Besides the Daily Mail, the AP is now reporting Proof of “God Particle” Found. They include the caveat:
But after decades of work and billions of dollars spent, researchers at the European Organization for Nuclear Research, or CERN, aren’t quite ready to say they’ve “discovered” the particle…
Senior CERN scientists say that the two independent teams of physicists who plan to present their work at CERN’s vast complex on the Swiss-French border on July 4 are about as close as you can get to a discovery without actually calling it one…
Rob Roser, who leads the search for the Higgs boson at the Fermilab in Chicago, said: “Particle physicists have a very high standard for what it takes to be a discovery,” and he thinks it is a hair’s breadth away.
which suggests that neither CMS nor ATLAS have quite managed to reach the 5 sigma threshold, and CERN remains dedicated to not discussing the obvious result of combining the data.
The AP report also has:
CERN spokesman James Gillies said Monday, however, that he would be “very cautious” about unofficial combinations of ATLAS and CMS data. “Combining the data from two experiments is a complex task, which is why it takes time, and why no combination will be presented on Wednesday,” he told AP.
From everything I’ve heard, my impression is that the reason no official combination is being produced is not because it would be technically impossible to do so on a time-scale of days, but because the decision not to do such a combination for ICHEP was made for reasons described here. The problem with this is that it may lead to a lot of confusing explanations like this in the AP report, which muddles how particle physics experiments are done and the obscure issue of 5 sigma/experiment or in combination:
experts familiar with the research at CERN’s vast complex on the Swiss-French border say that the massive data they have obtained will essentially show the footprint of the key particle known as the Higgs boson — all but proving it exists — but doesn’t allow them to say it has actually been glimpsed…
Roser compared the results that scientists are preparing to announce Wednesday to finding the fossilized imprint of a dinosaur: “You see the footprints and the shadow of the object, but you don’t actually see it.”
Better for CERN to just announce discovery and break open the champagne…
Update: Weird. The AP seems to have changed their title from “Proof” to “Evidence”. This may be the first time in history that a media headline about particle physics is incorrectly pessimistic (“Evidence” usually means a 3-sigma signal, which existed last December, “Proof” would be a better way to describe a 5+ sigma signal, if that’s what the combined CMS/ATLAS data shows).
Update: Curiouser and curiouser. I’m hearing that per-experiment combinations are around 5 sigma or above. Very unclear why the AP report is indicating no completely conclusive discovery announcement. Maybe the CERN administration is playing a game with us, downplaying expectations…
Update: As pointed out in a comment, Matthew Chalmers at Nature has
The ATLAS and CMS experiments are each seeing signals between 4.5 and 5 sigma, just a whisker away from a solid discovery claim.
Update: The Atlantic covers the best blogs you should be reading to follow the Higgs story. They miss Resonaances and a few others. About me, they have:
If the Higgs boson was a dead celebrity, Woit would be your TMZ — first to the scene, first to break it, and have it be right.
I have a bad feeling that this is just going to frustrate and confuse the public even more. I understand the arguments for and against officially combining the data to claim a discovery, and I see the merits of both sides, but given all the media buzz, it’s going to seem now like a real let-down to much of the public if they come out speaking of it as simply “strong” evidence. Perhaps a compromise wording such as “we found it beyond a reasonable doubt, but we are going to continue to refine our analysis” might be in order. Of course, I’m not a particle physicist, and all this might be moot anyway, depending on how accurate this and other stories are.
Of course, all this might be moot
After the faster-than-light shambles CERN is playing it safe, and rightly so. It’s taken nearly fifty years to get from prediction to validation, and it would be poor champagne indeed that couldn’t wait a few more months to be uncorked.
Let the media stew. When mankind looks back at 2012 five hundred years hence, the discovery of the Higgs Boson is all they’ll care about.
True enough. By the way, I found this article from Discovery News that is more clearly written (IMO) than the AP story about this issue:
The argument that there must be at least two experiments so that one can act as check to the other is understandable. In this case, however, I think it may as well be turned around. They won’t combine CMS and ATLAS data yet, because they want to have a winner (like UA1 with the W boson), and the loser relegated to the role of verification (like UA2). Whatever. If both experiments see a 4+ signal separately, as far as I’m concerned is a discovery. And a simultaneous one, so a technical draw.
Pingback: God Particle: “Proof” now changed to “Evidence” | Uncommon Descent
Here’s what the journal Nature has to say:
Putting a Higgs “discovery” at exactly 5 sigma is a little arbitrary, isn’t it?
Well, you have to pick some number, and any such choice will be kind of arbitrary. The 5 sigma convention comes from wanting to pick a number so high that you’re sure it can’t possibly be a mistake. What CERN may be doing here though is taking that standard as a starting point and insisting on something much stronger: the 5 sigma can’t come from combinations of the various different channels and different experiments.
If a single experiment were to report a 4 sigma signal at a certain mass in a certain channel, one might reasonably say that this wasn’t absolutely certain to be a new particle. If you see several signals this size in two completely distinct experiments, all at exactly the same mass, then it’s absurd to claim that doubt remains.
Pingback: Live-Blog zu den Higgs-igen Enthüllungen im Juli « Skyweek Zwei Punkt Null
Could someone explain in simple terms how we know that this particle is Higgs (the same question Nature seems to be asking)? Is it due to very particular decay characteristics? How long will it take to confirm that it behaves as Higgs predicted?
The standard model gives very specific predictions for the behavior of the Higgs, which are fixed by the distinctive properties of the Higgs field (spontaneous breaking of the SU(2) gauge symmetry fixes how it interacts with gauge fields, giving mass to fermions fixes its interaction with them). These predictions have for a long long time shown that if the Higgs mass is 125 GeV, it could not be observed until the energies and luminosities of the LHC were available, and then it should be seen with certain specific signal sizes in certain specific decay channels.
What the LHC experiments are seeing is, after decades of no such signal in these and similar channels, signals appearing in just the right channels, with roughly the right signal size. If this isn’t a Higgs, it’s something very like it.
The big question now will be seeing if the signal sizes in the various channels agree with the SM predictions. Initially the measurements of these sizes will be crude, but they’ll get better with more data. The Tevatron signal today was 2 +/- .7 times the prediction. On Wednesday we should see numbers for 3 different channels from the LHC. Rumors are that some of them are larger than expected, one (WW) is smaller. These signal sizes though are just getting big enough to be sure they are there, so we’re a ways away from a precise measurement of their amplitude. If these numbers don’t agree with the SM, then you’ve got something that acts a lot like a Higgs, but is doing something somewhat different than expected, which would be quite exciting to find.
In medicine if treatment is shown a couple of trials work to an accuracy of 3 sigma (< 0.3% chance of it being a statistical fluke), its unquestionably and uncontroversially regarded to be the discovery of a cure.
A 4-sigma signal would have a chance of less than about 6 in 100 thousand of being a statistical fluke, while a 5-sigma signal would have a chance of less than 6 in 10 million.
Peter is right in describing the press' uncharacteristic pessimissm as "weird".
The particle physicist’s use of statistics here may be somewhat over-optimistic, neglecting things like the “look-elsewhere” effect. But that’s why the insistence on 5 sigma, otherwise 3 sigma really would be enough to claim discovery. In any case though, the numbers I’m hearing from the experiments correspond to a very, very high degree of certainty that they’re seeing something at 125 GeV.
What’s especially weird is the change in the AP headline. It looks like someone got upset at the initial “proof” headline (OK, “proof” is only really in math, but 5 sigma is a pretty close analog in science). I’m curious who decided to change it to “evidence”, which is kind of seriously wrong here, since “evidence” is the standard term used for a 3 sigma signal, and we’re well beyond that.
In the annals of physics, what is the highest sigma signal that turned out to be spurious?
Probably the best known example in HEP is the “pentaquark”, where there were claims of 4-5 sigma observations, and this is probably the sort of thing CERN is worried about. I’m no expert on that story, but it seems to me somewhat different. For one thing, I don’t think there’s ever been a reliable theory predicting pentaquarks or how they should behave. So, if you start looking for the things, there’s a monster “look-elsewhere” effect since you can look for them in all sorts of types of collisions and all sorts of channels. The scale of effort going into the LHC experiments and care taken in the analyses is way beyond that of the much smaller scale experiments that claimed to see pentaquarks.
By the way, I looked for the 1984 Rubbia top quark discovery paper to see what he was claiming. It’s in Physics Letters B vol 147, page 493 (1984). I gather by the time they submitted the paper in October, their evidence had weakened from the July 4 time of the CERN announcement and press release. There’s a reference to “a clear signal” but a caveat that “more statistics are needed to confirm these conclusions and the true nature of the effect observed”. No quantitative statistical significance is quoted. It’s clear that a statistical measure would be kind of irrelevant anyway, since they have a very small number of events and the problem is clearly whether they’ve actually understood those events, not exactly how many they are. All in all, a very different kind of analysis than the modern ones at CMS and ATLAS.
The Empire (Fermilab) strikes back, according to the New York Times:
In terms of sigma, remember DAMA have a continuing claim to have 10-sigma evidence for dark matter detection which essentially no-one believes. And recently OPERA had 6-sigma-ish evidence for faster than light neutrinos.
However the thing to remember with counting sigma is that all these experiments are big complicated beasts with lots of people working on them and lots of systematics that may or may not be understood. So when evaluating number of sigma we need more than just statistics – there also needs to be a sanity check about whether there are hidden systematics that may not be understood, or a bug in the code that leads to rubbish being spewed out.
So with DAMA, there is a 10-sigma evidence for annual modulation in their event rates – and this isn’t doubted. However sceptics will point out that lots of potential systematics (such as the temperature) also modulate annually. So the significance of the claim relates to the ability of the experiment to control all annual modulations – and the judgement on this has nothing to do with `statistical’ significance.
With the Higgs, (one of) the main channels is Higgs to 2 photons. The signal here appears as a bump on an otherwise smooth and falling distribution. It is hard to see how systematics could reproduce such a bump – particularly across two experiments and also with other supporting decay channels – making the `number of sigma’ significance credible.
For further clarification: the bump in Higgs to photon-photon should be sharp and narrow, as the Higgs has small width. Such a sharp spike on an otherwise smooth distribution is very hard to reproduce by a systematic.
What are the implications of a 125 GeV higgs for Susy? Does the LHC have enough energy to potentially find its Susy partner?
a Higgs below 135 Gev was a firm prediction of the MSSM. So a 125 GeV Higgs (with somehow SM-like behavior) fits perfectly. However, it s basically impossible to derive the masses of SUSY particles, in particular scalar tops, from it. Anything between about 200 GeV up to (in principle) multi-TeV or even higher remains possible.
Thank you Sven.
I thought because the higgs is relatively light it has to have a massive partner.
To supplement Sven’s answer, the problem with SUSY is that you have to break it somehow, and you can get a huge range of possibilities depending on how you do this. So, even once you know the mass of the Higgs, there are no definite predictions about other particles related to it by SUSY.
The interesting thing we’ll learn soon is, at least crudely, what the cross-section x branching ratios are for several channels of Higgs decay. If these differ from the SM values, that will be very interesting, and surely there will be attempts to explain this in terms of SUSY. Note that, as far as I know, there are no SUSY predictions one way or the other about this.
The Split A1 was an effect over 5 sigma.
It was most likely the result of experimenter bias.
Although the look elsewhere effect will now be greatly reduced, because the interesting region was already defined prior to accumulation of the new data.
But now the danger once again experimenter bias. It will be interested to hear how the experiments dealt with that problem… whether they were blinded and whether or not they applied changes after unblinding.
Looks like the Tevatron `hint’ is all in H>b bbar. Curious, because the LHC saw better `hints’ in H>gamma gamma… must be the terrific LHC em calorimeters.
The difference in channels between the Tevatron and LHC experiments really comes down to the fact that the LHC is pp, while the Tevatron is ppbar. In a ppbar collider the dominant Higgs production is qqbar->W(Z)->H W(Z) where the W(Z) essentially radiates a Higgs. These events are easy to trigger on because of the high energy leptons or missing energy associated to the W(Z) decay. A light SM Higgs primarily decays to bbbar, so this an easy thing to look for. Events with one or two leptons, lots of missing energy, and two b-tagged jets.
At the LHC where you have pp collisions, the dominant Higgs production is gg->H, where you get the Higgs and nothing else. The Higgs will still primarily decay to bbbar, but there’s also ALOT of other QCD processes that will give you two b-tagged jets, so it’s hard to “pick out” these Higgs events from the background processes. A light Higgs rarely decays to gamma gamma, but despite the low branching ratio this channel still gives the best signal to background discrimination at the LHC (at least in early data).
Leaked CERN video confirming a new particle:
“News: Cern Higgs boson announcement: we have observed a new particle”
I love the Atlantic article. Right on the button. Well done Alexander Abad-Santos.
I havn’t stopped laughing since reading the Atlantic article. I have already renamed some of my links as….well you know if you have read the article.
The signal for the `sub-millisecond pulsar’ (1989B) that was `discovered’ in the remnant of SN1987A by Pennypacker et al., being a simple frequency measurement, had an enormously high statistical significance. It was between 11 and 37 sigma depending on when the observations were made: it was an absolutely beautiful peak, way above the noise.
I remember that hundreds of suspect theoretical papers were written after the fact explaining how such an object could possibly exist. All of the ex post facto theory was very dubious because rotational velocities at the surface of such a quickly rotating neutron star would need to be very significant fractions of the velocity of light, and it was almost possible to imagine a high density equation of state soft enough to allow the thing to exist, and also allow for the supernova explosion. But that didn’t stop the theoretical astrophysicists!
“The frequency of the pulsar during the Jan. 18 observation was tracked by dividing the data into independent half-hour runs; the statistical significance during these runs ranged from 11 to 37 standard deviations. The frequency exhibits a sinusoidal modulation; the 15 frequency measurements were within 5 percent (rms) of a sine function with a central value of 1968.629 Hz (barycentric), amplitude 3 x 10E-3 Hz (peak-to-peak) and period of 8 hr.”
All theory aside, the `signal’ turned out to be due to a resonance in a faulty radio camera: the experiment had two cameras: camera A and camera B, and it turned out the pulsar was seen in camera A but not in camera B. Pennypacker had to publicly retract the observation, but I don’t seem to recall many retractions having been made by the theorists who explained it!
But the Higgs observation isn’t likely to be in the same ballpark: the signal can’t be misproduced so easily at CMS and ATLAS.
Nice answer, thanks, tdb. Why is Higgstrahllung is not yet discernible at the LHC?
I’m far from an expert on the Higgs analyses, but I believe associated production, or Higgstrahllung, is about ~100 times smaller production cross section than gluon fusion; but bbbar decay for a ~125 GeV Higgs is ~1000 times more likely than gamma gamma decay. Of course then you have very different experimental uncertainties (b-tag efficiency vs. photon ID efficiency, etc.). To be honest I’m not sure how much data is needed for a Higgstrahllung channel to become a viable analysis at the LHC, but it’s certainly more than the ~10 fb^-1 being analyzed now.