Tuesday, 13 December 2011

Visual on Higgs

Today CERN presented the latest Higgs search results. But, first things first. The live podcast of the seminar was stalling, choking, or breaking down completely. Well into the 21st century a huge international lab doing cutting edge science (and boasting of inventing the world wide web) is unable to provide a smooth transmission of its most important public presentation of the year. One may call it ironic, or repeat Didier Drogba's famous words after the 2009 Champions League semi-finals.

OK, after venting my anger I can talk some physics.
  • The main news, correctly rumored on blogs before, is that there is a significant excess of Higgs-like events corresponding to the Higgs mass of ~125 GeV. More precisely, the local significance of the ATLAS excess is 3.6σ, or 2.5σ sigma when the "look-elsewhere effect" in the 110-146 GeV mass range is taken into account. For the CMS the significance is somewhat smaller: 2.6/1.9 with/without the look-elsewhere effect. Separately, these excesses would be shrugged off; combined, they are very suggestive that we're seeing the real thing at last.












  • In ATLAS, only the H→γγ and H→ZZ*→2l+2l- channel have been updated with the full 2011 dataset. Those are the ones where statistics is limited due to the small Higgs branching fractions. However, in these final states the 4-momentum of the Higgs can be fully reconstructed to a good precision, offering a very good mass resolution of order 2 GeV. CMS updated all main channels, also those that do not provide a lot of mileage near 125 GeV at this point.
  • The excess is seen by both experiments and in each of these channels. The excess in H→γγ peaks around 124 GeV it CMS, and around 126 GeV in ATLAS, which I guess is perfectly consistent within resolution. In the 4-lepton channel, ATLAS has 3 events just below 125 GeV, while CMS has 2 events just above 125 GeV. On top of that there's the long-standing excess in the H→WW*→l+l-2ν channel, which however is not the driving force anymore. It's is precisely this overall consistency that makes the signal so tantalizing.











  • ATLAS observes slightly more events than expected from the Standard Model Higgs at that mass. In ATLAS the best fit to the Higgs cross section corresponds to roughly 1.5 the Standard Model value. Thus, one way or another, the 125 GeV thing is a fluke. It may have been good luck if these events are due to Higgs, or bad luck if they are due to other Standard Model backgrounds...










  • CMS now excludes Higgs down to 127 GeV. The ATLAS limits slightly worse around 130 GeV, but thanks to lucky background fluctuations they happened to exclude the low mass region between 112.7 and 115.5 GeV. The latter is a spectacular confirmation that what the ALEPH experiment saw back in 2001 was a genuine fundon (fundons are elementary particles produced in high-energy collider near the end of the budgetary cycle).
So much for now. The ATLAS and CMS combination notes are already out.

21 comments:

muon said...

CMS was not lazy at all. ;) You can get the official note on the combined physics results here: http://cdsweb.cern.ch/record/1406347?ln=en

In fact there is a whole release of information here: http://cms.web.cern.ch/news/cms-search-standard-model-higgs-boson-lhc-data-2010-and-2011

Finally, you stated that the ZZ and WW channels were not updated with the full data set. This is true of ATLAS, but not true of CMS.

John Peacock said...

Allowance for "look elsewhere" seems too generous. It's probably OK for each experiment individually, but you only want to correct once in the combined result. The LE corection seems to amount empirically to removing about 0.6 sigma. So I'd add the raw results in quadrature: 3.6 and 2.5 gives 4.4 combined, so probably about 3.8 allowing for LE. Not 5sigma, but 3.8sigma would be good enough in most fields, barring some unknown systematic.

Kea said...

Wow! Fairies exist! Amazing work from ATLAS and CERN, and thanks for the early report.

Anonymous said...

Will Atlas have updated the other channels by January?

Anonymous said...

Let's face it, if that fluke is a ordinary SMhiggs, then we have nightmare scenario, don't we? Even no Higgs at all would boost theoretical physics further. And now, 2012 nobel prize in physics will be the last one for HEP ever. i've always thought 125 gev higgs will prove SM cannot be valid above few Tev, and yet it is a single bump and it fits SM higgs crosssec., prod. ratio just well.
DesertUpToPlancScale

Anonymous said...

Hi Jester,

thanks for the rant. This was very annoying, and should be highly embarrassing for a lab that advertises itself as the place where the WWW was invented.

Anonymous said...

CERN invented the Web, has a huge collider with 10-digits budget, owns a next-gen high speed computing grid, but in 2011 cannot properly broadcast some standard resolution video over the internet?

Kind of sad... or maybe HEP just became mainstream while we weren't looking :)
(The Sheldon Cooper Effect lol)

Jester said...

Thanks Michael, sorry for the CMS channels wrong. (It would probably help if I could hear more than 20% of the broadcast...)

Anonymous said...

God may not play dice, but He's played a good joke on the designers of LEP and the LHC with this mass.

Jester said...

and the best joke is that LEP was originally designed to run at 250 GeV :-)

Luboš Motl said...

John Peacock, exactly my words! The look-elsewhere punishment can't be done for each detector separately if they nontrivially agree on the Higgs mass value (approximately). When it's done correctly, we still end up with a 4-or-so sigma signal which may not count as an official discovery but it would really be bold to bet against it.

Anonymous said...

No broadcast problem for me... Maybe the problem was at your end ?
Anyway, 125 GeV or so was predicted a long time ago by SUSY people :-)

Jester said...

No, I had a fast broadband connection, and the broadcast was working perfectly until 1 minute before the start. Everybody I know who's not in Europe had the same problem.

Anonymous said...

so maybe the problem was in the connection to the US? Slovenia was fine.

John Peacock said...

Lubos: glad to find someone else who is not statistically timid to an extent that I find inexplicably common. Last week in Stockholm, the Nobel prize was awarded for supernova cosmology, based on 1998 work that probably was less formally significant than we now have for Higgs (or for something, at least). If you've done your statistics properly (and a 10-billion-Euro experiment should have done), then 4sigma is a definite discovery unless some unknown systematic wrecks it. But such a systematic would also wreck a 5sigma or a 20sigma result. So why wait to claim it?

Luboš Motl said...

Dear John Peacock,

in this case I agree. It's the single most important search and 4-sigma after look-elsewhere correction is a huge evidence that can realistically be only reverted by a qualitative human error - and such a human error could destroy stronger signals, too.

Still, in general, I wouldn't adopt the 4-sigma-is-enough rule for other situations. More precisely, if one studies a topic that is much more special than the search for the most important particle, such as "the search for pentaquarks using collisions of X21 and Y654" [fake names], then this is not a special topic and there are hundreds or thousands of them.

So there should be a special punishment for "look elsewhere" where "elsewhere" also includes all equally (un)important papers. Because there are hundreds of them, the probability of a fake positive in one of them is really 100+ times higher than if we assumed that there is only one.

If someone could get 4-sigma even after this correction, then OK, but I don't think we know any. So my inclination to accept 4-sigma evidence as evidence that mostly settles the story is limited to the "exceptional" situations that I was watching as being the most important one(s), and I will avoid this rule for "unexpected signals at random places" which may still occur by chance, and most 3-sigma signals of this type will still go away.

Of course, a rational estimate whether the signal goes away is also affected by theoretical expectations. I think that I can separate them from the experimental evidence - and I have done so so far - but on the other hand, I am in no way ashamed of the theoretical understanding we have and the impact it makes on my expectations. The Higgs should really be there, it shouldn't be too heavy because of high-precision tests, and it's excluded above 127 GeV, so chances are that it has to be in the 116-127 GeV window. This argument may be added on top of the cold numerical positive evidence for the Higgs to create a really strong picture.

At least at the beginning, the evidence for a nonzero C.C. was probably weak, too. But I guess it's much stronger than 5-sigma today.

Cheers
LM

Cheers
LM

mfb said...

The broadcast had problems in Europe as well.

@Jester: Be careful with your comparison with/without look-elsewhere effect. As far I understood the talks, ATLAS included the full mass range in the calculation, while CMS used the (not excluded) low mass region alone.

I perfer the CMS way here (fluctuations in a region which is already excluded are not a sign of a SM higgs), but as long as both methods are different the values have a different meaning.

Anonymous said...

Assuming this does pan out, would it torpedo theories wherein the Higgs is a top composite, since 125 GeV < 175 GeV?

Jester said...

mfb, for both ATLAS and CMS the LLE corrected significance I quoted refers to the low mass range. They also give the LLE correction for the entire mass range, but that I left out (I don't think it's relevant anymore).

Anon, not sure what models are you referring to.

Anonymous said...

Jester: brilliant coverage, as usual. (btw: no problems with the broadcast in Spain, played at a packed seminar room without relevant breakdowns.)

I can't but buy the view that the relevant numbers are those after LEE in the non-excluded mass range. n.b. the LEE will politely become irrelevant once decent statistics is available in all relevant channels.

to those discussing (often at the formal level of cocktail bar chats) whether 5sigma requirements are fair or just a masochistic overkill:

statistics is paramount in HEP experiments nowadays --- and careful, conservative statistics at that. if we witnessed an alien spaceship landing --- well, we wouldn't be discussing evidence on the basis that there is only one event. but here we're talking subtracting tiny effects from a background whose systematic uncertainty we control poorly. yes: SM predictions _are_ poor for many reasons; refinements are necessarily data-driven; and proper, safe statements are contrived. people should think twice about calling off painful analyses by those better acquainted with the problem, based (at the best) on vague recollections of a third-year statistics course.

one way to avoid much of the misunderstanding: put in LEE from the very beginning, then compute CL's for the estimate of the significance itself, finally quote the "error of the error" --- as some old LEP analyses did. so: instead of "3.5sigma, turned 2.4sigma after LEE", say "significance between 1.8sigma and 3.0sigma at 95% CL".

but maybe I'm naive and people would get equally rowdy even then...

to theorists: if confirmed, just tells us that some relatively narrow scalar state is appearing in radiative corrections in the expected ballpark. a few years will be needed to study its nature. thus, few implications at this point for most BSM work lines --- apart from, perhaps, constrained SUSY. but don't worry: 100+ soft breaking parameters can helpfully come to the rescue...

Anonymous said...

Now let's find the other four Higgs.