A binary system in the S cluster close to the supermassive black hole Sagittarius A* Authors: Florian Peißker, Michal Zajaček, Lucas Labadie, Emma Bordier, Andreas Eckart, Maria Melamed, Vladimír Karas Published: 2024 ABSTRACT AbstractHigh-velocity stars and peculiar G objects orbit the central supermassive black hole (SMBH) Sagittarius A* (Sgr A*). Together, the G objects and high-velocity stars constitute the S cluster. In contrast with theoretical predictions, no binary system near Sgr A* has been identified. Here, we report the detection of a spectroscopic binary system in the S cluster with the masses of the components of 2.80 ± 0.50 M⊙ and 0.73 ± 0.14 M⊙, assuming an edge-on configuration. Based on periodic changes in the radial velocity, we find an orbital period of 372±3 days for the two components. The binary system is stable against the disruption by Sgr A* due to the semi-major axis of the secondary being 1.59±0.01 AU, which is well below its tidal disruption radius of approximately 42.4 AU. The system, known as D9, shows similarities to the G objects. We estimate an age for D9 of 2.7−0.3+1.9×106 yr that is comparable to the timescale of the SMBH-induced von Zeipel-Lidov-Kozai cycle period of about 106 yr, causing the system to merge in the near future. Consequently, the population of G objects may consist of pre-merger binaries and post-merger products. The detection of D9 implies that binary systems in the S cluster have the potential to reside in the vicinity of the supermassive black hole Sgr A* for approximately 106 years.Subject terms: Stellar evolution, Transient astrophysical phenomena Introduction The central parsec around the supermassive black hole (SMBH) Sgr A* contains a large number of stars that constitute the Nuclear Star Cluster (NSC)1, which is one of the densest and most massive stellar systems in the Galaxy. These stars vary in terms of their ages, masses, sizes, and luminosities2. In the vicinity of Sgr A* of about 40 mpc, there is a high concentration of stars3 that orbit the black hole at velocities of up to several thousand km/s4,5 inside the S cluster. The presence of stars close the Sgr A* is not surprising because it was expected that old and evolved stars would gradually descend towards Sgr A* due to the cluster relaxation timescale of ~1010 yr6. This is because star formation is significantly inhibited by tidal forces and high energetic radiation in the vicinity of the SMBH. In fact, a cusp of late-type stars with stellar ages of >3 × 109 yr was identified7. Interestingly, these late-type stars coexist with massive early-type S cluster members that exhibit an average age of ~4−6 × 106 yr8,9, resulting in the formulation of the “paradox of youth”10. Until now, no companions have been identified for these young B-type stars11, although binary rates close to 100% have been proposed12. Therefore, the presence of binary systems in the S cluster is a crucial question to investigate the dynamical evolution of stars in the vicinity of Sgr A*13,14. Given that the evolution of high-mass stars is altered by their binary interactions15, it is important to understand the prevalence of putative binary systems in this cluster. In this work, we present the detection of a spectroscopic binary in the S cluster. Based on the photometric characteristics of the binary system, known as D9, it can be considered to be a member of the G-object population16,17. The age of the system is ~2.7 × 106yr, which is comparable to the von Zeipel-Lidov-Kozai cycle period of approximately 106 years. The dusty source D9 is most likely composed of a Herbig Ae/Be star associated with the primary. The lower-mass companion can be classified as a T-Tauri star. In the near future, the binary may undergo a merging event due to the ongoing three-body interaction of the system with Sgr A*. The uncertain nature of the G objects can thus be resolved, at least in part, thanks to the binary system D9 whose imminent fate appears to be a stellar merger. Results Observations Using archival data observed with the decommissioned near-infrared integral field unit (IFU) of Spectrograph for INtegral Field Observations in the Near Infrared (SINFONI, mounted at the Very Large Telescope)18,19 in the H+K band (1.4−2.4 μm) between 2005 and 2019, we investigate the blue-shifted Brackettγ (Brγ) emission of the source D9 (Fig. 1), which is part of the G-object population in the S cluster16,17,20. In addition, we include recent Enhanced Resolution Imaging Spectrograph (ERIS) observations carried out by the ERIS Team as part of the commissioning run in 202221. For the analysis of the three-dimensional data cubes that consist of two spatial and one spectral dimension, we perform standard reduction steps (flat-fielding, dark, and distortion corrections). We obtain single barycentric and heliocentric corrected data cubes that are stacked for each year individually to construct a final mosaic of the entire S cluster region. Based on the best-fit Keplerian solution, we obtain an estimate of the periapse distance of the D9 system from Sgr A* of 29.9 mpc (0.75 arcseconds) adopting MSgrA* = 4 × 106 M⊙ and 8 kpc for the mass and the distance of Sgr A*, respectively22,23. Furthermore, we find a close to edge-on orbital inclination of (102.55 ± 2.29)°. With an eccentricity of 0.32 and a semi-major axis of 44 mpc, D9 qualifies as an S cluster member with orbital parameters comparable to other S stars3,24. Due to the orbit of the B2V star S2 (S0-2) that intercepts with the trajectory of D9, we focus on the data set of 2019 to identify a continuum counterpart in the H and K band to the Brγ line-emitting source. Detection of the D9 system close to Sgr A* in 2019. a Shows the Doppler-shifted Brγ line map extracted from the H+K SINFONI data cube with a corresponding wavelength of 2.1646 μm (vacuum wavelength 2.1661 μm). b, c Shows the near-infrared H (1.6 μm) and K (2.1 μm) band data observed with SINFONI. d Denotes the mid-infrared L (3.76 μm) band observation carried out with NIRC2. Sgr A* is marked with a × , D9 is encircled in every plot. Due to its main sequence character, the marked close-by star S59 can only be observed in the H and K bands. On the contrary, the brightest K band source of the S cluster, S2/S0-2 can be observed in every shown infrared band. To increase contrast, an image sharpener is applied suppressing expansive point spread function (PSF) wings. To emphasize the astrometric robustness of the image sharpener, we adapt the lime-colored contour lines from the non-sharpened data. The contour line levels in b are at 10–80% of the peak intensity of S2, increasing in 5% steps. c The contour lines are set at 20–100% of the peak intensity of S2, separated by 10%. d The contour lines are set to 85%, 90%, 95%, and 100% of the peak intensity of S2. The labels of the axis indicate the distance to Sgr A* located at ΔRA = 0.00” and ΔDEC = 0.00''. In any plot shown, north is up, and east is to the left. Magnitudes To increase the photometric baseline, we incorporate Near-infrared Camera 2 (NIRC2, mounted at the Keck telescope) L band imaging data from 2019 to cover the near- and mid-infrared25. The science-ready data was downloaded from the Keck Observatory Archive26. Due to the high stellar density of the S cluster27, dominant point spread function (PSF) wings are a common obstacle that hinders confusion-free detection of fainter objects such as G128, DSO/G229, or D920. Therefore, we used an image sharpener on the continuum data of 2019 to reduce the impact of the challenging crowding situation in the S cluster (Supplementary Fig. 1 and Supplementary Table 1). With this procedure, we enhance fainter structures but preserve the photo- and astrometric aspects of the input data. To emphasize the robustness of the image sharpener, we invoke the contour lines of the input data as a comparison, as demonstrated in Fig. 1. Analyzing the displayed extinction corrected data (Supplementary Table 2), we find H−K = 1.75± 0.20 and K−L = 2.25 ± 0.20 colors for D9 suggesting photometric similarities with D2 and D2320. The latter two sources are believed to be associated with young T Tauri or low-mass stars16,30,31. Due to these photometric consistencies (Supplementary Fig. 2), we tested the hypothesis using a Spectral Energy Distribution (SED) fitter. Spectral energy distribution The SED fitter32 applies a convolving filter to the individual values to reflect on the response function of the instrument filter. Because the photometric system of SINFONI is based on the Two Micron All Sky Survey (2MASS) data base, we select the corresponding filters “2H” and “2K”. For the NIRC2 MIR data, we use the United Kingdom Infrared Telescope (UKIRT) L’ band filter because it is based on the Mauna Kea photometric system33. With these settings, the fitter compares models with the input flux (Fig. 2) where we limit the possible output that satisfies Δχ2 ≤ 3. These models represent young stellar objects (YSOs) and are composed of a stellar core, an accretion disk, and a dusty envelope. These typical components constitute a YSO and can be traced in the near- and mid-infrared parts of the spectrum. As input parameters, we used the H (0.8 ± 0.1 mJy), K (0.3 ± 0.1 mJy), and L (0.4 ± 0.1 mJy) band flux densities estimated from the continuum detection presented in Fig. 1. Considering common YSO models, the H and K band emission traces the core components of the system, whereas the L band emission can be associated with a dusty envelope. Based on a photometric comparison with 104 individual models, the best-fit of the SED fitter results in a stellar temperature of 1.2 × 104 K and a corresponding luminosity of approximately 93 L⊙, which are associated with a stellar mass of 2.8 ± 0.5 M⊙ (see Table 1). Spectral energy distribution of the D9 system. The extinction corrected data points refer to the flux density values in the H, K, and L band observed with SINFONI and NIRC2. We use 104 individual models to find the best-fit of the data shown with gray lines. The final best-fit result is depicted with a black line. Based on the shown fit, the properties of the primary of the D9 binary system are derived and listed in Table 1. The uncertainties of the data points are estimated from the photometric variations along the source. Based on the reduced χ2 value of ~2, the displayed best-fit solution was selected. Table 1Best-fit parameters of the D9 systemSecondary Keplerian ParameterPD9b [year]1.02 ± 0.01eD9b0.45 ± 0.01ωD9b [deg]311.75 ± 1.65aD9b [au]1.59 ± 0.01iD9b [deg]90.00msin(iD9b) [M⊙]0.73RVoff [km s−1]−29.19 ± 3.00χν20.31rms [km s−1]16.38Keplerian parameter for D9 orbiting Sgr A*eD9a0.32 ± 0.01iD9a [deg]102.55 ± 2.29aD9a [mpc]44.00 ± 2.42ωD9a [deg]127.19 ± 7.50ΩD9a [deg]257.25 ± 1.61PD9a [yr]432.62 ± 0.01Radiative transfer modeliintrinsic [deg]75.0 ± 19.0R [R⊙]2.00 ±l 0.13log(L/L⊙)1.86 ± 0.14log(TD9a[K])4.07 ± 0.05MD9a [M⊙]2.80 ± 0.50MDisk [10−6 M⊙]1.61 ± 0.02We list the orbital parameters for the binary of D9 together with the motion of the system around Sgr A*. In addition, the best-fit stellar properties based on the SED fitter are included. The uncertainties of the binary parameter and the radiation transfer model are based on the reduced χ2. For the Keplerian elements, we use MCMC simulations to estimate the uncertainty range. Since the inclination of the secondary is assumed to be iD9b = 90°, no uncertainty for msin(iD9b)=0.73 M⊙ is given. Best-fit parameters of the D9 system We list the orbital parameters for the binary of D9 together with the motion of the system around Sgr A*. In addition, the best-fit stellar properties based on the SED fitter are included. The uncertainties of the binary parameter and the radiation transfer model are based on the reduced χ2. For the Keplerian elements, we use MCMC simulations to estimate the uncertainty range. Since the inclination of the secondary is assumed to be iD9b = 90°, no uncertainty for msin(iD9b)=0.73 M⊙ is given. Periodic pattern While finalizing the analysis of D9, a pattern of radial velocity came to our attention. By inspecting the SINFONI mosaics that depict every observed night between 2005 and 2019, we found a clear periodic signal shown in Fig. 3 between −67 km/s and −225 km/s using the Doppler-shifted Brγ emission line with respect to its rest wavelength at 2.1661 μm. A comparison of the periodic pattern of D9 with the Doppler-shifted Brγ emission line of D23 demonstrates that the signal is not an artifact (Supplementary Fig. 3). From the orbital fit and the related inclination of i = (102.55 ± 2.29)°, we know that D9 is moving on an almost edge-on orbit with a proper motion of vprop = 249.43 ± 5.01 km/s. Since S2 (S0-2) moves with a proper motion of almost 800 km/s34, the comparable slow velocity of D9 implies that the intrinsic RV baseline vbase of the system, estimated with (vmin+vmax)/2, will not change significantly between 2005 and 2019. We normalize all observed velocities vobs to this baseline with vobs-vbase to obtain vnorm, which is the input quantity for the fit of the binary system performed with Exo-Stricker35. Due to the poor phase coverage before 2013, we split the data to perform an independent sanity check. The fit displayed in Fig. 3 resembles the epochs between 2013 and 2019, where we used a false-alarm probability of 10−3 similar to that used by ref. 14. The data baseline between 2005 and 2012 represents a non-correlated parameter to the Keplerian model of the binary provided by Exo-Striker, which is in agreement with the fit that is based on the epochs between 2013 and 2019 (Fig. 3). With a similar motivation, we incorporate the ERIS observations from 2022 that show a satisfactory agreement with the RV model and the expected LOS velocity of the binary, consisting of a primary and a secondary. Regarding the possible impact of a variable baseline vbase (i.e., the LOS velocity vobs of D9 increases), we measure a difference of ± 15km/s between 2013 and 2019, which is consistent with the estimated uncertainty of ± 17km/s from the fit. We conclude that a variation of vbase over the complete data baseline is inside the uncertainties and does not impact the analysis significantly. However, a forthcoming analysis of the binary system D9 should take this adaptation into account because it is expected that an alteration of the intrinsic LOS velocity will exceed the uncertainty range of the individual measurements of the periodic signal within the next decade. Radial velocity of D9 between 2005 and 2022 observed with SINFONI and ERIS. a–c We display three selected nights to show the variable Brγ emission line with respect to the rest wavelength at 2.1661 μm. The top three plots correspond to the same colored boxes as in the radial velocity evolution model shown in (d). We have indicated the exact data point using magenta color. d The SINFONI data is indicated in green, and the two ERIS observations from 2022 are highlighted in black. Due to the decommissioning, no high-resolution spectroscopic data are available between 2020 and 2021. In addition, the usual observation time for the Galactic center at Cerro Paranal (Chile) is between March and September, which explains the limited phase coverage. All data points in the radial velocity (d) correspond to a single night of observation. The velocities in the left y axis are related to the observed blue-shifted Brγ emission lines. Due to data processing, these values are shifted and arranged to an estimated zero-velocity baseline (see the right y axis). The uncertainties of the individual data points are calculated from the root-mean-square (RMS) deviation (see Table 1). In the subsequent analysis, we will refer to the primary as D9a, whereas the secondary companion will be denoted as D9b. With the binary orbiting Sgr A*, this three-body system is divided into an inner and outer binary. The inner binary describes D9a and D9b, while the outer one represents the D9 system orbiting Sgr A*. The best-fit result includes an offset of vbase with RVoff = −29.19±3.00 km s−1 due to the eccentricity of the secondary eD9b of 0.45 ± 0.01, which causes an asymmetric distribution of the LOS velocity around the baseline. With this offset, we obtain vmod=vnorm+RVoff as displayed in Fig. 3. The related Keplerian parameters of the secondary orbiting its primary are listed in Table 1. From the fit but also evident in the periodic RV data points (Fig. 3), we find an orbital period for the secondary of PD9b = 372.30±3.65 days = 1.02 ± 0.01 yr, which can be transferred to a total mass of the system of about Mbin = 3.86 ± 0.07 M⊙, considerably above the derived D9 (i.e., the primary) mass of MD9a = 2.8 ± 0.5 M⊙. The difference in mass for MD9a and Mbin cannot be explained solely by the uncertainty range. However, inspecting msin(iD9b)=0.73M⊙ and the assumed inclination of the secondary of 90° results in the maximum mass of the companion. The assumed inclination of the secondary is motivated by an almost edge-on orbit of D9 (Table 1). Although the circumprimary disk does not necessarily have to be aligned with the orbit of the binary as is found for T-Tauri systems36, surveys of Herbig Ae/Be stars suggest a tendency towards a coplanar arrangement37. Assuming that the orbit of the secondary is approximately aligned with the circumprimary disk with an intrinsic inclination of the primary D9a of iintrinsic = (75±19)° (Table 1), we are allowed to transfer the related uncertainties to msin(iD9b). Following this assumption, we find a mass for the secondary of MD9b = 0.73 ± 0.14 M⊙ consistent with the derived primary mass of MD9a = 2.8 ± 0.5 M⊙ and the total mass Mbin = 3.86 ± 0.07 M⊙ of the system. Discussion Radiation mechanism Taking into account the periodic variation of Brγ emission, we want to highlight three different scenarios as a possible origin of the periodic Brγ signal. Firstly, the emission of the Brγ line is solely the result of a combination between the gaseous accretion disk and stellar winds of the primary38,39. In this scenario, the secondary disturbs this emission by its intrinsic Keplerian orbit around the primary. Secondly, a possible origin of the Brγ line could be the presence of a circumbinary disk around the D9 binary system enveloping the primary and the secondary. In this case, the interaction between the primary with the secondary allows inward gas streams from the circumbinary disk resulting in the observed periodic Brγ line40. The third and foremost plausible scenario is the interaction between two accreting stellar objects. It is well-known that especially Herbig Ae and T-Tauri stars exhibit prominent Brγ emission lines associated with accretion mechanisms41,42. For instance, a radial shift of the accretion tracer has been observed for the DQ Tau binary system43. It has been proposed that this resonance-intercombination may be explained by stellar winds of the secondary44. Due to Keplerian shear, line photons can escape the optically thick material and produce the RV pattern, as observed for the D9 binary system45. Stellar types of primary and secondary Considering the presence of a primary and its companion, it is suggested that stellar winds interact with the Brγ emission of the accretion disk(s) of the binary system38,46 that gets periodically disturbed by the presence of the secondary47,48. Alternatively, the Brγ emission line is produced by both the primary and secondary as it is observed for the Herbig Ae star HD 104237 with its T-Tauri companion49. Comparing MD9a with the total mass of Mbin = 3.86 M⊙ of the system suggests that the secondary does not contribute significantly to the photometric measurements analyzed in this work. If it were not the case, the estimated mass for the primary of the D9 system of MD9a = 2.8 ± 0.5 M⊙ would be lower, while MD9b = 0.73 ± 0.14 M⊙ should be increased. Considering the estimated mass of the primary MD9a and the fixed upper limit of Mbin based on the observed period, the secondary can be classified as a faint low-mass companion, suggesting a classification as a T-Tauri star50. Considering the stellar mass, radius, and luminosity of the primary (Table 1), the system may be comparable to the young Herbig Ae/Be star BF Orionis, which is speculated to also have a low-mass companion51. On the basis of observational surveys, it is intriguing to note that most Herbig Ae/Be stars exhibit an increased multiplicity rate of up to 80%37,52. Another result of the radiative transfer model is the relatively small disk mass MDisk of (1.61 ± 0.02) × 10−6 M⊙, which could be interpreted as an indicator of the interaction between D9a and its low-mass companion D9b. Possibly, this ongoing interaction, but most likely the stellar winds of the S stars53, will disperse the disk of D9a in the future54–57. Using the derived luminosity and stellar temperature of D9a together with the evolutionary tracks implemented in PARSEC58, we estimated the age of the system of 2.7−0.3+1.9×106 yr. Migration scenario A potential migration scenario has been proposed by ref. 59 and can be described as the triple-system hypothesis. In this scenario, a triplet system migrates towards Sgr A* 60–62, where the two companions are captured to form a binary. It is possible that the third companion may be ejected from the cluster and subsequently become a hyper-velocity (HV) star, as postulated by refs. 63,64. A consequence of the disruption of the initial triplet is the resulting high eccentricity of the captured binary system close to unity65. Since the derived outer eccentricity of the D9 system is eD9a = 0.32 ± 0.01 (Table 1), we consider a migration channel different from the triple-system hypothesis. As proposed by refs. 60 and 61, molecular clouds can migrate towards the inner parsec and consequently close to Sgr A*. Speculatively, the D9 system could have formed during such an inspiral event. An additional implication based on the age estimate is the presumably evaporated circumbinary disk that enveloped the primary and secondary. The authors of ref. 66 found that the timescales for dismantling the circumbinary disk scale with the separation between the primary and secondary. The relation can be formulated with tdis.time≤106yr