Higgs limits at the Tevatron: <1.4 x SM !! August 23, 2007Posted by dorigo in news, physics, science.
With pleasure I visited the public web page of the CDF experiment today, finding a fresh new combination of most of the searches for Standard Model Higgs bosons ongoing at the CDF and D0 experiments. And it is an exciting piece of news!
As I discussed recently here, CDF has produced a combination of its results. D0 did the same, and then, with a technically very complicated but conceptually very simple operation, the two results have been combined.
So let us have a look at the combination, and then discuss it briefly. It is shown in the plot below:
As usual, the y axis shows the 95% CL limit in units of Standard Model cross section for Higgs production, for each value of the Higgs mass (the x axis). A limit at a value y=2.5 for a value x=145 GeV means that the Tevatron experiments exclude, with 95% confidence, that the Higgs boson has a production cross section at the Tevatron larger than 2.5 times the SM prediction IF its mass is 145 GeV.
The yellow band (yellow has become a classic for LEP exclusion regions) “cuts away” the part of the plane where the Higgs cannot be, having been excluded by the LEP II experiments.
Then there are the results. Let us forget about individual expected limits by CDF and D0 (red and blue dashed lines) and concentrate instead on the black ones. The dashed black line shows the 95% CL limit that the Tevatron experiments were supposed to be able to set, given the datasets analyzed in each analysis and every small feature of each. You get that black dashed line only after combining the results of combinations of pseudoexperiments, taking care of nuisance parameters, integrated luminosities, cross correlations, correlated efficiencies, statistical effects of any kind, and other devilish details. But that is only an expected limit: it tells you what is the real power of the experiments.
The real limit is the one shown with the continuous black line. And here we get to see that, had the Lepton-Photon conference (for which this plot was finalized) been held in october or november this year, the Tevatron could have had a chance to start crying “No higgs at 160 GeV” on newspapers throughout the world. In fact, the integrated luminosity of the analyses most contributing at 160 GeV (H->WW decay searches) were of about 1.9/fb in each experiment, but both CDF and D0 have about 60% more data already in their pockets, and only lacked the time to process them, validate them, and include them in the analyses.
The Tevatron has been a bit lucky there though: as you can see in the plot, the limit set at 160 GeV is about twice as good as the one they expected to set. Lucky ? A debatable point indeed. However, the first point to bring home is definitely that the CDF and D0 experiments are starting to bite in the Higgs hunt business.
Then, you get to see something strange. What is going on at 120 GeV ? The limit there is twice as bad as pseudoexperiments predict!
That fact only highlights the amount of variability of the results that the Tevatron can display, with respect to their “expected limits”. Indeed, the thin dashed black line (and the other dashed lines too) is a misrepresentation of reality: instead of a thin line you should be looking at a wide band – a 1-sigma contour of expectations. The lack of the band is certainly only due to lack of time and hurry in producing the plot in time for summer 2007 conferences, and I think this plot will get more detail in a month or two. In any case, in the absence of the band, I am unable to speculate on how likely it was to get such a lousy limit at 120 GeV if the Higgs was not there.
Indeed, a unwitty soul could be looking at the discrepancy between dashed and continuous black lines above the mark at 120 GeV and claim that what the Tevatron experiments are observing at 120 GeV is exactly what they would be seeing if there was a Standard Model Higgs boson produced with a larger-than expected cross section sitting right there. Because the lousier limit there means that the experiments saw more candidate events at 120 GeV than they expected to find from background sources…. An excess ? No way! A fluctuation, and be sure, a mild one.
When the +-1-sigma bands of expected results will be released, we will be able to quantify just how unlucky – or unlikely – the result is at 120 GeV, if the Higgs boson is not there. In the meantime, the second point to bring home from the plot is that that 95% CL line at 120 GeV is still at 8.5 times the expected Higgs cross section.
That underlines how hard the Higgs boson search is at low mass – something the LHC experiments have already started to dread. The ATLAS and CMS experiments, I wish to remind you, will be in great embarassment if the Higgs mass happens to be just a few GeV higher than what LEP II excluded. And that not because they would be taking data next year only by virtue of having shut down the LEP II experiment only months short of discovering the Higgs boson: but because the Higgs at 120 GeV is tough, very tough, and at the LHC life is even tougher than at the Tevatron. Proof be the wagons of money the experiments have invested in exquisite electromagnetic calorimeters, to detect a blip in the diphoton invariant mass from the H->gamma gamma decay!
UPDATE: in a comment below, an anonymous visitor asks why the actual limit at 120 GeV is so much worse than the expected one, given that the combination by CDF (which I posted a few days ago) is not terribly above their own expectation. I mentioned that the combination of limits is a tricky business, and that one could not really guess until one had all the data available. So here is the D0 combination by itself, the missing piece:
One clearly sees that D0, too, got a larger exclusion at 120-140 GeV than they expected. If you combine this with the CDF limit, I have no reason to doubt that the combination will be worse than expectations in that range.
But let us look at numbers a bit more carefully: at 120 GeV, D0 expected to set a limit at 7xSM and got 10xSM, and CDF did exactly the same: 7xSM expected, 10xSM obtained. At 140 GeV, though, D0 expected 7.5xSM and got 10xSM, CDF expected to set limits at 7.5xSM and actually got 5.8xSM (or so I read off the plot – I am too lazy to fish out the numbers from internal CDF documentation, and I would still be unable to diffuse it here!).
What we get is that a combination of those numbers yields, according to the CDF-D0 group who did the exercise, at 120 GeV a 5xSM expected limit and x8 obtained, while at 140 GeV a expected 4xSM and an obtained 3xSM. Do these numbers make sense ? I think that simple statistical arguments justify those figures, and I thus have to dismiss the rather annoying hypothesis put forward by the commenter that “for the _expected_ overall curve, WHlvbb accidentally used its 68% CL expected, rather than 95%, in its contribution to overall, resulting in a lower than intended overall expected curve“.
UPDATE II: as I posted the above update, a suspicion arose in my mind. Indeed, the numbers for 120 GeV make more sense to me than those at 140 GeV, when a 10xSM combined with a 5.8xSM yield a 3xSM – suspicious. So I take back what I wrote above for the case of the point at 140 GeV. Rather, let us give a closer look at the “point” at 140 GeV in the combination plot, which I paste here magnified for your convenience:
One observes something weird: while there appear to be kinks in the red and blue dashed lines at 140 GeV (the red one is evident, the blue one less so, but still is there), one sees no kink for the black dashed line at the same abscissa.
A suspicion might arise that the black dash was only computed at 130 and 160 GeV, and a straight line drawn between those points. That would explain the slightly better-than-expected combined result at 140!
I will investigate.
UPDATE III (yes, they keep coming): I just revised the title to “1.4xSM”, since that is the actual value at 160 GeV.
UPDATE IV: I checked the issue with the Higgs conveners in CDF and basically, what I can do here is mention a few facts:
- the combination is preliminary
- it is based on preliminary results from the individual analyses
- a better combination, with 1- and 2-sigma bands on the expected limit, and with more data points will be produced soon.
All the above is to say that there is no real mystery, and that one should be looking at the combined plot with a grain of salt. Let us wait for a better version of the plot, which is red hot from the press ….