Higgs: evidence growing stronger

The news of the day is the update on Higgs searches released by the LHC and Tevatron at the Moriond conference. The most important updates are:

• The significance of the low mass Higgs excess at the Tevatron increased much more than expected, given a tiny amount of data added since the previous combination. As explained at more length in my previous post, most of the steam comes from the improved b-tagging efficiency in CDF, which greatly helps to pick the H → bb signal. All in all, the Tevatron now reports a ~2.5 sigma excess in the interesting mass region between 120 and 130 GeV, although the poor mass resolution does not allow them to clearly favor any particular mass in that range. The excess is somewhat larger than predicted by the Standard Model Higgs, with the best fit value around 1.5 standard cross sections, however the standard cross section is within the 1 sigma band.
• It is also interesting to have a look at the separate H → bb search channel where for a few more months the Tevatron is still superior to the LHC. In that channel alone the excess is even larger than in the full combination, corresponding to a ~3 sigma local significance. You should not pay too much attention to the fact that the best fit cross section on the right peaks at 135 GeV. Due to the poor mass resolution, 120 and 135 GeV Higgs looks about the same to the Tevatron, however the heavier Higgs the smaller the predicted event rate so one needs a larger boost to explain the same number of observed event. This can be better seen on the CDF plot of the p-value on the left: the significance of the excess over the background is about 3 sigma for any Higgs mass hypothesis between 125 and 135 GeV. What is more interesting is that near 125 GeV Higgs the observed signal strength is noticeably higher than the one predicted by the Standard Model, with the best fit around twice the Standard Model cross section. And that is mildly exciting.
• ATLAS updated more search channels with the full 5fb-1 dataset (in December they updated only the 2 most sensitive γγ and ZZ→4l channels). Even if neither of the new WW, bb and ττ channels taken separately has a sufficient sensitivity yet, combined they do have some impact. Thanks to that, they are now able to exclude at 95% confidence level the Higgs boson up to 122.5 GeV (except for the probably irrelevant blip at 118). Together with CMS, they are leaving only a small allowed window between 122.5 and 127.5 GeV. We're almost there! From now on if anybody mentions the look-elsewhere effect the correct response is "what elsewhere?" ;-)
• Amusingly, the new ATLAS channels have not only shrunk the allowed Higgs mass range, but also decreased by almost 1 sigma the combined significance of the excess at 125 GeV. You can see an example in this table that shows the expected and observed number of events in different categories of the H → WW → 2l2v search channel. Some categories show a small excess, some a small deficit, but nothing jumps in your face although it might have. More data will tell if this is just bad luck on the part of ATLAS, or something interesting is going on...
• Although CMS fired most of their guns in December (having updated most channels to the full 5fb-1 statistics back then) they keep making small improvements in the analysis. One noticeable improvement presented today is in the diphoton search channel, thanks to a new fancy multivariate analysis whose expected sensitivity (red) is much better than that of the previous cut-based analysis (blue). Thanks to that the diphoton signal around 125 GeV grew a bit stronger and now corresponds to a p-value of almost 3 sigma. At the same time they are able to exclude the Higgs mass in the windows of 117.5-­120.5 GeV and 128.5-­132.0 GeV.
• CMS also presented a nice graphics showing how the 125 GeV Higgs signal splits between different search channels. We see that we're getting a bit too much signal in the diphoton channel, and a bit too little signal in the WW and ZZ channel, but otherwise there is a decent agreement with the expectations from the Standard Model Higgs. If ATLAS provide an analogous plot it would look similar, except they have a somewhat stronger signal in ZZ and weaker in WW. The LHC has does not yet have enough sensitivity to the Higgs decaying to fermions (unlike the Tevatron), and thus the b-bbar and tau-tau channels currently do not provide any meaningful insight.

For more reports on today's Higgs updates, see Tommaso, Philip, Matt and again Matt.

Higgs: teaser from CDF

The Moriond conference is now unfolding up on the slopes of Mont Blanc. The real action -- the Higgs update from the LHC and the Tevatron -- will happen only on Wednesday. However, as in any other 007 movie, there is a teaser before the opening credits. On the previous week another conference known as the Italian Moriond took place in the same room of the same hotel on the same ski slope. It included the talk by Homer Wolfe that uncovered many details of the upcoming CDF Higgs analysis, although stopping short before the final shootout, that is before giving away the new Higgs limits.

There hasn't been much expectation about the upcoming Tevatron Higgs update . That's because the previous Tevatron combination is based on up to 8.6 fb-1 of data and doesn't show much in the interesting mass range: only a broad 0.5 sigma excess near 125 GeV and the limits on the Higgs production rate at the level of twice the Standard Model cross section. Naively, adding 1 more crappy femtobarn should not change much, unless dramatic improvementd in the analysis. have occurred.

Now, CDF claims that they do have a dramatic improvement!

The most revamped analysis is the one where the Higgs boson is produced in association with a W or Z boson and decays into a pair of b quarks. Much as at the LHC, W/Z+H is not the main production channel at the Tevatron. However, the charged lepton and/or missing energy from W/Z decay give a chance to trigger on the event, and thus pick the H → b bbar signal from the overwhelming QCD background. Luckily enough, W/Z+H is also the last channel where the Tevatron is still highly competitive with the LHC (the latter has to fight a much larger QCD background in this channel, and only later this year it will reach a sensitivity comparable to that of the Tevatron). The news is that between now and last summer CDF has largely improved their b-tagging, with the efficiency gains reaching 40%. This is of double importance in low mass Higgs searches, as signal events contain two b-jets. Combining that with other smaller improvements, CDF expects to see twice as many signal events in the interesting mass range compared to the previous analysis. Correspondingly, the expected sensitivity for exclusion is now at about 1.4 Standard Model cross sections near 130 GeV, tantalizingly close to where the signal should pop up.

Actually, CDF went as far as to showing the distribution of the invariant mass of the b-jet pairs in the events that pass the latest selection in the W+H channel. In theory, that distribution should have a peak at the true value of the Higgs mass, although the peak will be badly smeared due to large uncertainties inherent in measuring the energy of b-jets. One thing that is clear from the plot is that the Tevatron won't teach us anything new about the Higgs boson mass: any excess is expected to span many mass bins and it may be hard to tell the 115 GeV signal from the 130 GeV one. Another thing that is visible, at least to a naive eye, is that the plot shows.... an excess of Higgs-like events in the interesting mass range. Without an insider analysis it's of course impossible to say what that really means in terms of the signal significance, but it's not impossible we will be pleasantly surprised on Wednesday...

CDF confirms CP violation in charm

The most interesting non-Higgs results coming from the LHC in 2011 was the observation CP violation in neutral D-meson decays delivered by the LHCb collaboration. Recall that LHCb studied the decays of the D and anti-D mesons into a pair of charged pions and a pair of charged kaons. For each D → π+π- and D → K+K- one can measure the decay rate, and subtract the analogous decay rate of anti-D (and normalize by the sum), thus obtaining the asymmetries A(π+π-) and A(K+K-). If either of the asymmetries is non-zero, that means the particular partial decay width is different for the particle and the anti-particle, which is a signal of CP violation. Since the total number of D and anti-D produced in LHC collisions is different and hard to estimate, LHCb quoted only the difference of the asymmetries ΔA = A(K+K-) - A(π+π-) where that uncertainty cancels out. They found ΔA = (-0.82 ± 0.24)%, significantly away from zero, which constituted the first evidence for CP violation in the charm sector at the level of 1%. What distinguishes this result from 1000 other flavor physics observables presented by the LHC, Tevatron, and B factories is that, until last year, theorists expected Standard Model processes to provide much smaller contributions to the CP asymmetry in D meson decays, estimates lying in the ballpark of 0.01-0.1%.

Exactly the same processes can also be studied at the Tevatron. CDF measured the same observable last year even before LHCb (they also gave separate CP asymmetries for the π+π- and K+K- final states), but at the time no statistically significant CP violation was observed. Now CDF has an update using the full Tevatron data set of almost 10 inverse femtobarn. They find ΔA = (-0.62 ± 0.23) % which is 2.7 sigma away from zero. This is an independent confirmation of the LHCb result. Combining all experimental measurements of CP violating observables in the D meson sector one arrives at the following graph
One can read off that the world average of the direct ΔA = (-0.67 ± 0.16)% is clearly non-zero, while there's currently no hint of the so-called indirect CP violation in D-Dbar meson mixing. Thus, we now know for sure CP is violated in decays of D mesons, an interesting fact in itself. However, we're still in the dark whether this result is a signal of new physics beyond the Standard Model, or just an unexpected enhancement of some difficult-to-calculate Standard Model contributions. I wish I could say that more data will clarify the situation, but even that is not clear at the moment...

See also the CDF talk in La Thuile.

Other Neutrino Anomalies

Many a time I have stressed on this blog that neutrinos are boring, though I should specify that they are boring from the point of view of a theoretical physicist. For experimentalists, on the other hand, neutrinos are first of all annoying. Indeed, taking part in a neutrino experiment seems the shortest path to trouble, because of weird anomalies affecting every other experiment. Although the most notorious anomaly - the phantom of OPERA - appears to be dead and buried, there remain several spooky effects observed by other experiments. Here is a list of the most interesting neutrino anomalies that to date have not been explained away.

LSND anomaly. The mother of all neutrino anomalies. The LSND experiment in Los Alamos was shooting a beam of 20-200 MeV muon antineutrinos and looking for an appearance of electron antineutrinos in a detector 30m away. They saw a clear excess of electron antineutrinos which at the time was being interpreted as the first evidence of neutrino oscillations. However, the LSND result would correspond to a neutrino mass difference of nearly 1eV, whereas subsequent experiments (Super-K, KamLAND, SNO) demonstrated that, although the 3 Standard Model neutrinos can indeed oscillate one into another, the mass differences are much smaller than suggested by the LSND. The LSND result thus cannot be explained with 3 active neutrinos, although it may fit in a larger framework with one or more sterile neutrinos.

MiniBoone anomaly. Conceived as the ultimate test of the LSND anomaly, with a larger baseline (540m), larger neutrino energies (0.5-3 GeV), but a similar L/E. The Miniboone experiment in Fermilab was also looking for an appearance of electron neutrinos in a beam of muon ones, moreover it had a button to choose between neutrino and antineutrino beams. One clear result was a refutal of the simple 2-neutrino oscillation interpretation of the LSND signal, as nothing was seen in the signal region in the neutrino mode. However, the experiment saw a weird excess of electron-neutrino-like events at low energies, in the neutrino mode below, and in antineutrino mode just above their analysis threshold. Again, there may be a more complicated framework with several sterile neutrinos where these results can be accommodated. In preparation is the MicroBoone experiment which will be the ultimate test of the MiniBoone anomaly.

Gallium anomaly. The least known of the neutrino anomalies. The GALLEX and SAGE collaborations were gallium-target radiochemical experiments focused on studying solar neutrinos. For calibration purposes, they measured the flux of electron neutrinos produced by artificial radioactive sources placed inside the detectors. They found the observed-to-expected flux ratio to be 0.86 ± 0.06, as if some of the neutrinos were vanishing by oscillating into sterile ones.

Reactor anomaly. An oven-fresh neutrino anomaly based on ancient experimental results. Over the years several detectors located near nuclear power stations all over the world measured the flux of electron antineutrinos emitted from the reactors. At the time none of them noticed anything unusual, on average the flux was maybe 1 sigma below the expected one. However, a recent progress in modeling nuclear reactions led to a reevaluation of the neutrino energy spectrum and emitted power, which increased the expected flux. Therefore, the short-baseline reactor neutrino experiments now face a deficit of the neutrino flux: the observed to expected rate is 0.943 ± 0.023, roughly a 2.5 sigma discrepancy. Again, are electron neutrinos rapidly oscillating into sterile ones, thus vanishing from our sight? The reactor anomaly will be soon tested by an experiment with the ominous name Nucifer, located 7m from the reactor in Saclay.

As you can tell, the neutrino anomalies described above are not as spectacular as breaking the speed of light, but for exactly this reason there is a better chance that at least one of them is trying to tell us something about the real world… Are they signals of unexpected new physics, or just a manifestation of difficult to control systematic effects inherent in neutrino physics? The latter is of course far more likely, but nevertheless it will be interesting to follow the updates.

I borrowed heavily from a seminar of Guillaume Mention, slides here.

More on the stop story

Last Tuesday a number of new exclusion limits for 3rd generation superpartners by ATLAS collaboration was made public. To a casual observer, the seminar presenting the results may have resembled a reenactment of the Saint Valentine's day massacre. Nevertheless I will argue here -- more out of contrariness than conviction -- that SUSY may be battered, bruised, covered in slime, and badly bleeding, but formally she is not yet dead. For most particle theorists, the death will be pronounced scalar partners of the top and bottom quarks are excluded up to ~500 GeV, the precise threshold depending on taste, education, and diet. That's because excluding light stops and sbottoms is going to squash any remaining hopes that supersymmetry can address the naturalness problem of electroweak symmetry breaking, for which purpose it was originally invented. However, the LHC collaborations have yet not presented any robust limits on the stop masses. Instead, here's what the LHC have taught us about SUSY so far:

• Since summer 2011 we know that in a generic case when all colored superpartners including stops have comparable masses and decay to the lightest supersymmetric particle (LSP) producing a considerable amount of missing energy then stops have to be heavier than about 1 TeV.
• Since last Tuesday we know if the only squarks below TeV are those of the 3rd generation then gluinos have to be heavier than about 600-900 GeV, depending on the details of the supersymmetric spectrum.

An example of the recent ATLAS bounds is shown on the right. The assumption here is that the gluino decays via an intermediate virtual stop into 2 top quarks + LSP. In that case, combination of searches in the 2 same-sign lepton channels, and the 1 lepton + 1 b jet channel excludes gluino masses all the way up to 700 GeV, a bit more if the LSP is nearly massless. Even stronger bounds are reported assuming gluinos decaying via a virtual sbottom squark into 2 bottom quarks + LSP. In that case gluino masses up to 900 GeV are excluded, see below.


However there still exists one important loophole. One can imagine that gluinos are very heavy, and only the 3rd generation squarks have masses below 1 TeV. In that case, the 3rd generation squarks have to be produced directly in gluon-gluon collision without a help from intermediate gluinos, in which case the production cross section is suppressed. Strangely enough, limits on direct stop production are currently almost non-existing. As explained in my previous post, there exist some theorist-level limits derived by reinterpreting other LHC searches which claim that stops need to be heavier than 200-300 GeV, depending on how they decay. Thus, for all we know, the stop might have a similar mass to the top quark without having been noticed at the Tevatron or LHC! Plugging this loophole is now the most important issue in SUSY phenomenology. For the direct production of sbottoms, who should come in package with stops due to the SU(2) symmetry of weak interactions, the current experimental situation is a bit better. An ATLAS search published last December excludes sbottoms below ~400 GeV for a very light LSP, although this search completely loses steam when LSP is heavier than ~100 GeV . Hopefully, we will learn much more at the Moriond conference 2 weeks from now.

To summarize, the year 2012 will surely go down in history as the Higgs year. But among more studious future historians of science it may also be remembered as the year when SUSY, at least in its natural form, was finally laid to grave...

See the new ATLAS notes here and here for more details on the searches for gluino mediated stop production. See also Peter's blog.

How to find a stop

Lately there's been a surge of interest in hypothetical scalar partners of the top quark, the stops in short. So it may be a good moment to sell a few technical details to a larger audience. For theorists, a stop is easy to distinguish experimentally: it looks like a top but with a twiddle on top. However experimentalists are not as smart, and they have to invest much more time and effort in order to identify stops at the LHC.

What it looks like depends first of all on how it decays. Even the minimal SUSY model offers countless possibilities. Leaving out the case of stable stops, in the MSSM stops ultimately decays to a number of known particles from the Standard Model plus the lightest supersymmetric partner (LSP) who is assumed to be a very weakly interacting particle showing up as missing momentum in a detector. Some possible decay chains are:
Stop → top + LSP, Stop → W + sbottom → W + b + LSP, Stop → bottom + chargino → bottom + W + LSP, etc.
The bottom line is that the MSSM stop should manifest itself at the LHC as an excess of events with:

• top and/or bottom quarks,
• significant missing energy due to the LSP.

Now, how to produce it. Being top partners, stops carry a color charge, hence they can be produced in gluon collisions which are easy to come by at the LHC. However, on the plot you see that production of stop pairs is far less frequent than that of gluinos and 1st generation squarks of similar mass. In physics jargon, s-channel production of scalar particles is velocity suppressed as a consequence of angular momentum conservation. This is the main reason why, as you'll see below, the LHC limits on stop pair production are so much weaker than those on gluinos. However, there is a trick to boost the stop production rate by producing them indirectly in gluino decays: gluino → stop + top, as long as the gluino is not much heavier than the stops. As a bonus, this production mode generates more junk in the detector that could be helpful in discriminating signal from background. For example, one can imagine the sequence:
pp → 2 gluinos → 2 stops + 2 tops → 4 tops + 2 LSPs
which leaves us with 4 top quarks in the final state. The 4-top production rate in the Standard Model is very small, therefore observation of such a final state at this point would be a clear sign of new physics. Another place where this sort of cascades could show up are the searches for same-sign top quarks.

What is the experimental situation so far? As far as I know, the LHC collaborations have not yet published any limits on direct stop production. On the other hand, gluino mediated stop production was targeted in this note based on 1 fb-1 of ATLAS data. The plot shows that gluinos decaying in the sequence:
gluino → top + stop → 2 tops + bottom + chargino→ 2 tops + bottom + W + LSP
cannot be lighter than 500 GeV. During the next 30 days leading to the Moriond conference many more searches based on larger data samples will be released, starting with the Valentine Day ATLAS talk.

For today, we can get some idea of the current LHC sensitivity from this paper, which compiles a large number of SUSY searches and recasts the results in terms of limits on stops. The LHC reach for direct stop production (right plots) is poor, corresponding to stop mass of only 200-300 GeV. For gluino mediated stop production (plots below) the limits are much better and extend to approximately 700 GeV gluino masses (though the precise limits may depend on details of the SUSY spectrum; it is probably possible to design spectra for which these limits are somewhat weaker). Amusingly, the dedicated ATLAS search does not seem to be the most sensitive probe of gluino mediated stop production. Instead, more stringent limits come from vanilla SUSY searches (decaying top quarks produce jets, b-jets, and/or leptons that can be picked up by these searches). We'll see very soon whether the coming experimental analyses will significantly improve these limits.

Update: see the slides of the Feb 14 ATLAS talk for more limits on stops.

Higgs: stronger and more exciting

Today the CMS and ATLAS collaborations dumped into public pages a dozen publications describing the Higgs searches in the 2011 LHC data. In the first approximation, these are the same results that were presented on December 13. But there is one surprise...

The CMS collaboration had every reason to think that life was unfair. For the last round of Higgs searches they made significantly more effort and analyzed more possible signatures than ATLAS. The latter updated only 2 channels to the full dataset, and in principle had worse sensitivity in the H→ZZ*→4l channel (due to slightly higher pT thresholds in the analysis). In spite of that, the significance of the Higgs-like excess near 125 GeV was much weaker in CMS than in ATLAS. Naturally, the CMS researchers have spent the last 2 months scouring their drawers for strayed Higgs events. And they found.

New interesting events are reported in the H → γγ channel. Compared to the December 13 presentation, CMS added a new category of events which, apart from 2 photons, contain 2 energetic jets in the forward (closer to the beam pipe) region of the detector. Such events could arise in the so-called vector boson fusion (VBF) process, where each of the 2 colliding quarks emits a W or Z boson which coalesce to create a Higgs boson (right graph). The 2 original quark get deflected and may be seen in a detector as two forward jets. On the other hand in gluon fusion (left graph), which is the dominant Higgs production process at the LHC, the 2 colliding gluons "vanish" and the final state rarely contains 2 forward jets. Also background processes are less likely to produce 2 photons in association with 2 such jets. Hence, by selecting diphoton events with 2 forward jets we can probe a distinct Higgs production process, with less signal (the VBF cross section is 10 times smaller than the gluon fusion one), but also with less background.

Now, in the VBF class CMS finds 7 diphoton events in a 1-GeV bin at the invariant mass 124 GeV, where only about 2 events would be expected from non-Higgs background. By itself it would be nothing, but together with the rest of events in the diphoton and 4-lepton channels it provides another support for the existence of the Higgs boson in the mass range 124-126 GeV. All in all, the local significance of the excess near 125 GeV in the combined CMS analysis is now over 3 sigma, very similar to that of ATLAS. While many small improvements have been made, my feeling is that significance was pumped up mostly by these additional VBF events.













But there is something intriguing here. Now in both ATLAS and CMS the best fit of the Higgs rate in the H → γγ channel alone is about twice the Standard Model rate, with the standard rate being over 1 sigma away in both cases. Actually, with the present amount of data CMS would not expect to see any VBF events, as rate predicted by the Standard Model is too small (see below that in this channel their fit is almost 4 times the standard rate, although with a large error). Could it be that we're seeing a non-Standard-Model Higgs boson with an enhanced decay rate and/or enhanced VBF production cross section? Of course, it is far too early for jumping into conclusions: the errors are still large and we may easily be observing an upward fluctuation. Besides,the combination of all channels doesn't show any dramatic enhancement of the Higgs rate. In any case we're free to speculate while waiting for more data (and a word from ATLAS on the VBF events).















See also Matt for more details and more caution.