Unprecedented Near-infrared Brightness and Variability of Sgr A*

The Galactic center hosts the closest supermassive black hole to the Earth, offering us a unique opportunity to study in detail the physical processes that occur in its vicinity. The Galactic black hole, Sgr A*, has been monitored extensively across many wavelength regimes (e.g., Balick & Brown 1974; Falcke & Markoff 2000; Baganoff et al. 2001; Hornstein et al. 2002; Yusef-Zadeh et al. 2006; Marrone et al. 2008; Sabha et al. 2010; Neilsen et al. 2013; Bower et al. 2015; Rauch et al. 2016; Capellupo et al. 2017; Ponti et al. 2017; Gravity Collaboration et al. 2018) For recent reviews see Genzel et al. (2010) and Morris et al. (2012). These observations have shown that the source luminosity is nine orders of magnitude below the Eddington luminosity and is highly variable (Narayan et al. 1998; Quataert et al. 1999; Quataert & Gruzinov 2000a, 2000b; Dodds-Eden et al. 2009; Eckart et al. 2012; Haubois et al. 2012; Yusef-Zadeh et al. 2012; Dibi et al. 2016; Mossoux et al. 2016; Ponti et al. 2017).

Observations of Sgr A* in the near-infrared are an effective way to monitor the variability of the black hole. First detected with adaptive optics (AO) images in 2003 (Genzel et al. 2003; Ghez et al. 2004, 2005), recent re-analysis of speckle imaging data has enabled detections of Sgr A* back to 1998, establishing a time baseline across two decades (Chen et al. 2019). The near-infrared flux variations have been characterized as a red-noise process that is correlated in time (e.g., Do et al. 2009). There have been several proposed models of the distribution of flux values over time: a single power-law model, a log-normal model (Witzel et al. 2012), a log-normal model with an additional tail at higher flux levels (Dodds-Eden et al. 2011), and a log–log-normal distribution (Meyer et al. 2014). Recently, based on a comprehensive analysis of over 13,000 observations from historical AO data (2003–2014) from the Keck Telescopes and Very Large Telescope (VLT) and space data at 4.5 μm from Spitzer (2014–2017), Witzel et al. (2018) found that the variability of Sgr A* in the near-infrared can be consistently described as a red-noise process with a single log-normal distribution for the flux variations. Chen et al. (2019) found that the model from Witzel et al. (2018) is also consistent with speckle data from the Keck Telescopes from 1995–2005.

While historical near-infrared observations can all be fit with a single model, Sgr A* has the potential to greatly change its luminosity and variability. For example, observations of X-ray light echos from 6.4 keV iron line emission suggest that in the past few hundred years, Sgr A* may have undergone a few relatively brief (up to ∼10 yr) luminosity excursions by factors up to 10⁵ (e.g., Koyama et al. 1996; Clavel et al. 2013; Terrier et al. 2018). In addition, the environment around Sgr A* is very dynamic, with stars and other objects passing near the black hole, which may affect its accretion flow. In 2018, the star S0-2 reached within 100 au of the black hole. Also, in recent decades, two dusty objects (G1 and G2) have shown signs of tidal interaction with the black hole (e.g., Gillessen et al. 2012; Eckart et al. 2013; Phifer et al. 2013; Witzel et al. 2014, 2017; Pfuhl et al. 2015; Plewa et al. 2017; Gillessen et al. 2019). There have been numerous suggestions that these sources may deposit gas or alter the accretion onto Sgr A*, changing its luminosity and accretion state (e.g., Loeb 2004; Schartmann et al. 2012).

Here, we report new Keck Telescope near-infrared observations of Sgr A* in 2019. These observations show unusually bright flux levels and variability, with peak fluxes exceeding twice the maximum historical flux measurements. In Section 2 we present the observations. Section 3 presents the light curves and analyses. Section 4 presents comparisons with the historical data, comparisons with models, and discussion of potential physical explanations for these observations. We conclude in Section 5.

We observed the Galactic center on four nights in 2019 with the Keck II Telescope using the narrow camera in the Near-Infrared Camera 2 (NIRC2) instrument with the Laser-guide Star Adaptive Optics (LGS AO) system (Wizinowich et al. 2006). We include all available 2019 observations from Keck having sampling duration >20 minutes and high enough image quality (full-width half-maximum <100 mas) to detect Sgr A*. We use the K' filter (2.12 μm) on three nights and a combination of K' and H-band (1.64 μm) filters on one night (Table 1). Individual K' images consist of 10 coadds of 2.8 s integration time each, while the H-band images consist of four coadds of 7.4 s each. On 2019 May 13, the observations alternated between six frames of K' images and six frames of H-band images, for a total of about 6 minutes spent on each filter before switching. Images centered on Sgr A* are shown in Figure 1. Standard image reduction methods were applied to the images including flat-fielding, sky subtraction, and cosmic-ray removal (e.g., Jia et al. 2019).

Zoom In Zoom Out Reset image size

Table 1.  Near-infrared Sgr A* Observations

Date (UT)	Nobs	Duration	Max Fobsa	Min Fobs	Max Fdereddenb	Min. Fderedden
		(minutes)	(mJy)	(mJy)	(mJy)	(mJy)
2019 Apr 19	35	147	0.48 ± 0.04	0.09 ± 0.01	4.6	0.85
2019 Apr 20	152	87	1.74 ± 0.04	0.07 ± 0.01	16.7	0.65
2019 May 13c	82	149	6.19 ± 0.08	0.08 ± 0.01	59.6	0.79
2019 May 23	109	213	0.70 ± 0.02	0.05 ± 0.01	6.7	0.48

Notes.

^aObserved fluxes are converted from K' to K_s filter. ^bFluxes are dereddened using an extinction of ${A}_{{K}_{s}}=2.46$ mag. ^cH-band observations are also made this night (Figure 1).

Download table as:  ASCIITypeset image

We extracted photometry for point sources in each individual image to construct light curves of Sgr A* and calibration stars. The brightness and position of all sources were measured using the point spread function (PSF)-fitting software StarFinder (Diolaiti et al. 2000). Details on running StarFinder on individual frames for Galactic center observations can be found in Do et al. (2009). Sgr A* is detected in almost every frame of observations on three nights. There are fewer detections of Sgr A* on 2019 May 19 due to lower-quality AO correction and faint flux levels for Sgr A*. We calibrate the point source photometry using reference stars defined by Gautam et al. (2019). We convert between K' magnitudes to absolute fluxes F, using the relationship ${F}_{K^{\prime} }=6.86\times {10}^{5}\times {10.0}^{-0.4K^{\prime} }$ mJy (Tokunaga & Vacca 2005). To convert from K' fluxes to K_s fluxes, we use the filter transformation ${F}_{{K}_{s}}=1.09{F}_{K^{\prime} }$ computed for the color of Sgr A* (Chen et al. 2019). We will mainly use observed fluxes in this work to avoid confusion with the value of extinction to apply. To de-redden the flux measurements, one can use the relationship ${F}_{\mathrm{deredden}}={F}_{\mathrm{obs}}\times {10}^{0.4{A}_{{K}_{s}}}$. To compare to fluxes from (Witzel et al. 2012), use ${A}_{{K}_{s}}=2.46$ (Schödel et al. 2010).

We estimate the Sgr A* relative photometric uncertainties by using the relationship between flux level and flux uncertainty of nearby stars. The photometric uncertainties for stars on each night were determined by using the standard deviation of their individual flux measurements. We find that the reference stars are stable during the nights with <3% photometric errors. Following Do et al. (2009), we fit a power-law to the relationship between flux level and flux uncertainty for stars within 1'' of Sgr A*. We then use this relationship to infer the flux uncertainty for the all flux measurements of Sgr A*. The Sgr A* relative photometric uncertainties are typically less than 5% at high flux levels and about 15% for faint flux values.

Our observations show Sgr A* to be highly variable in 2019. On 2019 May 13, Sgr A*'s flux level changed by a factor of 75 within 2 hr (from 6.19 ± 0.08 mJy to 0.08 ± 0.01 mJy). The maximum observed fluxes occurred during the beginning of the observations, suggesting that Sgr A* was likely even brighter earlier in the night. On the night of 2019 April 20, Sgr A* flux also shows large variations with measurements ranging from 1.74 ± 0.04 mJy to 0.07 ± 0.01 mJy. The two other nights of our study (2019 April 19 and 2019 May 23) show less variation at the level of one to three magnitudes of change during the night. Figures 1–3 show the observed light curves.

Zoom In Zoom Out Reset image size

Observations on 2019 May 13 also include H-band observations, which allows us to constrain the H − K' color of Sgr A* on this night. We find that the H-band and K' light curves can be matched using a color of H − K' = 2.45 mag. With this color shift, the H-band and K' points generally transition smoothly between the two filters. The exception may be during the large drop in flux at 11:42 UT. While outside of the scope of this Letter, we plan to characterize possible color changes during this night in a future work (G. Witzel et al. 2019, in preparation).

In addition to high flux values, two of the nights also show large drops in brightness over very short timescales. On 2019 May 13 at around 11:42 UT, the brightness of Sgr A* dropped from K' = 14.5 mag to K' = 15.8 mag, a factor of 3 in flux, within 7 minutes. An even larger change occurred on 2019 May 23 12:15 UT, when Sgr A*'s brightness changed from K' = 15.2 mag to K' = 17.6 mag, a factor of 9 in flux, within 2 minutes (corresponding to a light travel of ∼3 Schwarzschild radii for a 4 × 10⁶M_⊙ black hole).

The Sgr A*-infrared observations presented here show peak flux levels that are unprecedented compared to the historical data. We use the distribution of flux variations from Witzel et al. (2018), which includes data spanning 10 yr and over 13,000 K_s flux measurements of Sgr A* from Keck and VLT (over 130 nights), to compare with the observations in 2019. The observations reported in Witzel et al. (2018) were made with integration times between 28 and 40 s per measurement. The observations reported by Chen et al. (2019) increase this time baseline to over 20 yr. We find that the peak flux levels from 2019 May 13 exceed the maximum observed historical flux (3 mJy) by a factor of 2 (Figures 3 and 4). On 2019 April 20, the peak flux levels are brighter than 99.7% of all historical data points.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We also find the flux variations observed during the four observing periods in 2019 to be significantly different than in the historical data from Witzel et al. (2018). Using a two-tailed Kolmogorov–Smirnov (KS) test, we find a KS-statistic of 0.146, which corresponds to a probability of ≪0.01% that a randomly drawn data set based on the historical Sgr A*-infrared distribution will produce a KS-statistic as extreme as, or more extreme than, the KS value derived when comparing the distribution of flux variations observed in 2019. The Anderson-Darling statistic (Scholz & Stephens 1987), which is more sensitive to the tails of the distribution, has a value of 57. This also corresponds to a probability ≪0.01% that a randomly drawn data set based on the historical Sgr A*-infrared distribution will produce such a statistic.

To better understand the variability of Sgr A* we use the statistical model of Witzel et al. (2018) to assess the probability of observing light curves similar to the four that are presented here. We use the Witzel et al. (2018) model for comparison because: (1) this model can accommodate the temporal correlation in the flux of Sgr A*, which is crucial for statistical analyses and Monte Carlo simulations (e.g., Do et al. 2009), and (2) this model was created using the largest sample of near-infrared data to date. The correlation of Sgr A*'s flux with time has an important consequences: the brighter Sgr A* is, the larger the changes in flux density will be. If Sgr A* is already at an elevated flux density level, the probability of observing an even brighter state is much larger than the time-averaged probability for such a bright state. For continuous samples with a monitoring duration that is not significantly longer than the correlation timescale (∼245 minutes), this can result in large deviations of the sample distribution of flux densities from the underlying distribution. In order to investigate the consequences of the flux correlations in time, we use the simulation approach presented in Witzel et al. (2018). We use the posterior of their model 3, i.e., a log-normally distributed, red-noise process with a characteristic break timescale that successfully describes the historic VLT and Keck data in K-band and eight full days of Spitzer/IRAC data at 4.5μm. From this posterior distribution, we sample 10,000 parameter combinations and generate one random light curve with the time sampling of the observed Keck data (∼30 nights of historical data and four nights of 2019 data) for each parameter set. We present the complementary cumulative distribution function (CCDF, or 1—CDF) of these 10,000 light curves in Figure 5 in the form of the median CDF and 1, 2, and 3σ credible intervals.

Zoom In Zoom Out Reset image size

Based on simulations drawn from the Sgr A* flux model, we can compute the probability of observing the light curves presented here. The simulations show that if we repeated our experiment 10,000 times with the time sampling of the 30 nights of historical Keck observations (from 2005 to 2013) of Sgr A*, and including the 2019 nights, there is a 0.3% probability that we observe a single night with flux levels as high as seen on 2019 May 13. These simulations are the most consistent with our current observations because they have very similar noise properties, observed duration, and timing. If we also include VLT nights for a total of over 130 nights, then the probability is less than 1.5% to observe flux values higher than 6.18 mJy on a single night. We can also consider the probability of randomly drawing 4 light curves similar to the observations in 2019. This probability is considerably lower at less than 0.05%. We also note that we have only observed the decay of the light curve on 2019 May 13, which suggests the actual maximum was likely even higher (Figure 3).

Here we examine two possibilities for explaining the very unusual brightness and variation of Sgr A* observed this year: (1) the statistical models need to be changed or updated, and (2) there is a physical change in the accretion activity of Sgr A*. Based on the statistical model of Witzel et al. (2018), with the four nights of observations, there is a probability of less than 0.05% to observe flux levels >6 mJy. The long tail of high flux levels, which occurred on multiple nights observed this year, is a strong indication that a log-normal distribution of flux variations may not be sufficient to describe the infrared activity of Sgr A*. Before these 2019 observations, Witzel et al. (2018) had shown that the near-infrared spectral properties at low flux densities can be explained by log-normally distributed flux densities in K- and M-band, though Dodds-Eden et al. (2011) had suggested including an additional component to describe high flux levels. Before this year, the highest flux levels deviated from the median model expectation by only <2σ (Witzel et al. 2018); the observations this year suggest the model should be re-derived to include the new data to determine if an additional component in the model is required. Most models also assume that Sgr A* is stationary (with no time dependence in the model parameters). With additional measurements, it will become possible to robustly differentiate between changes in the physical state of Sgr A* and a stationary model (e.g., Dodds-Eden et al. 2011; Meyer et al. 2014). We can also study potential changes in the activity of Sgr A* in the past several years by including data from Keck and VLT from 2014–2018 into an analysis similar to that of Witzel et al. (2018), which was based on data from 2003 to 2013. With additional data, we can also re-examine the timing characteristics and break timescale for light curves observed this year to look for deviations from the current model (e.g., such as a shorter break timescale) and to better understand the time evolution of accretion on Sgr A*.

Another possibility is that these observations are an indication that Sgr A* is experiencing an increase in activity due to either a change in its accretion state or accretion rate. As mentioned above, the passage of windy stars through their periapses (most notably, S0-2, in 2018 May) could potentially cause an extended rise in the accretion rate (Loeb 2004), although Ressler et al. (2018) argued that the effect of S0-2 on the structure of the accretion flow should be negligible. The S-star cluster has no known stars more massive than S0-2, so there are not any obvious high-mass-loss-rate candidate sources close to Sgr A*, including mass-losing giant stars. An alternative mechanism for causing an accretion rate increase is simply the irregular gas flow toward Sgr A*, which is likely to be quite lumpy and variable (Cuadra et al. 2008). Considerable work has recently been applied to this possibility by those interpreting the G sources as gas clumps orbiting toward Sgr A* (e.g., Schartmann et al. 2012; De Colle et al. 2014; McCourt et al. 2015). Among them, Kawashima et al. (2017) predicted a radio and infrared brightening in the ∼2020 time frame.

We predict that if the activity level of Sgr A* is indeed higher, then observations at other wavelengths should also show increased flux levels. Witzel et al. (2018) can explain the log-normal nature of lower near-infrared flux densities by a radiative model that is dominated by an exponential synchrotron cooling cutoff within or near the near-infrared band. However, the authors note that this limits the flux densities to a range below 2 mJy. In order to reach higher flux densities (such as observed this year), the overall brightness of the synchrotron spectrum needs to scale accordingly. Once the emission is dominated by the synchrotron spectrum, there should be strong correlation in the fluxes at other wavelengths (see also Eckart et al. 2012). This would predict large flux variations at radio, sub-millimeter, and X-ray wavelengths that would be observable and can be directly used to test this model.

Our recent observations of the Galactic center have captured Sgr A* in an unprecedented bright state in the near-infrared. Even more so, three of the four nights show Sgr A* in a clearly elevated state. The brightest flux levels observed in 2019 are over twice the peak flux value ever observed in the near-infrared from Keck and VLT. The distributions of flux variations from the four nights are also very unusual compared to the historical data, showing significant deviations from the model that was previously able to describe all historical Keck, VLT, and Spitzer measurements (Witzel et al. 2018; Chen et al. 2019).

The 2019 measurements push the limits of the current statistical models. These models may need to be revised to gain a better understanding of the probability of observing very high flux levels. In addition, the statistical models for Sgr A* variability should be expanded to provide more robust tests for changes to the Sgr A* accretion properties over time.

The major question is whether or not Sgr A* is showing increased levels of activity, and if so, how long it will last. Additional data, preferably multi-wavelength observations, throughout 2019 and beyond will be necessary to study the nature of its current variability.

We thank the anonymous referee for helpful comments. We thank the staff and astronomers at Keck Observatory especially Jim Lyke, Randy Campbell, Sherry Yeh, Greg Doppmann, Cynthia Wilburn, Terry Stickel, and Alan Hatakeyama. Support for this work was provided by NSF AAG grant AST-1412615, the W. M. Keck Foundation, the Heising-Simons Foundation, the Gordon and Betty Moore Foundation, the Levine-Leichtman Family Foundation, the Preston Family Graduate Fellowship (held by A.G.), and the UCLA Galactic Center Star Society. R.S. has received funding from the European Research Council under the European Union's Seventh Framework Programme (FP7/2007–2013)/ERC grant agreement No. 614922. R.S. acknowledges financial support from the State Agency for Research of the Spanish MCIU through the "Center of Excellence Severo Ochoa" award for the Instituto de Astrofísica de Andalucía (SEV-2017-0709). This research was based on data products from the Galactic Center Orbits Initiative, which is hosted at UCLA and is a key science program of the Galactic Center Collaboration. The W.M. Keck Observatory is operated as a scientific partnership among the California Institute of Technology, the University of California, and the National Aeronautics and Space Administration. The Observatory was made possible by the generous financial support of the W.M. Keck Foundation. The authors wish to recognize that the summit of Maunakea has always held a very significant cultural role for the indigenous Hawaiian community. We are most fortunate to have the opportunity to observe from this mountain.