Friday, October 4, 2013

The Race for Gold*

Gold ring. Credit: P.Edmonds
Astronomers know a lot, but they’re not certain how the gold in this ring was made. There are two leading ideas, both containing at least one destructive event. One idea is that it was created in the collapse of a massive star followed by a powerful explosion. These supernova explosions form neutron stars – dense objects with extraordinary properties – or black holes. In the extreme conditions produced by the explosion, gold and other heavy elements may form. This has been the default story for many people, including astronomers like Neil Tyson**.

The other idea is even more spectacular. Here, we follow the story of particles that were squashed down into a neutron star following a supernova. Normally they would stay locked away for countless billions of years, but in a few special cases they can be liberated. If the neutron star has a companion that also explodes to form a neutron star or black hole, and if these two collapsed objects are close to each other, they can form a kind of deadly dance. As they spin around each other, energy leaks away via ripples in space-time and the orbit gets tighter until the two objects merge and explode, forming a black hole. In the aftermath of the explosion, neutrons slam into the stellar remains to form gold and other heavy elements like platinum.

video
Animation showing a merger of two neutron stars resulting in a gamma-ray burst. CreditDana Berry, SkyWorks Digital, Inc.

That's quite a story, involving three powerful explosions and the warping of space-time. In this post I'll explain some details about the closely contested academic race that uncovered evidence for this exotic process, and two different publicity approaches that followed. I’ll focus on the aggressive one, because it’s interesting to see how boundaries were pushed in several ways.

I didn't know the merger story was a serious contender for making heavy elements like gold until this paper appeared on the arXiv: "Smoking Gun Or Smoldering Embers? A Possible R-Process Kilonova Associated with the Short-Hard GRB 130603B". A short-hard gamma ray burst (GRB) is thought to occur when the merger described above occurs, producing a powerful explosion and a burst of gamma rays. A chain of nuclear reactions called the "r-process" should then occur, producing heavy elements like gold and a second source of radiation called a "kilonova", a few days after the explosion.

In the paper mentioned above, Edo Berger and colleagues from Harvard-Smithsonian Center for Astrophysics (CfA; in full disclosure, where I work) used observations from the Hubble Space Telescope (HST) to argue that an infrared source appearing at the position of a GRB might be a kilonova. They also pointed out an alternative explanation was feasible, when particles slam into the surrounding environment giving an afterglow.

Their paper made the kilonova claim carefully, using "possible" in the headline and other cautious statements, like: "If true, the kilonova interpretation provides the strongest evidence to date that short GRBs are produced by compact object mergers, and places initial constraints on the ejected mass."

What followed really caught my attention, as someone who works in astronomy publicity. This paper was the basis for a CfA press release and press conference on July 17th. The press release framed the result as possibly explaining the origin of the gold found on Earth and in the universe, in the merger story described earlier. Bling was mentioned, and the result received a lot of press coverage. This coverage was heavily driven by the CfA publicity, as there was a clear gap between the content of the paper and that of the press release and press conference. The paper did not mention the debate about the source of heavy elements. Also, the paper did not explicitly mention gold, so no description of the expected yield from this possible kilonova was given, and no details of the extrapolation to calculate the total amount of gold in the universe were given, including the uncertainties.

Edo Berger at the CfA press conference announcing possible detection of a kilonova. Credit: CfA/Edo Berger
The paper was not accepted for publication at the time of the publicity. The authors did their analysis extremely quickly, as the HST data they used was available on June 13th and their paper was put onto the arXiv on June 17th. The authors could have delayed submission of the paper to allow more details to be included in it or publicized a follow-up paper with these details. An extra benefit of this approach is that more data could have been included, to check for expected fading of the source.

A competition emerges

Why the rush? I can see only one answer: competition. Another team, led by Nial Tanvir of the University of Leicester, originally proposed for the data that Berger et al. used, which was Director's Discretionary Time (DDT) and was publicly available straight away. Tanvir et al. submitted their own paper, titled "A search for kilonova emission associated with GRB 130603B: the smoking gun signature of a compact binary merger event" only 3 days after Berger et al submitted their paper. An Astrobites blog post by Elisa Chisari gives more details about the astrophysics.

The competition aspect is obvious because Tanvir et al. finished their abstract by saying
"We note that we felt compelled to submit this provisional report of our work, despite our HST DDT program being incomplete, due to other authors having already posted an analysis of the publicly available first epoch data." 
That’s not a statement you see every day on the arXiv, and I look at it regularly. The good news is that they also argued a kilonova may have been detected, but they cautioned that "we cannot yet say whether the light is transient in nature".

They soon cleared up this uncertainty when a second set of HST observations from the DDT program came in. This time the data was not made public straight away, as it usually is. Observers can ask for a short proprietary period if they have a good reason. Tanvir told me: 
"We didn't ask in the first instance, but after our experience with the first epoch, we argued that it was important to give sufficient time to do a careful analysis of the data without being rushed into publication." 
So, Tanvir and his team were given one month to work on the data exclusively, before it went public and they came out ahead in the next stage of the race with Berger et al. On Saturday August 3rd, the day that the second set of data went public, they had a paper published in Nature with Advance Online Publication: "A ‘kilonova’ associated with the short-duration γ-ray burst GRB 130603B" and posted it online at the arXiv. This paper incorporated the second epoch of HST data and new ground-based data and reported clear dimming of the source, as shown in the press release image put out by Space Telescope Science Institute (STScI) that accompanied the Nature paper. The publicity emphasized the strong evidence that the kilonova detection provides for the origin of short GRBs.
 
Publicity images from the Hubble Space Telescope for the Nature paper by Nial Tanvir et al. Credit: NASA, ESA, N. Tanvir (University of Leicester), A. Fruchter (STScI), and A. Levan (University of Warwick)
The Nature paper was received on July 16th and accepted on July 26th. I asked Tanvir what argument they made to Nature for such rapid publication and he said: 
"Nature were aware that there was a likelihood that the result of the second epoch would be published/publicised first elsewhere if their publication was not sufficiently quick. Of course, the normal refereeing process had to be adhered to, which could have delayed things, but thankfully didn't."
Their Nature paper is clearly very different from the "provisional report" they submitted to the arXiv earlier, not surprisingly because of the extra HST data. Ironically, their paper addresses the production of elements more directly than the Berger et al. paper by saying "it is speculated that this mechanism may be the predominant source of stable r-process elements in the Universe" and "If this simplest interpretation of the data is correct, it provides […] (ii) confirmation that such mergers are likely sites of significant r-process production."

However, Berger et al. weren't finished. They put a revised version of their paper on the arXiv on the same day that the Tanvir et al. paper was published in Nature. Along with the extra HST data they added data from Magellan, a ground-based telescope. Now the title of their paper makes a clear statement: "An r-Process Kilonova Associated with the Short-Hard GRB 130603B". Berger told me that they obtained the HST data from the archive as soon as it was available (August 3rd) and immediately worked on the data and then submitted their revised paper to the arXiv about 3 hours later. This is a remarkably fast turnaround, though I note that the PDF of the paper is dated August 6th, so they may have revised this quick submission.

Both groups ran a strong race, as they worked very quickly on their papers and got what seems to be the right interpretation when they only had some of the data. In a sense this was Olympic-level research in the race for gold's origin. The authors knew something about the other team's work because they both submitted reports to the Gamma-ray Coordinates Network (GCN) describing early progress. Tanvir et al's report came first on June 13th followed by Berger et al's more detailed report only 7 hours later.

Aggressive publicity

There were several striking features about the Berger team's publicity, as noted earlier: the heavily framed story about gold, the short schedule for the analysis of the data, the unfinished peer review and the incomplete data for the initial submission to the arXiv. It's the combination of these features that makes this publicity aggressive. Contrast this with the Tanvir et al. team who took the classical approach by not pursuing publicity until the full set of data came in and the paper was accepted in a peer-reviewed journal.
 
An artist's impression of two neutron stars merging. Credit: Dana Berry, SkyWorks Digital, Inc.
I have some experience with heavily framed stories, and my opinion has evolved with time. We've dug stories out of papers before, but I'm less interested in adopting this approach now. It's gratifying to find a hidden feature in a story, but issues with transparency and accessibility can result, both for science writers and for others who look at the paper after being directed from the release or news articles. Here, several key details of the gold calculation are missing, for example. I explain why this might be important below.

Research on transient phenomena is clearly a high-pressure field, where analysis and paper writing can occur at a very fast rate. One example is an interesting object called SN 2009ip. This object was originally thought to be a supernova, but when another outburst was observed it was reclassified as a supernova impostor. A subsequent outburst from SN 2009ip was claimed to be a real supernova, as I wrote in a blog post last year. The team who made this claim submitted a paper on the same day that some of the data were obtained (they provided some details later about how this was done). They clearly wanted to beat their competition in submitting a paper on this event (though note that they did not pursue publicity until their paper was accepted several months later). Subsequently, at least 4 papers have appeared disputing the claim that a supernova occurred. As time passes it will become clear if the star was destroyed and a supernova really occurred. If it didn’t, I’ll correct my blog posts and consider writing a new one. Although I’m impressed with the skill and drive shown by astronomers in these fields, it's not clear that extremely rapid analysis and submission consistently delivers the most reliable work.

What about the importance of peer review? In my work at the Chandra X-ray Center we have been advised that NASA only does press releases on papers that have been accepted after peer review (although their Communication Policy makes no mention of this requirement). Peer review isn’t infallible, but it is a useful filter and I think NASA’s policy is sound. (Note that CfA is not part of NASA and weren’t bound by their policy on peer review, though I believe they usually follow it.) It’s not common for major, publicity-killing revisions to occur in the papers that we track as they undergo peer review, but it's not unheard of either. It is very common to see comments along these lines: “we’ll recommend this paper be accepted if certain changes are made”.

In the case of Berger et al. the press conference was held one month after the paper was submitted and about two weeks before the second epoch of HST data became available. Because of the latter, I’ll speculate that the most reasonable reaction from a referee – who very likely would have known about the second epoch – would be to say “I’ll review the paper after you add the extra data in”. The potential was to turn those guarded statements in the first version of the paper into much stronger ones, by waiting just a few weeks and doing some extra analysis. At that stage only Tanvir et al. knew what the follow-up HST data were showing.

To summarize, I think the most aggressive aspect of the publicity was the timing – the paper that led to the CfA press conference was based on incomplete data. It was provisional work. In the revised paper they were soon able to make a much stronger case for a kilonova having been detected.

How much gold was made?

Both teams concluded that a kilonova was observed, a cosmic first. However, after reviewing the numbers for the Berger paper and publicity, I noticed a couple of potential discrepancies. The calculated ejecta mass was ~0.01 solar masses in the original Berger et al. paper – as quoted in the press release – and it was 0.03–0.08 solar masses in their revised paper (using the same ejecta velocity as the original paper it was 0.03). A factor of 3 isn't enormous for astronomy, especially for groundbreaking work, but I am curious about whether re-analysis with the new data was the reason for the difference (I asked Berger and a co-author but they did not answer).

The ejecta mass is used to estimate the gold yield, where there's potentially a bigger problem with the numbers. The amount of gold wasn't quoted in either version of the Berger et al. papers, but in the CfA press release a value "as large as 10 moon masses" was given. However, a Washington Post article by Joel Achenbach, quotes an estimate of 20 Earth masses of gold given by astrophysicist Daniel Kasen, an expert on kilonovas. This value is 200 times larger than the CfA release's estimate. I suspected a typo was made in the Washington Post article, but Kasen promptly confirmed this number and provided some details of his calculation when I contacted him (I asked Berger and a co-author about the discrepancy, but they did not answer). Taken at face value, these numbers suggest gold might be much more common than the authors thought, assuming the same neutron star merger rate. This is where some details about the gold calculation might have been valuable.
Daniel Kasen from Berkeley Astronomy Department. Credit: Berkeley and Daniel Kasen
This uncertainty detracts from the confident claims made in the press conference, including "This material also contains several moons worth of gold, suggesting that neutron star collisions are the primary producers of gold in the universe" and "we can in principle account for the amount of gold that we see in the Milky Way and the universe, so there's no real need to invoke another mechanism" [where this other mechanism is the supernova idea]. This statement was qualified somewhat with: "This doesn't mean that supernovae themselves do not contribute some fraction for the gold in the universe but I would argue that this is the dominant formation mechanism."

This confidence carried over into much of the press coverage. For example a USA Today story began with: "The gold glinting on your wedding band was likely born in a cataclysmic merger of two exceedingly exotic stars, astronomers report Wednesday". "A strange glow in space has provided fresh evidence that all the gold on Earth was forged from ancient collisions of dead stars, researchers reported Wednesday,” said the Associated Press.

A more cautious viewpoint came in an excellent article in New Scientist by Lisa Grossman, who concluded: 
"Daniel Kasen of the University of California, Berkeley, thinks the idea is intriguing, but he points out that we don't know for sure how common neutron star mergers are. 
"For reasonable rates, if they're ejecting that much mass, it's plausible or likely that they're ejecting a big chunk of the heavy elements of the universe," he says. "But it would be helpful if they get more data points.""
That sounds like excellent advice, and I'm sure both the Berger and Tanvir teams – and others – will be searching hard for more cases. In the revised version of their paper Berger et al. admit that the rate of compact binary mergers is "poorly known", and I don’t know how quickly it will improve.

In the meantime, it will be interesting to follow reactions to the work of Berger and Tanvir. I suspect that the supernova idea won't be easily tossed aside, because there are some strong supporters of it. For example, both Berger et al. and Tanvir et al. cite a paper by Metzger et al. (2010) that discusses the r-process and kilonovas in detail. Metzger et al. (2010) mentions the astrophysical origin of the r-process remaining one of the "great mysteries in nuclear astrophysics” and quotes a paper by Qian and Wasserburg (2007) for a recent review. One might guess that such a paper would heavily support neutron star mergers as a way to make heavy elements like gold. However, in a 52-page paper Qian and Wasserburg dismiss the idea of neutron star mergers in a 15-line footnote on page four, and don't otherwise discuss it! I didn’t ask their opinion about this study, but it would be interesting to hear their reaction. (Side comment: Wasserburg lists his address as: "The Lunatic Asylum, Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA 91125, USA." I haven't heard of this department at Caltech.)

Who won?

Having described this academic competition as a race, it’s fair to ask who won? Overall, I don’t think there was a clear winner. The Berger et al. team certainly obtained more publicity than Tanvir et al. but the latter team had their paper published first. They also adopted a more conservative – and I think more robust – approach to the publicity.

To their credit, the Berger et al. team showed courage and confidence in publicizing their initial submission. They took a gamble by saying a kilonova was likely to have been observed and they were vindicated when the extra data came in. I know astronomers who won’t submit papers to the arXiv until they’ve been accepted, so posting a paper and then publicizing it before acceptance would definitely be considered too brave.

I am interested in hearing people’s reactions to this story, including the competition that took place and the aggressive publicity tactics that were employed. What do scientists think? What about science communicators and public information officers?


(*) This post was originally meant to be a follow-up to my earlier post It's been done before, or it's wrong. Part I. It was going to be Part II, corresponding to the "it's wrong" reference, but then the story developed.
(**) A hat-tip to Joe Hanson for posting about the Neil Tyson clip.

3 comments:

  1. it is informative post for human.This post publish new important news.This post article is important and helpful.Thanks a lot for share your great post. fifa world cup

    ReplyDelete
  2. Excellent article Peter. You should work this into an opinion piece and submit it to Science or Nature.

    ReplyDelete
    Replies
    1. Thanks very much for the positive comment. It was a fascinating story to follow and write about.

      Delete