign that took place in 2007 can be summarized as
Simultaneous VLA and VLBA observations indicate that flare emission from Sgr A* at 43 GHz
arises from within the scattering size of Sgr A* which is ∼ 0.3 × 0.7mas (Bower et al. 2004)
or within the inner 30×70 Rsof Sgr A*.
We show the evidence of varying spectral index values when weak and bright NIR flares are
compared. In addition, the NIR flare statistics indicate that the probability of flare emission
– 28 –
is proportional to the inverse of the flux density. Simulations of the histogram of such flares
assuming uniform distribution of peak flare emission is consistent with observations. The
significance of the probability of flare emission is inversely proportional to the flux of flare is
In addition to a powerful X-ray flare with a NIR counterpart and 11.8µm upper limit on 2007
April 4 that had been reported earlier by Porquet et al. (2008), Dodds-Edden et al. (2009),
and Trap et al. (2009), we show evidence of three new X-ray flares with NIR counterparts.
The origin of X-ray production is explained in the context of ICS employing the structure
details of the Sgr A* emitting region inferred from intrinsic size measurements.
picture, the seed photons associated with flares in NIR wavelengths are upscattered by the
sea of electrons that are responsible for the quiescent emission of the Sgr A* in radio and
submillimeter wavelengths. A prediction of this model is a time delay between the peaks of
X-ray and NIR flare emission.
The comparison of the light curves at multiple wavelengths indicated time delays implying optically
thick emission. We also argue a tantalizing radio flare three hours after the strongest NIR and
X-ray flare detected on 2007 April 4. These measurements are consistent with an adiabatic
expansion of hot plasma. Although these measurements weaken the hot spot model of flare
emission, we can not distinguish whether there is jet activity associated with the observed
time delays or the expansion of hot plasma that is bound to Sgr A*.
This work is partially supported by the grant AST-0807400 from the National Science Foundation.
Some of the data presented here were obtained from Mauna Kea observatories. We are grateful to
the Hawai’ian people for permitting us to study the universe from this sacred summit. Research
at the Caltech Submillimeter Observatory is supported by grant AST-0540882 from the National
Science Foundation. Research grants are also given by Australian Research Council (DPO986386)
and Macquarie University. The SMT is operated by the Arizona Radio Observatory (ARO), Stew-
ard Observatory, University of Arizona. The XMM-Newton project is an ESA Science Mission with
instruments and contributions directly funded by ESA Member State and the USA (NASA).
Alexander, T. 1997, MNRAS, 285, 891
Baars, J. W. M., Genzel, R., Pauliny-Toth, I. I. K., & Witzel, A. 1977, A&A, 61, 99
Baars, J. W. M., Martin, R. N., Mangum, J. G., McMullin, J. P., & Peters, W. L. 1999, PASP,
Baganoff, F. K. 2003, in Bulletin of the American Astronomical Society, Vol. 35, Bulletin of the
American Astronomical Society, 606
Baganoff, F. K., Maeda, Y., Morris, M., Bautz, M. W., Brandt, W. N., Cui, W., Doty, J. P.,
Feigelson, E. D., Garmire, G. P., Pravdo, S. H., Ricker, G. R., & Townsley, L. K. 2003, ApJ,
Balick, B., & Brown, R. L. 1974, ApJ, 194, 265
– 29 –
B´ elanger, G., Goldwurm, A., Melia, F., Ferrando, P., Grosso, N., Porquet, D., Warwick, R., &
Yusef-Zadeh, F. 2005, ApJ, 635, 1095
Bittner, J. M., Liu, S., Fryer, C. L., & Petrosian, V. 2007, ApJ, 661, 863
Blandford, R. D., & Begelman, M. C. 1999, MNRAS, 303, L1
Blundell, R. 2004, in Proceedings of the 15th International Symposium on Space Terahertz Tech-
nology, April 27-29, 2004, p. 3, 3
Bower, G. C., Falcke, H., Herrnstein, R. M., Zhao, J.-H., Goss, W. M., & Backer, D. C. 2004,
Science, 304, 704
Broderick, A. E., & Loeb, A. 2006, ApJ, 636, L109
Do, T., Ghez, A. M., Morris, M. R., Yelda, S., Lu, J. R., Hornstein, S. D., & Matthews, K. 2008,
Journal of Physics Conference Series, 131, 012003
Dodds-Eden, K., Porquet, D., Trap, G., Quataert, E., Haubois, X., Gillessen, S., Grosso, N., Pantin,
E., Falcke, H., Rouan, D., Genzel, R., Hasinger, G., Goldwurm, A., Yusef-Zadeh, F., Clenet,
Y., Trippe, S., Lagage, P.-O., Bartko, H., Eisenhauer, F., Ott, T., Paumard, T., Perrin, G.,
Yuan, F., Fritz, T. K., & Mascetti, L. 2009, ApJ, 698, 676
Doeleman, S. S., Weintroub, J., Rogers, A. E. E., Plambeck, R., Freund, R., Tilanus, R. P. J.,
Friberg, P., Ziurys, L. M., Moran, J. M., Corey, B., Young, K. H., Smythe, D. L., Titus, M.,
Marrone, D. P., Cappallo, R. J., Bock, D. C.-J., Bower, G. C., Chamberlin, R., Davis, G. R.,
Krichbaum, T. P., Lamb, J., Maness, H., Niell, A. E., Roy, A., Strittmatter, P., Werthimer,
D., Whitney, A. R., & Woody, D. 2008, Nature, 455, 78
Eckart, A., Baganoff, F. K., Morris, M., Bautz, M. W., Brandt, W. N., Garmire, G. P., Genzel,
R., Ott, T., Ricker, G. R., Straubmeier, C., Viehmann, T., Sch¨ odel, R., Bower, G. C., &
Goldston, J. E. 2004, A&A, 427, 1
Eckart, A., Baganoff, F. K., Morris, M. R., Kunneriath, D., Zamaninasab, M., Witzel, G., Sch¨ odel,
R., Garc´Ia-Mar´In, M., Meyer, L., Bower, G. C., Marrone, D., Bautz, M. W., Brandt,
W. N., Garmire, G. P., Ricker, G. R., Straubmeier, C., Roberts, D. A., Muzic, K., Mauerhan,
J., & Zensus, A. 2009, ArXiv e-prints
Eckart, A., Baganoff, F. K., Sch¨ odel, R., Morris, M., Genzel, R., Bower, G. C., Marrone, D., Moran,
J. M., Viehmann, T., Bautz, M. W., Brandt, W. N., Garmire, G. P., Ott, T., Trippe, S.,
Ricker, G. R., Straubmeier, C., Roberts, D. A., Yusef-Zadeh, F., Zhao, J. H., & Rao, R.
2006a, A&A, 450, 535
Eckart, A., Sch¨ odel, R., Garc´Ia-Mar´In, M., Witzel, G., Weiss, A., Baganoff, F. K., Morris, M. R.,
Bertram, T., DovVciak, M., Duschl, W. J., Karas, V., K¨ onig, S., Krichbaum, T. P., Krips,
M., Kunneriath, D., Lu, R.-S., Markoff, S., Mauerhan, J., Meyer, L., Moultaka, J., MuVzi´ c,
K., Najarro, F., Pott, J.-U., Schuster, K. F., Sjouwerman, L. O., Straubmeier, C., Thum,
C., Vogel, S. N., Wiesemeyer, H., Zamaninasab, M., & Zensus, J. A. 2008, A&A, 492, 337
Eckart, A., Sch¨ odel, R., Meyer, L., Trippe, S., Ott, T., & Genzel, R. 2006b, A&A, 455, 1
Edelson, R. A., & Krolik, J. H. 1988, ApJ, 333, 646
– 30 –
Eisenhauer, F., Genzel, R., Alexander, T., Abuter, R., Paumard, T., Ott, T., Gilbert, A., Gillessen,
S., Horrobin, M., Trippe, S., Bonnet, H., Dumas, C., Hubin, N., Kaufer, A., Kissler-Patig,
M., Monnet, G., Str¨ obele, S., Szeifert, T., Eckart, A., Sch¨ odel, R., & Zucker, S. 2005, ApJ,
Falcke, H., Goss, W. M., Matsuo, H., Teuben, P., Zhao, J.-H., & Zylka, R. 1998, ApJ, 499, 731
Falcke, H., & Markoff, S. 2000, A&A, 362, 113
Falcke, H., Markoff, S., & Bower, G. C. 2009, A&A, 496, 77
Genzel, R. 2000, ArXiv Astrophysics e-prints
Genzel, R., Sch¨ odel, R., Ott, T., Eckart, A., Alexander, T., Lacombe, F., Rouan, D., & Aschenbach,
B. 2003, Nature, 425, 934
Genzel, R., & Townes, C. H. 1987, ARA&A, 25, 377
Ghez, A. M., Becklin, E., Duchjne, G., Hornstein, S., Morris, M., Salim, S., & Tanner, A. 2003,
Astronomische Nachrichten Supplement, 324, 527
Ghez, A. M., Salim, S., Hornstein, S. D., Tanner, A., Lu, J. R., Morris, M., Becklin, E. E., &
Duchˆ ene, G. 2005, ApJ, 620, 744
Ghez, A. M., Salim, S., Weinberg, N. N., Lu, J. R., Do, T., Dunn, J. K., Matthews, K., Morris,
M. R., Yelda, S., Becklin, E. E., Kremenek, T., Milosavljevic, M., & Naiman, J. 2008, ApJ,
Ghez, A. M., Wright, S. A., Matthews, K., Thompson, D., Le Mignant, D., Tanner, A., Hornstein,
S. D., Morris, M., Becklin, E. E., & Soifer, B. T. 2004, ApJ, 601, L159
Gillessen, S., Eisenhauer, F., Quataert, E., Genzel, R., Paumard, T., Trippe, S., Ott, T., Abuter,
R., Eckart, A., Lagage, P. O., Lehnert, M. D., Tacconi, L. J., & Martins, F. 2006, ApJ, 640,
Gillessen, S., Eisenhauer, F., Trippe, S., Alexander, T., Genzel, R., Martins, F., & Ott, T. 2009,
ApJ, 692, 1075
Goldston, J. E., Quataert, E., & Igumenshchev, I. V. 2005, ApJ, 621, 785
Goldwurm, A., Brion, E., Goldoni, P., Ferrando, P., Daigne, F., Decourchelle, A., Warwick, R. S.,
& Predehl, P. 2003, ApJ, 584, 751
Hawley, J. F., & Balbus, S. A. 2002, ApJ, 573, 738
Hornstein, S. D., Matthews, K., Ghez, A. M., Lu, J. R., Morris, M., Becklin, E. E., Rafelski, M.,
& Baganoff, F. K. 2007, ApJ, 667, 900
Kreysa, E. 1990, in Liege International Astrophysical Colloquia, Vol. 29, Liege International As-
trophysical Colloquia, ed. B. Kaldeich, 265–270
Li, H., Dowell, C. D., Kirby, L., Novak, G., & Vaillancourt, J. E. 2008, Appl. Opt., 47, 422
Liu, S., & Melia, F. 2001, ApJ, 561, L77
– 31 –
—. 2002, ApJ, 566, L77
Liu, S., Melia, F., & Petrosian, V. 2006, ApJ, 636, 798
Liu, S., Petrosian, V., & Melia, F. 2004, ApJ, 611, L101
Markoff, S., Falcke, H., Yuan, F., & Biermann, P. L. 2001, A&A, 379, L13
Marrone, D. P., Baganoff, F. K., Morris, M. R., Moran, J. M., Ghez, A. M., Hornstein, S. D.,
Dowell, C. D., Mu˜ noz, D. J., Bautz, M. W., Ricker, G. R., Brandt, W. N., Garmire, G. P.,
Lu, J. R., Matthews, K., Zhao, J.-H., Rao, R., & Bower, G. C. 2008, ApJ, 682, 373
Marrone, D. P., & Rao, R. 2008, in Presented at the Society of Photo-Optical Instrumentation
Engineers (SPIE) Conference, Vol. 7020, Society of Photo-Optical Instrumentation Engineers
(SPIE) Conference Series
Melia, F. 1992, ApJ, 387, L25
Melia, F. 2002, in Bulletin of the American Astronomical Society, Vol. 34, Bulletin of the American
Astronomical Society, 951
Meyer, L., Do, T., Ghez, A., Morris, M. R., Witzel, G., Eckart, A., B´ elanger, G., & Sch¨ odel, R.
2008, ApJ, 688, L17
Meyer, L., Eckart, A., Sch¨ odel, R., Duschl, W. J., MuVzi´ c, K., DovVciak, M., & Karas, V. 2006,
A&A, 460, 15
Moneti, A., Stolovy, S., Blommaert, J. A. D. L., Figer, D. F., & Najarro, F. 2001, A&A, 366, 106
Narayan, R., Mahadevan, R., Grindlay, J. E., Popham, R. G., & Gammie, C. 1998, ApJ, 492, 554
Narayan, R., Yi, I., & Mahadevan, R. 1995, Nature, 374, 623
Nayakshin, S., & Sunyaev, R. 2003, MNRAS, 343, L15
Porquet, D., Grosso, N., Predehl, P., Hasinger, G., Yusef-Zadeh, F., Aschenbach, B., Trap, G.,
Melia, F., Warwick, R. S., Goldwurm, A., B´ elanger, G., Tanaka, Y., Genzel, R., Dodds-
Eden, K., Sakano, M., & Ferrando, P. 2008, A&A, 488, 549
Porquet, D., Predehl, P., Aschenbach, B., Grosso, N., Goldwurm, A., Goldoni, P., Warwick, R. S.,
& Decourchelle, A. 2003, A&A, 407, L17
Reid, M. J., & Brunthaler, A. 2004, ApJ, 616, 872
Scargle, J. D. 1982, ApJ, 263, 835
Sch¨ odel, R., Ott, T., Genzel, R., Hofmann, R., Lehnert, M., Eckart, A., Mouawad, N., Alexander,
T., Reid, M. J., Lenzen, R., Hartung, M., Lacombe, F., Rouan, D., Gendron, E., Rousset, G.,
Lagrange, A.-M., Brandner, W., Ageorges, N., Lidman, C., Moorwood, A. F. M., Spyromilio,
J., Hubin, N., & Menten, K. M. 2002, Nature, 419, 694
Shklovskii, I. S. 1960, Soviet Astronomy, 4, 355
– 32 –
Trap, G., Goldwurm, A., Terrier, R., Dodds-Eden, K., Gillessen, S., Genzel, R., Pautin, E., Lagage,
P., Ferrando, P., Belanger, G., Porquet, D., Grosso, N., Yusef-Zadeh, F., & Melia, F. 2009,
Advances in Space Research, submitted
van der Laan, H. 1966, Nature, 211, 1131
Yuan, F., Markoff, S., & Falcke, H. 2002, A&A, 383, 854
Yuan, F., Quataert, E., & Narayan, R. 2003, ApJ, 598, 301
Yusef-Zadeh, F., Bushouse, H., Dowell, C. D., Wardle, M., Roberts, D., Heinke, C., Bower, G. C.,
Vila-Vilar´ o, B., Shapiro, S., Goldwurm, A., & B´ elanger, G. 2006a, ApJ, 644, 198
Yusef-Zadeh, F., Roberts, D., Wardle, M., Heinke, C. O., & Bower, G. C. 2006b, ApJ, 650, 189
Yusef-Zadeh, F., Wardle, M., Heinke, C., Dowell, C. D., Roberts, D., Baganoff, F. K., & Cotton,
W. 2008, ApJ, 682, 361
Zylka, R., Mezger, P. G., & Lesch, H. 1992, A&A, 261, 119
This preprint was prepared with the AAS LATEX macros v5.2.
– 33 –
Table 1. Spectral Index Distribution Using NICMOS
Event F(1.45µm ± σ) F(1.70µm ± σ)
β ± σ
Table 2. Measured Time Lags with 1-σ Errors
— 2.64 (-0.67, +0.5)
aThe X-ray data is time-shifted by 5.25 hours and time-stretched by 3.5
– 34 –
Fig. 1.—: A schematic diagram showing 13 telescopes joining the 2007 April campaign. The width
of individual observing period is not scaled.
– 35 –
observations on 2007 April 1–7. No background flux is removed from these plots. (b) The HST
1.70µm light curves of Sgr A*, the star S0-2, and a region of background emission. The constancy
of the S0-2 and background light curves strongly suggest that the variability of Sgr A* emission is
intrinsic to Sgr A*.
(a) The light curves of Sgr A* at 1.45µm and 1.70µm for the seven windows of HST
– 36 –
Fig. 3.—: The background subtracted light curves of Sgr A* and S0-2 for each of the seven HST
observing windows, with flare events labeled. The data points at 1.70 and 1.45 µm are sampled at
144 sec intervals.
– 37 –
Fig. 4.—: This plot shows all the 64-sec sampled data for the 16 periods that flares are identified.
The 16 periods are stacked on the top and bottom panels. The x-axis is the elapsed time - each flare
episode occupies a 45-minute slice within the 8-hour axis. Each flare event is labeled, as defined in
Figure 3, The Sgr A* and S0-2 light curves are on the upper and lower portion of the panel. The
data show only the 1.70µm photometry sampled at intervals of 64 sec but the light curves for each
flare are laid side by side.
– 38 –
Fig. 5.—: Light curves of SgrA* for April 1-6 at H (1.66 µm), in red, Ks(2.12µm), in blue, Ks
in polarimetric mode in blue and L’ (3.8µm), in black, bands. are taken from Dodds-Eden et al.
(2009). There are a total of seven periods of flaring activity reported in these observations. The
brightest flares occurred on 2007, April 4 and April 5.
Fig. 6.—: Light curves of all the X-ray data taken with the XMM-Newton during the 2007 April
observing campaign (Porquet et al. 2008). The data are averaged over a 144sec sampling. Five
X-ray flares are detected, four of which had simultaneous coverage with the VLT and HST and
showed NIR counterparts.
– 39 –
obtained with VLA observations taken during 2007, April 1-4. The sampling time is 87s for Sgr
A* and 90 sec for the calibrator at the bottom of each panel. The Sgr A* uv data is restricted
to >100kλ in order to suppress the contribution of extended emission. (b - Right) Similar to (a)
except that only the light curves of Sgr A* are shown with a sampling time of 300 sec.
(a - Left) Light curves of Sgr A* and the calibrator 17444-31166 at 43 GHz data
Fig. 7.—: (c) The light curves of Sgr A* on 2007, April 4 are shown in the top and bottom panels
at 43.185 GHz and 43.535 GHz, respectively. The light curves are shown with a 30s sampling
time.The data points with large error bars correspond to a small number of data points in a given
sample. The data corresponding to minimum and maximum uv baselines are selected between 110
and 125 kλ.
– 40 –
Fig. 8.—: (a - Top ) Light curves of Sgr A* on 2007, April 1, 5 and 11 using VLBA at 43.22 GHz.
The sampling time is 60 sec. (b - Bottom Left) The light curve of Sgr A* observed with VLBA on
April 2, 2007 at 22 and 43 GHz. The sampling time is 300 sec. (c - Bottom Right ) The light curve
of Sgr A* observed with VLBA on April 10, 2007 at 14 and 43 GHz. The sampling time is 300 sec.
– 41 –
Fig. 9.—: (a - Left) The light curve of Sgr A* observed with VLA and VLBA on April 1, 2007.
The sampling time is 300 sec. The center frequencies of the VLA and VLBA light curves correspond
to 43.34 GHz 43.22 GHz, respectively. The selected VLA uv data is >100kλ. (b - Right) Similar
to (a) except that the observations are carried out on April 2 at 43 GHz. VLBA observations on
April 1 (bottom panel) are sampled continuously unlike those made on April 2. VLBA plots are
shown in red (bottom) whereas VLA plots are shown in blue (top).
– 42 –
with a sampling time of 15 minutes. (b) Similar to (a) except at 350µm on 2007, April 4-5 with a
sampling time of 15 minutes.
(a) Light curves of Sgr A* at 450µm based on CSO observations on 2007, April 1-3
– 43 –
(c) Similar to (a) except at 850µm on 2007, April 6 using a sampling time of 10
Fig. 11.—: Light curves of Sgr A* at 230 GHz using SMA on 2007 April 1, and 3-5. The sampling
times are 27 min, 6.5 min, 8 min and 10 min for the April 1, 3, 4 and 5 light curves, respectively.
– 44 –
Fig. 12.—: Light curves of Sgr A* and the calibrator G34.3 and 1757-240 (blue) at 230 GHz using
SMT on 2007 April 1 and April 2-4, respectively, with a sampling time of ∼7 minutes.
– 45 –
Fig. 13.—: Light curves of Sgr A* at 240 GHz (1.25mm) taken with IRAM on 2007, April 3-4 with
sampling time of ∼10 minutes. For each time sample, there are two data points estimating the flux
of SgrA* from repeated pointing measurements.
Fig. 14.—: Light curves of Sgr A* taken with the NMA. The top panels show the light curves of
Sgr A* at 90 GHz and 102 GHz on April 3 whereas the bottom two panels show simultaneous light
curves at 2.23mm (134 GHz) and 146 GHz on 2007, April 4. The flux of the calibrator 1744-312 is
shown at the bottom of each panel. The sampling time is 3 minutes.
– 46 –
Fig. 15.—: Light curves of Sgr A* and the calibrator 1730-130 obtained with CARMA at 90 GHz
with a sampling time of 300 sec on 2007, April 2-5. The uv data > 20kλ are used to make the Sgr
A* light curves.
simultaneous single Gaussian fit and power law fits to both the noise and the flares. The dotted
lines show the Gaussian and power law fits.
A histogram plot of the detected signals and the noise at 1.70µm as well as the
– 47 –
Fig. 17.—: Similar to Figure 16 expect that the 2004 histogram of flare activity (Yusef-Zadeh et
al. 2006) is plotted at 1.60µm.
– 48 –
Time (arbitrary units)
Flux, (arbitrary units)
Flux, (arbitrary units)
Fig. 18.—: (a) (Left) A synthetic light curve constructed from the sum of 100 Gaussian profiles
with peak positions and, standard deviations drawn uniformly between -1.5 to 101.5 and 0 to 0.5
time units respectively; the probability distribution of the peak fluxes are distributed as 1/(peak
flux) between 0.01 and 1 flux units. (b) (Right) Distribution of uniformly sampled flux values in
the simulated flares. The dashed line indicates a slope of 1/Sν.
– 49 –
log F145 (mJy)
log F170 (mJy)
Fig. 19.—: A log-log plot of NIR fluxes in the F170 and F145 filters of NICMOS at 1.70µm and
1.45µm, respectively. The thick line in red shows the spectral index β=0.6. The thin dotted lines
to the right and left of the β = 0.6 line correspond to β = −2,−4 and β = +2,+4, respectively.
observations and the corresponding power spectrum of the residual flux of Sgr A*, respectively.
The dashed lines show the significance of the power spectrum at 99% and 99.9% confidence levels.
We explain the significance of the peak in the text.
20.—: The top and bottom boxes show the light curve of 2007, April 4 based on HST
– 50 – Download full-text
Fig. 21.—: (a - Left) The top panel shows the light curves of two near-IR flares identified as 4A ad
4B measured with NICMOS on 2007, April 4 with a sampling time of 64s at 1.70µm. The bottom
panel shows the X-ray counterpart to these flares with a sampling time of 300 sec. These X-ray
flares are identified as flare #4 and #5 by Porquet et al. (2008). (b - Right) Similar to (a) except
for a NIR and X-ray flare that occurred on April 2, 2007.