Here are list of words I am going to use for the higgs talk giving at La Thuile "Search for Low Mass Higgs Boson at the Tevatron" 1)Title page: Good aternoon, I will talk about Search for the SM Higgs Boson at the Tevatron on hehalf of CDF and D0 collaboration 2) In this talk, I will briefly review the Higgs search strategies. I will show you the recent results reported at EPS and some future prospects. More details can be found in these web pages. 3) The higgs boson is the last unobserved particle postulated in the standard model which help explain the origin of mass in the universe. Observation of the Higgs boson are longstanding key objectives to probe the mechanism of electroweak eymmetry breaking. There is a window of opportunity for Tevatron. With full dataset and improved analysis Tevatron could add crucial information about H->bb that is most difficult to detect at LHC. Recent data from LHC leave significantly less room for the Higgs boson to hide. 4) What we know about the Higgs Boson mass so far: The plot on left show the theoretical allowed region of Higgs mass as function of the scale where the new physics could occur by requiring the vacuum stability. The plot on right show the allowed region from LEPII, Tevtron and LHC direct search and indirect search from electroweak precision data. 5) The complex of tevatron collider is shown at the Fermilab near Chicago. It produces proton and anti-proton collision at the center mass of 1.96 TeV. Tevatron is doing very well and has delivered close to 12 fb-1 to each experiment with the record luminosity >4.3x1032 and is scheduled to shutdown by the end of Sept this year. Most results presented today are based on 8 fb-1 data and full dataset updates are coming in the winter conferences. 6) These are schematic view of CDF and D0 detectors. Both of them are general-purpose Detectors with excellent lepton identification, tracking, silicon vertexing for btagging, superior jet and missing et resolution. 7) The dominant Higgs production processes at the Tevatron are gluon-gluon fusion, and the associated production with W and Z boson. The upper plot show the production cross vs the Higgs mass at the Tevatron and the bottom plot shows the Higgs decay branching ratio as function of Higgs mass. For Higgs mass above 135, it will decay predominately into WW* and ZZ For Higgs mass below 135 GeV, the Higgs predominantly decays into bbbar, which makes the associated production with W and Z semileptonic decay most assessible while direct production gg->H->bbbar is limited by multi-jet QCD background. 8) While Higgs events are rare, the backgrounds are copiously produced in many order of magnitude, shown in the plot on right, that ttbar, W+jets, single top will give similar final states. The challenge is how to separate the small signal from the huge background using advanced analysis techniques. Recent observations of single top as shown in the bottom right and diboson provide a solid ground that these advanced tools do work to separate small signal from large backgrounds. 9) At the Tevatron, the Higgs search strategies are quite similar for the corresponding CDF and D0 analyses. Depends on Higgs decay, the searches are divided in many different channels. We select each sample that maximize signal acceptance by utilizing as much of detector as possible, improving lepton ID and trigger efficiencies. The samples are further divided into multiple sub-channels according to S/B in terms of lepton type, btag and jet multiplicity. Carefully model all background and cross check using control regions in data. Use advanced multivariate analysis tools to further improve S/B based on the full event kinematics. Our goal is to leave no Higgs events behind. Best sensitivity is obtained through the combination of many independent search channels and both CDF and D0 data. 10)For low mass higgs signatures we look for dijet mass resonance in associated with W/Z decays where W decays into lepton + neutrino gives the final states: one lepton+ neutrino + 2b, Z->nunu or W->lnu gives the final states: two neutrino + 2b, Z->ll gives the final states: two lepton + 2b. These are dijet mass after requiring double btagging in WH, ZH and ZH->nunubb where the data are consistent with sum of various background expections. Note the Higgs signal has been scaled by a factor of 10 for Higgs mass at 115 GeV. 11)For high mass Higgs signiture, we look for Higgs decay into WW pair that decay into dilepton events with large missing Et. The events are further divided into 0, 1, 2 jets bins that have very different background composition. In 0 jet bin, the predominat background is WW, which has very different dR than the one from Higgs decay as shown in bottom left. In 2jet bin, the predominat background is ttbar shown in the bottom right. 12) Since we know exactly what we are looking for, we can be bit more aggressive to employ the advanced multivariante techniques to suppress the background. For example, the leading order matrix element are used to calculate event probabilities based on set of observed inputs and likelihood ratio respects to the rest of backgrounds. Or they can be feed into an artificial neural network or boosted decision tree and find a discriminant variable. The typical improvement is about 25% respect to use a single variable, such as dijet mass. The primary gains in recent years mainly from improved signal acceptance: more triggers, loose lepton ID, better b-tagging... 13) Now, I will go over each individual analysis in some detail. One of gold channels for low mass higgs search is Higgs production association with a W boson where the W decays semileptonicly and Higgs decays into bbbar. We select events with one isolated high Pt lepton, large missing et and 2jets. Apply btagging and split events into double (Tight+loose) and single tags. For multivariante discriminate, CDF uses BNN in W+2jet and ME in W+3jets while D0 uses Random Forest decision trigger for both 2 and 3 jets. The bottom left plot show the BNN output in double tight tagged W+2jet at CDF and the plot on bottom right show the Random Forest BDT output in double tagged w+2jets from D0. Both data are consistent with background expectations. The expected Higgs signals are also shown in red, but rescaled with a large factor. 14) Since there is no excess of signal, we can set a limit on the Higgs production cross section times BR as function of Higgs mass. The left plot shows the observed and expected limits as function of Higgs mass in cdf. The solid curve is for the observed limit and dash is the expected. The right is for D0 limit. For Higgs mass 115 GeV, CDF set a limit at 2.65(2.6) x SM while D0 set a limit at 4.6(3.5) x SM. Not competitive for a single chaannel, need to combine all other channels and both cdf and d0. 15) Another interesting channel to look for higgs is in Higgs production associated with Z boson where the Z boson decay into a charged lepton pair and Higgs decay into bbbar. This channel has a low event rate due to a small branching fraction of Z to ee or mumu, but provides a clean signature. We select two high Pt leptons from Z+2jet. Apply btagging and split events into 1 or 2b-tags. Then we use a kinematic fit to improve dijet mass and neural network to separate ZH from top and Z+bbbar backgrounds. The plot on left show the NN output 10% slice along Z+jets vs ZH in double tags from CDF. The plot on right show the RF output in double tags from D0. Again, the data agree well quite with the background expectation. CDF has few candidates very Higgs-like, but it's not statistical significant. CDF is able to set a limit 4.8(3.9) x SM while 4.9(4.8) x SM for D0. 16) We also have looked for Higgs in ZH channel where Z decay into two neutrino, or wh where the lepton is missing. It has large rate, but large QCD multijet, and much more difficulty. However, the final state is relative clean, containing two b-jets and large missing et. We require missing ET >50 GeV and two b-tagged jets. Furthermore, we use \delta phi between the missing met and close jet and track met with charged tracks to build a multivariant discriminant to reduce the multi-jet QCD backgrounds. The final discriminant is obtained by combining dijet mass, track met and other kinematic variables, shown in the right. Again there is no Higgs signal, CDF set limit at 2.3 x SM at 95% CL, compared to 4.0 xSM expected. D0 set at 3.4 xSM with 4.2 x SM expected for Higgs mass @ 115GeV 17) Now I move on the high mass higgs searches for Higgs decaying into WW* where both w decay semileptonicly. The final state signiture consists of two opposite-sign high pt leptons, large missing et and containing zero or 1jet. The key of the analysis is to maximize lepton acceptance. The main background is WW*, which can be further reduced using the angular correlation between two leptons from a scaler Higgs decay. Both CDF and D0 divide the analysis into many sub-analysis based on the lepton purity and the number of jets. Then further train NN to seperate signal from background. The left plot show the NN output from OS 0jet, tight lepton pairs. The right show the BDT outpur from emu+met+0jet. The points are data are consistent with background expectations. Then we fit to extract the Higgs limit by combine other sub-channels together. 18) Other searches are also being considered. If the SM is correct, these are not as sensitive. But every litter bit helps amd nature could be different. The plot on left shows the limit for H->tautau the plot on right shows the limit for H->gg 19) There are many mutually exclusive final states. The plots show each individual limit from each channel as function of Higgs mass. Both cdf and D0 are in good agreement in all channels and we combine them statistally to improve the Higgs sensitivity. 20) To check the consistency between data and expectations, we rebin the final discriminant from each channel in terms of s/b The data with similar S/B may be added with no lose in sensitivity The plots on left show the events in each log(s/b) bins for mh=165 GeV, points are data, light blue is background and red is expected sm signal. The right plot shows data after background subtraction, compared to the signal. 21) Similar plots for Higgs mass 140 GeV. Again overall the data agree with background well, no excess of events observed in the highest S/B bins. 22)Similar plots for Higgs mass 115 GeV. Again overall the data agree with background well, no excess of events observed in the highest S/B bins. 23) We checked the search sensitivity using log-likelihood ratio. The plot shows the combined distributions of the log-likelihood ratio for different hypothesis as function of Higgs mass. The black dot is for the background-only hypothesis, the red dot is for signal-plus-background hypothesis, and the solid curve is for the observed data. The sizes of one and two sigma bands indicate the width of the LLR background only distribution. The separation between the background only and signal + background provides a measure of the search sensitivity, which is about 1.5 sigma at low mass and slight more than 3 sigma at mh=165 GeV 24)The plot shows the tevatron combination after combining cdf and D0 together. We are able to exclude the Higgs mass at the high end between 156bb combined limit as function of Higgs mass. 27) The plot show Tevatron H->WW combined limit as function of Higgs mass. There seems one sigma access between 120 and 150 GeV. 28) With the absence of H->WW signal, we can use it to constrain some new physics models such as 4th generation, which may exist in nature with their masses much higher than the current experiment limit. reinterpret H->WW search, we set 95% CL limit 124< MH<286 GeV GeV. In the fermiophobic model, cdf is able to exclude MH<115 GeV @ 95% CL. 29) The plot on left shows the higgs sensitivity obtained over time, which improves better than 1/sqrt(L) over time. The sensitivity has been improved more than a factor of 2 since 2005. The yellow band is what we expected with future improvements. The plot on right shows the analysis luminosity required to achive expected n sigma as function of the Higgs mass With 10 fb^-1 data, Tevatron could exclude significant fraction of the low mass Higgs. 30) Conclusion