Ok, having heard the talks, had a lot of conversations with colleagues, and looked a bit more carefully at the plots from today’s talks at CERN, I have a few revised comments, somewhat more optimistic than my initial reaction, with more details to follow.
Exclusion in new regions:
- The CMS experiment has excluded a new chunk of the Higgs mass range, from 127 or 129 GeV (depending on how aggressive you want to be) to about 600 (previously the two experiments together excluded 141 to 476. [Look to CMS for the official numbers]
- ATLAS did not change their result for high mass but excludes the lightweight Higgs down to 131.
- Both experiments are seeing effects in the 120-130 GeV range that are preventing them from excluding as much as they would have expected to do on average.
Hints from ATLAS:
- ATLAS, using Higgs –> 2 photons and Higgs –> ZZ –> two leptons and two antileptons, shows an excess at about 126 GeV, of a local significance of 3.6 sigma (but only 2.3 sigma with most conservative form of the look elsewhere effect; the more complete number lies somewhere between).
- If there were a Standard Model Higgs particle at 125-126 GeV, the expected local significance would be 2.4 sigma. The signal seen is therefore compatible with a Higgs signal of this type.
- It is still compatible with a large fluctuation in the background.
Hints from CMS:
- CMS, using 5 different measurements — Higgs –> 2 photons, Higgs –> ZZ –> two leptons and two antileptons, Higgs –> WW –> lepton, antilepton, neutrino, antinuetrino, Higgs –> taus AND Higgs –> bottom quark/antiquark pairs — claims an excess at about 120 GeV and another at 124 GeVwith a local significance of 2.6 sigma (but only 1.9 sigma with the most conservative form of the look elsewhere effect; the better number lies somewhere between).
- If there were a Standard Model Higgs particle at 125-126 GeV, the expected local significance would be 2.6 sigma. The signal seen is therefore very compatible with a Higgs signal of this type.
- The claim CMS wants to make is that the ratios of the excesses in the 5 different channels is quite consistent with a Higgs signal at around 124 GeV. I am not sure I believe this argument, but it is certainly a legitimate one to make. To believe it, I would need a much more careful explanation of how the three more subtle analyses were done.
- It is still quite compatible with a moderate fluctuation in the background.
Are they compatible?
- There are some weird features in the data that could be due to low statistics, but roughly speaking the answer seems to be that the hints at ATLAS and CMS are reasonably compatible with each other.
Given this,
- What we saw today is probably compatible with a Standard Model-like Higgs at about 125 GeV.
- What we saw today is also probably compatible with a large but not extraordinary fluctuation in the backgrounds, perhaps combined with a subtle technical problem in one or another analysis.
- And the only way to find out which of these two is the truth is to gather a lot more data in 2012. Period.
30 Responses
Brout and Englert did not have the boson.
Matt,
You might want to take a look at the live chat at Nature, involving Gordon Kane and Bill Murray (of ATLAS).
http://www.nature.com/news/live-qa-the-hunt-for-the-higgs-1.9642
Murray says: ” We have no proof, but I polled 10 tof the leaders of the ATLAS search and my wife last night if they thought we had found it and they ALL said yes. But that is not proof, it could be over-excitement.”
Kane says: “for a higgs to be meaningful it must be part of a supersymmetric theory”
“Recently we have published string theory calculations that imply the higgs boson mass is 125 GeV so if it is there are strong implications for connecting string theory to the real world, and for what the higgs discovery implies. we did that before the data.”
“string theories are now well enough understood to predict higgs physics,”
As you and Tommaso both said, Peter, it depends what you ask. I was asking whether the data presented constituted strong evidence. If instead you ask people if they think they have found it, they say that there’s a good chance of it — but that the evidence is still very slim. And certainly not “firm”.
Personally I think there is quite a decent chance that they have found it. I can make a strong argument in its favor. I can also see places where I can pull on a loose end and the whole thing unravels.
As for Gordy — do you, of all people, really think he can calculate the Higgs mass from string theory? If not, why did you mention it?
It’s well known what I think of Gordy’s claim, but it might be a good idea if HEP theorists in general express publicly an opinion about it. He’s going on a publicity campaign for this, including another such live chat at Science magazine tomorrow.
Oh, for heavens’ sake. What an atrocity. He’s an old friend of mine; is he going to force me into this?
Dear Prof. Strassler,
as every independent observer from the outside can see from the comment section below Your newer post “Theory / Experiment Discussion at CERN about Higgs” for example, it is Mr. Peter Woit (why do You belief everything this gentleman says?) who repetedly and persistenly tries to force You into an ugly discussion and NOT Your old friend Prof. Kane. What Prof. Kane says or discusses at other places in the internet, chat boxes, etc, has not automatically to be Your business if You dont want it.
Just to say and put certain things straight …
Cheers
Hi Matt
ATLAS does not just use 2 channels… there is also WW->lnu lnu and tau tau in the 126 GeV combination. The WW is also contributing to the excess (and this could be consistent with what was seen in July).
Leonard, the individual bins are Poisson processes (I don’t know who would challenge that) but the expected rates are correlated in a very detailed way. It’s not simply a product of a bunch of Poisson where the means are all fixed a priori. Also, the background is definitely NOT assumed to be known 100%… all the hard work goes into quantifying the uncertainties on signal and background rates. I don’t know where these two very inaccurate characterizations are coming from, but they should not propagate further.
Lastly, one should not use the “number of sigma” (p-value under background-only hypothesis) in the different channels (eg 5 in one channel and 3 in another) if they are consistent with each other (or consistent with the SM hypothesis). Instead, the experiments show the best fit “signal strength” with error bands which is a proper estimator for the rate. One can see from those plots that the error bars are quite large and it is consistent at the ~1sigma level. One can do a little better by directly measuring ratios in which case some systematics cancel, but that hasn’t been shown.
Matt, Tony Smith observes that the gamma gamma channel contains an ATLAS excess against a CMS deficit. What would you say about this?
I cannot answer yet. I am currently quite confused about the CMS data. This is related to my misinterpretation yesterday. I will try to figure out the correct interpretation today.
what’s wrong with CMS’s interpretation of their data in terms of statistical significance?
I agree with Matt, I would give 20-30% the current data is an SM Higgs. I think the odds Tommaso is offering (3:1 on) are then very good .
First, I do not take the least likelihood/combined analysis that are so much in vogue these days, after the LEP1999 saga, to be all they are cracked up to be. Firstly, they assume data in individual bins are uncorrelated Poisson processes -where is the justification for that? The bins must be correlated, though in ways that are hard to quantify. But leaving this aside, another weakness is that the test is between two scenarious, pure background or background + SM Higgs. There are pitfalls here. The background is assumed known 100%. Yet from the ZZ data, it is clear in this channel the theoretical background is not very good. Hence its 3 events at 130GeV contributing 2.4 sig are not really worth that. If you look at this channel over all CM energies that is clear, but the ‘blind’ combined analysis can not see that. The problem from there being just two scenarios can be easily demonstrated: if you get, say, 5sig in favor of a SM Higgs in the gam/gam, but a 3sig exclusion in the WW, it is clear we are seing something, but 95% its not the SM Higgs. Yet the combined analysis does not have the latter as a possibility, and will favor a SM Higgs on such data.
My take on the data is that there is definitely something in the gam/gam channel, but what? The peak has a strange ‘saw-tooth’ shape, instead of the slanted Gaussian expected – and this is not a statistical fluke – we saw exactly the same shaped peak 4 months ago when there was much less data. What does this mean? Is it some sort of technical glitch?
Then there is the WW channel. There is really not much more data here than 3 months ago, when we had 1.8fb-1. The signal in this channel has the same few humps we saw then – when the hump at 140GeV was touted as a possible Higgs. Now the hump at 130GeV is concentrated on. If CERN had waited till the analysis in this channel included all 4.8 fb-1 then this channel should have provided 3sig in favor of a SM Higgs now. This would have been the icing on the cake. Yet CERN did not release this data. Why? Is it telling a different story?
To Henry: I have some explanation of the “sigma” used in the Higgs talks at my site, in a post specifically meant for that (http://www.science20.com/quantum_diaries_survivor/fundamental_glossary_higgs_broadcast-85365).
To Matt: I am happy that upon reading better the information (but still lacking the more detail you would need about the individual analyses, as you point out) you are now more positive about the genuine nature of the signal.
I would like to point out that large statistical fluctuations do occur, but are indeed rare – that the two results by ATLAS and CMS are consistent with each other (let alone with the expected SM H rate) is a quite strong observation. We should be careful when we discuss fluctuations to make the point that what has happened the most frequently in past false signals in HEP was that systematics were mis-evaluated, not that the data took so often the liberty of going all over the place.
Now, regardless of flukes: you said earlier that you give this signal as a slightly less than 50:50 chance. Are you willing to bet it at 2:1 odds ? 3:1 ? I can offer you 300$ if it is proven that the Higgs does not exist in that range, against your 100$ if instead the signal will gain in strength with added data. Let me know.
Cheers,
T.
It would interesting to note about Higgs mechanism and how the whole thing came about.
Idea was influenced to a large degree by theory of superconductivity, Bardeen, Cooper and Schrieffer, (BCS theory).
and the work of Yoichiro Nambu. Further contributions came from Guralnik, Hagen and Kibble, Englert – Brout, and of course our Peter Higgs. Lets hope that today’s conference is the first step in this direction of either
a confirmation of Higgs mechanism or possibly something else. Needless to say we wish both CERN teams the best !
In my public talks I always mention Anderson and Nambu, and the actions of the 6 Higgs-particle inventors.
Is it true that if this is a statistical fluctuation, then the Higgs isn’t there? In which case we’re in very new and exciting territory? Should we be rooting for a fluke?
we’re only talking right now about the STANDARD MODEL Higgs particle, the simplest possible one, and the region 115-127 GeV still hasn’t been ruled out. OF course it would be great if the simplest possible Higgs isn’t the one present in nature.
Hi Matt
I was wondering if you could shortly explain the sigma system that they seem to use when giving results which seem to indicate the probability of the results being down to random fluctuations in the data?
Not today — start with reading http://en.wikipedia.org/wiki/Standard_deviation
There is a good explanation of what we mean by “sigma” and the “look-elsewhere effect” for the general public here: http://www.perimeterinstitute.ca/images/outreach/interpreting_the_cern%20results.pdf
Yes, but what are the sigmas based upon: i e what is the null hypothesis in this current observation? Is it = The Higgs does exist, but elsewhere than here (this GeV range)?
Is the Higgs assumed to exist?
No, the Higgs is not assumed to exist.