CDMS is an experiment located underground in the Soudan mine in Minnesota. It consists of two dozens of germanium and silicon ice-hockey pucks cooled down to 40 mK. When a particle hits the detector it produces both phonons and ionization, and certain tell-tale features of these two signals allow the experimenters to sort out electron events (expected to be produced by mundane background processes) from nuclear recoils (expected to be produced by scattering of dark matter particles, as the apparatus is well shielded from ordinary nucleons). The last
analysis, published early 2008, was based on a data set collected in the years 2006--2007. After applying blind cuts they saw zero events that look like nuclear recoils, which allowed them to set the best limits so far on the scattering cross section of a garden variety WIMP (for WIMPs lighter than 60 GeV the bounds from another experiment called XENON10 are slightly better).By now CDMS must have acquired four times more data. The new data set was supposed to be unblinded some time last autumn, and the new improved limits should have been published by now. They were not.
And then...
- Fact #1: CDMS submitted a paper to Nature, and they were recently accepted. The paper is embargoed until December 18 (embargo is one of these relics of the last century that somehow persists until today, along with North Korea and Michael Jackson fans) - the collaboration is not allowed to speak publicly about its content. Consequently, CDMS has canceled all seminars scheduled before December 18.
- Fact #2: A film crew that was supposed to make a reportage from unblinding the CDMS data was called off shortly before the scheduled date. They were told to come back in January, when the unblinding will be restaged.
- Theory #1: The common lore is that particle physics papers appearing in Nature (the magazine, not the bitch) are those claiming a discovery. It is not at all impossible that the new data set contains enough events for an evidence or even a discovery. If the zero events in the previous CDMS paper was a downward fluke, several WIMP events could readily occur in the new data. Furthermore, in some fancy theories like inelastic dark matter, a large number of WIMP scattering events is conceivable because the new data were collected in summer when the wind is favorable.
- Theory #2: Data-starved particle theorists once again are freaking out for no reason. There is no discovery; CDMS will just publish their new, improved bounds on the scattering cross section of dark matter. CDMS is acting strangely because they want to draw attention: the experimental community is turning toward noble liquid technologies and funding of solid-state detectors like the one in CDMS is endangered.
Important update: I just received this in an email:
It is still true that the new CDMS data are scheduled to be released on December 18th, and there will be presentations in a number of labs around the world. But if there's no Nature paper then theory #2 becomes far more likely.I was alerted to your blog of yesterday (you certainly don't make contacting you easy). Your "fact" #1, that Nature is about to publish a CDMS paper on dark matter, is completely false. This would be instantly obvious to the most casual observer because the purported date of publication is a Friday, and Nature is published on Thursdays. Your "fact" therefore contains as much truth as the average Fox News story, and I would be grateful if you would correct it immediately. Your comments about the embargo are therefore, within this context, ridiculous. Peer review is a process, the culmination of which is publication. We regard confidentiality of results during the process as a matter of professional ethics, though of course authors are free to post to arxiv at any point during the process (we will not interfere with professional communication of results to peers). Dr Leslie Sage Senior editor, physical sciences Nature
Surely it is highly unlikely that they can claim 5 sigma significance of anything unless the fluctuation was strongly downwards to start with..?
ReplyDeleteHowever I don't completely understand how they can do their statistics. They define a 'signal region', does that mean that everything inside the region is 100% certain to be caused by a dark matter particle? I would think not. They have some inevitable background such as 'surface events' which was estimated in 2008 to give 0.6 +- 0.5. With 4 times as much data this should produce 2 or 3 background events.
It is a nice exercise to work out how many events they would have to see to be '3 sigma' or '5 sigma' confident that it is not a background fluctuation away from 2.4 expected events. Well, I think they could get 3 sigma.
Either that or they re-estimate the background downwards! That would be interesting...
CDMS always has ~ 1 bkg event expected. They can revise their cuts (in a blind fashion) before opening the box in order to keep the expected bkg at 1 event, with some loss of exposure. I don't have any information on this particular run, but if they continue to do that, 5 events would only happen 0.4% of the time (I think, just quick estimate).
ReplyDeleteFrom my understanding they tighten their selection criteria as they increase their exposure so that they alway predict 0.5 events in each release.
ReplyDeleteI've heard there will be some number of events, but "consistent" with their previous results. Best guess: 3 or 4 events with an expectation of 0.5.
And the saddest thing of all is... because Nature's behind a paywall, and because I'm not at an institution that pays, I'm not gonna see the thing even when the embargo lifts.
ReplyDeleteHere's hoping the post it on arxiv.org on the 18th... if Nature even lets them.
Actually, I heard rumors that there will be an arxiv paper on Dec 18. If not, there are at least two CDMS seminars scheduled on that day, and the slides should be available soon thereafter.
ReplyDeleteCDMS saw 3 recoil-like events before muon veto cuts. Even if all survived, this is consistent with background expectations (but would generate equal or weaker limits than their previous release). I.e., no grounds for a claim. If no events survived the cut, the cumulative >400 kg-day would generate an improved limit perhaps worthy of submission to Nature. Specially true from an experiment with a technical review over the horizon and massive budget requirements for continuity. Theory folks should take a chill pill.
ReplyDeleteI recently took the tour at FermiLab, and the tour guide said, "And as it turns out, Neutrinos do have mass. It should be published before the end of the year."
ReplyDeleteDark matter is nothing but the gray matter from all the Boltzmann brains.
ReplyDeleteWhy is it in Sudan?
ReplyDeleteThe Soudan Underground Laboratory is the leading deep underground science and engineering laboratory in the United States today.
ReplyDeleteSoudan.umn.edu
Of course, it's always possible that with more data, they found some systematic that required more analysis time, pushing back the whole schedule.
ReplyDeleteOf course, it doesn't explain all the hush-hush or the Nature submittal.
Weird date -- Dec 18th.
ReplyDeleteIt's a Friday. Nature comes out on Thursday (the 17th). Embargoes are lifted at 12:01a Thursday, which means Nature-related science news often hits the US on Wednesday evening (the 16th).
So it's not clear how the 18th would be tied to a Nature-related publication.
@Bob
ReplyDeleteGood point, unless it was an advanced online publication
So people were talking about an arXiv paper on December 18, not that the Nature publication appears on that day. This would not be a complete nonsense - PAMELA, for example, posted their arXiv paper some time before their paper appeared in Nature. But now Nature says that they are NOT about to publish any CDMS paper. So the whole story is moot because the "facts" weren't facts at all.
ReplyDeleteAccording to the editor's statement, couldn't CDMS still post a paper to the arXiv and then submit to Nature?
ReplyDeleteCDMS has found indirect proof for WIMPs. Is there any other interpretation of that announcement ?!?
ReplyDeleteYes. Theory #2.
ReplyDeleteSo why are they waiting until Friday, 18? Why the mistery in the CERN seminar announcement? And the "filtrations" of rumors? Just for the show?
ReplyDeleteIMHO, this would only be somewhat justified if they have some interesting signal... Otherwise they are going nuts!
Everything is starting to make sense. Their bckg expectation has been revised to 0.5 events in 300 kg-days (this is the exposure that survives cuts, down from 600 kg-day raw). It used to be 0.6+/-0.5 in 120 kg-days (last paper). They saw a maximum of three events, what was there before muon veto cuts. If all three survived, it is almost a 3 sigma effect. Now, the credibility of this is essentially nil (esp. due to the downward revision of the expected bckg), but they need to make a big noise about it, just in case.
ReplyDeleteOne Leslie Sage said: "Your "fact" therefore contains as much truth as the average Fox News story,"
ReplyDeleteBeneath the dignity of the office of "senior editor" -- dragging one's personal politics into a rumour debunking.
What evidence is there that CDMS is "going nuts"? Given that the Nature rumor probably came from a failed game of telephone, I'm not sure why people think that the collaboration is doing anything strange...
ReplyDeleteWhatever has been found it can't be a DM particle, because it doesn't exist in nature - I'm sure.
ReplyDeleteWhat a nasty email from Nature! It's not as though you were slandering anyone, after all.
ReplyDeleteI think Leslie owes you an apology, or at least a box of cookies.
Hey, comparing me to Fox News might have been a bit harsh, but given that the story of the Nature paper is false I fully deserved it.
ReplyDeleteThe Nature comment was a wholly unnecessary way to correct a factual error which was posted in good faith.
ReplyDeleteThe content of this post was neither slanderous nor defamatory to any person or organisation. The effect of the post has in fact been to increase interest in an important experiment which is about to release a new set of data. This is positive.
If the above speculation is correct (3 events on a background of 0.5), a quick calculation gives a probability for a background fluctuation to 3 or more events of 1.4% - well short of 3 sigma significance. Including an uncertainty in the background will further decrease the statitistical significance. Such a result is well short of a "discovery", but more than sufficient to generate dozens of theory papers by Christmas.
ReplyDeleteJester, if your original story has been as accurate as an average FoxNews story, then it seems that it was almost exactly accurate.
ReplyDeleteOn the other hand, given the fact that Mr Sage needs to compare the scientific information with a TV network that is inconvenient for him politically, I have to urge everyone to take Mr Sage's words with a grain of salt - or, which may be even more reasonable, to ignore them as noise.
So it remains unknown whether there will be a paper in Nature because the only source that gave an unambiguous answer is untrustworthy.
Nature's reaction is to strong. Why would they care about what a blogger write? God knows there is a lot going around in the blogosphere. A reaction as strong as the one showed in the email hints that something is actually being hidden.
ReplyDeleteBest, Per
I even have to read conservative-bashing in letters from the Editor of Nature...about particle physics???
ReplyDeleteI think the reaction of the Nature editor is caused by the comments about the "embargo", which give a bad image about the journal and seem to be not true.
ReplyDeleteAbout "CDMS going nuts" (if finally they don't have nothing really interesting), the major evidence now would be format of the seminar announcement ([...] I will report on what we saw when we “opened the box”, whether we have seen WIMPs or not[...]). Why not reveal it before the seminar? This is too strange...
Because they haven't seen anything :-)
ReplyDeleteIf they have, what form do we get the info in? A cross-section?
Dear Anonymous,
ReplyDeleteMr Sage may be speculatively right that there won't be any CDMS paper in Nature about dark matter in December, but that can't change anything about the "image" of Nature when it comes to embargoes.
Jester has just referred to the official embargo policy of Nature which is very real and very well-known to Jester. He has the indisputable right to consider it an artifact of the previous centuries. And Nature surely follows it with other papers, even if this particular paper doesn't exist.
Best wishes
Lubos
Science seems to publish on Fridays.
ReplyDeleteThat seems like a pretty unprofessionally written e-mail to send. I'm just a grad student, but I've been alarmed by seeing these things more and more as I get more involved in the physics community and I find it more than a little disturbing. We're supposed to be professionals, not whiny know it all 14 year olds.
ReplyDeleteOh come on guys, stop it, who cares? The only relevant thing is the number and the recoil energies of the CDMS events.
ReplyDeleteNature (the journal), as always attempting to endear itself with us in the physical sciences. This is a journal tending mostly to those who have yet to find a cure to the common cold, and not worthy of this much attention here.
ReplyDeleteWith Dec. 18 seminars posted or rumored at CERN, Fermilab, LBL, SLAC, and UCSB (+ others?), it certainly sounds like CDMS thinks they have something interesting to say. If not, then WTF???
ReplyDeleteThe CERN abstract claims a background of 0.5 events - my bet would be they either have one "golden" event with very low BG probability (if that even makes sense for CDMS) or a handful of candidates.
If it looks like SUSY DM, you can be sure that the other dark matter experiments, as well as the Tevatron experiments, are going to be looking hard to see if they somehow missed it in their data. That loud cry of frustration will be from CERN...
It would be great to see the SM finally spring a leak!!
Nature does not wish to hinder communication between scientists. For that reason, different embargo guidelines apply to work that has been discussed at a conference or displayed on a preprint server and picked up by the media as a result. (Neither conference presentations nor posting on recognized preprint servers constitute prior publication.)
ReplyDeleteOur guidelines for authors and potential authors in such circumstances are clear-cut in principle: communicate with other researchers as much as you wish, but do not encourage premature publication by discussion with the press (beyond a formal presentation, if at a conference).
Where's the problem?
It seems some other guys in Europe are doing same search as CDMS, and released data recently, check arXiv:0912.0805
ReplyDeleteMy e-mail could indeed have been worded more diplomatically, and I did not expect it to be posted -- I apologize for seeming harsh, but given that we had not even been approached by the group and it was presented as "fact" that a paper had been accepted, I was in a rather bad mood to have to debunk something not relevant to us.
ReplyDeleteAnyone who has concerns about embargoes should contact me directly. Our policies in the physical sciences are in fact quite flexible (especially given that we have always allowed posting to arxiv at any time).
Leslie Sage
Nature
l.sage@us.nature.com
It could be that there is a Nature paper, it is just not imminent (I am parsing the Nature editor comments above). In this scenario, the only thing that is imminent is the arXiv version appearing on the 18th.
ReplyDeleteI should also mention a perfectly plausible political reason for the multiple talks on the same day. The result is important whether there is a discovery claim or not, and the CDMS collaborators might not want to give the honor of giving the first talk to only one person.
ReplyDeleteTo my mind (I am a competing experimentalist in the field), the CDMS behavior is consistent with either a positive (but low statistical significance) signal or just a factor of a few better limit. In either case, it is to CDMS's advantage to garner as much attention as possible prior to announcing the result, and to announce the result in a carefully planned, coherent manner.
But if I were to bet, I would bet 5:1 that there are 2-3 events in the WIMP signal box. It's bound to happen occasionally when you are aiming for 0.5 background events, just from fluctuations from background (which are not necessarily only statistical). Remember that this is a blind search, and CDMS will feel obligated to report events in their WIMP acceptance box, even if they find after opening the box that their cuts weren't optimal for reaching 0.5 background events.
I still fail to see why the collaboration itself has behaved strangely. If you were finished with a result before Christmas, then (a) there wouldn't be a convenient conference at which to give the talk, and (b) lots of people would be gone the following week, so you might pack any seminars in on the same day.
ReplyDeleteIt is not uncommon for these experiments to first present the results at a conference. The last CDMS result (in 2008) was first presented at a dark matter conference with the arxiv paper appearing the same day (or maybe the day before). There is a dark matter conference going on in Santa Barbara, so waiting to post their results until the day of the talk would not be unusual.
ReplyDelete"In either case, it is to CDMS's advantage to garner as much attention as possible prior to announcing the result, and to announce the result in a carefully planned, coherent manner."
ReplyDeleteHow is it advantageous for them to look like fools? There's no such thing as free publicity...
whats the use of finding a particle that doesn't interfere with normal matter ??
ReplyDeleteit could be a whole set of particles / atoms but none we could touch or see
tom o'bulls: Why would they look like fools? Just because people are getting over-excited about unsubstantiated rumors shouldn't reflect on the CDMS collaboration itself. The only thing they should be judged on is the quality of the work they present, signal or no signal.
ReplyDeleteSomebody reads your blog...or doesn't.
ReplyDeletehttp://www.dailymail.co.uk/sciencetech/article-1234473/Dark-matter-discovered-Scientists-key-evidence-mine.html
"Why would they look like fools? Just because people are getting over-excited about unsubstantiated rumors shouldn't reflect on the CDMS collaboration itself. "
ReplyDeleteThey schedule talks all over the place, announce in advance that they are putting a paper on the arxiv, behave generally in a manner guaranteed to trigger off a tsunami of rumours.... they had better have something big to announce.
The following statement now appears on the CDMS web site:
ReplyDeleteCDMS Statement - December 10, 2009
The CDMS collaboration has completed the analysis of the final CDMS-II runs, which more than doubled the total data from all previous runs combined. The collaboration is working hard to complete the first scientific publication about these new results and plans to submit the manuscript to arXiv.org (http://arXiv.org) before the two primary CDMS talks scheduled for Thursday, December 17, 2009 at Fermilab and at SLAC. Jodi Cooley, the CDMS analysis coordinator and a professor from Southern Methodist University, will present the talk at SLAC at 2 p.m. PST, and Lauren Hsu, a scientist from Fermilab, will present the talk at Fermilab at 4 p.m. CST. A Web cast of Cooleys talk will be available on the CDMS Web site (http://cdms.berkeley.edu/)
Jest, blogging is as good as journalism nowadays, so remember the old mantra about source checking: "If your mother says she loves you, check it out." this said, it seems to me that Nature is very much in an aggressive defense position...
ReplyDeletegood post nonetheless, many of us not-so-dark people were unaware of this business.
Let's not call things "facts" unless they are consistent with the definition of the similar word, 'facts'. E.g., "fact #1" is false. No paper has been submitted to Nature. The data was not unblinded last autumn, it was unblinded only very recently.
ReplyDeleteA reliable source has hinted that CDMS has three signal events against a background of 0.5, as already reported above. These events are in the energy ranges between 10 and 20 keV.
ReplyDeleteWill noone deal with the opening phrase of this post?
ReplyDelete"The essence of blogging is of course spreading wild rumors."
So it was all a plot to find out whether there are more than half a dozen attention-paying readers of your blog ;)
I didn't know that Nature editors were so ill-tempered...
ReplyDeleteouch, shamed even by SciAm... ;-)
ReplyDeletehttp://www.scientificamerican.com/article.cfm?id=dark-matter-cdms
Yes, it appears that the best way to boost popularity is to sink credibility :-|
ReplyDeleteThe Nature comment was a wholly unnecessary way to correct a factual error which was posted in good faith.
ReplyDeleteI would think not. They have some inevitable background such as 'surface events' which was estimated in 2008 to give 0.6 +- 0.5. With 4 times as much data this should produce 2 or 3 background events.
ReplyDelete