Here’s why LB-1 can’t Contain a 70-Solar-Mass Black Hole

A tale of spectral analysis gone terribly wrong

In late November, a team of astronomers reported that the binary system LB-1 contained a black hole that was far more massive than theories of stellar evolution predict. Several weeks later, we found out they were completely wrong.

Astronomers have the most advanced stellar evolution codes and simulations ever made at their disposal. In addition to centuries of observations, we can now travel forward and backward in time in the life of a star, determining how it was born and how it will die. Thanks to powerful computers, we have a more detailed understanding of the lives of stars than ever before.

This is why a paper published in late November 2019 caused such a furor. It claimed that a star system called LB-1 contained a black hole 70 times the mass of the Sun. Now, even making generous assumptions about the composition and history of the star that formed that black hole, that result is completely at odds with our understanding of stellar evolution. Most models predict black holes substantially less massive, and if the result was true, it could throw our models into disarray. You might remember hearing about this; the Cosmic Companion wrote an article on it shortly after the paper was published in Nature.

A Hubble image of R136, an open cluster 160,000 light-years away that contains some of the most massive stars known. The heaviest weigh in around 300 solar masses — about the same required to form the black hole in LB-1, according to the Chinese team who first measured its mass. Many of the stars in R136, however, are losing mass through violent stellar winds, and will end up as black holes that are much lighter — closer to the actual mass of the BL-1 black hole, which astronomers have demonstrated to be much less massive that first reported. Image credit: NASA, ESA, F. Paresce (INAF-IASF, Bologna, Italy), R. O’Connell (University of Virginia, Charlottesville), and the Wide Field Camera 3 Science Oversight Committee.

Naturally, astronomers were skeptical. One of the pillars of the scientific method is reproducibility, and so several groups independently reanalyzed that data and checked the paper’s assumptions. Several weeks after the discovery was announced, the consensus was that the black hole was, in fact, nowhere near 70 solar masses.

Where did the first team of scientists go wrong? How could their calculated mass be so high — about ten times as high as what other groups found? What actually lies in LB-1? The answer these questions, we have to first understand the assumptions made by the astronomers behind the discovery and look at them with a critical eye. Let’s dive in.

A Theoretical Impossibility

The impetus for the paper came from a radial velocity survey performed by LAMOST, a Chinese optical telescope, beginning in 2016. The survey consists of follow-up observations of a number of Kepler fields, focusing specifically on spectroscopic binary stars. One of the over 3000 targets surveyed was named LB-1, whose luminous component appeared to be a main sequence B star. It interested scientists because it featured periodic radial velocity shifts, along with a strong, broad Hα line.

A team of Chinese astronomers (Liu et al. 2019) performed further observations in late 2017 and early 2018 using the OSIRIS spectrograph at the Gran Telescopio Canarias and the HIRES spectrograph at the Keck Observatory, which confirmed the radial velocity variations. The optical spectra enabled the group to measure the B star’s metallicity, surface temperature and surface gravity, and from there to use stellar models to determine its mass, radius and age.

Figure 2, Liu et al. On the left is a plot of fits to the radial velocity curves for both the B star and its dark companion, based on periodic spectral shifts of the Hα line. Notice that the companion’s motion appears less dramatic, implying a higher mass. On the right is a plot of the line itself, as a function of velocity shift. Keep this “wine bottle” shape, as the authors referred to it, in mind for later.

The star’s companion was not luminous, suggesting a compact object like a neutron star or black hole. Fits to the radial velocity curve placed a lower bound of 6.3 solar masses on its mass, ruling out the former option. Now, this curve was based on measurements of the absorption lines of the star, tracing its motion. The astronomers noticed that the Hα line could be an indicator of the companion’s motion — assuming it arose from a disk around what now surely seemed to be a black hole, and indeed, other disk configurations could not produce the required Doppler broadening.

Using the Hα radial velocity curve, the team calculated that the black hole should have a mass 68 times that of the Sun — plus or minus about a dozen solar masses. This is substantially higher than any known stellar-mass black hole, and the group were unable to produce such a system using stellar evolution models unless they made extreme assumptions about the mass loss of its progenitor, or treated the black hole as the remnant of the merger of two lower-mass black holes. This is the feature of interest that garnered LB-1 so much attention.

Extended Data Figure 7, Liu et al. This is a plot of the mass of the black hole based on different possible values for the mass of the progenitor once it enters the main sequence. The different curves show different mass loss rates based on different assumptions about the star’s winds. Only for high initial masses and very weak winds — and some possibly unrealistic assumptions about the resulting supernova — can the star retain enough mass to form such a massive black hole.

Unfortunately, in addition to the discrepancy with evolutionary models, there was one more problem with the results. Given the expected luminosity of the B star, the distance to the system would have to be 4.23 kpc — about twice the value determined by the telescope Gaia, which has produced detailed astrometric measurements of over one billion stars. The team hand waved this away by assuming that the wobble of the binary meant that the Gaia value was wrong, though without much of an argument in support.

Flaws in the Evolutionary Codes

As is the case with any extraordinary new scientific claim, many astronomers around the world scrutinized the results. The paper was published in late November, but by the middle of December, several groups had independently found flaws in Liu et al.’s methodology. One a team of theorists specializing in evolutionary models of binary star systems (Eldridge et al. 2019). Their code, the BPASS grid, formed a number of systems that fit the observations well but only required lower-mass black holes, many below 10 solar masses. This fits in with our prior understanding of how massive binary systems evolve.

Figure 1, Eldridge et al. These plots detail several models of the evolution of the system over its lifetime, covering several million years. On the far left is a Hertzsprung-Russell diagram, showing some familiar evolutionary curves. The other panels show radius and mass evolution; notice that neither star is overly massive, in contrast with the Liu et al. models.

The other key result of the paper was that the distance found by Gaia actually appeared quite reliable. They noted that Liu et al.’s argument about binary wobble rested on the assumption that Gaia’s measurements were performed at specific points in the system’s orbit, which appeared quite unlikely given the satellite’s observation log. Furthermore, calculations of the wobble showed that it should not contribute enough to the system’s parallax to justify the larger claimed distance. Other concerns about Gaia’s parallax-position biases were shown to be unfounded.

Why didn’t the Chinese group find the lower-mass black hole systems which the BPASS grids were able to synthesize? Eldridge et al. suggested that the problem was methodological: Liu et al. had used rapid population synthesis codes, which have less detailed physics and are unable to produce low-mass black holes. Therefore, their models were only capable of making relatively massive stellar black holes — which in this case proved to be a mistake.

Absorption, not Emission

The other paper I’d like to talk about (El-Badry & Quataert 2019) is a lesson in how to properly interpret astronomical data. Recall that Liu et al. thought that the Hα measurements showed an emission line from an accretion disk around the black hole. Two astronomers at UC Berkeley showed that this was an incorrect assumption, and that the data showed a Hα absorption line from the B star superimposed on a broad Hα emission line from a source elsewhere in the binary system.

Figure 1, El-Badry & Quataert. This toy model shows how a moving absorption line on top of a stationary emission line can mimic a moving emission line. The black curve shows the combined lines when the B star is moving tangentially to the line of sight. The red curve shows the line when the star is moving at 53 km/s; notice that absorption at 53 km/s appears to produce an emission on the opposite side — creating the “wine bottle” shape seen by Liu et al. The blue curve shows the line when the B star is on the other side of its orbit, moving at -53 km/s.

The logic was as follows. Say you have a circumbinary structure of some kind in the system — a disk, an outflow, etc. Perhaps this structure will produce Hα emission. Now, the B star should contribute absorption lines to the system’s spectrum, including an Hα absorption line, which moves in phase with the star. The combination of the two, if improperly modeled, will appear to show an Hα emission line in antiphase with the star — which might then mimic a source around the black hole.

Figure 4, El-Badry & Quataert. This plot shows two sets of data points; one, in red, contains the values Liu et al. found when they neglected to account for absorption by the B star. The other, in black, contains the values the Berkeley group computed when correctly accounting for absorption. These points predict a far less drastic radial velocity curve, and thus a much lighter dark companion.

El-Badry and Quataert showed that the astronomers had failed to account for this complicated blend of emission and absorption. When the effect was corrected for, it showed only a small radial velocity shift coming from the emission line, indicative of a circumbinary structure. The mass of the black hole would presumably be the in the range of 6 to 20 solar masses, which, as with Eldridge et al.’s results, fits comfortably with theory.

The Berkeley astronomers were not the only ones to show that Liu et al.’s spectral analysis had been improperly performed. A Belgian group (Abdul-Masih et al. 2019), almost simultaneously, made the same argument, based on new data they gathered using the HERMES spectrograph on the Mercator Telescope in La Palma. They suggested that a companion mass of about 4 solar masses was more likely; while this rules out a neutron star, it does not preclude the unknown object from being a rapidly rotating main sequence star, like the luminous B star, or a helium star. Though the final possibility may be ruled out on spectral grounds, the rapid rotator scenario cannot be until finer spectra are obtained.

When these preprints are considered together, it seems clear that LB-1 does not, in fact, contain a 70-solar mass black hole. It was simply a case of mistaken analysis, incorrect assumptions, and inadequate stellar models. While the slew of rebuttals in December might be disappointing — and it presumably was, for Liu et al. — the episode is still a lesson in scrutinizing results, especially some of the more surprising ones.

Graham is an undergraduate studying astrophysics and doing work on high-energy spectroscopy. He writes about deep sky objects and space exploration.

Did you like this article? Subscribe to The Cosmic Companion Newsletter!