jump to navigation

DZERO refutes CDF’s multimuon signal… Or does it ? March 17, 2009

Posted by dorigo in news, physics, science.
Tags: , , , , ,

Hot off the press: Mark Williams, a DZERO member speaking at Moriond QCD 2009 -a yearly international conference in particle physics, where HEP experimentalists regularly present their hottest results- has shown today the preliminary results of their analysis of dimuon events, based on 900 inverse picobarns of proton-antiproton collision data. And the conclusion is…

DZERO searched for an excess of muons with large impact parameter by applying a data selection very similar, and when possible totally equivalent, to the one used by CDF in its recent study. Of course, the two detectors have entirely different hardware, software algorithms, and triggers, so there are certain limits to how closely one analysis can be replicated by the other experiment. However, the main machinery is quite similar: they count how many events have two muons produced within the first layer of silicon detector, and extrapolate to determine how many they expect to see which fail to yield a hit in that first layer, comparing to the actual number. They find no excess of large impact parameter muons.

Impact parameter, for those of you who have not followed this closely in the last few months, is the smallest distance between a track and the proton-antiproton collision vertex, in the plane transverse to the beam direction. A large impact parameter indicates that a particle has been produced in the decay of a parent body which was able to travel away from the interaction point before disintegrating. More information about the whole issue can be found in this series of posts, or by just clicking the “anomalous muons” tab in the column on the right of this text.

There are many things to say, but I will not say them all here now, because I am still digesting the presentation, the accompanying document produced by DZERO (not ready for public consumption yet), and the implications and subtleties involved. However, let me flash a few of the questions I am going to try and give an answer to with my readings:

  • The paper does not address the most important question – what is DZERO’s track reconstruction efficiency as a function of track impact parameter ? They do discuss with some detail the complicated mixture of their data, which results from triggers which enforce that tracks have very small impact parameter -effectively cutting away all tracks with an impact parameter larger than 0.5cm- and a dedicated trigger which does not enforce an IP requirement; they also discuss their offline track reconstruction algorithms. But at a first sight it did not seem clear to me that they can actually reconstruct effectively tracks with impact parameters up to 2.5 cm as they claim. I would have inserted in the documents an efficiency graph for the reconstruction efficiency as a function of impact parameter, had I authored it.
  • The paper shows a distribution of the decay radius of neutral K mesons, reconstructed from their decay into pair of charged pions. From the plot, the efficiency of reconstructing those pions is excessively small -some three times smaller than what it is in CMS, for instance. I need to read another paper by DZERO to figure out what drives their K-zero reconstruction efficiency to be so small, and whether this is in fact due to the decrease of effectiveness with track displacement.
  • What really puzzles me, however, is the fact that they do not see *any* excess, while we know there must be in any case a significant one: decays in flight of charged kaons and pions. Why is it that CDF is riddled with those, while DZERO appears free of them ? To explain this point: charged kaons and pions yield muons, which get reconstructed as real muons with large impact parameter. If the decay occurs within the tracking volume, the track is partly reconstructed with the muon hits and partly with the kaon or pion hits. Now, while pions have a mass similar to that of muons, and thus the muon practically follows the pion trajectory faithfully, for kaons there must be a significant kink in the track trajectory. One expects that the track reconstruction algorithm will fail to associate inner hits to a good fraction of those tracks, and the resulting muons will belong to the “loose” category, without a correspondence in the “tight” muon category which has muons containing a silicon hit in the innermost layer of the silicon detector. This creates an excess of muons with large impact parameter. CDF does estimate that contribution, and it is quite large, of the order of tens of thousands of events in 743 inverse picobarns of data! Now where are those events in the DZERO dataset, then ?

Of course, you should not expect that my limited intellectual capabilities and my slow reading of a paper I have had in my hands for no longer than two hours can produce foulproof arguments. So the above is just a first pass, sort of a quick and dirty evaluation. I imagine I will be able to give an answer to those puzzles myself, at least in part, with a deeper look at the documentation. But, for the time being, this is what I have to say about the DZERO analysis.

Or rather, I should add something. By reading the above, you might get the impression that I am only criticizing DZERO out of bitterness for the failed discovery of the century by CDF… No, it is not the case: I have always thought, and I continue to think, that the multi-muon signal by CDF is some unaccounted-for background. And I do salute with relief and interest the new effort by DZERO on this issue. I actually thank them for providing their input on this mystery. However, I still retain some scepticism with respect to the findings of their study. I hope that scepticism can be wiped off by some input – maybe some reader belonging to DZERO wants to shed some light on the issues I mentioned above ? You are most welcome to do so!

UPDATE: Lubos pitches in, and guess what, he blames CDF… But Lubos the experimentalist is not better than Lubos the diplomat, if you know what I mean…

Other reactions will be collected below – if you have any to point to, please do so.


1. Luboš Motl - March 17, 2009

Congratulations for having such sane remote collaborators at the same collider. 😉


2. ervin - March 17, 2009


Does this mean an end to the speculations linking the CDF anomaly to Dark Matter? Or is it too soon to say?…

Ervin Goldfain

3. gordonwatts - March 17, 2009

Dude!!! Public document?!? WTF? If that was public why don’t you just link to the DZERO web site’s version!? What you have there is an internal note that was never designed to be out there. Its review depends on lots of other supporting documents and internal knowledge at DZERO (the jargon and plots were never designed for someone outside DZERO to look at). A “conference” note, just as in the case with CDF, is meant to be read externally. And it has not been released yet (is in review now). I know the s**t storm I would be in if I ever posted an internal note of CDF’s like this; please, live by mutual respect and take it down. Link to the conference note as soon as it is up.

4. Luboš Motl - March 17, 2009

Dear Gordon, I will happily change my link and/or remove it and replace by rumors-only 😉 if you tell me what to do.

Tommaso, it seems to be an Italian week. There’s a great new Italian book here:


and some more Italian words about it


and I just received a video explaining how Italians measure time:

5. Kea - March 17, 2009

Sounds dubious. I’ll wait for the published version.

6. Luboš Motl - March 17, 2009

Dear Gordon Watts,

I am really confused by your implicit statement that the PDF file is not a public document. If it were not public, why it is listed and linked on the official web page of a QCD conference


under Tuesday morning? The D0 website is not the only public place in the world. If you didn’t want Williams to make these things public, you should have prevented him from speaking about it at a conference with a public availability. 😉

Moreover, I think that you should be grateful to Tommaso for helping you to show that CDF are, compared to D0, irresponsible amateurs. 🙂

Best wishes

7. dorigo - March 17, 2009

Hi Gordon,

I am quite sorry, I now removed the link.

I got that document from a DZERO colleague of yours, who had mentioned at a CMS meeting that DZERO had published a refutation of the CDF signal. I had not understood it was private… It is his fault, not mine. Of course, he was just kind to share it, but he should have mentioned it was not approved material.


8. Luboš Motl - March 17, 2009

Dear Gordon, I am sorry, I’ve changed my mind. Please consider the link in #6 as my path towards the file that I consider a fully public result from a conference, regardless of the opinion of people at D0. I will not remove the link on my blog. Best wishes, Lubos

9. Anonymous - March 17, 2009

This is all very amusing.

10. Luboš Motl - March 17, 2009

Hi Tommaso,

in the Williams.pdf file, the biggest difference between CDF and D0 I can see so far is on page 4/15, timing.

CDF has no timing constraints. D0 only picks if t(A) as well as t(C) is less than 10 ns. That could throw away a baby, couldn’t it?


11. Gordon Watts - March 17, 2009

Sorry Lubos — the result isn’t private — it is just the analysis note that T had linked above. The official version of the conference note – the write-up – will be put out soon. The talk that everyone is linking to should definately get picked over and lots of questions asked!! It was the writeup that was an internal document.

In future, if you can’t find the PDF on the official D0 web site, your eyebrows should go up! Time to use your anonymous blog in that case!! 😉 Thanks a bunch for taking it down!! I appreciate it!

12. onymous - March 17, 2009

Lubos, 1 ns = 1 foot, and the impact parameters in question are centimeters.

13. Haelfix - March 17, 2009

This is certainly unexpected. Everyone I know felt that the CDF excess would be explained away, but not that it would dissappear entirely! Whats going on over there, is D0’s trigger that much tighter than CDFs?

14. Luboš Motl - March 17, 2009

Gordon: I see! Thanks for the clarification of the secret documents.

Tommaso: sorry for the mistake, I was silently reading “ps”, like in those “20 ps” lifetimes. So in that case, I don’t see any “qualitative” difference between the CDF and D0 approaches. D0 is morally looking at the very same question with pretty much the same parameters, using their tools. I don’t believe it is quite a coincidence that they get an agreement.

Imagine that what you’re producing is a new metastable particle, and you’re producing hundreds of thousands of them. The collisions at D0 are the same as those at CDF, so they’re producing hundreds of thousands of them, too. Now tell me how it exactly happens that any traces of those 500,000 particles disappear if they’re looking at qualitatively identical ratios as you do.

It even seems to me that they will agree with me, and not with you who wrote long poems about the “completely” different experiments. By the way, I am not understanding your thought about diplomacy. Not sure whether I am a better experimenter or diplomat – it’s hard to compare. Both are excellent. Surely, any diplomat who would be worse than excellent would fail to be nominated as the EU ambassador to the U.S., like I was. 🙂

15. dorigo - March 17, 2009

Dear Lubos,

I do not expect that you can appreciate the subtleties of these measurements. In fact, it is a very hard thing to do even for a die-hard experimentalist, let alone an amateur.

But you are a very intelligent amateur, so I will take the time to give you a few hints. What you do with them, is up to you. It is just a random act of kindness.

DZERO measures the efficiency of their “tight” silicon requirement from J/Psi decays entirely contained in the beam pipe.

CDF measures the corresponding number with J/Psi and Y, and finds a number; that number is then averaged, taking into account the sample composition (measured in a different PRD) with the SIGNIFICANTLY LOWER efficiency they get for bbbar events. 25% plus 23% times some factors equals 24.4%. Now, that number is the one which gets extrapolated.

Now: imagine you are a bbbar event. You have a 10% chance of not making it into the tight sample, because one of your two B hadrons might be neutral AND decay beyond the beampipe. However, DZERO extrapolates you using a higher efficiency, determined from J/Psi alone. Now, N(tight)/(eff_T/L) is the prediction. Since a larger-than-needed denominator is used, a prediction too small by 10% is expected.

Instead, DZERO sees nothing. So they do not even find the 10% of excess from bbbar that their extrapolation should obtain!

Now, add one thing: it is the fact that their “loose” sample is entirely dominated by Drell-Yan events. You can see it from a figure in the conference note: 50% of the sample is Upsilon decays. In CDF, it is some 6-7%.

Now, add a third thing. with 740/pb, and with a tight cut on muon rapidity, CDF has over half a million events. DZERO, with 900/pb, and with a muon rapidity extended from within |\eta|<0.56 to |\eta|<1.0, finds 170,000. Note that muon acceptance has to be squared to compute number of dimuons. So they see one tenth of the rate that CDF sees. Yet, their rate of Y decays is higher by a factor of five.

Now, add a fourth thing: they show a very small efficiency for tracks with large impact parameter: it is clear by noting that the reconstruction efficiency for K \to \pi \pi decays is of the order of 7%.

To me, the above means only one thing: that DZERO has a very small efficiency for large-impact-parameter tracks, and their extrapolation to the loose sample finds no excess because there cannot be any, under such circumstances.

Otherwise, one should understand how DZERO does not find some tens of thousands of bbbar events failing the tight cuts, nor why they fail to find any contributions from decays in flight.


16. Luboš Motl - March 18, 2009

Dear Tommaso, thanks, kind of understood. Can’t one simply calculate, assuming the SM (perhaps plus some particular new physics), what is the efficiency of both detectors to reveal large IP tracks?

Alternatively, aren’t you doing something wrong with the Y subsample – where a difference seems to be hiding according to your description?

17. dorigo - March 18, 2009

Hi Haelfix,

I think that yes, the dzero dataset is much tighter, for not well understood reasons. For sure reconstruction efficiency plays a role, but with the information available it is not yet clear to me whether the largest effect is a drop with impact parameter or the 3/5 request for silicon hits (CDF has 3/7). Also, DZERO might also have fewer non-muon backgrounds like punch-throughs -their detector is thicker.

I believe the final version of their conference note will address these points, now that we raised them 😉


18. ZEPHIR - March 18, 2009

I’m NOT considering these results a “background”. By my opinion CDF is on something and it may be pretty dangerous for terresterial life.

19. dorigo - March 18, 2009

Hi Lubos,

sure, one can try to determine the efficiency as a function of IP, with some inventiveness. However, this was not done by DZERO -at least not in the documentation I got so far- and so we are left in the dark on the origin of the effects I outline in my previous comment.

And Ervin, #2, sorry for the late answer: but yes, it is too soon to say, although I am indeed willing to bet any sum of money up to one year of salary that this is background, 10 to 1.


20. La partícula Y(4140) descubierta en el Fermilab podría ser un error de cálculo « Francis (th)E mule Science’s News - March 19, 2009

[…] DZERO del Fermilab, en lugar de confirmar el descubrimiento, lo desmienten. La nueva partícula puede que en realidad no exista, que sea un mero error de cálculo en el “complicado” fondo de partículas en el que se […]

21. Homer J Simpson - March 23, 2009

Well, after 8 years running, CDF Run II has discovered that its simulation does not describe muon punchthroughs and main authors of this paper (Dr. Giromini & Co.) mistook this obvious screw-up for manifestation of new physics. The rumor is that the paper has been rejected by PRL, correct?

Detector simulation of D0 detector must be more reliable!

I wonder if CDF will call the paper off to spare further embarassment.

22. dorigo - March 23, 2009

Dear Homer,

you chose an appropriate name for the nonsense of your comment.

Nobody mistook anything for anything else. CDF found out they could not understand all sources of backgrounds in dimuon-triggered events, and found it interesting enough to publish, because it resolved three inconsistencies of past measurements -the too high bb cross section when measured with muons, the integrated mixing probability, and some features of sequential decays.

The paper is submitted to PRD, and not to PRL.

You really do not understand the first thing here. D0 is not using the simulation by the way.


23. Qué ha pasado con “El Tevatrón encuentra una nueva partícula en contra del Modelo Estándar” « Francis (th)E mule Science’s News - June 29, 2009

[…] sido encontrada nueva física más allá del Modelo Estándar? No todos lo creen así. Por ejemplo, Tommaso Dorigo, coautor del paper del CDF en ArXiv, opina que el análisis de los datos de DZERO no es […]

24. Qué ha pasado con "El Tevatrón encuentra una nueva partícula en contra del Modelo Estándar" | Phasing - June 29, 2009

[…] sido encontrada nueva física más allá del Modelo Estándar? No todos lo creen así. Por ejemplo, Tommaso Dorigo, coautor del paper del CDF en ArXiv, opina que el análisis de los datos de DZERO no es […]

25. G. Sardin - June 30, 2009

Those interested in NEW PHYSICS may read the following article:

Fundamentals of the Orbital Model of Elementary Particles http://arxiv.org/ftp/hep-ph/papers/0102/0102268.pdf

And :

Structural and dynamical significance of the proton g-factor http://personales.ya.com/sardin/articles/proton%20g-factor.pdf

And have a look at the following clip:


Sorry comments are closed for this entry

%d bloggers like this: