Today at the Moriond conference ATLAS and CMS updated their diphoton resonance searches. There's been a rumor of an ATLAS analysis with looser cuts on the photons where the significance of the 750 GeV excess grows to a whopping 4.7 sigma. The rumor had it that the this analysis would be made public today, so the expectations were high. However, the loose-cuts analysis was not approved in time by the collaboration, and the fireworks display was cancelled. In any case, there was some good news today, and some useful info for model builders was provided. 
Let's start with ATLAS. For the 13 TeV results, they now have two analyses: one called spin-0 and one called spin-2. Naively, the cuts in the latter are not optimized not for a spin-2 resonance but rather for a high-mass resonance (where there's currently no significant excess), so the spin-2 label should not be treated too seriously in this case. Both analyses show a similar excess at 750 GeV: 3.9 and 3.6 sigma respectively for a wide resonance. Moreover, ATLAS provides additional information about the diphoton events, such as the angular distribution of the photons, the number of accompanying jets, the amount of missing energy, etc. This may be very useful for theorists entertaining less trivial models, for example when the 750 GeV resonance is produced from a decay of a heavier parent particle. Finally, ATLAS shows a re-analysis of the diphoton events collected at 8 TeV center-of-energy of the LHC. The former run-1 analysis was a bit sloppy in the interesting mass range; for example, no limits at all were given for a 750 GeV scalar hypothesis. Now the run-1 data have been cleaned up and analyzed using the same methods as in run-2. Excitingly, there's a 2 sigma excess in the spin-0 analysis in run-1, roughly compatible with what one would expect given the observed run-2 excess! No significant excess is seen for the spin-2 analysis, and the tension between the run-1 and run-2 data is quite severe in this case. Unfortunately, ATLAS does not quote the combined significance and the best fit cross section for the 750 GeV resonance.
For CMS, the big news is that the amount of 13 TeV data at their disposal has increased by 20%. Using MacGyver skills, they managed to make sense of the chunk of data collected when the CMS magnet was off due to a technical problem. Apparently it was worth it, as new diphoton events have been found in the 750 GeV ballpark. Thanks to that, and a better calibration, the significance of the diphoton excess in run-2 actually increases up to 2.9 sigma! Furthermore, much like ATLAS, CMS updated their run-1 diphoton analyses and combined them with the run-2 ones. Again, the combination increases the significance of the 750 GeV excess. The combined significance quoted by CMS is 3.4 sigma, similar for spin-0 and spin-2 analyses. Unlike in ATLAS, the best fit is for a narrow resonance, which is the more preferred option from the theoretical point of view.In summary, the diphoton excess survived the first test. After adding more data and improving the analysis techniques the significance slightly increases rather than decreases, as expected for a real particle. The signal is now a bit more solid: both experiments have a similar amount of diphoton data and they both claim a similar significance of the 750 GeV bump. It may be a good moment to rename the ATLAS diphoton excess as the LHC diphoton excess :) So far, the story of 2012 is repeating itself: the initial hints of a new resonance solidify into a consistent picture. Are we going to have another huge discovery this summer?
Thanks for the good news, Jester.
ReplyDeleteIt may be the turning point everyone is hoping for. I keep my fingers crossed the signal is real and not some hidden fluke in data analysis or some artifact similar to the 17 keV neutrino claim.
We still have a few minutes left for a group prayer to Saint Patrick to make it real ;)
ReplyDeleteSo far it's been discussed that the signal is only seen in diphoton. How strong is the hypothesis that 750GeV signals are absent in the other channels? i.e. is it feasible that we could see a surging signal in another channel after a long and successful year of data taking?
ReplyDelete@Ervin: A statistical fluctuation is possible, but I really don't see how a problem with data analysis would lead to such a peak in the diphoton mass spectrum. It is a really clean channel experimentally. If you do the analysis wrong, you reduce your sensitivity, but you don't get fake peaks. Guillet shoulder would be an exception, but that is somewhere at 100-150 GeV with the absolute pT cuts and not present with relative cuts.
ReplyDelete@Anonymous: 2016 will certainly allow more channels to get a visible 750 GeV signal. The expected ~30/fb at 13 TeV will render all previous datasets negligible in comparison. Basically all searches based on an invariant mass pay attention to the region around 750 GeV now. If there is a visible signal in some "reasonable" channel, we won't miss it.
Looking forward to the summer.
I remember that some of the strongest arguments in favor of it being a fluctuation were the tension with run 1 (now essentially gone, I gather) and the relatively small global significance (now improved substantially for CMS and also for ATLAS if the rumor is true). I know it's still not nailed down, but can we start saying "holy shit" yet?
ReplyDeleteYou previously estimated the probability of a discovery at 10%, If I recall correctly. How about now?
Xezlec, 20%.
ReplyDeleteAnon, to complete the answer: every model predicts signals in other channels, at least in some diboson channels (ZZ,Z\gamma,WW). The fact that we haven't observed any bump in other channels puts some constraints on model building, but does not require any unnatural gymnastics yet. That last statement will change if nothing is observed in other channels by the end of the year.
I am a layperson. Please help clarify the following. My reading of the first graph (Spin 0) is that the 750 GeV peak corresponds to a total of 9 excess events. Did I make a mistake? If so, what is the actual number of excess events?
ReplyDeleteA bit more, but that's about right. In the December analysis ATLAS had 23 events in the 2 bins around 750 GeV, with the predicted background of 11, so 12 excess events. I haven't yet digitized the new data, but I suppose these numbers haven't changed.
ReplyDeleteAnother layperson, unfortunately. I've been following the GeV bump since the beginning and know the RS Graviton has, off and on, been a theoretical favorite (at least among the media). So reading Alessandro Strumia's slide from Moriond today struck me as interesting, noting that the peaks disfavor the predictions of an RS Graviton, thereby disfavoring the theory. This, along with as far as I can tell strong Spin 0 numbers, seems to point in a different direction.
ReplyDeleteDo I have that all horribly wrong? Or is this all simply guess work for everyone until more data is taken?
Has anyone done a naive combination of the significances of both experiments and applied LEE? And the same if the ATLAS 4.7sigma analysis rumour is accurate? I'm not familiar enough with statistics to do that myself.
ReplyDeleteIs this new analysis of ATLAS something CMS might be doing too? And does all this redoing of analysis constitute a form of selective evidence (i.e. redoing the analysis until you get an answer that you like) that should reduce significance, or is it a predefined process?
@ mfb
ReplyDeleteMy statement on flukes in data analysis is on rather general grounds. As the second LHC run explores un-chartered territory, it is possible (although not very likely) that some of the assumptions that underlie data analysis might have to be revised. I agree that the channel is clean, but there may be surprises in our modeling of diphotons emerging from events at larger energies.
Hi Jester,
ReplyDeletenothing about the width?
to me it looks that atlas 8tev data has no peak. the slight excess is a fluke and at the wrong energy too. at the same time the 750gev looks real to me, which would imply that the 750gev boson is simply not produced directly by hadron collisions but by some transient superheavy entity in the multi TEV regime?
ReplyDeleteOn the high-mass sideband (tail above the peak) of the ATLAS plot there are 11 events above the shown background; there are 4 underneath. This background is underestimated (as usual with the two-photon plots that we are seeing, dropping standards?).
ReplyDeleteFelipe
Anon, fitting by eye is tricky. Doing the math one finds that the background curve is a very good fit at the sidebands, see 1601.07330.
ReplyDeleteAnon-2, decays from a heavier parent resonance to 750 GeV is a valid possibility. I personally find it less likely, but we'll see what the data will say.
Akidbelle, nothing really new: large width still somewhat favored by ATLAS, but not by CMS.
Anon 2:46, yes, the original Randall-Sundrum model with all SM matter localized on the TeV brane is disfavored because it predicts a large dilepton signal.
ReplyDeleteAnon 3:54, indeed, sculpting cuts is always a danger, as was proven by a "discovery" of a 115 GeV Higgs at the early stage of the ATLAS experiment. However, experimentalists are well aware of this and have developed safety protocols. They certainly don't apply random cuts to the data and pick the ones that make the peak larger - the usefulness of cuts has to be first demonstrated on simulated data.
"CMS also cleaned up their run-1 data"
ReplyDeleteFalse: run-1 data were used directly without "cleaning".
thx, fixed
ReplyDeleteJust a note of pessimism, everyone is looking at the 750Gev bump and getting excited about the things that confirm it and not so excited about the things that don't.
ReplyDeleteSo a spike of "I see it too" in the data analysis is nearly as expected as the 750 "confirms my theory" papers on the arxiv...
As has been said already, the real story is, we need to wait for 30 fb^{-1}.
Jester, there's a combined run-1 and 2 significance for CMS but not for ATLAS (unless I missed something) which is at 3.9 with only run-2 data. Would it imply that when updated run-1 numbers are added it could go above 4.0?
ReplyDeleteWe've been running a question about the diphoton resonance (and the probability of it turning out to be new physics) at Metaculus:
ReplyDeletehttp://www.metaculus.com/questions/41/
It's been steadily trending up, but I have no idea how many predictions are from people who actually know the details of the experiments. Anyone willing to make their own prediction (in terms of probability percentage) is more than invited to do so! Extra points for giving your reasoning.
We also put one up about Backovic's 'ambulance chasing' paper's prediction that predicts the total number of citations of the ATLAS paper:
http://www.metaculus.com/questions/178/
it will be interesting to see if the new data bumps up the paper-writing rate and breaks Backovic's model.
RBS, yes, naive combination of run-1 and run-2 gives 4.3 sigma significance in the large width case. The true significance may be a bit smaller, but probably not much smaller.
ReplyDeleteYou know how sometimes people write physics papers about really wacky, outlandish ideas like chameleon fields and such? Do you think there will ever be a paper that examines the spin-1 case? What would be the reason that no physicist on earth would ever want to write a paper about an obvious topic that still hasn't been covered in teh literature? Seems like an easy way to rack up some of the almighty citations. I'm pretty much looking for a follow up on this:
ReplyDeletehttp://arxiv.org/abs/1211.2288
I know obviously it can't have spin-1 but I'm looking for some source material for a sci-fi story I might write.
In the current Standard Model we has SU(3)xSU(2)xU(1). Every attempt I know of for BSM derives this from some higher symmetry. Has anyone noticed the bottom up approach and looked for ways this can be built up from something more fundamental. We have a 1, 2, 3 here. I hypothesize a new strong force of SU(4) based on absolutely nothing except this observation if this bump really is a particle hiding a strong force.
ReplyDeleteThe fact that roughly the same bump appears in CMS and ATLAS at this significance suggests that while there is a real possibility that the signal is not real, suggested by that absence of a signal in other channels, that if it is not real, the source must also surely be some shared systematic error (probably either in the background expectation calculation or in the instrumentation) rather than a statistical fluke.
ReplyDeleteCan someone comment on the fact that local went up but global went somewhat down? If the data has more 'spikes' then this is more likely to happen somewhere (?) If I just look at the global significance, how can someone get excited??!!
ReplyDeleteThanks for the article explaining the latest results! ATLAS and CMS hyperlinks in the beginning of the article don't work (they have extra atlas: and cms:).
ReplyDeleteThanks! Fixed.
ReplyDelete>> If I just look at the global significance, how can someone get excited??!!
ReplyDeleteThe global significance of combined ATLAS and CMS is large, over 3 sigma before the Moriond update, and probably over 4 sigma after the update and including run-1. It can still be a fluctuation but it's a huge fluctuation, which justifies our excitement.
@andrew: how could you possibly get a systematic issue that produces a peak in the diphoton mass spectrum? The calorimeter elements do not know about invariant masses, so detector effects are ruled out (also by the tens of cross-checks done). A wrong background shape could have some influence, but you see the peak even by eye, without any fit - it can certainly not explain most of the excess.
ReplyDeleteWhen CMS and ATLAS published their Higgs-observation in the conference in 2012, they presented both somehting around 5.0 sigma. was this local or global significance, i.e. did these values take into account the LEE?
ReplyDeleteAs far as I understand, if one combines the statistics, it is only neccesary to take into account the LEE once, right? Something like: Oh, i have a fluctuation of global 3sigma at 750GeV; what does the other experiment have AT THAT SPECIFIC VALUE?
Thanks!
continuing the post above: In particular, what would be required to claim an observation? Two independent experiments need to see the same thing. That would mean, two experiments need to see a global fluctuation of >5sigma, right?
ReplyDelete@Andreas: The 5 sigma for the Higgs in July 2012 was the local significance, but at that level global and local significance become similar (because a fixed trials factor doesn't chance the sigma value so much for large significances).
ReplyDeleteTaking the LEE into account in combinations is actually a bit more complicated (e.g. what happens if the peak is not exactly at the same mass value, or how to combine different resolutions?), but as a rough estimate, taking just that of one experiment is not so bad. So yes, you can take the 1-2 sigma global significance from one experiment and ask "what does the other one see there?".
" what would be required to claim an observation?" - well, technically, whatever the experiments decide to do. As this is not a particle predicted by typical theories, I guess the experiments will be quite careful. If both experiments reach about 5 sigma local significance at compatible mass values in August (=extrapolation from the current excess strength), then the full 2016 dataset should prove the existence of something new beyond reasonable doubt. So my prediction, if (!!) that excess is actually new physics: "significant excess" in summer, "observation of new particle" end of 2016.
"@andrew: how could you possibly get a systematic issue that produces a peak in the diphoton mass spectrum? The calorimeter elements do not know about invariant masses, so detector effects are ruled out (also by the tens of cross-checks done). A wrong background shape could have some influence, but you see the peak even by eye, without any fit - it can certainly not explain most of the excess."
ReplyDeleteExamples, not exhaustive:
1. Both experiments may be relying on the same source for some component of the calculation of the expected number of background events that erroneously omitted a factor of two somewhere.
2. Due to a wiring or placement error, a hit on a couple of the photo receptors also triggers two neighboring photo receptors when the incoming photon has a high energy.
3. A broken magnet, while not preventing photons from being detected, throws off the calibration of some other part of the event detection system in a way that causes events that would have been removed from the data due to cuts based on the calibrated system to not be removed.
Of course, the usual course of events when a systemic error produces a seemingly way too large error, is that it is usually subtle and unexpected, because everyone has checked the obvious things. But, it can take a long time to work out the particular problem which once it is discovered looks really obvious and stupid.
1. There is no deeper calculation for the expected background. The background estimate is purely from a fit to the observed spectrum. And as you can see, it does fit, it is not a factor 2 too low.
ReplyDelete2. That would not lead to a peak in the invariant mass spectrum. It could mess around with the energies for the individual photons, but that does not produce a peak.
3. That does not produce a peak either.
There are hundreds of ways a pT spectrum of some objects can get wrong, but none of them produce a peak in the invariant mass of two isolated photons.
>> The global significance of combined ATLAS and CMS is large, over 3 sigma before the Moriond update, and probably over 4
ReplyDelete>> sigma after the update and including run-1. It can still be a fluctuation but it's a huge fluctuation, which justifies our excitement.
I'm sorry, can you do the algebra for me? Let's do with the narrow bump. I see CMS 1.6 sigma global (2012+2015) and Atlas (2015) 2.0 sigma global. How do you get 4+ combined? Are you assuming Atlas 2012+2015 is somewhat bigger?
In fact CMS 2015 only has a <1sigma global! I'm sure there are many other places where they see such bumps. (I can spot a few by eye...)
That being said, is there an example of such a 'large' combined sigma that went away in recent history? educate me, I'm young :)
You cannot combine the global significances. That would ignore that both peaks are at a very similar position. As rough estimate, you can combine the global significance of one experiment with the local significance of the other at the same place (because the first experiment already tells you where to look).
ReplyDelete"is there an example of such a 'large' combined sigma that went away in recent history?"
LHCb had a 3.5 sigma fluctuation in delta A_CP (covered in a previous blog post: http://resonaances.blogspot.de/2011/11/lhcb-has-evidence-of-new-physics-maybe.html ). Only one measured value, so no look-elsewhere-effect on the analysis level. With more data it just vanished.
I don't know what happened to 3.9 sigma dimuons, I don't rememer seeing updates: http://resonaances.blogspot.de/2011/07/d0-4-sigma-like-sign-dimuon-anomaly.html
There were some more deviations, but most of them in channels where systematic uncertainties are much more likely to pop up (and those usually have much larger tails than statistical uncertainties). OPERA's "6 sigma" could have been 100 sigma and they still would haven't seen new physics, but just a loose cable.
Thanks mfb!
ReplyDeleteIf I understood correctly:
a) At this point we may as well just quote the ATLAS local significance as our level of confidence (a 1.6 sigma global from CMS doesn't change much) namely between 3-4 sigma.
b) We have seen this level of (local) significance go away in recent history.
However, is there a situation where *two* independent measurements had a combined 4 sigma significancy, yet the signal went away? I've heard the wise talk about the 17 kev neutrino, leptoquarks and whatnots (e.g. DAMA)... But given the recent experience with the Higgs, isn't this a duck? Incidentally, how uncomfortable should we be with hte lack of counterparts? Z+photon?
thanks for the very educational blogging/discussions!
@anon: it doesnt matter whether the 4-sigma fluctuation is measured in one experiment, or is shared via two and combinined via combining their results.
ReplyDeleteAndreas, I agree that may be true statistically but clearly not sociologically speaking ;)
ReplyDeleteImagine ATLAS had ~ 4 sigma but CMS <~ 1. Would you be equally excited even if the combined could be about the same we have right now? After all, in such scenario, CMS could be seeing a downward statistical fluctuation, no? I am guessing many people would feel less comfortable being a real deal. I totally agree with you in terms of the numbers (we have a series of events who cares where are they coming from) but you know, this ambulance seemed to have much more traction than the previous ones, there's even a theory prediction out there for the number of papers :)
Well, combining two experiments can have advantages and disadvantages. In general it reduces the probability of a stupid mistake (-> OPERA cable), but you can get correlated systematics that are not covered in the oversimplified combination described above. Both should not matter here as the uncertainties are completely dominated by statistics, then the number of experiments is not relevant.
ReplyDeleteZgamma leptonic has a lower Z branching ratio*efficiency by more than a factor 10, I wouldn't be worried by the lack of events there. Zgamma hadronic has a better branching ratio but even worse efficiency and higher background. I don't have exact numbers, but it doesn't look like the Zgamma search would be sensitive (i.e. more than 1 sigma) even to a 1:1 branching ratio fraction compared to diphotons.
Thanks for very interesting blogging about possible new physics in LHC. I've been following Cern experiments since the first runs years ago (even before that accident w/ magnets that delayed everything by a year atleast) and this spring and the whole year 2016 will hopefully become as facinating and scientifically significant as the year when the Higgs discovery was announced. The sigmas look very promising for new physics in 750 GeV diphoton channel and let's hope that Atlas and CMS will be able to announce a discovery or at least 'significant excess' for new particle to open up new physics beyond Standard Model.
ReplyDeleteA recent example where multiple experiments saw a large signal that vanished after a careful analysis was the Pentaquark of 2002-2004.
ReplyDeleteJester, I am surprised about the rumor. How can 'looser cuts' increase the statistical significance of an excess by that much? Wouldn't one expect exactly the opposite?
ReplyDeleteAs I understand it, 'looser cuts' effectively increases the sample size by admitting larger number of events, and if the fluctuation is a real particle, increasing the sample size will increase the statistical significance of the fluctuation.
ReplyDeleteAvelino: this is indeed surprising (which is why ATLAS is very cautious). As far as I know, the main feature of the loose cuts analysis is that it allows more forward photons. Since the SM background is strongly peaked in the forward direction, and energy resolution degrades in the forward parts of the calorimeter, one would expect that the loose cuts would lead to a *loss* of significance. Apparently, the result is opposite... This could mean many things: one interpretation is that there's an experimental problem in ATLAS, and another is that the BSM signal is also strongly peaked in the forward direction, which would be the case for a higher-spin 750 GeV resonance. The problem with the latter interpretation is that there does not seem to be a forward enhancement of the signal in CMS.
ReplyDeleteThanks for the detailed answer, Jester. Let us wait (and speculate) then!
ReplyDeleteThe fact that the "loose cuts" ATLAS analysis boosts the significance of the excess may be caused by a high-degree of directionality in the photon beam. This, in turn, would suggest beam coherence, which is typical for radiation emitted by confined cavities (like quantum dots). This explanation stands only if CMS sees the same effect with a similar magnitude.
ReplyDeleteWait... rumors are now discussed as if they would be actual science. Okay, whatever. But taking the absense of very specific rumors as evidence that CMS does not see something? Come on.
ReplyDeleteAddendum: A higher significance with looser cuts (IF ATLAS has this, which is a big IF) is not unusual. If you know exactly what you are looking for, you tune the selection to give the highest expected significance - both loosening and tightening would then decrease the expected significance. The actual significance still has some statistical fluctuations, so it can be a bit higher in either direction. If you do not know in advance what exactly you are expecting, you'll rarely hit exactly the right spot for whatever pops up (if there is something). Maybe the signal is very clean and tighter cuts are better, maybe the signal has some messy other stuff going on and looser cuts are better.
ReplyDeleteIt's not surprising, but it's suspicious.
ReplyDeleteLooser cuts do not necessarily reduce the significance. But, usually, in data analyses, the cut levels are chosen in order to maximize significance. If their loose-cut analysis gives a higher significance, then why did they set up tighter cuts at default?
mfb, I agree, the secret ATLAS analysis could be just riding a random fluke (like the 115 GeV Wuon resonance in ATLAS that appeared after relaxing photon quality cuts). Concerning CMS, what I wrote is based on public data: if the BSM signal were indeed strongly peaked forward, one would expect a larger signal in their EBEE category, as compared to the signal in the EBEB category.
ReplyDeleteAnon, the cuts are chosen to maximize sensitivity to a simulated monte carlo signal, not to the one in the real data.
@Anonymous: If that excess is a particle, then it is certainly an unexpected one. The 13 TeV analyses were focused on 1.5+ TeV, with maybe a few signal events over a background of nearly zero. It is hard to tune cuts to things you do not even expect.
ReplyDelete@Jester: well, I don't know what exactly you heard as rumor, but in general selection and forward direction don't have a 1:1 correlation.
So according to Pauline Gagnon on Quantum Diaries, for the combined run-1 and run-2 data, the CMS global significance actually went *down* with this update (from 1.7 to 1.6 sigma), even as the combined local significance went up. How is that possible?
ReplyDeleteIs 20:1 a fair bet?
ReplyDeletehttp://www.science20.com/a_quantum_diaries_survivor/want_to_bet_201_that_the_750_gev_resonance_is_real-168606
Thanks Chris (re pentaquark).
ReplyDeleteXezlec, this is indeed hard to understand. Probably the two numbers 1.7 and 1.6 were obtained using different methods.
ReplyDeleteAnon, 20:1 is fair, but what would I do with all these wine bottles? If vodka is on the table then I could take the bet.
"Probably the two numbers 1.7 and 1.6 were obtained using different methods."
ReplyDeletecould it be they got more data that is more 'spiky'? If that's the case it'd be more 'likely' to have a bump somewhere, reducing the global number? Otherwise this makes no sense...
If there are multiple resonances there, how long would it take the get enough data to distinguish them? Is the next year likely to be enough?
ReplyDelete@Global significance: They added more data and changed the analysis a bit, which could give more ways to look elsewhere.
ReplyDelete@20:1 bet: It just goes in the wrong direction. If it is a new particle, wine for celebration won't be an issue, and if it is not a new particle, you even lose some wine. Related xkcd comic: https://xkcd.com/955/
@Anon, multiple resonances: Depends on the resonances and their decay modes. Different spin and different spin-sensitive decay modes would make it relatively easy (different decay modes then show different angular distributions). Different decay modes and different width could work as well. Different spin and the same decay modes would give a notable effect with a larger statistics, probably the run 2 dataset or something like that. Same spin, width and decay modes: uh oh. Some theory model might be first, if it predicts various properties of the observed resonance(s) properly.
Any thoughts on Motl's suggestion of a Zy bump at 350 GeV in ATLAS and CMS data, albeit not very pronounced.
ReplyDeletehttp://motls.blogspot.com/2016/03/z-gamma-excess-near-375-gev-in-new-and.html
"Any thoughts on Motl's suggestion of a Zy bump at 350 GeV in ATLAS and CMS data"
ReplyDeleteMotl sees a bump that will probably reveal supersymmetry about twice a week. Executive summary: he's full of it.
andrew, also note that a comment on that article pointed out that 375 is 3 x 125. So we may have a resonance at 3x Higgs and another at 6x (and possibly another at 12x). Can that mean something or are we numerologists at this point?
ReplyDelete"340 is close to 375", seriously? That is several standard deviations away. ATLAS 13 TeV has some excess at 340 and absolutely nothing at 375. CMS 8 TeV has some excess at 375 and a deficit at 340 GeV. Clear evidence of ... what?
ReplyDeleteAdam you're day-dreaming ;-).
ReplyDelete"In summary, the diphoton excess survived the first test. After adding more data and improving the analysis techniques the significance slightly increases rather than decreases, as expected for a real particle."
No they didn't! What counts is clearly the global significance, right?! The global significance remained the same for ATLAS and decreased for CMS, as expected for a random fluctuation.
"The signal is now a bit more solid: both experiments have a similar amount of diphoton data and they both claim a similar significance of the 750 GeV bump."
No they don't. Wile ATLAS claims a moderate global significance of 2 \sigma, CMS now claims a global "significance" of 1 \sigma (down from 1.2 \sigma). Haha.
Both experiments see the strongest fluctuation at nearly the same (and compatible) masses. You cannot just look at global significances, that would ignore this point completely. We probably won't get an official combination as waiting for more data is better, but you can have a look at the unofficial combinations: the combined global significance exceeds the local significances of the individual experiments.
ReplyDeleteIf this di photon signal is real,could this be a composite particle? Can you have a Boson made of more fundamental particles, the Higgs appears not to be. Could this be another scalar Boson?
ReplyDeleteJester, these are all big and bigger ifs, but suppose (fingers crossed) there's indeed a new force and possibly more in store at higher energies. So we keep finding new force carrier bosons but the number of fundamental fermions remains the same: three generations each of leptons, neutrinos and quarks. Would such a pattern, if the trend continues cause at some point a real tension with supersymmetry - or are there reasons why most fermions would end up with very large masses while bosons spread over lower energy spectrum?
ReplyDelete@mfb
ReplyDeleteI am not denying what u wrote. But the question Adam's post addresses is different, it is: "did it get better with the new analysis or worse?" And to answer this question u shouldn't just
stare at the local values, you have to take into account how the background changed, too.
Therefore clearly you have to consider the global values here, and they got worse...
@Anonymous: A new state as composite particle has been discussed in the blog already. Yes, it is possible. Mesons are an example for non-elementary bosons, and something meson-like is indeed a candidate for this particle (if it is a particle).
ReplyDelete@Maurice: Global values are always a bit arbitrary. Take the search range, for example: do the experiments search for a peak in the range of 500 GeV to 1.5 TeV (where a graviton-like signal would make some sort of sense), or in the range of 200 GeV to 6 TeV (where a peak could appear in general)?
ReplyDeleteDo you just search for narrow resonances, or do you search for broader ones as well (and if yes, up to which width)?
Those choices are completely arbitrary, but they will have a strong influence on medium-sized global significances.
I am not refusing what you said my question is that did i get better with new analysis?? And to answer this question you shouldn't look at the local values but you need take into account global values also......
ReplyDelete