(UPDATE*) The super-Earth in the news. Same one in a paper last year. Who found it first?
Here’s a little episode - in this world of instant e-publishing, open-access-everything, and large international teams of scientists rolling their data out lickety split – that seems worth a few followup calls from any astronomy reporter interested in the sociology, collaboration, and competition of science.
In case you missed it, we received a very interesting comment on the post that ran a few days ago about GJ 667Cc, the apparent super-Earth in the habitable zone of a red dwarf star 22 light years away. It is from Markus Pössel, an astronomer and also a press officer for the Max Planck Institute for Astronomy in Heidelberg. He wonders whether we have a fine example of the ambiguity of discovery. Our post on the recent news is just a few below this one, but here is his comment again in full:
——————–
I do see your point about “mass” vs. “size. But apart from that, there’s also an elephant in the room. It’s this E-print here by Bonfils et al.: http://arxiv.org/abs/1111.5019v2 – published on astro-ph in November 2011.
It’s based on the HARPS data also mostly used by the UCSC/Carnegie group who are now reported in the popular press as the discoverers GJ 667Cc (in fact, the article is by the people who built HARPS and actually took the data). The preprint identifies GJ 667Cc, gives its orbital period, gives a mass estimate not dissimilar to the one by Anglada-Escudé et al., and hypothesizes that it could be a habitable planet, since it gets about 90% the amount of radiation the Earth gets from the Sun.
So who’s the discoverer of the new planet? The preprint by Bonfils et al. was out in electronic form earlier, but the article is not accepted in any journal. It also references, for GJ 667Cc, an article that is still “in preparation”. On the other hand, it does give the basic parameters, and shows that Bonfils et al. evidently found GJ 667Cc in their HARP data. Anglada-Escudé et al. cite this preprint, although if you only browse their article cursorily, you might miss the fact that “similar to one of the candidates reported here” refers to GJ 667Cc that everyone is now making such a fuss about.
There are some subtleties here involving priority in the age of electronic publishing; the way the story is now widely reported seems blatantly unfair, though. And so far, I haven’t come across anyone reporting this. / Markus Pössel
————————————
I pasted up there top right one of the pertinent grafs in the November paper on the arXiv watering hole server for physics and astronomy reports that haven’t yet and some of which never will hit a journal. It is 70+ pages long and covers a huge range of data from many stars. Reporters may be forgiven for not poring through it for any habitable-zone super-Earths. One thinks, again, somebody in the reporting business ought to call around now, if they have not already, and find out if there is an issue here.
Notably, the authors of these two papers include some of the towering figures in extrasolar planet history. The paper to which Dr. Pössel refers includes among its authors Michel Mayor and Didier Queloz, the first – in 1995 – to announce discovery of a planet, a “hot Jupiter,” using the Doppler shift method of planet detection. Last week’s paper’s authors include Paul Butler of the Carnegie Institute, one half of the team (Geoff Marcy the other ) that was neck and neck with Mayor and Queloz in developing the method and went on within months to become its most prolific practitioners. Another of the authors in this week’s paper, Steve Vogt of UC Santa Cruz and Lick Observatory, worked closely with Marcy and Butler in those days on spectroscopy advances. So, there are bigshots in the mix.
I forwarded the comment to Butler and Vogt.
This shows why it can pay off big to call outsiders about a prominent team’s research. A savvy reporter would have known that the Lick-Carnegie team that was in the news this week is the descendant of the earlier Marcy-Butler planet-finding factory. It wouldn’t have taken inspiration to have contacted Mayor or Queloz or others in their group in Europe to ask for its take on the new report’s significance. It would have been a delightful moment for such hypothetical, diligent reporter to get in reply something like “Oh, that super-Earth. We already published on it.”
Late Addition: A little bird tells us that one press release has been updated to say the new paper confirms the previous one: Planetary Habitability Laboratory Press Release ;
*UPDATE: In the European magazine Ciel & Espace, David Ditch has a conversation with the lead author of the November paper, Xavier Bonfils of the Astrophysics Laboratory in Grenoble. The link goes to the Google translation of the article and it’s not bad for robot work. Bonfils exclaims, in this rendering, “We know the existence of Gliese 667Cc for several months!” The story’s lede holds that the American team merely rediscovered it. Thanks to commenter Daniel Fischer for leading us to the link.
- Charlie Petit
February 9th, 2012 at 8:00 pm
If we must point out anomalies in this “case”, how about these interesting factoids in the paper trail (pun intended):
- The abstract of the Bonfils et al. paper lists Gliese 667Cc as one of “9 planets already announced by our group”, yet the only reference for this planet given is “Delfosse (2011, in prep)”. No prior announcement of planet c existed in the literature, and the ‘quoted’ paper is apparently still in preparation, even today. “Already announced” …?
- A new paper from the same group hit arxiv just 2 days ago, http://arxiv.org/abs/1202.1564, and on page 9 it speaks of only *one* planet found with HARPS at Gliese 667C, again to be discussed in a paper not yet in existence, now “Delfosse et al., 2012, in preparation”. It doesn’t even say whether this one planet is b or c – and what happened to the other one.
- Anglada-Escudé et al. actually reference a previous claim of planet b by the HARPS group at a conference in 2009 and also quote the 2011 HARPS paper with its sketchy report of a *candidate* planet c. They use only publicly available HARPS data and add RV measurements from two other telescopes. They thusly get a clearer signal for c than HARPS alone, actually finish a paper and find numerous institutions to issue press releases about it.
Looks generally like a fair procedure to me, although Bonfils is now crying foul in http://www.cieletespace.fr/node/8549 … Ah, and that PHL press release always had the reference to the 2011 paper, the “update” mentioned in its header just refers to the addition of further illustrations.
February 10th, 2012 at 1:34 pm
@Daniel Fischer: Sure. As I mentioned in my comment, there are arguments on both sides, and one of the points to consider is the unsusual “in preparation” paper.
I agree that the new paper mentioning only one planet is also weird.
We don’t need to apply exegetic tools to the word “candidate”, though. The Bonfils 2011 paper does give the False Alarm Probability, after all (a measure of the probability the planetary signal in the periodogram is, in fact, just noise). It’s 0.012, falling just short of the authors’ own criterion of <1%. Which still leaves the question whether it makes more sense to talk about "discovery" and later "confirmation" in such a case rather than – well, what would you even call the first action, if only the second one is "the discovery"?
Btw, calling the Bonfils report "sketchy" seems rather misleading to me. It describes the observational techniques, the statistical methods used, shows the periodogram for the detection of Cb and the residual periodogram for Cc, gives the necessary chi-square and FAP values and discusses additional features such as variability. That's a sizeable amount of information.
And, finally getting back to the original issue, why should those ambiguities and subtleties mean that it's OK to ignore the Bonfils 2011 paper altogether in the press release texts and in subsequent coverage?
February 11th, 2012 at 3:03 pm
Digging deeper I sent a mail to Mr Delfosse Friday night, asking for their view of the situation – and today I already have detailled responses by both him and Mr Bonfils in my mailbox, plus the Delfosse et al. paper that will appear on Arxiv next Monday. In a nutshell:
- The HARPS teams now firmly insists on its priority of the discovery and consider it established beyond doubt with the Bonfils et al. paper appearing on Arxiv last year. In their view the one paragraph there gave all the important facts.
- They also say that they found their competitors’ paper “stunning and shocking” and that its analysis was actually inferior to their own (this goes into various technicalities I still have to check).
- The Delfosse et al. paper not only discusses planets Cb and Cc but even the possibility of third planet, Cd. That the Feb. 7 paper had only one planet was a mistake; the author was not involved in the actual HARPS planet hunt and had overlooked the announcement(s) for Cc.
Most importantly with respect to science journalism, the appearance of Delfosse et al. on Arxiv will also be accompanied by a press release that will center of the occurence rate of habitable planets – but also claim the discovery of Cc for the HARPS team. Things are getting interesting now …
February 11th, 2012 at 6:33 pm
@Daniel Fischer: Very interesting indeed. I’m looking forward to reading the new e-print on Monday.
It’s weird that the HARPS team would insist on having established Cc “without doubt”, though, since, as mentioned above, in the preprint, with false alarm probability 0.012 > 1%, it falls short of their own criteria.
Is “one paragraph” their words or your own, though? In the Bonfils paper, key information about Cc is contained in the various tables and figures, after all.
February 12th, 2012 at 11:40 pm
This is a very interesting discussion. It would be useful to hear something
from the Anglada-Escude team, as there may be some important details that
we’re missing.
Based on the public evidence, I’m also having trouble understanding why
Bonfils et al. would say they’ve established the presence of Cc “without
doubt”. Maybe they are referring to information present in the upcoming
Delfosse et al. paper, but from an outsider’s point of view this paper does
not exist.
From experience I can say it’s deceptively challenging to write a good
press release. You want to be interesting, concise and accurate and also
to publicize the work done by your institution or observatory. Unlike with
a science paper, it generally isn’t practical or appropriate to refer to
other work. One can argue that this case represents a gray area, but the
Bonfils et al. paper presents only marginal statistical evidence for Cc,
and the paper has not yet been accepted to a journal.
The way to *guarantee* that your work is mentioned in a press release is to
issue one yourself. It appears Bonfils et al. weren’t yet ready to do this
and the Anglada-Escude team beat them to it.
On the other hand, it’s easy to understand some frustration from the
Bonfils et al. team, since they’ve obviously done a great deal of work in
this area, and may feel that this is really their result. In particular,
the use of the HARPS data by the Anglada-Escude team might have
stung. However, the data was public and Anglada-Escude et al. did use
additional data along with a new analysis technique.
This case has some relevance to the ongoing discussion about Open Science.
The Bonfils et al. paper first appeared on the arXiv on Nov 21, with a note
that it had been submitted to A&A. I’m speculating here, but the paper’s
hint about a planet being detected in the habitable zone might have
provided motivation to the competing team. It also might have given their
competitors some time to work with while this long paper was being
refereed. Did this give a tactical advantage to Anglada-Escude et al., who
kept their paper off the arXiv until their press release was issued?
Finally, I wondered why Bonfils et al. appear to use two slightly different
names for the planet: “GJ 667Cc” and “Gl 667Cc”? They generally use the
latter while Anglada-Escude et al. use only the former.
February 13th, 2012 at 9:48 am
Posted for Guillem Anglada, Steve Vogt, and Paul Butler, who replied to my email about this affair:
Hi Charlie
Yes, the GJ 667Cc planet was already mentioned in the Bonfils preprint back in November and we duly and properly cited it in our paper.
So, if you ask who first publicly claimed there was such a candidate; it is true they said that publicly before. By the way, this also applies to the ‘less’ important GJ 667Cb. It was announced at a 2009 conference whose proceedings we could not find anywhere.
Of course, we were a bit taken aback when we saw their preprint back in November because we already had a detection and a half-written paper. We were very excited about it and were preparing a manuscript to Nature. That preprint changed our plans but we still thought it could make a nice ApJ Letter and maybe a press release if our announcement paper came before theirs. Even now, their November’s preprint has not yet turned into a peer-reviewed publication in a scientific journal. Also, the ‘discovery paper’ has not appeared, not even as a preprint.
In this collaboration, the Keck team has been tracking GJ 667C since 2000, roughly 4 years before the Swiss started obtaining data on this system. We were studying the 7-day candidate they described at a 2009 meeting (but never actually published) to better determine its parameters and to look for other planets in this system. We strongly suspected about other planets in this system since April, 2011, and about this 28-d candidate in particular since early September. So, in the last year, we have spent significant effort to collect more data, both at Keck and at Magellan, reanalyze those HARPS spectra and then to prepare a solid and detailed analysis that could be accepted in a refereed Journal.
Bonfils remarked recently that “they’ve known about this signal for several months”. It is not clear when the HARPS team first noticed this 28-d signal, but it is very telling that the uncertainty they cite for the 28-d period of GJ 667Cc is 0.5 days. Such an uncertainty is just too large, amounting over the four years to a full orbital period, and bespeaks of a result that was preliminary at best. It is not something you’d expect from a carefully prepared peer-reviewed discovery claim. Indeed, their own short discussion in the massive Bonfils 2011 preprint describes their great uncertainty in modeling this system, dealing with a highly eccentric 180-d orbit, and also a strong 90-d signal. They did not know if either or both were real planets, or are merely sampling alias effects. So they simply “filtered” these signals out and then extracted the 28d. They further admitted that the “false-alarm-probability” of their 28-d signal was only about 1%, not up to their usual standards for publishing.
It must also be said that we are not simply “USING” HARPS archived velocities here. The HARPS data are archived in a way such that the velocities can have unpredictable offsets, making them not reliable by any other than HARPS team individuals with access to their latest reduction pipeline codes. Instead, we started with their archived extracted spectra and derived our own radial velocities measurements from scratch. For this we developed our own analysis software which is radically different from their standard algorithm. A paper describing how it works is already been reviewed (Anglada-Escude & Butler, submitted to ApJ Supplement Series) and we should make it public in a matter of days or a few weeks. It turned out, happily, that our velocities derived from their HARPS raw spectra actually have higher precision than the velocities the Swiss team derives from those same spectra. That higher precision also helped to add welcome significance to our 28-d signal detection. The final killer to confirm the candidate and overcome the confusion in modeling the system was the incorporation of the additional PFS and Keck data to the new HARPS measurements. Only at this point is when we had great proof for a a great discovery and we thought it deserved a sound publication.
It then took two months to go through the proper referee process. Even though the Bonfils preprint was not peer-reviewed at the date, we followed proper protocol and duly cited it. That was considered sufficient by both the editor and the referee at ApJ Letters. In other disciplines, such as medical research or other applied sciences, making unfounded claims can have dangerous repercussions. Of course, neither the life of anybody nor any patent is in jeopardy here. Still, making unreviewed claims is neither right nor fair to other colleagues. I think that all the parts understand this and this is the reason why the issue about discovery rights has not been taken to public discussion.
As a final note, somebody made us realize that only one planet around GJ 667C is mentioned in the recent preprint from the HARPS group http://arxiv.org/abs/1202.1564 (made public on Feb 9, 2012; this is 4 days ago and after our press release on Feb 3). We find it very intriguing that two planets were ‘discussed’ before, and now one of them is omitted when the same ‘discovery paper’ is cited. I think this further indicates that they were not convinced to say the least.
The bottom line is this. If I were a journalist I would be very cautious in guessing what any group is ready to claim in front of the community. Of course, asking opinions cannot hurt but, unless there is an official complain from one of the participants, I think it is wiser leaving it as it is.
Guillem, Steve and Paul