Wednesday, November 21, 2012

Have We Seen This Cosmic Jerk Before?

It's easy to underestimate how difficult it is to write a press release or a popular-level description of a science result. The goals are to be accurate, interesting and concise, while avoiding anything that's misleading. From experience, that final goal can be especially challenging because it's difficult to anticipate how people might interpret a description. Other challenges include deciding whether to include previous work and how skeptical you should be about the result you are describing. Here, I will explain a new result in astrophysics to discuss some of these challenges. The result is interesting and represents an important advance for the field of cosmology, but I will argue here that the publicity effort contained some misleading information.

Early last week, results from the Baryon Oscillation Spectroscopic Survey (BOSS) were released by Busca et al. with the first measurement of the cosmic expansion rate just three billion years after the Big Bang. This is an extremely impressive accomplishment, pushing back to an era when the universe was only about 20% of its current age at a redshift of 2.3.

Figure 1: From the SDSS web-site: "An illustration of the concept of baryon acoustic oscillations, which are imprinted in the early universe and can still be seen today in galaxy surveys like BOSS
(Illustration courtesy of Chris Blake and Sam Moorfield)."


Two papers in the late 1990s showed that the expansion of the universe is accelerating, a result that was so important and surprising that it led to Nobel Prizes in physics for Adam Riess, Brian Schmidt and Saul Perlmutter in 2011. This acceleration started about five to seven billion years ago and is thought to be caused by a mysterious form of energy called dark energy. The new BOSS results show that when the universe was only three billion years old - almost 11 billion years ago - dark energy was relatively unimportant and the expansion of the universe was decelerating.

The two discovery papers by Riess et al. (1998) and Perlmutter et al. (1999) used Type Ia supernovas as standard candles to trace the expansion history of the universe. This is a different technique from baryon acoustic oscillations (BAO), which uses a standard ruler. For more details about BAO, you may read the two press releases put out by the Sloan Digital Sky Survey (SDSS) and Berkeley and a short primer on BAO, including a useful illustration (Figure 1). I also wrote a blog post that mentioned the technique.

A key element of Busca et al. is explained by a graph from the SDSS press release showing how the expansion rate of the universe varies with time (see Figure 2). The red point is the one from the new paper and the others are from previous papers using the same technique. The data points agree with a model where the expansion was rapid soon after the Big Bang but then slowed down until, remarkably, it started speeding up again when the repulsive effects of dark energy kicked in. This changeover from deceleration to acceleration can be called a "jerk", since this is the physics term for a change in acceleration, although the press releases don't use it.

Credit: Zosia Rostomian, LBNL; Nic Ross, BOSS Lyman-alpha team, LBNL
Figure 2: From the SDSS press release: "A new measurement from SDSS-III BOSS reaches twice as far into the past as previous direct measurements of the expansion rate (illustrated by the red dot on the left-hand side). The expansion rate of the Universe was slowing down 10 billion years ago, and started to speed up due to the influence of dark energy 6 billion years ago." 
** P.Edmonds comment: I think this figure is very well done, because it clearly shows deceleration followed by acceleration, and it includes the data and error bars, plus a model. It also highlights the data point provided by the new paper. However, I think it contains an error. Please add a comment if you see a problem.
The SDSS press release describes this result well, and then down in the 9th paragraph it says:

"What is fascinating about the new result is that, for the first time, we see how dark energy worked at a time before the Universe's current acceleration started."

The Berkeley press release says, in the 2nd paragraph:

""No technique for dark energy research has been able to probe this ancient era before, a time when matter was still dense enough for gravity to slow the expansion of the universe, and the influence of dark energy hadn’t yet been felt," says BOSS principal investigator David Schlegel, an astrophysicist in the Physics Division of the U.S. Department of Energy’s Lawrence Berkeley National Laboratory (Berkeley Lab)."

Consistent with these strong claims, a Reuters article by Chris Wickham includes the following quote about the evolution of the universe:

""Only now are we finally seeing its adolescence... just before it underwent a growth spurt," said Mat Pieri at the University of Portsmouth in Britain, one of the authors of the study."

Scientific American blog article by Kelly Oakes quotes another author:

""It’s the first measurement when the universe was slowing down," says Schlegel. "Which is pretty neat.""

What Does the Literature Say?

So, the reader can conclude that evidence for early deceleration, followed by acceleration - a cosmic jerk - has never been seen before, right? Not so fast. Using Type Ia supernovas to measure cosmic expansion - the same objects used to discover cosmic acceleration - Adam Riess and colleagues published a paper in a 2004 issue of The Astrophysical Journal (ApJ) that said:

"We have discovered 16 Type Ia supernovae (SNe Ia) with the Hubble Space Telescope (HST) and have used them to provide the first conclusive evidence for cosmic deceleration that preceded the current epoch of cosmic acceleration."

That's how they begin their abstract. Shortly after, they say:

"A purely kinematic interpretation of the SN Ia sample provides evidence at the > 99% confidence level for a transition from deceleration to acceleration or similarly, strong evidence for a cosmic jerk."

In a later paper, published in 2007, Adam Riess and colleagues report an extension of this work with more data and find even "stronger evidence" for a cosmic jerk. So, in case you're not keeping score, that's now REALLY strong evidence.

There is also a paper by David Rapetti and colleagues that used Chandra X-ray Observatory data combined with supernova studies, to conclude: "we obtain clear statistical evidence for a late time transition from a decelerating to an accelerating phase".  [Note, I work for Chandra doing publicity.]

Researching a New Result

What's going on here? At this stage I could just email the authors and ask, but for now I'm going to speculate because it gives some insight into the detailed and methodical way I research a story when doing publicity. A lot can be done independently, using straightforward, but powerful tools like the Astrophysics Data System (ADS) and Google.

Is it possible Busca et al. weren't aware of these earlier papers? On a professional level this is unlikely because, according to ADS, the papers received at least 2416, 918 and 85 citations respectively. There are 63 authors for Busca et al., so some of them will have heard of this work. What about coverage by the press?  Using Google I searched for "cosmic jerk dark energy" and came up with several articles, including one by Dennis Overbye at the New York Times that is titled "A 'Cosmic Jerk' That Reversed the Universe". The article is dated from October 11, 2003, and was likely based on data that went into Riess et al. (2004).

It's possible Busca et al. don't think these earlier results are reliable, but it's worth noting that the author lists of these papers are full of brilliant astronomers, in two cases led by Adam Riess who won the Nobel for work using the same techniques. This doesn't prove they're right, but it does show they should be taken seriously. Also, I haven't heard that their work has been discredited in any way. To back this up, I checked the first paper cited in the introduction of the paper by Busca et al., which is a review paper titled "Observational Probes of Cosmic Acceleration", led by one of the co-authors, David Weinberg, with Adam Riess as a co-author. This review paper mentions the two papers by Riess et al., along with a third paper by a different group, and concludes that these results "confirm the expectation that the universe was decelerating at high redshift".

I have no idea how the two press releases were prepared, and how carefully the paper by Busca et al. was reviewed by the PR specialists responsible for the releases. I'll note it took some digging to find the reference to the supernova work included above and, for preparing the press release it would have been better if Busca et al. had provided direct references to previous work in their paper. Of the previous papers giving evidence for a cosmic jerk, only one was directly referred to in the Busca et al. paper and that was Riess et al. (2007). This paper is cited only once, to explain that the expansion rates of Riess et al. (2007) were not being shown in a figure and to mention that the numbers in the 2007 paper are model dependent. No mention is made of the strong evidence for a cosmic jerk reported in the 2007 paper or any other paper.

This seems like an notable omission to me. I'll note that the Busca et al. paper isn't in its final form as the version on the arXiv has only been submitted to the journal Astronomy and Astrophysics, and has not been accepted yet. Presumably the authors have not responded to comments from a referee yet. I would be surprised if they are not encouraged to add explicit mention of previous evidence for a cosmic jerk, especially if the referee does cosmology work using supernovas. This shows one of the values of peer review. It can be extremely valuable to get comments from an outside expert, especially when evaluating the context and significance of a new paper.

There's no scandal here, as the authors may have been pre-occupied with correctly describing all of the details about their result and comparing it with previous work using BAO. For example, in the conclusion they say the following about their result:

"It represents both the first BAO detection deep in the matter dominated epoch and the first to use a tracer of mass that is not galactic".

I have no doubt that this description is correct, but I'm arguing that it's incomplete.

I also noticed that Busca et al. provide no references to the supernova work that discovered cosmic acceleration and motivated BAO studies in the first place. No reference is given to the seminal Riess et al. (1998) and Perlmutter et al. (1999) papers, which have each received over 6500 citations at a rate that is steadily increasing each year (see Figure 3). Supernova studies aren't mentioned at all in their introduction and are only briefly mentioned later in the paper. The authors may be proud to be presenting results that are mostly independent of the supernova studies, but I think it's appropriate to attempt some historical completeness and give appropriate credit to the two papers that motivated so much follow-up work.


Figure 3: Citations for Riess et al. (1998) (top) and Perlmutter et al. (1999) (bottom) papers taken from the ADS. This figure shows two effects: (1)  that work related to cosmic acceleration is increasing strongly with time and (2) authors like to refer to these two papers.
A similar issue of lack of completeness applies to the press releases. For example, this quote from David Schlegel is not incorrect:

"No technique for dark energy research has been able to probe this ancient era before, a time when matter was still dense enough for gravity to slow the expansion of the universe, and the influence of dark energy hadn’t yet been felt".

That's because the BOSS measurements apply to a time about 3 billion years after the Big Bang, but the supernova results extend no earlier than about 4.2 billion years after the Big Bang. So, it's true that no other dark energy technique has probed cosmic expansion as early as Busca et al. However, I think it's a misleading quote because it gives the impression that no other technique has probed the era when deceleration was occurring, and that's incorrect. Only subtle changes are needed to arrive at wording that is even more problematic, such as the text in the other release: "What is fascinating about the new result is that, for the first time, we see how dark energy worked at a time before the Universe's current acceleration started."

Again, I don't know how the press releases were prepared, but it's easy to imagine that the incomplete description in the paper may have led to the misleading information in the press releases. A game of "telephone" might have occurred, leading to problems with the publicity, especially if the release wasn't reviewed externally.

Can We Learn Anything Useful Here?

I don't want to give the impression that my colleagues and I have never made mistakes in describing results, but we try to avoid them by adopting the following tactics. We only do press releases for papers that have been accepted for publication, a tactic that clearly wasn't followed here. We also have an extensive review process, beginning with the authors and continuing with reviews involving several scientists not involved with the study and publicity specialists. This isn't a fast process, but it can and does uncover errors and potentially misleading statements. There's no guarantee that these tactics prevent problems or mistakes in any given situation, but we know from experience that they often do. It's unclear whether they would have helped here.

Even after doing all of this work and, in some cases, asking questions about the details in a paper and its significance until the authors are probably sick of seeing emails from me, we can still have problems if the findings themselves are flawed. All it can take is one error to make all of the results in a paper worthless. If these problems slip by without being caught by the referee or referees, and any other experts we might consult with, there's not a lot that we can do apart from posting a correction afterwards. That's why it's important to work hard at the aspects of communication that we can control, such as accurately describing a new result and carefully checking its significance and implications.

I hope my comments are interpreted by the communication teams and the scientists as constructive criticism. As an aside, I have rarely made comments on work done by other publicity teams. One memorable attempt involved a University press release about a Chandra result, where one of the scientists on the team had perfectly reasonable objections to the wording in the University-written press release. I agreed with the scientist and since the release hadn't yet gone out and it used Chandra data, I thought I could help out. However, the two PIOs were completely uninterested in examining, let alone revising, the wording and instead devoted multiple emails to explaining how many years of experience they had and speculating about the motivation of the scientist who was complaining. All I was looking for was some sort of relevant discussion and when it became clear that I wasn't going to get it, I gave up.

It's extremely useful for people who communicate science with the public to hear feedback about their work. We get a reasonable amount of feedback from the public through comments on our web-site and socal media, and through emails, but we don't get nearly as much feedback from scientists & science communicators, other than the ones we work with on a release. By giving feedback I'm openly encouraging feedback about our work. It's not quite the same as George W. Bush saying "bring 'em on", but you get the idea.

Another thing I've learned from years of working in publicity is that there are sometimes important and surprising issues affecting publicity that an outsider - like me in the case of this BOSS publicity - is unaware of. I plan to make a few inquiries to see if I'm missing anything important or if I've made mistakes in my interpretation of the research. If I find out anything interesting or important, and it's not potentially embarrassing to the people involved, I'll add a follow-up post.

Parting Messages

The scientists involved with the Busca et al. paper have a lot to be proud of, as they are the first team to confirm early deceleration and the existence of a cosmic jerk using a technique that is independent of Type Ia supernovas. Using BAO they have also surpassed the supernova technique in the ability to trace cosmic expansion as close as possible to the Big Bang, and in doing this they have mastered a new way to apply BAO. These results fully deserved to be publicized. To the best of my knowledge, most of the statements in the press releases are accurate and avoid being misleading, with the few exceptions listed above. The overall impression given by the media articles was sound.

Meanwhile, the astrophysics news-of-the-day has gone on past this story, to rogue planets and incredibly distant galaxy candidates and a sharp decline in the birth rate of stars and rumors of a big discovery on Mars. However, after finishing most of this post I noticed that Charlie Petit wrote about media coverage of this result at the Knight Science Journalism Tracker. Petit has done a terrific job as a tracker covering physics and astronomy stories for years and his semi-retirement has unfortunately caused a clear dip in the coverage of these stories.

Petit thought that the press releases and the media coverage did not do a good job explaining the details of the result, criticism that I don't really agree with considering the complicated nature of this story. I think the releases and articles describe the technical details of the result well, and Figure 2 is especially helpful. Anyone who's read this far will have noticed that I didn't even attempt to explain the technical details of this result.  Charlie also made some general comments about the story here:

"And somehow, the news reports tell us, from this the astronomers are starting to see how dark energy 11 billion years ago was already resisting gravity, keeping the universe expanding a little faster than one might expect. About all that had been known from earlier surveys, mostly of supernovae behavior as seen at various distances, is that around seven billion years ago the deceleration of the Big Bang's expansion's turned to acceleration - throwing glactic clusters farther apart at an ever increasing pace and into a far future of cold, infinite void."

I think the first sentence here is a misread of the result, and the second sentence may or may not involve a lack of recognition of the previous supernova work, depending on how you interpret it. My final message is: this stuff is hard!

4 comments:

  1. This is a great explanation of the context of this particular work and also of the tension between accuracy and hype in publicizing new results. Thanks for posting!

    ReplyDelete
  2. Also re: the error in the expansion rate figure: The y-axis is "expansion rate" and the line and data points are appropriate for that. But it's incorrect to put "speeding up" and "slowing down" on the y-axis as shown -- what they should have put there is "fast" and "slow" or similar. The *rate of change* of the expansion rate is the key factor here and that has to do with the slope of the expansion rate curve. The red point is roughly at the same level on the y-axis as the other points, so it's not that it's showing slowing down by its vertical position. It's just that the curve is going down there, whereas it's going up later on.

    I didn't notice this when I first glanced at the figure -- well spotted!

    ReplyDelete
    Replies
    1. Well done Katie. You spotted the error and explained it well. Thanks for reading the post too. I'm planning to do a shorter & quicker post next time!

      Delete
  3. QUITE interesting. Just discovered your site. Very well-done. Even as a financier, I can understand it. I am myself writing a book on finance and the many ways they use to "cheat(?)" on us at our own expenses. To be understood, I do it alike a novel.

    And yours appears to me to be a clear Universe novel way. Thanks for posting.

    ReplyDelete