Os resultados do projeto de mestrado de Roberta Duarte, atual doutoranda no Grupo de Buracos Negros, foi destaque no Jornal da USP. A pesquisa consistiu da aplicação pioneira de inteligência artificial para simular um buraco negro interagindo com o seu meio ambiente.
Our paper, “VLT/SINFONI study of black hole growth in high redshift radio-loud quasars from the CARLA survey”, has been accepted for publication in the Monthly Notices of the Royal Astronomical Society, and the pre-print appeared on astro-ph today. The paper was led by former graduate student Murilo Marinello, and this work formed a major part of his PhD dissertation published in April this year.
The new study focused on 35 distant, radio-loud quasars, the majority of which were selected from the Clusters Around Radio-Loud AGN (CARLA) survey. These quasars were known to have large black hole masses, emit luminous radio emission, and tend to be found in dense regions of the early universe. Therefore, they are believed to be good candidates for the distant progenitors of massive (elliptical) galaxies that dominate the universe today. The masses of their supermassive black holes had previously been estimated using the virial black hole mass method applied to their SDSS spectra. Due to their high redshifts, however, the only emission line available to make these measurements in these optical SDSS spectra was the CIV line. This line is known to be affected by non-gravitational effects (winds or outflows) and is thus not optimal for the virial black hole mass estimate. In this project, we therefore re-observed the quasars in the near-infrared using the SINFONI spectrograph on the Very Large Telescope in Chile. This allowed us to access the redshifted Ha broad emission line, and thus determine the black hole masses more accurately. This makes a big difference, as can be seen in the figure below showing the nice symmetric Ha line on the top and the distorted CIV line at the bottom for one of our quasars:
Together with a determination of the accretion rates of the quasars, which can be estimated from their luminosities, the new black hole masses were used to also derive the growth histories of these supermassive black holes. One major finding was that if these quasars had always been accreting at the same rates as measured at the current time, it would not have been possible for them to obtain their high observed masses within the cosmic time available since the Big Bang. The logical conclusion is thus that these quasars must have experienced a phase of much faster growth in the past. This can be nicely illustrated in the following figure:
The red points are the CARLA quasars from this study. The black solid lines show the growth tracks we found to be the ones describing their most likely histories. These tracks consist of two phases: a rapid growth phase starting from a one thousand solar mass black hole seed at z ~ 20, growing at the Eddington limit to a hundred million or more solar masses at z ~ 6, followed by a second, slower phase at the observed lower Eddington ratios until z ~ 2-3. As such, it is possible that the CARLA quasars are direct descendants of the luminous quasars found at z ~ 6-7.
In the local universe, there is a strong correlation between the masses of the supermassive black holes and the masses of their host galaxies. Since the more massive galaxies are also found, on average, in more massive dark matter halos, there is an indirect connection between the mass of the black hole and the mass of its halo. We therefore also tested whether the black holes in the CARLA quasars already “know” that they are located in dense galaxy environments:
We found a weak, low significance correlation between the black hole masses and the surface density of galaxies that surround them (the latter is a measure of the environment or halo mass of the CARLA quasars), and therefore do not find strong evidence that the most massive CARLA quasars are also in the most dense environments or massive halos. However, these galaxy surface densities had been previously determined with the Spitzer Space Telescope as part of the CARLA project, and are not very precise. In the future, we will therefore focus on trying to obtain more precise measurements of the environments of the CARLA quasars, and test again for possible correlations between black hole mass and environment.
Over the last decade more than five thousand gamma-ray sources were detected by the Large Area Telescope (LAT) on board Fermi Gamma-ray Space Telescope. Given the large positional uncertainty of the telescope, nearly 30% of these sources remain without an obvious counterpart in lower energies; these are called unassociated gamma-ray sources (UGSs). This motivated the release of several catalogs of gamma-ray counterpart candidates and several follow up campaigns in the last decade.
Majority is dominated by blazars
Between the associated sources, the large majority is composed by blazars, divided into BL Lacs, with a characteristic lineless spectrum (see figure below), and flat spectrum radio quasars (FSRQs), with broad emission lines and radio spectral index α < 0.5 (defined by the flux density S_ν ∝ ν^−α). In this sense, some of the most successful catalogs of gamma-ray candidate counterparts are the WISE Blazar-Like Radio-Loud Sources (WIBRaLS) catalog and the Kernel Density Estimation selected candidate BL Lacs (KDEBLLACS) catalog, both selecting blazar-like sources based on their infrared colors from the Wide-field Infrared Survey Explorer (WISE).
In this work we characterized these two catalogs, clarifying the true nature of their sources based on their optical spectra from Sloan Digital Sky Survey (SDSS) data release (DR) 15, thus testing how efficient they are in selecting true blazars. If a WIBRaLS2 or KDEBLLAC source is a true blazar, its spectrum may look like the following:
Typical FSRQ spectrum.
Typical BL Lac spectrum.
Based on the optical SDSS spectra, we found that at least ~30% of each catalog is composed by confirmed blazars, with quasars (QSOs) being the major contaminants in the case of WIBRaLS2 (~58%) and normal galaxies in the case of KDEBLLACS (~38.2%). We found that specially in the case of KDEBLLACS, the contaminants are mainly concentrated in the edges of the WISE color-color diagram (see figure below) and can be easily separated from the spectroscopically confirmed BL Lacs.
At least 30% of KDEBLLACS is composed by blazars.
The major contaminants in KDEBLLACS are QSOs and normal galaxies and they are concentrated mainly in the edges of the WISE color-color diagram.
Some sources in the Fermi-LAT catalogs are considered blazar candidates of uncertain type (BCUs) because the adopted association methods select a counterpart that satisfies at least one of the following conditions: i) An object classified as blazar of uncertain or transitional type in Roma-BZCAT. ii) A source with multiwavelength data indicating a typical two-humped blazar-like spectral energy distribution (SED) and/or a flat radio spectrum. BCUs are divided into three sub-types:
– BCU I: the counterpart has a published optical spectrum which is not sensitive enough for classifying it as FSRQ or BL Lac.
– BCU II: there is no available optical spectrum but an evaluation of the SED synchrotron peak position is possible.
– BCU III: the counterpart shows typical blazar broadband emission and a flat radio spectrum, but lacks a optical spectrum and reliable measurement of the synchrotron peak position.
In 4FGL, 1155 sources are considered as BCUs. Our analysis based on the optical spectra available in SDSS DR15 allowed us to give a conclusive classification for 11 of them: 2 BL Lacs, 4 BL Lacs with spectra dominated by the host galaxy, and 5 FSRQs. The SDSS spectral analysis also allowed us to find 25 new BL Lac objects which will be included in future releases of Roma-BZCAT.
This work contributes to a better understanding of the γ-ray sky in the Fermi-LAT era. In particular, the community will benefit from the characterization of WIBRaLS2 and KDEBLLACS in population studies of blazars and in subsequent programs of spectroscopic follow-up needed to confirm the nature of the UGSs.
The fact that some image of an event horizon would be announced on April 10th was not a surprise to our group. What caught us by surprise—since we are not members of the Event Horizon Telescope Collaboration (EHTC)—was the fact that the black hole imaged was in M87 and not in our Galactic Center (for more information, check out the six papers outlining this incredible, game-changing discovery).
Anyhow, a couple of days after the results were announced, while I was reading the first paper of the series, I was struck by a particular paragraph at the end of Section 6:
I immediately realized that I could produce a more precise estimate of the black hole spin of M87* based on the power of the relativistic jet. Kerr black holes can be completely described by only two numbers: the mass and the spin. There have been several measurements of the mass of M87* with a better than 10% uncertainty on the mass, for example using stellar or gas dynamics or the size of black hole shadow. Getting the spin however is a completely different story and much more difficult. Measuring the spin from the shadow is currently out of question because the images are not sharp enough to the degree that would allow us to get confident estimates. And many other methods in the literature suffer from issues such as large uncertainties in the data or model parameters.
I thought that perhaps I could contribute an interesting estimate of the spin of M87*. Wouldn’t that be a nice—and hopefully quick—paper? Over the next couple of days, I devoted myself entirely to getting this estimate right and assessing whether it would be worth of publishing. I was nervous because if I got that the spin a* is very low or consistent with zero then the result would not be very interesting and not worth writing a paper about it. What I found surprised me. And led me to write my quickest paper up to date: it took me two weeks from the beginning of the analysis up to having a manuscript submitted to ApJL.
Before digging into the results, what were the observables? The observables I chose were the total power carried by the jet coming from M87* (the jet power) and amount of mass being fed to the black hole—the mass accretion rate which I will also refer to as Mdot. If I have reliable measurements of these two numbers, then I could use current ideas about how black holes produce jets to tie the observations to an estimate of the black hole spin.
Where do these observables come from? The jet power was estimated by Russell et al. (2013) using Chandra X-ray observations of the hot gas around M87*. From the temperature, density and an idea of the volume of such gas, Russell et al. was able to quantify the average power dumped by the jet over a period of about one million years: about 1E43 erg/s. For reference, the black hole in the center of M87 is putting out ten billion times more energy in its environment than the Sun radiates per second.
I should mention that there are several different ways of estimating the jet power in M87* (for reviews, cf. EHTC paper V). I prefer the estimate from X-ray bubbles because it is more robust against time variability.
The mass accretion rate was estimated by Kuo et al. (2014) based on a clever idea originally proposed by Dan Marrone in the context of Sagittarius A* (Sgr A*). Marrone et al. (2006) figured out that if they have a good measurement of the amount of polarization that radiation suffers when it leaves the surroundings of the black hole, then using simple assumptions one can estimate the gas density near the black hole and Mdot. Concretely, one needs to measure the Faraday rotation measure (RM) and assume that one is observing synchrotron radio emission coming from the inner parts of the accretion flow and that his radiation is polarized by the accretion flow itself as it travels outwards and eventually reaches the observer. Kuo et al. (2014) measured the M87* RM and found that Mdot < 9E-4 Msun/year. In other words, Kuo et al. measured an upper limit to Mdot. This means that in one year, M87* eats up one Jupiter worth of mass (actually, less than that).
Importantly, if the observed polarization in M87* is not due to the accretion flow as a “Faraday screen”, then this will affect the estimates of Mdot. I will return to this point further below.
OK, so we have some amount of energy flowing out of the black hole—the jet power—and some upper limit on the amount of energy flowing into it—the Mdot. How can we put this together and estimate the rotation frequency of spacetime?
It turns out that black holes (BH) are in many ways similar to a car engine. If you wanted to reverse-engineer the energetics of an engine, you would just need to observe its fuel consumption, how much is lost in the exhaust and how much speed it delivers. Then you would have a good idea of how efficient the engine is and start working out how it could achieve such levels of fuel consumption. For a black hole it works exactly in the same way. If you know how much power it produces by accelerating particles in a jet, and if you know how much gas is being fed to the BH, you can work out how “green” it is. Why is this related to the title of this blog post? Because the level to which a BH is economical is related to how fast it rotates. The BH spin is the turbo in the engine: the larger the value of a* is, the larger is the amount of jet power produced by the BH for a given fixed Mdot.
To begin with, I used a model that specifies the efficiency of jet production η as a function of a*. This model is called Blandford-Znajek mechanism named after the researchers that solved Maxwell equations to first order in the curved spacetime of a BH almost forty years ago, and figured out how BHs can power jets (Blandford & Znajek 1977). The Blandford-Znajek model has a couple of free parameters and I needed to anchor these values otherwise I would not learn much about M87* from applying it to my data. I fixed the fudge factors in the model by using a series of advanced numerical simulations of how magnetized plasmas near event horizons behave as time progresses, which have the technical name of general relativistic magnetohydrodynamic (GRMHD) simulations of accretion onto Kerr BHs.
I based my models on the numerical results of my collaborator Sasha Tchekhovskoy, who is an assistant professor ar Northwestern University. The figure below summarizes how efficient BH engines are at producing jets according to the Blandford-Znajek model and GRMHD simulations.
There are two numbers that control the jet efficiency. The first one is the spin, of course. There is a second number as well: the magnetic flux on the event horizon of the BH. Because jets are powered by a helical twisting of magnetic field lines anchored in the event horizon, the power also depends on the magnetic field. Therefore, by modeling M87*’s data we can learn something not only about the spin but also the magnetic field near the BH.
The two main results of the paper are the following:
I get a robust lower limit on the black hole spin in M87* from the observations: a* > 0.5. This means that the black hole must be rotating at least at half of the maximal possible rotation frequency allowed by general relativity.
I find lower limits on the amount of magnetic flux threading the event horizon, 𝜙 > 5 (𝜙 in dimensionless units typical of GRMHD works). This means that the magnetic fields surrounding the BH are quite strong. This disfavors a whole category of accretion flow models known as “SANE” for M87*.
If these bounds were to be violated, then the BH would not be able to pump enough energy into the jet to be consistent with the observed power. Combining these results with the constraints from EHT observations—something that I have not done—should reduce even further the parameter space allowed for M87*.
A few more details about result #1. I actually considered both the cases in which the BH could rotating in the same (prograde) or opposite (retrograde) direction as the accretion flow (however, the angular momentum vectors must be parallel or antiparallel). If M87* is prograde, then the lower limit on the spin is |a∗| ≥ 0.4, otherwise it is |a∗| ≥ 0.5. I was not able to distinguish between the prograde or retrograde scenarios based only on the data available. Hopefully, the upcoming EHT polarimetric observations will shed more light on these issues.
What is the meaning of the spin parameter that I talked about above? The maximal possible rotation frequency allowed by general relativity corresponds to max(a*) = 1. At the maximal spin, the equator of the black hole would be rotating at the speed of light. Above that limit, one interpretation is that the black hole would break-up due to centrifugal forces and a naked singularity would be revealed. Nobody has figured out how to do that—even in theory.
Accretion rate and Faraday rotation measure
I should thank the referee because he/she really helped to improve the quality of the manuscript thanks to the thoughtful comments. One of the interesting points made by the referee was the following:
The upper limit on density relies on the model used by Kuo et al. to relate rotation measure to accretion rate. There are large uncertainties in this estimate! The RM depends not only on density, but also magnetic field strength and geometry, […] along a highly inclined line of sight. […]
This made me think about the underlying assumptions behind the Mdot estimate by Kuo et al. 2014. The idea goes back to Marrone et al. (2006) and requires a model for the density and magnetic field in the accretion flow, relying on the following assumptions:
the Faraday rotation is caused by the hot accretion flow in front of a source of synchrotron emission
the accretion flow is roughly spherical and characterized by a power-law radial density profile
the magnetic field is well ordered, radial, and of equipartition strength
Of course, real accretion flows are messy and turbulent. GRMHD simulations indicate that their magnetic fields are predominantly toroidal rather than radial (e.g. Hirose et al. 2004 and many other works). Marrone et al. argues that the assumption of a radial magnetic field should give only a small error. The outer radius used in the estimate of Mdot should depend on the coherence of the magnetic field. Kuo et al. assume rout=rBondi. If rout<rBondi, then Mdot will be even less than estimated, thereby increasing the lower limits on a* and phi.
Quantifying the impact of the line of sight on Mdot and hence on our estimates of a* and phi is also difficult. Given that the jet in M87 has a low inclination angle to the observer, one possibility that cannot be completely ruled out is that the RM originates from the jet sheath, with the line of sight of the observer not passing through the RIAF. This scenario was explored by Moscibrodzka et al. 2017 using GRMHD simulations. If that is the case, Moscibrodzka et al. concluded that the RM would be consistent with a higher Mdot than we considered (and potentially much higher). This would lower the spin and magnetic flux, as discussed in the letter.
Interestingly, in the models by the EHTC in paper V did not obey the Mdot constraints of Kuo et al. Mdot in their models is a free parameter that is tuned to reproduce the observed compact mm flux. The exact value of Mdot in the general relativistic ray-tracing simulations employed in paper V depends on the electron thermodynamics and spans a wide range.
One feature of current GRMHD simulations tackling the amount of Faraday rotation in RIAFs such as Moscibrodzka et al. (2017) is that they have a very small torus extending up to about 60M. Furthermore, they do not have a high enough spatial resolution to resolve MRI in the outer regions of the disk—which is a crucial ingredient for reliable RM estimates—and do not have long enough durations to establish inflow equilibrium in the outer parts of the disk. While a jet-originated RM is possible, I am afraid that some GRMHD simulations might be underestimating the polarization effects of the RIAF at larger scales and this might impact their conclusions on the amount of Faraday rotation. In conclusion, both models for M87*’s RM—the simple analytical RIAF model a la Marrone and the current round of GRMHD simulations—are incomplete. There is definitely a lot of space for improvements in these calculations.
This work was supported by FAPESP (Fundação de Amparo à Pesquisa do Estado de São Paulo) under grant 2017/01461-2.
Ontem foi um dia muito especial para o nosso grupo de pesquisa, com a divulgação da primeira imagem do horizonte de eventos de um buraco negro da história, pela colaboração Event Horizon Telescope (EHT) (press releases da NSF e ESO).
Este resultado científico é um grande divisor de águas na astronomia. A partir de agora, na astrofísica de buracos negros, vamos nos referir à era “antes do EHT” e “depois do EHT”. Buracos negros, que eram antes entidades abstratas—comumente ilustradas em filmes e desenhos animados, mas das quais tínhamos somente observações indiretas—agora tem uma imagem concreta.
A ilustração publicada no site xkcd revela bem as gigantescas escalas do astro. O horizonte de eventos em M87 tem 38 bilhões de km de diâmetro, que é um pouco maior que o nosso Sistema Solar! Apesar disso, lá dentro há uma quantidade de matéria seis bilhões de vezes maior que a massa do Sistema Solar. São números quase inimagináveis para quem não é da área.
É interessante mencionar que, ao olharmos para a mancha escura no centro da imagem acima, estamos nos defrontando com um imenso vazio cósmico. A massa de seis bilhões de Sóis que existe dentro do buraco negro de M87 está totalmente concentrada num ponto central chamado de singularidade. E entre a singularidade e o horizonte de eventos—nome que damos para superfície absolutamente negra do buraco negro—não há nada. É literalmente um coração das trevas, como o título do romance de Joseph Conrad.
A auréola dourada mostra a radiação eletromagnética com comprimento de onda de 1.3 mm, emitida pelo gás nas partes internas do turbilhão espiralando na direção do horizonte de eventos—chamado de disco de acreção—pouco antes de cair dentro do buraco negro e se perder para sempre do nosso universo.
We studied the Akira galaxy, which was named after the Akira manga by Edmond Cheung. Its companion galaxy is called Tetsuo. Akira is an interesting galaxy because it hosts a supermassive black hole fed at quite low rates—we call it a low-luminosity active galactic nucleus.
The black hole seems to be ejecting gas quite vigorously. In fact so vigorously that the BH outflow is capable of quenching star formation in the galaxy. Cheung et al. called this galaxy a “red geyser”.
We observed the nucleus of the galaxy (where the black hole is located) with Gemini integral field spectroscopy (IFU) in order to characterise the black hole outflow. This is a powerful technique because it gives us high-spatial resolution information on several emission and absorption lines.
Below is the money plot of the paper. It tells us that the outflow coming from the black hole is changing its orientation as it propagates away from the galactic nucleus! How to interpret this?
First of all, we do not think we are seeing a jet because this galaxy does not show any extended radio structures. We think this is a subrelativistic, uncollimated wind as shown in the illustration below. We interpreted this as a precessing wind, with the likely cause of the precession being a misalignment between the accretion disk and the BH spin aka Lense-Thirring precession.
Let’s congratulate Ivan on his brilliant masters dissertation defense. His dissertation’s title is “Winds and feedback from supermassive black holes accreting at low rates”. In this work, Ivan performed a suite of hydrodynamical simulations of hot accretion flows with a large dynamical range and long durations (comparable to the viscous timescale), aiming at better understanding black hole wind production and feedback in low-luminosity AGNs hosted by quiescent galaxies.
We have a paper coming out soon, where we will report the results of this work. Stay tuned!
The evaluation committee was composed of Thaisa Storchi Bergmann, Roderik Overzier and Diego Falceta Gonçalvez.
This work was funded by a FAPESP scholarship, grant number 2016/24857-6. It has made use of the following computing facilities:
Laboratory of Astroinformatics (IAG/USP, NAT/Unicsul; FAPESP grant 2009/54006-4)
Aguia cluster, HPC resources of Universidade de São Paulo
Recently, we embarked in the adventure of porting a “Monte Carlo radiative transfer in curved spacetimes” code from multi-threaded CPUs to GPUs. This work is important for three reasons:
It is fun!
It is important because most general relativistic radiative transfer procedures currently available neglect inverse Compton scattering, which is particularly important to the group since we want to understand the nonthermal emission of active galactic nuclei
Radiative transfer can be slow; on the other hand, it can usually be made massively parallel with a couple of algorithmic improvements
These are some of the reasons why we decided to port a radiative transfer code to GPUs. We have reasons to believe that we can achieve speed-ups of a factor of ~100 times compared to a parallel, OpenMP (CPU) version of the code when using a modern GPU such as a GTX 1080 Ti. This can be a game-changer to allow faster modelling of the radiation from accreting black holes.
We are collaborating with colleagues from the computer science department at the university (Alfredo Goldmann and Matheus Tavares Bernardino). This is work in progress and we have some exciting preliminary results, which we unfortunately cannot publicly share yet. We hope to soon have a paper reporting these results and share the GPU-accelerated black hole radiative transfer code on Github.
In the meantime, we have one software deliverable from this project which may be useful for other researchers: we partially ported some functions of the GNU Scientific Library (GSL) to CUDA. More specifically, we ported the following functions:
gsl_ran_dir_3d: generate random vectors in the 3D space
gsl_sf_bessel_K0_scaled_e, gsl_sf_bessel_K1_scaled_e, gsl_sf_bessel_Kn: different sorts of “exotic” Bessel functions
gsl_ran_chisq: generate random numbers drawn from the chi-square distribution
Our group just obtained a 2 million CPU-hours allocation time at the Santos Dumont supercomputer. We will use this allocation grant to perform our numerical simulations of black hole accretion disks and their outflows, in order to understand how they impact their galaxies.
The PhD students of the group working on gamma-ray observations—Fabio and Raniere—spent the last two weeks in Washington DC and surroundings. They went to the Fermi LAT Collaboration meeting at George Washington University, where they interacted with gamma-ray astronomers in the Fermi Collaboration. Raniere presented his ongoing analysis of the gamma-ray emission of a population of nearby AGNs.
Following the Collaboration meeting, the students presented their research at the Fermi Symposium in Baltimore. Raniere presented a poster about his work on the pulsar populations in Milky Way globular clusters—which is about to be submitted for publication—while Fabio gave a talk describing his analysis of the gamma-ray emission from the Galactic Center on constraints on Sgr A* physics.
After the symposium, Fabio and Raniere spent a couple of days visiting NASA Goddard Space Flight Center to discuss their research with GSFC scientists.
Their visit was possible thanks to NASA funds, grant xxxxxx.