New Higgs limits from the ZH final state by D0 November 5, 2007Posted by dorigo in news, physics, science.
Another nice Higgs search result was just produced by the D0 collaboration last week. The analysis is an update of previous results, based on more data – just short of one inverse femtobarn, 0.93/fb to be precise – and a more refined experimental technology. It involves looking for associated production of a Higgs boson and a Z boson, where the former decayed to a b-quark pair and the Z boson decayed to two neutrinos: .
Why two neutrinos ? We all know that observing the decay to charged leptons of the Z boson is much, much simpler: two electrons (E is energy)are easy to detect at the Tevatron, because almost nothing mimicks that signature; and two muons (P is momentum) are just as easy to spot. Even the decay can be dug out of large backgrounds, but neutrinos ?
Neutrinos disappear from the interaction region leaving no trace: the probability of yielding an interaction with the hadronic matter of the D0 detector is in fact ridiculously small. And indeed, nobody has ever seen a process in hadron collisions, because – well – you have nothing to detect. However, if the Z boson is produced together with something else – a hadronic jet, for instance – you can infer its decay to neutrinos by the large amount of missing transverse energy the pair leaves behind.
The concept above is more subtle than I managed to make it sound. When physicists talk about Z production at a hadron collider, what they mean is a collision between a quark from a proton with its exact antiparticle from an antiproton, which annihilate into a Z boson. This is a very well studied process called Drell-Yan mechanism after their first investigators. Of course, the remainder of the two projectiles plays no role for the hard collision, and these fragments produce what we call underlying event – a soft flux of energy rather uniform in the detector (if viewed as a function of a quantity called pseudorapidity – stuff for another post).
However, the picture above is what we call a “Leading Order” approximation of the process, which in reality is something more complex. Indeed, the initial state quarks before annihilating may (and usually do) emit QCD (quantum-chromo-dynamical) radiation, which manifests itself as hadronic particles detected somewhere in the detector, balancing in the plane transverse to the beam direction their total momentum with that of the Z boson, as conservation of impulse dictates. When QCD radiation is strong enough, it produces real jets of hadrons, otherwise it just manages to materialize into a few soft hadrons. Only in the former case, if the Z decays into neutrinos, one sees a unbalanced energy flow in the transverse plane, and with some fantasy can then infer that a Z boson was produced, decayed to neutrinos, and left a momentum imbalance – missing transverse energy!
Computing with precision the differential spectrum of Z boson transverse momentum is something very difficult, since it involves complex calculations involving infinite sums. A technique (soft gluon resummation) to sum the dominant terms in these infinite sums has been developed in the last fifteen years or so, and the contribution of multiple soft gluon emission from the initial state of Drell-Yan production (and similar other processes) is now very well understood. Check out, for instance, how D0 understands the transverse momentum spectrum of Z bosons (blue points) with the RESBOS (red curve) calculation below.
Ok. I think I explained how a single Z boson decaying to neutrinos is hard to see unless it recoils against something visible. Now, what happens when the recoiling particle is not a bunch of gluons but a single Higgs boson ? Of course, one sees the Higgs decay products recoiling against something which leaves missing transverse energy in the detector. The Higgs deacy products that D0 looks for in the analysis discussed here are a couple of b-quark jets. That is indeed the most probable final state of Higgs disintegration if the mass of the boson is below about 135 GeV.
It only remains to explain that taking the pains of understanding their ability to measure missing transverse energy with high precision is not something the D0 collaboration does for purely academical reasons. Indeed, the Z boson preferentially decays to neutrinos! One gets about twice as many neutrino decays than charged lepton decays. Moreover, the imperfect hermeticity of the detector to the charged leptons is an advantage for once: it affects the rate of visible and decays, while it actually increases the amount of detectable Higgs bosons!
“Wait a moment,” you might say. “True, I might fail to see a charged lepton from the Z decay, so I would catalog the event as a final state. But is it frequent to fail to see both, inflating the number of event candidates ?”
Well, as Oscar Wilde puts it, losing a relative is a tragedy, losing both smells of negligence… Indeed, D0 almost never fails to see both neutrinos. However, there is another process, , which has only one charged lepton, and if that is missed, the event will be cataloged as a with double neutrino decay of the Z! Now that is plain luck: they look for a particular process to detect the Higgs, and they get some bonus events from another one…
In reality, there is indeed a lot of cross-talk between the and final states. The search just happens to be on the receiving end, and so it is a little bit more sensitive to Higgs production than it would otherwise… All this is of course well known and studied. The overlaps of different searches are it in fact a problem when results in different final states need to be combined. But that is another story.
Ok, now I need to explain the analysis. I am rather exhausted by the long discussion above, and I bet you are too. So I will just summarize here what is another quite nice analysis in fact.
D0 selects events with large missing transverse energy and two b-quark jets. The latter are identified by an algorithm which reconstructs the secondary vertex produced when the long-lived hadrons containing the original b-quarks finally decay (but we are talking about a trillionth of a second!) into lighter particles. Then a few additional selection cuts are devised to repel some of the nastiest backgrounds – in particular, top pair production events where both top quarks decay into a lepton-neutrino-b-quark triplet. That is the so-called dilepton final state of top pair production, featuring two b-quark jets, lots of missing transverse energy, and two leptons. Now, if leptons are not identified and vetoed (and this may happen frequently when they are taus, because these particles are harder to deal with than electrons and muons), one may be looking at a very similar signature to that of the searched signal. Fortunately, requiring that the total transverse energy of all final state bodies (leptons, missing energy, jets alike) is below a certain value discards most of the top events while keeping the signal almost intact.
In the end the data contains a lot of background -ZZ,WZ production, top pair production, single top, QCD events with fake missing energy- and hopefully a teeny-tiny amount of real ZH events. How to proceed ? With a neural network classifier, of course. The kinematics of the various concurring background processes is different from that of genuine ZH decays, as the plots below show.
In each of the histograms, the ZH signal (in red) shows a slightly different distribution from backgrounds (top in blue, Z plus jets in cyan, W plus jets in yellow, and QCD in white). Left to right on the top row are the leading jet transverse momentum, the second jet transverse momentum, the mass of the b-quark-jet pair (which clusters around the true Higgs mass for the signal events), the missing transverse energy. On the second row you can (barely) see three other discriminating variables. More details can be obtained from the conference paper available from the D0 web site.
These distributions are fed to a neural network, whose output is a number from zero to one – zero being a very likely background event, one is an event with a sizable probability of being the coveted signal. Below you can see the distribution of the data on the neural network variable, for four different test values hypothesized for the Higgs boson mass (left to right, top to bottom: 105, 115, 125, 135 GeV):
The signal distribution overlaid in red to the backgrounds and to the experimental date (black points with error bars) demonstrates the amount of events one would see if ZH production were 50 times more frequent than it is expected to be. So there is no real chance to observe a signal, and the data is used to set a limit on the cross section. More specifically, the limit is obtained on the number of times the predicted Standard Model rate, as has become customary for Tevatron Higgs searches: at 120 GeV, the limit says D0 excludes anomalous ZH production with rate fifteen times higher than SM predictions.
The limit above (the black curve) by itself means very little. However, by combining it with tens of other similar results obtained by CDF and D0 as a function of Higgs mass, it will soon be possible for the Tevatron experiments to actually exclude a Standard Model Higgs boson in a meaningful range of mass values. And, with the LHC experiments still fiddling with their startup schedules, one cannot but sympathise with these coherent, continuative, strong effort by a bunch of physicists faithful to their good old glorious detectors. Long live D0 and CDF!