Contents lists available atScienceDirect
Physics
Letters
B
www.elsevier.com/locate/physletb
Search
for
CP violation
in
the
phase
space
of
D
0
→
π
+
π
−
π
+
π
−
decays
.
LHCb
Collaboration
a
r
t
i
c
l
e
i
n
f
o
a
b
s
t
r
a
c
t
Articlehistory:
Received15December2016
Receivedinrevisedform22March2017 Accepted27March2017
Availableonline4April2017 Editor:W.-D.Schlatter
A search for time-integrated CP violation in the Cabibbo-suppressed decay D0→
π
+π
−π
+π
− is performed using an unbinned, model-independent technique known as the energy test. This is the first application of the energy test in four-body decays. The search is performed for P -evenCP asymmetriesand, for the first time, is extended to probe the P -odd case. Using proton–proton collision data
corresponding to an integrated luminosity of 3.0 fb−1 collected by the LHCb detector at centre-of-mass energies of √s=7 TeV and 8 TeV, the world’s best sensitivity to CP violationin this decay is obtained. The data are found to be consistent with the hypothesis of CP symmetry with a p-value
of
(4.6 ±0.5)% in the P -evencase, and marginally consistent with a
p-valueof
(0.6 ±0.2)% in the P -oddcase, corresponding to a significance for CP non-conservation
of 2.7 standard deviations.
©2017 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Funded by SCOAP3.
1. Introduction
The decay D0
→
π
+π
−π
+π
− (charge-conjugate decays areimplied unless stated otherwise) proceeds via a singly Cabibbo-suppressedc
→
dud transitionwithanadmixture fromac→
uggluonic-penguintransition.WithintheStandardModel(SM),these amplitudeshave different weak phases, and the interference be-tweenthemmaygiverisetoa violationofthecharge-parity(CP) symmetry in the decay. Anothernecessary condition for this di-rect CP violation to occur is interference of at least two ampli-tudes with different strong phases. The strong phase differences areknowntobe sizeableincharm decaysandcan enterthrough theresonances that abundantly contribute to the four-body final states. The sensitivity to CP violation is usually best for decays wherethestrongphasesbetweeninterferingresonanceshavelarge differences.A rich spectrumof resonancescontributes tothe de-cayD0
→
π
+π
−π
+π
−,whichaccordingtotheamplitudeanalysisperformed by the FOCUS collaboration [1], is dominated by the amplitudesforD0
→
a1
(
1260)
+π
−witha1(
1260)
+→
ρ
0(
770)
π
+andforD0
→
ρ
0(
770)
ρ
0(
770)
.Inthe SM, violation of the CP symmetryin the charm sector isexpectedatorbelowthe
O
(10−3)level[2].Contributionsfrom particlesthat are proposed to exist inextensionsof theSM may participateinhigher-orderloop contributions (penguindiagrams) andenhancethe levelof CP violation. Multibodydecays, such asD0
→
π
+π
−π
+π
−,allowtheCP asymmetriestobeprobedacrossthephasespaceofthedecay,andtheselocalCP asymmetriesmay belargerthanglobalCP asymmetries.
The analysis of D0
→
π
+π
−π
+π
− is primarily sensitive todirectCP violation.InadditiontothisdirectCP violation,the time-integrated CP asymmetry in D0
→
π
+π
−π
+π
− decays can alsoreceiveanindirectcontributionarisingfromeither D0–D0 mixing
orinterferencebetweendirectdecaysanddecaysfollowingmixing. While directCP asymmetrydependsonthedecaymode, indirect
CP violationisexpectedtobe universalwithin theSM.The time-dependentmeasurementsofD0
→
π
−π
+, K−K+decaysconstrain theindirectCP asymmetry belowtheO
(10−3) level[3],whichisbeyondthesensitivityofthisanalysis.
Previously, the most sensitive search for CP violation in the
D0
→
π
+π
−π
+π
− decaywasperformedby theLHCb collabora-tionwithdatacollectedin2011[4].Abinnedχ
2 technique (SCP)
was used to exclude CP-violating effects at the 10% level. Four-body decays requirefiveindependent variables to fullyrepresent thephasespace. Consequently,binned analyseswillhavea trade-offbetweenminimisingthenumberofbins,inordertomaximise thenumberofeventsper bin,andretainingsensitivityto the in-terferencebetweenallcontributingresonances.Themeasurement presentedhereincludesdatacollected in2012,resultingina sig-nal sample that is aboutthree times larger than inthe previous LHCb analysis. The method exploitedhere, known as the energy test,isanunbinnedtechnique,whichisadvantageousinthe anal-ysisofmultibodydecays.
The energy test was applied for the first time to search for
CP violation in decays of D0
→
π
−π
+π
0 [5]; here we presentits first applicationto four-body decays. The energy test is used toassessthe compatibilityofthe observeddatawithCP
symme-try.ItissensitivetolocalCP violationinthephasespaceandnot
http://dx.doi.org/10.1016/j.physletb.2017.03.062
0370-2693/©2017TheAuthor(s).PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/).Fundedby SCOAP3.
to global asymmetries,which may also arise from different pro-duction cross-sectionsof D0 and D0 mesons at a proton–proton collider.The method identifies the phase space regions inwhich
CP violation is observed. Being model-independent, this method neither identifies which amplitudes contribute to the observed asymmetry nor measures the actual asymmetry. Consequently, a model-dependentamplitude analysisof D0 and D0 decayswould
berequiredifevidenceforanon-zeroCP asymmetryisobtained. Theanalysispresentedhereprobesseparatelyboth P -evenand
P -odd CP asymmetries. The P -even test is performed through the comparison of the distribution of events in the D0 and D0
phasespaces,characterisedbysquaredinvariantmasses. Addition-allycharacterising the eventsusinga triple productof final-state particlemomenta[6–8]givessensitivityto P -oddamplitudes,and thusallowsthefirsttestforP -oddCP asymmetriesinanunbinned model-independenttechnique.
2. Detectorandreconstruction
The LHCb detector [9,10] is a single-arm forward spectrome-tercoveringthepseudorapidity range2
<
η
<
5,designedforthe studyofparticlescontaining b orc quarks. Thedetectorincludes a high-precision trackingsystem consistingof a silicon-strip ver-tex detector surrounding the pp interaction region, a large-area silicon-stripdetectorlocatedupstream ofa dipolemagnetwitha bendingpower of about4 Tm, andthree stations ofsilicon-strip detectorsandstrawdrifttubesplaceddownstreamofthemagnet. Thetrackingsystemprovidesameasurementofmomentum, p,of chargedparticleswitharelativeuncertaintythat variesfrom0.5% atlowmomentumto1.0%at200 GeV/
c.Theminimumdistanceof atracktoaprimarypp interactionvertex(PV),theimpact param-eter(IP),ismeasuredwitharesolutionof(
15+
29/
pT)
μ
m,wherepTisthecomponentofthemomentumtransversetothebeam,in
GeV
/
c.Differenttypesofchargedhadronsaredistinguishedusing informationfromtworing-imagingCherenkovdetectors.Photons, electronsandhadronsareidentifiedby acalorimetersystem con-sistingofscintillating-padandpreshowerdetectors,an electromag-neticcalorimeterandahadroniccalorimeter.Muonsareidentified byasystemcomposedofalternating layersofironandmultiwire proportionalchambers.Thisanalysisusesthedatafrom pp collisionscollected bythe LHCbexperimentin2011and2012correspondingtointegrated lu-minositiesof,respectively,1 fb−1and2 fb−1atcentre-of-mass en-ergiesof7 TeV and8 TeV.Thepolarityofthedipolemagnetis re-versedperiodicallythroughoutdata-taking.Theconfigurationwith themagneticfieldpointingupwards(downwards),bendspositively (negatively)chargedparticles inthehorizontalplane towardsthe centre of the LHC. Similar amounts of data were recorded with eachpolarity,whichreducestheeffectofcharge-dependent detec-tion andreconstruction efficiencies on results obtained fromthe fulldatasample.
In the simulation, pp collisions are generated using Pythia 8[11]withaspecificLHCb configuration[12].Decaysofhadronic particles are described by EvtGen [13]. The interaction of the generatedparticles withthedetectorandits responseare imple-mentedusingthe Geant4toolkit[14]asdescribedinRef.[15].
Theonlineeventselectionisperformedbyatrigger,which con-sists of a hardware stage, based on high-pT signatures fromthe
calorimeterandmuon systems, followed by a two-levelsoftware stage.Eventsarerequiredtopassbothhardwareandsoftware trig-gerlevels.Thesoftwaretriggeratitsfirstlevelappliespartialevent reconstruction. It requires atleast one good quality track associ-atedwithaparticlehavinghighpTandnotoriginatingfromaPV.
A second-level software trigger, optimised forfour-body had-ronic charmdecays,fully reconstructs D0
→
π
+π
−π
+π
−candi-dates coming from D∗+
→
D0π
+s decays. The charge ofthe soft
pion (
π
s) tags theflavour of the D mesons atproduction,whichis needed as
π
+π
−π
+π
− is a self-conjugate final state acces-sible to both D0 and D0 decays. The trigger selection ensuresthesuppressionofcombinatorialbackgroundwhileminimisingthe distortion oftheacceptanceinthe phasespaceofthedecay. The triggerrequiresafour-tracksecondaryvertexwithalltracksbeing ofgoodquality andpassing minimummomentumandtransverse momentum requirements. The pions from the candidate D0
de-cay are required to have a large impact parameter significance (
χ
2IP) withrespect to all PVs, where
χ
IP2 is definedas thediffer-enceinthevertex-fit
χ
2 ofaPVreconstructed withandwithoutthe considered track. A part of the data collected in 2011 was takenwitha differentsecond-leveltriggerselection.Inthis selec-tion only D0
→
π
+π
−π
+π
− candidates are reconstructed whilethe D∗+ reconstructionisperformedonlyduring theoffline selec-tion. For a part of the data collected in 2012, additional events were selected in the trigger from the partial reconstruction of
D0
→
π
+π
−π
+π
− candidatesarisingfromD∗+→
D0π
s+decays, usingonlyinformationfromoneπ
+π
−pairandasoftpion.3. Offlineeventselection
Intheofflineselection,signalcandidatesarerequiredtobe as-sociated to candidates that passed the onlineselection described in theprevious section. In addition,the offlineselection imposes morestringentkinematiccriteriathanthoseappliedinthetrigger. The D∗+ and D0 candidates must have pT
>
500 MeV/
c.All thecandidatepiontracks,those fromthe D0 decayproducts andthe
π
smesons,arerequiredtohavepT>
350 MeV/
c andp>
3 GeV/
cto reduce thecombinatorialbackground.The candidate D0 decay products must form a good quality vertex. As a consequence of the non-negligible D0 lifetime, the D0 decay vertexshould typ-ically be significantly displaced from the PV; this is ensured by applyingaselectiononthesignificanceofthe D0 candidateflight
distance.Charmmesonsfromb hadrondecayshavelargerIPsdue tothecomparativelylongb hadronlifetimes.Thissecondarycharm contributionissuppressedby imposinganupperlimitonthe
χ
2 IPoftheD0 candidate.BackgroundfromD0
→
K−π
+π
−π
+decays, with a kaon misidentified as a pion, is reducedby placing tight requirementsontheπ
±particleidentificationbasedonthe ring-imaging Cherenkov detectors. The contribution of the Cabibbo-favoured D0→
K0S
π
+π
− decayisfoundtobe belowthepercentlevel.As investigatedin Ref.[4],partiallyreconstructed or misre-constructedmultibodyD(s)decays(e.g.,decayswithamissingpion orakaonmisidentifiedaspion)donotgive risetopeaking back-groundsundertheD0
→
π
+π
−π
+π
− signal.Constraints on the decay kinematics are applied to improve massandmomentumresolutions.ThefourpionsfromtheD0
de-cayareconstrainedtocomefromacommonvertexandthedecay vertex of the D∗+ candidate is constrained to coincide with its PV[16].Theseconstraintsareappliedinthedeterminationofthe massdifference,
m,betweentheD0 andtheD∗+.TheD0is con-strainedtoitsnominalmassinthedeterminationofthe kinemat-icsinthe D0
→
π
+π
−π
+π
− decay.Arequirementon fitquality for the D∗+ vertex fitsefficiently suppresses combinatorial back-ground. This requirement also suppresses the contribution fromD∗+ mesons originatingfrom long-lived b hadrons. The remain-ing componentintheanalysisfromthissourceisnot sensitiveto
CP asymmetries inb hadrons astheflavour tag isobtainedfrom thechargeofthe
π
sinthedecayoftheD∗+meson.The
π
sisalow-momentumparticle,withtheconsequencethatthelargedeflectioninthemagneticfieldleadstodifferent accep-tancesforthetwocharges.Consequently,thesoftpionisrestricted to the region where the detection asymmetry is small. This is
Fig. 1. Distributionofm withfitoverlaidfortheselectedD∗+ candidatesinthe 2012data.Thedatapointsandthecontributionsfromsignal,background,andtheir totalobtainedfromthefitareshown.
achievedthrough theapplicationoffiducialcutsonthesoftpion momentum,followingRef.[17].Asthekinematicsoftheslowpion are largely uncorrelatedwith the D0 phase space, the
π
sdetec-tionasymmetry wouldresultinaglobalasymmetrytowhichthis analysisisnotsensitive.Thereare,however,differencesinthe de-tectionefficienciesoftheD0andD0daughtersthatmayintroduce additionalasymmetrieslocalisedinthephasespaceofD0 decays,
andwhicharediscussedindetailinSect.7.
ThesignalregionintheD0invariantmassisdefinedas1852
<
m
(
π
+π
−π
+π
−) <
1882 MeV/
c2, corresponding to a full rangeof about four times the mass resolution. The signal yield is es-timated from the
m distribution, which is shown in Fig. 1 for the2012 data.These
m distributionsare modelled by thesum ofthreeGaussian functionsforsignalandasecond-order polyno-mialmultipliedbyathresholdfunction
√
1−
mπ/
m, wheremπisthe pion mass,describing combinatorial andrandom soft-pion backgrounds.Theselectedsamplescomprise320,000and720,000 signalcandidatesinthe2011and2012datawithpuritiesof97% and96%,respectively.Thefinalsignalsampleisselectedrequiring
|
m−
145.
44|
<
0.
45 MeV/
c2, which corresponds to selecting aregionwithawidthroughlytwicetheeffective
m resolution.
4. Descriptionofthephasespace
Fivecoordinatesarerequiredforafulldescriptionofthephase spaceoffour-body decays [18]. Incontrast tothree-body decays, thereisnostandardorcommonlypreferredchoiceofcoordinates. Two-bodyandthree-body invariant masscombinationsofthe pi-onsareusedascoordinateshere.Theenergytestperformedhere isa statisticalmethodcomparing thedistributions of D0 and D0
candidates in phase space (see Sect. 5). Therefore, it is sensitive to the position of an event in phase space and to the choice of coordinates spanning this phase space. The choice of coordi-nates influences the sensitivity of the analysis as it will change the distance between events in the phase space. Furthermore,
D0
→
π
+π
−π
+π
− decayscontaintwoπ
+ andtwoπ
− mesons.Thepionsofthesamechargecanbeinterchanged;asaresulteach decaycanbeplacedatfourpointsinphasespace.Theenergytest issensitivetosuchpioninterchange.Toobtainbothaunique out-putandoptimalsensitivityfromtheenergytest,anorderingofthe inputvariablesoftheenergytestisdefinedinthefollowing.
The order of the charges of the four pions in the D0 decay
π1π2π3π4
is fixed toπ
+π
−π
+π
−.1 There are four two-bodycombinations in which resonances can be formed andthese are the
π
+π
−pairs:π
1π2,π
1π4,π
3π2,andπ
3π4.Likewise,thereare fourthree-bodycombinations:thetwoofpositivecharge,π
1π2π3and
π
1π3π4,andthetwoofnegativecharge,π
1π2π4andπ
2π3π4. Theinvariantmassesofallpossibleπ
+π
−pairsarecalculatedand sorted foreach event. Theπ
+π
− pair withthe largestinvariant massisfixed tobeπ
3π4
,whichfullydeterminestheorderofallfourpions.Asonlyasmallfractionofthe
ρ(
770)
resonance,either produceddirectlyfromtheD0 decayorthrougha1
(
1260)
decays,contributestothelargestm
(
π
+π
−)
,theπ
3π4 combinationis ex-cluded fromthecoordinates used.Whileanycombinationof five variables covers thefull phase space, the choicemade hereisto keep variables sensitive to the presence ofthe main resonances. Thetwo-bodyandthree-bodymasscombinationsthatdonot con-tain theπ
3π4 pairare used fortheenergytest coordinates. This results in five invariant masses,π
1π2,π
1π4,π
2π3,π
1π2π3 andπ1π2π4
,whichareexpectedtocovermostoftheintermediateres-onancecontributions.
The choice of using only invariant masses has a limitation. Invariant masses are even under parity transformation, and a comparison of the D0 and D0 samples probes only P -even CP
asymmetries. In four-body decays, however, P -odd amplitudes can also be present. In D0
→
π
+π
−π
+π
− decays, there isonly one significant P -odd amplitude, the perpendicular helic-ity ( A⊥) of D0
→
ρ
0(
770)
ρ
0(
770)
decays. Alternatively, in thepartial-wave basis, it is the amplitude corresponding to the P-wave D0
→
ρ
0(
770)
ρ
0(
770)
decays. Its contribution to the totalD0
→
π
+π
−π
+π
− widthis about6% [19].The defaultapproachisextendedtomakeacomplementarytestofthe P -oddCP
asym-metry, which may arise from interference between P -odd and
P -evenamplitudes.ThisisdiscussedfurtherinRef.[8].
Tripleproducts,whicharebydefinition P -odd,canbe usedto probeP -oddCP violation[20].Theseasymmetriesareproportional tothecosineofthestrong-phasedifferencebetweenthe interfer-ing partial waves [6], and thus will be enhanced where P -even CP asymmetries,proportional tothe sineofthe strong-phase dif-ference, lack sensitivity. A triple product CT
=
p1· (
p2×
p3)
isconstructedfor D0 decays,wherepionmomentaaremeasured in
the D0 rest frame. Here
p
1,
p2 and p3 are the vector momentaof
π
1,π
2 andπ
3 sortedasdescribed above(i.e.,π
4 isexcluded).Thecorrespondingtripleproductforthe D0 decaysisobtainedby
applyingtheCP transformation,CP
(
CT)
= −
C(
CT)
= −
CT.TheCTobservableisconstructedbychargeconjugatingthepionsentering
CT (i.e.,theexcludedpioninCT isthe
π
+inthelargestm(
π
+π
−)
combination).Thetotalsampleisdividedintofoursubsamples ac-cordingtotheD0 flavourandthesignofthetripleproduct:
[
I]
D0(
CT>
0),
[
II]
D0(
CT<
0),
[
III]
D0(
−
CT>
0),
[
IV]
D0(
−
CT<
0).
(1)SamplesI andIII arerelatedbythe CP transformation,andso areII andIV.TheyieldsofthesefoursamplesarelistedinTable 1. ThetestforthepresenceofP -evenCP asymmetryisperformedby comparingthecombinedsampleI
+
II withthecombinedsample III+
IV.ThiscorrespondstotheintegrationoverCT andisthede-faulttest,inwhichD0andD0 samplesarecomparedinthephase
space spannedwith invariant massesonly. Similarly, the test for a P -oddCP asymmetryisperformedbycomparing thecombined sample I
+
IV withthecombinedsample II+
III.Thiscomparison1 InaD0decaytheorderofπ
1π2π3π4ischarge-conjugated,π−π+π−π+,with respecttothatoftheD0decay.
Table 1
Theyieldsofsignaleventsinthefoursamplesthatobtainedfromfitstothem
distribution.
Sample I II III IV
Yields 256 466±629 246 629±519 258 274±574 246 986±607
isperformed inthe samephase spaceasthe default P -even ap-proachandallowstheP -oddcontributiontotheCP asymmetryto be probed, since the P -even contribution cancels [8]. No triple-product asymmetry measurements exist for D0
→
π
+π
−π
+π
−decays and the previous LHCb study [4] was performed in the phasespacebasedontheinvariantmassesonly.Consequently,this isthefirsttime a P -oddCP asymmetryisinvestigatedinthis de-caymode.
5. Energytest
Model-independentsearchesforlocalCP violationaretypically carried out using a binned
χ
2 approach to compare therela-tive density in a binof phase space of a decay withthat of its
CP-conjugate.Thismethodwas usedinaprevious studyof D0
→
π
+π
−π
+π
−decays[4].Asdiscussedintheprevioussection,five coordinates are required to describe four-body decays. A model-independent unbinned statistical method called the energy test was introduced in Refs. [21,22].The potential for increased sen-sitivity of this method over binnedχ
2 analyses in Dalitz plotanalyseswas showninRefs.[8,23] anditwas firstapplied to ex-perimentaldatainRef.[5].
This Letter introduces the first application of the energy test technique to four-body decays, where it isused to compare two eventsamplesintestsofbothP -evenandP -oddtypeCP violation.
The P -evenenergytestseparateseventsaccordingtotheirflavour, andthencompares these D0 and D0 samples.The P -odd energy testseparateseventsusingboththeirflavourandsignofthetriple product,asdescribedintheprevioussection.
Ateststatistic, T ,isused tocomparetheaveragedistances of eventsinphasespace.ThevariableT isbasedonafunction
ψ
i j≡
ψ(
di j)
whichdependsonthedistancedi j betweeneventsi and j.Itisdefinedas T
=
n i,j>iψ
i j n(
n−
1)
+
n i,j>iψ
i j n(
n−
1)
−
n,n i,jψ
i j nn,
(2)where the first and second terms correspond to an average weighteddistancebetweeneventswithinthen eventsofthefirst sample andbetween then events of the second sample, respec-tively. The third term measures the average weighted distance betweeneventsinthefirstsample andeventsinthesecond sam-ple.Ifthedistributionsofeventsinbothsamplesare identical,T
willrandomlyfluctuatearoundavalueclosetozero.
Thenormalisationfactors inthedenominatorsoftheterms of Eq.(2)removetheimpactofglobalasymmetriesbetweenD0 and
D0 samples. In the P -odd test, subsamples of both D0 and D0
samples are combined. Consequently, any global asymmetries in these could result in local asymmetries in the samples used for theP -oddtest.Therefore,forthe P -oddtesttheglobalasymmetry between D0 and D0 is removed by randomly rejecting some of the D0 candidatesto equalisethesize ofthesamples ofthetwo
flavoursbeforecombiningtheeventsintothetwosamplesthatare compared.
The function
ψ
should decrease with increasing distance di jbetweenevents i and j,inordertoincreasethesensitivityto lo-calasymmetries.AGaussianfunctionischosen,
ψ(
di j)
=
e−d2
i j/2δ2,
withatuneableparameter
δ
(seeSect.6)thatdescribesthe effec-tiveradiusinphasespacewithinwhichalocalasymmetryis mea-sured.Thus,thisparametershouldbelargerthantheresolutionofdi j butsmallenoughnottodilutelocallyvaryingasymmetries.The
distancebetweentwopointsisobtainedusingthefivesquared in-variantmassesdiscussedintheprevioussectionandcalculatedas
d2i j
= (
m212,j−
m212,i)
2+ (
m142,j−
m214,i)
2+ (
m223,j−
m223,i)
2+ (
m2123,j−
m2123,i)
2+ (
m2124,j−
m2124,i)
2.
(3)In the caseof CP violation, the average distances entering in the third termofEq.(2)are larger than inthe other terms.Due to the characteristics of the
ψ
function, this leads to a reduced magnitudeofthisthirdtermrelativetotheotherterms.Therefore, largerCP asymmetriesleadtolargervaluesofT .Thisistranslated into a p-valueunderthehypothesis of CP symmetryby compar-ing the T value observed in data to a distribution of T valuesobtainedfrompermutationsamples.Thepermutationsamplesare constructedbyrandomlyassigningeventstoeitherofthesamples, thus simulatinga situation withoutCP violation. The p-valuefor the no-CP-violationhypothesis isobtainedasthe fractionof per-mutationT valuesgreaterthantheobservedT value.
Forscenarios wheretheobserved T valuelieswellwithin the range ofpermutation T values,the p-valuecan be calculatedby simply countinghowmanypermutation T valuesarelarger than the observed one. Iflarge CP violation isobserved, the observed
T value is likely to lie outside the range of permutation T
val-ues.Inthiscasethepermutation T distributioncanbefittedwith a generalised-extreme-value (GEV) function, as demonstrated in Refs. [21,22] andused inRef. [5]. The p-valuefrom the fitted T
distribution can be calculated as the fraction of the integral of the functionabove the observed T value.The uncertaintyon the
p-value is obtained by randomly resampling the fit parameters within their uncertainties, taking into account their correlations, andby extractinga p-valueforeachofthesegenerated T
distri-butions.Thespreadoftheresultingp-valuedistributionisusedto set68%confidenceintervals.A90%confidence-levelupperlimitis quoted where no significantly non-zero p-value can be obtained fromthefit.
The number of permutations is constrained by the available computingtime.Thedefaultp-valueextraction,definedbefore ob-tainingtheresultfromthedata,usesthecountingmethodaslong asatleastthreepermutationT valuesarefoundtobelargerthan theobservedT value.Otherwise,thep-valueisdeterminedby in-tegratingthefittedGEVfunction.The p-valuespresentedhereare basedonover 1000permutationsforthedefaultdataresultsand on100permutationsforthesensitivitystudies(seeSect.6).
Avisualisation ofregions ofsignificantasymmetry isobtained byassigning anasymmetrysignificancetoeachevent.The contri-butions ofasingle eventinone sample, Ti,andasingleeventin
theothersample,Ti,tothetotalT valuearegivenby Ti
=
1 2n(
n−
1)
n j=iψ
i j−
1 2nn n jψ
i j,
(4) Ti=
1 2nn−
1 n j=iψ
i j−
1 2nn n jψ
i j.
(5)Having obtainedthe Ti and Ti values for all events,the
per-mutation method is also used here to define the significance of eachevent.Foragiveneventi theexpectedTi distributioninthe
caseofCP symmetryisobtainedbyusingthepermutationmethod. The distributions ofthe smallestnegative andlargest positive Ti
valuesofeachpermutation,TiminandTimax,areusedtoassign sig-nificancesofnegativeandpositiveasymmetries,respectively.Iffor
Table 2
OverviewofsensitivitiestovariousCP-violationscenariosin sim-ulation.A andφdenote, respectively,therelativechangein magnitudeandthechangeinphaseoftheamplitudeofthe res-onance R.TheP-wave ρ0
(770)ρ0
(770)is a P -oddcomponent. The phase changein this resonanceis tested with the P -odd CP-violationtest.Resultsforallotherscenariosaregivenwiththe standardP -eventest.
R (partial wave) (A,φ) p-value (fit) a1→ρ0π(S) (5%,0◦) 2.6+−31..47×10−4 a1→ρ0π(S) (0%,3◦) 1.2+−31..62×10− 6 ρ0ρ0(D) (5%,0◦) 3.8+2.9 −1.9×10−3 ρ0ρ0(D) (0%,4◦) 9.6+24 −7.2×10−6 ρ0ρ0(P) (4% ,0◦) 3.0+1.2 −0.9×10− 3 ρ0ρ0(P) (0% ,3◦) 9.8+4.4 −3.8×10− 4
realdatatheTivaluefallsintheright(left)tailofthedistribution
of Timax (Timin) containing 32%, 5% or0.3% of thevalues itis as-signedapositive(negative)significanceof1,2or3
σ
,respectively. Thesameprocedureisappliedtothe Ti distribution,leadingtoaphasespacewithan invertedasymmetry pattern. Thismethodis illustratedinSect.6.
The practical limitation of this method is that the number of mathematical operations scales quadratically with the sample size.Furthermore,asignificantnumberofpermutationsisrequired to get a sufficient precision on the p-value. In this analysis, the methodisimplementedusingparallelcomputingongraphics pro-cessingunits(GPUs)[24].
6. Sensitivitystudies
Tointerprettheresults,astudyofthesensitivityofthepresent datasampletodifferenttypesofCP violationisrequired.The sen-sitivityisexaminedbasedonsimplifiedsimulationsamples gener-atedaccordingtoa preliminaryversion ofthe modelinRef. [25] basedonCLEO-c measurements.The generationis performed us-ingMINT,asoftwarepackageforamplitude analysisofmultibody decays that has also been used by the CLEO collaboration [26]. Giventhe highpurity obtainedin theselection, backgrounds are neglectedinthesesensitivitystudies.
The variation ofthe selection efficiency across phase spaceis takenintoaccountinthesestudies.Thisefficiencyismeasured us-ingasampleofeventsbasedonthefullLHCb detectorsimulation. Theefficiencyvariesmainlyasafunctionofthemomentumofthe lowestmomentumpionintheeventinthe D0restframe,andthis
efficiencyvariationisparameterised. However,thedependenceof theefficiency on thisparametrisation is relatively weak. For fur-therstudies the efficiency, based on theparametrisation, is then appliedtothesimplifiedsimulateddatasets.
Various CP asymmetries are introduced by modifying, for a chosen D0 flavour, either the magnitude or the phase of the
dominant amplitude contributions: a1
(
1260)
+→
ρ
0(
770)
π
+ (inS-wave)and
ρ
0(
770)
ρ
0(
770)
(both P-waveandD-wave). There-sultingsensitivitiesare shownin Table 2.The p-values,including their statisticaluncertainties,are obtainedfromfits ofGEV func-tions.
The asymmetry significances for each simulated event are shown in Fig. 2 for P -even and P -odd CP-violation tests and projected onto invariant masses of the selected three-pion and two-pion subsystems. The CP violation is introduced as a phase differenceineitherthea1
(
1260)
+ orP-waveρ
0(
770)
ρ
0(
770)
am-plitudes(see Sect.6). The shapesof the regions with significant asymmetry thatare visiblein theone-dimensional projectionsin Fig. 2 cannot be easily interpreted in terms of the amplitudes
and phase differences of the contributing resonances; the two-dimensional spectra are not easy to understand either. However, repeating thisexercise fordifferent scenarios ofCP violation, the observed features are found to be sufficiently distinguishable to helpidentifytheoriginofanyCP asymmetryinthedata.
Thesensitivityofthemethodalsodependsonthechoiceofthe effectradius
δ
.Studiesindicategoodstabilityofthemeasured sen-sitivityforvaluesofδ
from0.
3 to1 GeV2/
c4,whicharewellabovethe resolution of the di j and small compared to the size of the
phasespace. Thevalue
δ
=
0.
5 GeV2/
c4 yieldsthebestsensitivity tomostoftheCP-violationscenariosstudiedandwaschosen,prior to thedata unblinding,asthe defaultvalue. The optimalδ
value mayvarywithdifferentCP-violationscenarios.Hence,thefinal re-sultsarealsoquotedforvaluesof0.
3 GeV2/
c4and0.
7 GeV2/
c4.7. Systematiceffects
There are two main sources of asymmetry that maydegrade or bias the energy test. One is an asymmetry that may arise from background andthe other is dueto detectionand produc-tion asymmetries that could vary across phase space. The effect of theseasymmetries is studied for both the P -even and P -odd CP-violationmeasurements.
Backgroundasymmetriesaretestedbyapplyingtheenergytest toeventsinthesidebandssurroundingthesignalregioninthe
m
vs. m
(
π
+π
−π
+π
−)
plane. These events are randomly split into 11subsamplescontainingthesamenumberofbackgroundevents asexpectedtocontributeunderthesignalpeak.The p-values ob-tainedare compatiblewithauniformdistribution andno signifi-cantasymmetryisfound; p-valuesrangebetween2% and96% (4% and 91%) for P -even ( P -odd). As the background present in the signalregionisfoundtobeCP symmetric,nocorrectionisapplied inthe T valuecalculationdiscussedinSect.5.Positivelyandnegativelychargedpionsinteractdifferentlywith matter;thedifferencesintheinelasticcross-sectionsare momen-tumdependent andsignificant forlow-momentumparticles [19]. Thepresenceofthe
π
+π
−pairsinthefinalstatemakesthe detec-tionasymmetriescanceltofirstorder.Residual localasymmetries mayremainincertain regions ofthephasespacewhereπ
+ andπ
− havedifferentkinematicdistributions.Theseeffectsaretested using the Cabibbo-favoured decay D0→
K−π
+π
−π
+ as a con-trolmode.Thischannelisaffectedbykaondetectionasymmetries, whichareknowntobelargerthanpiondetectionasymmetriesand thus should serve as a conservative test. The data sample is ob-tainedwiththesamekinematicselectioncriteriaasforthesignal channelandimposingrequirementsonthecandidateK±particles identifiedusinginformationfromthering-imagingCherenkov de-tectors.Thecontrolsampleissplitintotensubsets,eachofwhich contains approximately the same amount of data as the signal sample. Theenergytest yields resultscompatiblewitha uniform distributionof p-valueswithvaluesbetween3% and87% (8% and 74%)forP -even( P -odd),whichisconsistentwiththeassumption thatthissourceofasymmetry isbelowthecurrentlevelof sensi-tivity.Asymmetriesinthesoftpiondetection,althoughlargely uncor-relatedwith the D0 phase space, arereduced usingfiducial cuts
(see Sect.3). Charmmesons produced inthe pp interactions ex-hibit a production asymmetry up to the percent level, which is slightly dependent on the meson kinematics [27,28]. More D∗−
particles areobserved than D∗+, givingriseto aglobal asymme-try,towhichtheappliedmethodisinsensitivebyconstruction.No significantlocalCP asymmetryisexpectedowingtothesmall ob-servedcorrelationoftheD∗momentumandtheD0 phasespace.
Fig. 2. (a,b)DistributionofpermutationT -valuesfittedwithaGEVfunctionforthesimulatedsampleandshowingthemeasuredT -valueasaverticalline,and(c,d,e, f) local asymmetrysignificances.LeftcolumnplotsareforaP -evenCP-violationtestwitha3◦phaseCP violationintroducedinthea1(1260)+resonance(seetext),projectedonto the(c)m(π1π2π3)and(e)m(π1π2)axes.Rightcolumnplotsarefor a P -oddCP-violationtest with3◦ phaseCP violationintroducedintheP-waveρ0(770)ρ0(770) resonanceprojectedontothesameaxes.Inplots(c,d,e,f) thegreyareacorrespondstocandidateswithacontributiontotheT -valueoflessthanone standarddeviation. Thepink(blue)areacorrespondstocandidateswithapositive(negative)contributiontotheT -value.Light,mediumordarkshadesofpinkandbluecorrespondtobetween oneandtwo,twoandthree,andmorethanthreestandard-deviationcontributions,respectively.(Forinterpretationofthereferencestocolourinthisfigurelegend,the readerisreferredtothewebversionofthisarticle.)
8. Resultsandconclusions
The application of the energy test to all selected D0
→
π
+π
−π
+π
− candidates using an effective radius ofδ
=
0
.
5 GeV2/
c4 yields T=
1.
10×
10−6 in the P -even CP-violationtest and T
=
2.
11×
10−6 in the P -odd CP-violation test. Thepermutation T value distributions are shown in Fig. 3 (a, b). By counting the fraction of permutationswith a T value above the nominalT valueinthedata,a p-valueof
(
4.
6±
0.
5)
% isobtained in the P -even CP-violation test and(
0.
6±
0.
2)
% in the P -odd CP-violationtest.ThecentralvalueoftheP -oddresultwould cor-respondto beingatleast 2.7standard deviations fromthe meanof a normal distribution. The significance levels of the Ti values
are shown in Fig. 3 (c, e) projected onto the m
(
π1π2π3
)
axis, and Fig. 3(d, f) projected onto them(
π1π2
)
axis.In the P -eventest, a small phase space region contains candidates with a lo-cal negative asymmetry exceeding 1
σ
significance. Furthermore, inthe P -oddtest,candidateswithalocalpositiveasymmetry ex-ceeding 2σ
significance are seen in a small phase-space region dominatedbytheρ
0 resonance,whichcanbecomparedwiththecorresponding plotsin Fig. 2.Varying the effectiveradius results inthe p-valueslistedinTable 3.Thecentralvaluesofthe P -odd
results for
δ
=
0.
3 GeV2/
c4 correspond to morethan three stan-dard deviations fromthe mean of a normal distribution but theFig. 3. (a,b)DistributionofpermutationT -valuesfittedwithaGEVfunctionandshowingtheT -valueofthedatatestsasaverticalline,and(c,d, e, f) localasymmetry significances.Leftcolumnplotsareforthe P -evenCP-violationtest, projectedontothe(c)m(π1π2π3)and (e)m(π1π2)axes.Rightcolumn plotsarefor the P -odd
CP-violationtestprojectedontothesameaxes.Inplots(c,d,e,f)thegreyareacorrespondtocandidateswithacontributiontothe T -valueoflessthanone standard deviation.IntheP -evenCP violationtestthepositive(negative)asymmetrysignificanceissetfortheD0candidateshavingpositive(negative)contributiontothemeasured
T value.Inthe P -oddCP violationtestthepositive(negative)asymmetrysignificanceissetforsampleI+IV havingpositive(negative)contributiontothemeasuredT
value(seeSect.5).Thepink(blue)areacorrespondstocandidateswithapositive(negative)contributiontotheT -value.Light,mediumordarkshadesofpinkandblue correspondtobetweenoneandtwo,twoandthree,andmorethanthreestandarddeviationcontributions,respectively.(Forinterpretationofthereferencestocolourinthis figurelegend,thereaderisreferredtothewebversionofthisarticle.)
Table 3
Resultsforthe P -evenandP -oddCP-violationtestsforthreedifferentvaluesoftheeffectiveradiusδ. Thep-valuesobtainedwithboththecountingandGEVfittingmethodsaregiven(seetext).The count-ingmethodisthedefaultmethod.
δ[GeV2
/c4] p-value P -even p-value P -odd
Counting GEV fit Counting GEV fit
0.3 (0.88±0.26)% (0.78±0.10)% (0.24±0.14)% (0.28±0.04)% 0.5 (4.6±0.5)% (4.8±0.3)% (0.63±0.20)% (0.34±0.05)% 0.7 (16±2)% (17±2)% (0.83±0.48)% (0.52±0.16)%
significance falls below this level when considering their uncer-tainties.
In summary, a search for time-integrated CP violation in the Cabibbo-suppresseddecayD0
→
π
+π
−π
+π
− isperformedusing a novel unbinned model-independent technique. This is the first applicationoftheenergytesttofour-bodydecaysandextendsthe approach toallow P -odd CP violation searches. Thisanalysis has thebestsensitivityfroma singleexperimentto P -even CPviola-tionandisthe firsttestfor P -oddCP violationinthisdecay.The dataarefoundtobemarginallyconsistent withthehypothesisof
CP symmetryatthecurrentlevelofprecision.
Acknowledgements
We express our gratitude to our colleagues in the CERN ac-celerator departments forthe excellent performance of the LHC. We thank the technical and administrative staff atthe LHCb in-stitutes. We acknowledge support from CERN and from the na-tional agencies: CAPES, CNPq, FAPERJ and FINEP (Brazil); NSFC (China); CNRS/IN2P3 (France); BMBF, DFG and MPG (Germany); INFN (Italy);FOMandNWO (TheNetherlands);MNiSW andNCN (Poland);MEN/IFA (Romania);MinES andFASO (Russia);MINECO (Spain);SNSFandSER(Switzerland);NASU(Ukraine);STFC(United Kingdom); NSF (USA). We acknowledge the computing resources that are provided by CERN, IN2P3 (France), KIT and DESY (Ger-many), INFN (Italy), SURF (The Netherlands), PIC (Spain), GridPP (UnitedKingdom), RRCKIandYandexLLC(Russia), CSCS (Switzer-land),IFIN-HH(Romania),CBPF(Brazil),PL-GRID(Poland)andOSC (USA).This workwas supported inpartby an allocation of com-puting time fromtheUniversity ofManchester andtheOhio Su-percomputerCenter.We areindebted tothecommunities behind the multiple open source software packages on which we de-pend.Individual groupsor members have received support from AvHFoundation(Germany),EPLANET,MarieSkłodowska-Curie Ac-tionsandERC(EuropeanUnion),ConseilGénéraldeHaute-Savoie, LabexENIGMASSandOCEVU,RégionAuvergne(France),RFBRand YandexLLC (Russia),GVA, XuntaGalandGENCAT(Spain), Herchel SmithFund, The RoyalSociety, RoyalCommission forthe Exhibi-tionof1851andtheLeverhulmeTrust(UnitedKingdom).
References
[1]FOCUScollaboration,J.M.Link,etal.,Studyofthe D0→π−π+π−π+decay,
Phys.Rev.D75(2007)052003,arXiv:hep-ex/0701001.
[2]Y.Grossman,A.L.Kagan,Y.Nir,NewphysicsandCP violationinsinglyCabibbo suppressedD decays,Phys.Rev.D75(2007)036008,arXiv:hep-ph/0609178.
[3] HeavyFlavorAveragingGroup,Y.Amhis,etal.,Averagesofb-hadron,c-hadron,
andτ-leptonpropertiesasofsummer2014,arXiv:1412.7515,updatedresults andplotsavailableathttp://www.slac.stanford.edu/xorg/hfag/.
[4]LHCbcollaboration,R.Aaij,etal.,Model-independentsearchforCP violationin
D0→K−K+π+π− andD0→π−π+π−π+decays,Phys.Lett.B726(2013)
623,arXiv:1308.3189.
[5]LHCbcollaboration,R.Aaij,etal.,SearchforCP violationinD0→π−π+π0
decayswiththeenergytest,Phys.Lett.B740(2015)158,arXiv:1410.4170.
[6]G.Valencia,AngularcorrelationsinthedecayB→VV andCP violation,Phys. Rev.D39(1989)3339.
[7]A.Datta,D.London,Triple-productcorrelationsinB→V1V2decaysandnew
physics,Int.J.Mod.Phys.A19(2004)2505,arXiv:hep-ph/0303159.
[8]C. Parkes, S. Chen, J. Brodzicka, M. Gersabeck, G. Dujany, W. Barter, On model-independent searches for direct CP violation in multi-body decays, arXiv:1612.04705.
[9]LHCbcollaboration,A.A.AlvesJr.,etal.,TheLHCbdetectorattheLHC,J. In-strum.3(2008)S08005.
[10]LHCbcollaboration,R.Aaij,etal.,LHCbdetectorperformance,Int.J.Mod.Phys. A30(2015)1530022,arXiv:1412.6352.
[11]T.Sjöstrand,S.Mrenna,P.Skands,AbriefintroductiontoPYTHIA8.1,Comput. Phys.Commun.178(2008)852,arXiv:0710.3820.
[12]I.Belyaev,etal.,HandlingofthegenerationofprimaryeventsinGauss,the LHCbsimulationframework,J.Phys.Conf.Ser.331(2011)032047.
[13]D.J.Lange,TheEvtGenparticledecaysimulationpackage,Nucl.Instrum. Meth-odsA462(2001)152.
[14]Geant4collaboration,J.Allison,K.Amako,J.Apostolakis,H.Araujo,P.A.Dubois, etal.,Geant4developmentsandapplications,IEEETrans.Nucl.Sci.53(2006) 270;
Geant4collaboration,S.Agostinelli,etal.,Geant4:asimulationtoolkit,Nucl. Instrum.MethodsA506(2003)250.
[15]M.Clemencic,etal.,TheLHCbsimulationapplication,Gauss:design,evolution andexperience,J.Phys.Conf.Ser.331(2011)032023.
[16]W.D.Hulsbergen,DecaychainfittingwithaKalmanfilter,Nucl.Instrum. Meth-odsA552(2005)566,arXiv:physics/0503191.
[17]LHCbcollaboration,R.Aaij,etal.,EvidenceforCP violationintime-integrated
D0→h−h+decayrates,Phys.Rev.Lett.108(2012)111602,arXiv:1112.0938.
[18]E.Byckling,K.Kajantie,ParticleKinematics,Wiley,NewYork,1973.
[19]ParticleDataGroup,C.Patrignani,etal.,Reviewofparticlephysics,Chin.Phys. C40(2016)100001.
[20]LHCbcollaboration,R. Aaij,et al.,Search forCP violation usingT -odd cor-relationsinD0→K+K−π+π− decays,J.HighEnergy Phys.10(2014)005,
arXiv:1408.1299.
[21]B.Aslan,G.Zech,New testforthe multivariatetwo-sampleproblem based onthe conceptofminimumenergy,J.Stat. Comput.Simul.75(2005)109, arXiv:math/0309164.
[22]B.Aslan, G. Zech,Statisticalenergy as atool for binning-free,multivariate goodness-of-fittests, two-sample comparisonand unfolding, Nucl. Instrum. MethodsA537(2005)626.
[23]M.Williams,Observing CP violation inmany-bodydecays,Phys. Rev.D84 (2011)054015,arXiv:1105.5338.
[24] NVIDIACorporation,Thrustquickstartguide,DU-06716-001(2014). [25]P. d’Argent, et al., Amplitude analyses of D0→π+π−π+π− and D0→
K+K−π+π−decays,arXiv:1703.08505.
[26]CLEO collaboration, M. Artuso, et al., Amplitude analysis of D0 →
K+K−π+π−,Phys.Rev.D85(2012)122002,arXiv:1201.5716.
[27]LHCbcollaboration,R.Aaij,etal.,MeasurementoftheD±production asym-metryin7TeVppcollisions,Phys.Lett.B718(2013)902,arXiv:1210.4112.
[28]LHCb collaboration, R. Aaij, et al., Measurement of the D+s −D−s
pro-ductionasymmetry in 7 TeV pp collisions, Phys. Lett. B 713 (2012) 186, arXiv:1205.0897.
LHCbCollaboration