The HitlinPorterHitlin Frank Porter Tomo Mayashita We have for many - PDF document

1 The Hitlin/PorterHitlin, Frank Porter To
The Hitlin/PorterHitlin, Frank Porter Tomo Mayashita, We have for many years specialized in heavy flavor physics. the study of heavy quarks and leptonsFor the last two decades, mainly with the experimentWe were the progenitors of the experiment, whose discovery of violation in the meson system led to the 2008 Nobel Prize forKobayashi and Maskawaand the 2016 Panofsky Prize of the APS, and play major roles to this dayWe now are working in the Mu2e experiment at FermilabWe are also designing the nascent LDMXexperiment at SLACHaving been central to the establishment of the Standard Model of HEP, we now focus on searching for evidence for physics beyond the Standard Model in two areas:Precision measurements in very rare processes involving quarks and leptonsSearches for carriers of nonravitational forces between no

2 rmal and dark matter ARwasdesignedtomeas
rmal and dark matter ARwasdesignedtomeasureviolationinmesondecays569publications,basicallyshapingthe\feldofheavyquarkandleptonstudies woareasofcurrentinterestinBSMphysicsARproducedevidenceofadeparturefromtheStandardModelleptonuniversalityprediction: R�D()
=B� B!D()� 
B� B!D()`� 2f;fromtheSM-valueis=4 American,July17,2017:\LawbreakingParticlesMayPointtoaPreviouslyUnknownForceintheUniverse"LauritsenLunchSeptember20,2017 ) Analysis Wearepursuingtwonewmeasurementsindierentchannels.Thisisadicultanalysiswithhighbackgroundsandsmallsignals,mustextractstatistically.UsesupervisedlearningmethodstooptimizestatisticalpowerStackedscoredensityprojections grouphasahistoryofinnovationinsophisticateddataanalysismethodsThisanalysisincludesseveralexamples,includingfas

3 tkerneldensityestimatorevaluation vergeo
tkerneldensityestimatorevaluation vergeofunblinding\frstanalysiswehaveunderwayLauritsenLunchSeptember20,2017 3.5 3.0 2.5 2.0 1.5 1.0 0.5 Z1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9stacked densityOptimized Proportions (a)Z1. Figure8.2:Stackedeventtypes. (a)Z1. Figure8.2:Stackedeventtypes. 2.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 2.0logit(logre D* score) 0 200000 400000 600000 800000 Dh D*h ** Comb Cont (a)Eventtypestackedhistograms. (b)Data-Simulationcomparison. (c)Eventtypedensities.Figure6.10: 6 8 10 12 14 16 18 20 22logN 101 100 101 102 103 104 105 106 milli secondsrunning time CPU direct CPU dual tree GPU direct GPU dual tree Figure7.1:Runtimebenchmark.isthenumberoftrainingandquerypointsandlogisinbase2.1approximatelywhenlog=202.Leafevaluation:ifthe\frstphasecannotjustifymakingtheapproximationsbeforereachingaleaf,itwill

4 beforcedtoperformthenaiveevaluationovert
beforcedtoperformthenaiveevaluationoverthepointsresidingataleafnode.Eventhoughitisguaranteedthatthenumberofleafevaluationswillbeboundedsuchthattheruntimecomplexityis),thehiddenconstantswillstronglydependontheactualnumberofleafevaluations,whichcanstillbelargewhentheextentofthekernelislarge.Onewaytospeedthisupistoobservethatthedirectevaluationlendsitselfverynaturallytoparal-lelization.Infact,thenaivewayofperformingthesekindsof-bodyproblemsisoneofthecanonicalwaysinwhichGPUmanufacturersmeasureandbenchmarktheimprovementsmadebetweendierentchipgenerations.Thisleadustoalgorithmalgorithm20],whichperformsthedirectevaluationonNVIDIAGPUs.Onemightnoticethatalgorithmdoesnotperformtheparallelizationstrategythatequationdirectlysuggests.Thatis,insteadofcreating)threadstoparallelizeeachterminthecomputationandthen\fnallydoing

5 aparallelaggregation,weinsteadassigneach
aparallelaggregation,weinsteadassigneachthreadthetaskofaggregatingallcontributionsforasinglequerypoint.ThisisduetotherelativespeedsofmemoryaccessesontheGPU;itturnsoutthatthecomputationaleciencythatwegainintakingadvantageofalltheavailableparallismisovershadowedbythenumberofglobalmemoryaccessesthatitrequires.7.1.3Resultsandbenchmarksbenchmarkstheruntimeofeachalgorithmicimprovementforkerneldensityestimation.Thedirect(dual-tree)algorithmisrepresentedbyanopen(closed)circle,whiletheCPU(GPU)evaluationisrepresentedusingblue(green)circles.Wecanclearlyseetheperformancegainobtainedbyapplyingthevariousstagesofalgorithmicimprove-ments.Byextrapolatingtheruntimeouttolog=20,weseethatthetotalperformanceimprovement the Dark ARhasbeenaworldleaderinsearchingforevidenceofdarkforcemediatorsManysearchescanandhavebeendone:Darkpho

6 tonDarkHiggsbosonDarkintermediatebosonsD
tonDarkHiggsbosonDarkintermediatebosonsDarkhadronsDarkscalar/pseudoscalarMuonicdarkforceLeptophilicdarkscalarSelf-interactingdarkmatter September20,2017 the Dark Searchisin2Dspaceofcouplingstrength(verticalaxis)anddarkparticlemass(horizontalaxis) rkphoton e+e�!\rA0;A0!invisible DarkZ0 darkforceexplanationfor(anomalyruledout September20,2017 EPJWebofConferences Figure1.FeynmanDiagramforproductionofgaugebosonbasedonthemodelatancollider.2DataandEventSelectionWeusedthedatacollectedbytheBABARdetector[4]withthetotalluminosityof514fb.Mostofthedataweretakenatthe)resonanceplusincludingabout28fbdataat)and14fbdataat)and48fbdataattheo-resonance.The)resonancedecaystoapairofBBBWeusedabout5%ofthedatasettovalidateandoptimizetheanalysismethod.Therestofthedatawasonlyexaminedafternishnalizingtheanalysismethod.Forthebac

7 kgroundstudywegeneratedsignalMonteCarlo(
kgroundstudywegeneratedsignalMonteCarlo(MC)samples.SignalMCeventsaregeneratedusingMadGraph5[6],whichcalculatesmatrixelementsforthesample.TheMCthenwereshoweredusingPythia6[7]forabout30dimasshypotheses.ThemainbackgroundcomesfromtheQEDprocesses.WegeneratethedirectprocessesofusingDiag36[8],whichincludesthefullsetofthelowestorderdiagrams.TheDiag36doesnotincludeinitialstateradiation(ISR)samples.TheeventsoftheprocessofaregeneratedusingBHWIDE[9]andtheMCeventsof)and)aregeneratedusingKK[10].Theo-resonancedatasamples,u,d,s,c),aresimulatedusingEvtGen[11].Theeventsprocessesofweregeneratedusingastructurefunctiontechnique[12,13].FinallythedetectoracceptanceandreconstructioneciencyaredeterminedusingMCsimulationbasedonGEANT4[14].Weselecteventscontainingexactlytwopairsofoppositelychargedtracks,consistentwiththetopol-ogyoftheproce

8 ss:nalstate.Themuonsareidentiedb
ss:nalstate.Themuonsareidentiedbyparticleidenticationalgorithmsforeachtrack.Werequirethesumofenergiesoftheelectromagneticclus-tersthatarenotassociatedtoanytrackmustbelessthan200MeV.Wenallyrejecteventsthatcomefromthe)and),wheredecaysifthedimuoncombinationiswithin100MeVofthe)wherepionsaremisidentiedasmuons.Thedistributionofthefour-muoninvariantmassafterallselectionsisshowninFig.2(left).Atthelowmassofthefour-muoninvariantmass,9GeV,iswellreproducedbytheMonteCarlosimulationincludingdirectdecaysof,however,theMonteCarlosimulationoverestimatesthefullenergypeakby30%andfailstoreproducetheradiactivetail.Theoveresti-matesimulationisexpectedbecausetheDiag36simulationdoesnotsimulatetheinitialstateradiation(ISR)events.ToestimatethepotentialISRemissionweselecteventsbyrequiringafour-muoninvariantmassdistribut

9 ionwithin500MeVofthenominalcenter-of-mas
ionwithin500MeVofthenominalcenter-of-massenergy.Wealsorequirethetrackstooriginatefromtheinteractionpointtowithinitsuncertaintyandconstrainingthecenter-of-massenergyofthesystemtobewithinthebeamenergyspread.Thefour-muoninvariantmassafterallowingthepotentialISRemissionisnowtasshowninFig.2(right). 5 cm LDMX LDMX willextend he search for the darkforce carriers that mediate between normal matter and the more abundant, but as yet unseen, dark matter Much interest now focusses on the low mass regime (below the MeV region), wheresearches are best done in fixed target experimentsThe LDMX experiment uses an electron beam (available at SLAC or JLab, and a new missing momentum technique, to achieve unsurpassed sensitivity, to the “thermal relic” limit Mu2e Mu2e, our main activity, is a search for Charged Lepton

10 Flavor Violation(CLFV) using the proces
Flavor Violation(CLFV) using the process of muonelectron conversion in the field of an aluminum nucleusNeutrino oscillations show that neutral lepton flavor is not conserved (a BSM effect)We therefore expect to also have a violation of charged lepton flavor, which is unobservablysmall in the Standard ModelMany BSM theories, however, predict CLFV at an observable level in to conversion (current limit 6x10) or the decay (current limit 5.7x10Mu2e improves search sensitivity by four orders of magnitude over existing experiments, to the range where observation is possible 5 cm Mu2e The leap in sensitivity is made possible by the high intensity pulsed muon beam at FermilabMu2e is currently under construction at Fermilab, with commissioning commencing in 2020The full dataset will provide an improvementin sensitivity o

11 f 10to a mass scale well beyond the reac
f 10to a mass scale well beyond the reach of LHC experimentsOne month of data will improve on the currentexperimental limitCaltech responsibilities in Mu2eElectromagnetic calorimeterInitial design configuration and optimizationCsIcrystals and photosensorsSource calibration systemData acquisition softwareSoftware triggerImproved muon stopping target 5 cm Mu2e The electromagnetic calorimeter consists of 1348 CsIcrystals, each read out by two 6element SiPMphotosensors, arranged in two annular disksIn concert with RenYuan Zhu’s crystal lab, we have been responsible|for studying many types of candidatecrystalsand photosensors, for light output,uniformity, radiation hardness, ……We are now preparing to fully set up and calibrate 50% of the CsiSiPMsWe have also developed a high energy gamma ray calibratio

12 n source usingneutron generator to produ
n source usingneutron generator to produce 6.13 MeV s from the reactionWe have built the prototype in the Isotope Lab behind Bridgeand are now designing the fullsystem for installation at Fermilab Mu2e We also contribute to the design and implementation of the trigger and data acquisition softwareas well as the optimization of the muon stopping targetTriggerDAQ Conclusions Our group is pursuing several promising avenues to find evidence for physics beyond thetheStandard ModelThe effect of new particles in loops on precisely predicted processes with dataSearches for dark force carriers in annihilation with a new acceleratorbased experiment LDMXSearches for charged lepton flavor violation in muonelectron conversion with Mu2eIn all these experiments, our goal is worldleading new physics limits or discovery sensiti