Search for new phenomena in events with a photon and missing transverse momentum in $\mathit{pp}$ collisions at $\sqrt{s}=13$ TeV with the ATLAS detector

P u b l i s h e d f o r S IS S A b y S p r i n g e r R e c e i v e d : A pril 6, 2016

R e v i s e d : M ay 18, 2016 A c c e p t e d : M ay 31, 2016 P u b l i s h e d : June 9, 2016

Search for new phenomena in events with a photon and m issing transverse momentum in p p collisions at

√ s = 13 T e V with the A T L A S detector

T h e A T L A S collaboration

E-m ail: atlas.publications@cern.ch

Ab s t r a c t:

R esults of a search for new phenom ena in events w ith an energetic ph oton and large m issing tran sv erse m om entum w ith th e ATLAS experim ent a t th e Large H adron Collider are rep orted. T he d a ta were collected in p ro to n -p ro to n collisions a t a centre- of-mass energy of 13TeV and correspond to an integ rated lum inosity of 3.2 fb - 1 . T he observed d a ta are in agreem ent w ith th e S ta n d ard M odel expectations. Exclusion lim its are presented in m odels of new phenom ena including p air produ ctio n of d a rk m a tte r candidates or large e x tra sp atial dim ensions. In a simplified m odel of d a rk m a tte r and an axial-vector m ediator, th e search excludes m ed iator m asses below 710 GeV for d a rk m a tte r can d idate m asses below 150 GeV. In an effective th eo ry of d a rk m a tte r production, values of the suppression scale M* up to 570 GeV are excluded and th e effect of tru n c a tio n for various coupling values is rep orted . For th e A D D large e x tra sp atial dim ension m odel th e search places m ore string ent lim its th a n earlier searches in th e sam e event topology, excluding M d up to a b o u t 2.3 (2.8) TeV for two (six) ad ditional sp atial dim ensions; th e lim its are reduced by 20-40% depending on th e num ber of ad ditio nal sp atial dim ensions when applying a tru n c a tio n procedure.

Ke y w o r d s:

H adron-H adron scatterin g (experim ents)

ArXi y ePr i n t:

1604.01306

O p e n A c c e s s, C opyright CER N,

for th e benefit of th e ATLAS C ollaboration.

A rticle funded by S C O A P 3.

d o i:10.1007/JH E P06(2016)059

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

Contents

1 I n tr o d u c tio n 1

2 T h e A T L A S d e t e c t o r 3

3 M o n te C a r lo s im u la tio n s a m p le s 4

4 E v e n t r e c o n s t r u c tio n 6

5 E v e n t s e le c tio n 7

6 B a c k g r o u n d e s t im a t io n 8

6.1 Z

and W

backgrounds 9

6.2

+ je ts background 9

6.3 Fake photons from m isidentified electrons 10

6.4 Fake photons from m isidentified je ts 10

6.5 B eam -induced background 10

6.6

F in al background estim atio n 11

7 R e s u lts 11

8 S y s t e m a t ic u n c e r ta in t ie s 12

9 I n t e r p r e t a tio n o f r e s u lts 14

10 C o n c lu s io n 17

T h e A T L A S c o lla b o r a tio n 24

1 In tr o d u c tio n

T heories of d ark m a tte r (D M ) or large e x tra sp atial dim ensions (L E D ) predict th e pro­

d u ctio n of events th a t contain a high transverse m om entum ( p t) p h o ton and large m issing tran sv erse m om entum (referred to as

+ E “ lss events) in pp collisions at a higher rate th a n is expected in th e S ta n d ard M odel (SM ). A sam ple of

+ ETpiss events w ith a low expected co n trib u tio n from SM processes provides powerful sensitivity to m odels of new phenom ena [1- 5].

T he ATLAS [

, 7] and CMS [

, 9] collaborations have rep o rted lim its on various m odels based on searches for an excess in

+ E™ ss events using pp collisions at centre-of-m ass energies of yfs = 7 and

TeV (LHC R u n 1). T his p a p e r rep o rts th e results of a search for new phenom ena in

+ ETpiss events in pp collisions at ^/s = 13 TeV.

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

A lthough th e existence of DM is well established [10], it is not explained by curren t theories. One c an d id ate is a weakly in teracting m assive particle (W IM P, also denoted by x ), which has an in teractio n stre n g th w ith SM particles near th e level of th e weak interaction. If W IM P s in teract w ith quarks via a m ediato r particle, th ey could be pair- produced in pp collisions at sufficiently high energy. T he xX p air would be invisible to th e d etecto r, b u t

+ ETpiss events can be produced via rad iatio n of an in itia l-state p hoton in

].

A m odel-independent approach to d a rk m a tte r produ ctio n in pp collision is th ro u g h effective field theories (E F T ) w ith various forms of in teractio n betw een th e W IM P s and th e SM particles [11]. However, as th e typical m om entum tran sfer in pp collisions a t th e LHC could reach th e cut-off scale required for th e E F T ap proxim ation to be valid, it is crucial to present th e results of th e search in term s of m odels th a t involve th e explicit pro du ction of th e in term ed iate sta te , as shown in figure 1 (left). T his p a p e r focuses on simplified m odels assum ing D irac ferm ion DM can didates produced via an s-channel m ed iator w ith axial-vector interactio ns [12- 14]. In th is case, th e in teraction is effectively described by five p aram eters: th e W IM P m ass m x , th e m ediato r m ass m med, th e w id th of th e m ediato r r med, th e coupling of th e m ed iato r to quarks gq, and th e coupling of th e m ed iator to th e d a rk m a tte r particle gx . In th e lim it of large m ed iator m ass, these simplified m odels m ap onto th e E F T operators, w ith th e suppression scale

M* linked to m med by th e relation

T he p ap er also considers a specific E F T benchm ark, for which n either a simplified m odel com pletion nor th e simplified m odels yielding sim ilar kinem atic d istrib u tio n s are im plem ented in an event g en erato r [16]. A dim ension-7 E F T o p e ra to r w ith direct couplings betw een DM and electrow eak (EW ) bosons, and describing a co n tact in teractio n of ty p e 77XX, is used [14]. T he effective coupling to photons is p aram eterized by th e coupling stre n g th s k

and k

, which control th e stre n g th of th e coupling to th e U(1) and SU(2) gauge sectors of th e SM, respectively. In this model, d a rk m a tte r pro d uctio n proceeds via

(right). T here are four free p aram eters in th is model: th e E W coupling stren g th s k

and k

, m x , and th e suppression scale A.

T he ADD m odel of LED [17] aim s to solve th e hierarchy problem by hypothesizing th e existence of n additio nal sp atial dim ensions of size R, leading to a new fun d am ental scale M d related to th e P lanck m ass, M Planck, th ro u g h M planck w M ‘D+nR n . If these dim ensions are com pactified, a series of massive graviton (G) m odes results. Stable gravitons would be invisible to th e ATLAS detecto r, b u t if th e graviton couples to photons and is produced in association w ith a photon, th e d etecto r signature is a

+ Em iss event. Exam ples of g raviton p ro d u ctio n are illu strated in figure

.

T he search follows a stra te g y sim ilar to th e search perform ed using th e

TeV d a ta collected durin g th e LHC R u n 1 [7] . D ue to th e increased centre-of-m ass energy, th e search presented here achieves b e tte r sensitivity for th e ADD m odel case w here direct com parison

1T he suppression scale, also referred to as A, is th e effective m ass scale of particles th a t are in tegrated o u t in an E F T . T he non-renorm alizable op erato rs are suppressed by powers of 1 / M *.

- 2 -

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

Figure 1. P ro d u c tio n of pairs of d a rk m a tte r p articles (x x ) via an explicit s-channel m ediator, m ed (left) an d p ro d u c tio n of p airs of d a rk m a tte r p articles (x x ) via an effective y y xX vertex (right).

Figure 2. G ra v ito n (G) p ro d u c tio n in m odels of large e x tra dim ensions.

w ith th e 8 TeV search result is possible, as is shown later. Different DM models, proposed in ref. [14], are also considered.

T he p ap er is organized as follows. A brief description of th e ATLAS d e te c to r is given in section 2 . T he signal and background M onte C arlo (MC) sim ulation sam ples used are described in section 3 . T he reco n struction of physics objects is explained in section 4 , and th e event selection is described in section 5. E stim atio n of th e SM backgrounds is outlined in section 6 . T he results are described in section 7 and th e system atic uncertainties are given in section 8 . T he in te rp re ta tio n of results in term s of m odels of new phenom ena including p air pro du ctio n of d a rk m a tte r can d idates or large e x tra sp atial dim ensions is described in section 9 . A sum m ary is given in section 10.

2 T h e A T L A S d e te c to r

T he ATLAS d e te c to r [18] is a m ulti-purpose particle physics a p p a ra tu s w ith a forw ard- backw ard sym m etric cylindrical geom etry and near 4n coverage in solid angle.2 T he inner track ing d e te c to r (ID) covers th e p seudorapidity range |n| < 2.5, and consists of a silicon

2ATLAS uses a right-handed coordinate system w ith its origin a t th e nom inal in teractio n point (IP) in th e centre of th e detecto r and th e z-axis along th e beam pipe. T he x-axis points from th e IP to th e centre of th e LHC ring, and th e y-axis points upw ard. Cylindrical coordinates (r, d) are used in th e transverse plane, d being th e azim uthal angle around th e beam pipe. T he pseudorapidity is defined in term s of th e p olar 6 angle as n = — ln [tan(6/2)].

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

pixel d etector, a silicon m icrostrip d etecto r, and, for |n| <

.

, a straw -tu b e tra n sitio n rad i­

atio n track er (T R T ). D uring th e LHC shutdow n in 2013-14, an ad d ition al inner pixel layer, known as th e insertable B-layer [19], was added aroun d a new, sm aller radius beam pipe.

T he ID is surrounded by a th in superconducting solenoid providing a 2 T m agnetic field. A high-granularity lead /liq u id -arg o n sam pling electrom agnetic calorim eter covers th e region

|n| < 3.2 and is segm ented longitudinally in shower d ep th . T he first layer, w ith high g ran u­

larity in th e n direction, is designed to allow efficient d iscrim ination betw een single photon showers and two overlapping photons originating from a n

decay. T he second layer col­

lects m ost of th e energy deposited in th e calorim eter in electrom agnetic showers in itiated by electrons or photons. Very high energy showers can leave significant energy deposits in th e th ird layer, which can also be used to correct for energy leakage beyond th e EM calorim e­

ter. A steel/scin tillato r-tile calorim eter provides hadronic coverage in th e range |n| < 1.7.

T he liquid-argon technology is also used for th e hadronic calorim eters in th e end-cap region 1.5 < |n| < 3.2 and for electrom agnetic and hadronic m easurem ents in th e forw ard region up to |n| = 4.9. T he m uon sp ectro m eter (MS) surrounds th e calorim eters. It consists of th re e large air-core sup erconducting toroidal m agnet system s, precision track ing cham bers providing accu rate m uon tracking o ut to |n| = 2.7, and fast d etectors for triggering in th e region |n| < 2.4. A two-level trig g er system is used to select events for offline analysis [20].

3 M o n te C arlo sim u la tio n sa m p le s

Several MC sim ulated sam ples are used to estim ate th e signal acceptance, th e d etecto r efficiency and to help in th e estim atio n of th e SM background contributions.

For all th e DM sam ples considered here, th e values of th e free param eters and th e event generation settings were chosen following th e recom m endations given in ref. [14].

Sam ples of DM p ro d uction in simplified m odels are generated via an s-channel m edi­

a to r w ith axial-vector interactions. T he gq coupling is set to be universal in q u ark flavour and equal to 0.25, gx is set to 1.0, and r med is com puted as th e m inim um w id th allowed given th e couplings and masses. A grid of points in th e m x - m med plane is generated. T he p a rto n d istrib u tio n function (P D F ) set used is NNPDF30_lo_as_0130 [21]. T he program M G 5_aM C @ N L O v2.2.3 [22] is used to generate th e events, in conjunction w ith

8.186 [23] w ith th e N N PD F2.3L O P D F set [24, 25] and th e A14 set of tu n ed p aram e­

ters (tune) [26]. A p h o to n w ith a t least 130 GeV of tran sverse m om entum is required in M G 5_aM C @ N L O . For a fixed m x , higher m med leads to h ard er p T and Em iss spectra. For a very heavy m ed iato r (> 10 TeV), E F T conditions are recovered.

For DM sam ples from an E F T m odel involving dim ension-7 op erato rs w ith a contact in teraction of ty p e

x x , th e p aram eters which only influence th e cross section are set to

= 1.0 and A = 3.0 TeV. A scan over a range of values of m x is perform ed. T he settings of th e generators, P D F s, underlying-event tu n e and generator-level requirem ents are th e sam e as for th e simplified m odel DM sam ple generation described above.

Signal sam ples for ADD m odels are sim ulated w ith th e

8 .186 g enerator, using th e N N PD F2.3L O P D F w ith th e A14 tune. A requirem ent of pTmin > 100 GeV, where

- 4 -

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

leading-order (LO) m a trix elem ents for th e

^

process to increase th e efficiency of event generation. Sim ulations are ru n for two values of th e scale p a ra m ete r M D (2.0 and 3.0 TeV) and w ith th e num ber of e x tra dim ensions, n, varied from two to six.

For W /Z

backgrounds, events containing a charged lepton and n eu trin o or a lepton pair (lepton is an e, y or t ), to g eth e r w ith a p h o ton and associated je ts are sim ulated using th e

2.1.1 g en erato r [27]. T he m a trix elem ents including all diagram s w ith th re e electrow eak couplings are calculated w ith up to th re e p arto n s a t LO and m erged w ith

p a rto n shower [28] using th e M E+PS@ L O p rescription [29]. T he CT10 P D F set [30] is used in conjunction w ith a dedicated p a rto n shower tu n in g developed by th e

Sh e r p a

autho rs. For

* /Z events w ith th e Z decaying to charged particles a requirem ent on th e d ilepton invariant m ass of m u > 10 GeV is applied a t g enerato r level.

E vents containing a p hoton w ith associated je ts are also sim ulated using

2.1.1, gen erated in several bins of p h oton p t from 35 GeV up to larger th a n 4 TeV. T he m atrix elem ents are calculated a t LO w ith up to th re e p a rto n s (lowest p T slice) or four parto n s and m erged w ith

p a rto n shower using th e M E+PS@ L O prescription. T he CT10 P D F set is used in conjunction w ith th e d edicated p a rto n shower tuning.

For W /Z + je ts backgrounds, events containing W or Z bosons w ith associated jets are again sim ulated using

2.1.1. T he m a trix elem ents are calculated for up to two p a rto n s a t NLO and four p a rto n s a t LO using th e Com ix [31] and O penLoops [32] m atrix elem ent generators and m erged w ith

p a rto n shower using th e M E+PS@ N LO pre­

scription [33]. As in th e case of th e

+ je ts sam ples, th e CT10 P D F set is used to g eth er w ith th e dedicated p a rto n shower tu ning. T he W /Z + je ts events are norm alized to NNLO cross sections [34]. These sam ples are also generated in several p T bins.

M ulti-jet processes are sim ulated using th e

8.186 generator. T he A14 tu n e is used to g eth er w ith th e N N PD F2.3L O P D F set. T he E v tG e n v1.2.0 program [35] is used to sim ulate th e b o tto m and charm hadron decays.

D iboson processes w ith four charged leptons, th re e charged leptons and one n eu trin o or two charged leptons and two neutrinos are sim ulated using th e

2.1.1 generator. T he m a trix elem ents contain all diagram s w ith four electroweak vertices. T h ey are calculated for up to one p a rto n (for eith er four charged leptons or two charged leptons and two neutrinos) or zero parto n s (for th re e charged leptons and one n eutrino) at NLO, and up to th re e parto n s a t LO using th e Com ix and O penLoops m a trix elem ent g enerators and m erged w ith

p a rto n shower using th e M E+PS@ N LO prescription. T he CT10 P D F set is used in conjunction w ith th e d edicated p a rto n shower tu n ing. T he g enerator cross sections are used in th is case, which are a t NLO.

For th e generation of t t and single to p quarks in th e W t and s-channel, th e

B o x v2 [36, 37] g en erato r is used, w ith th e CT10 P D F set used in th e m a trix elem ent calculations. For all to p processes, to p -q u ark spin correlations are preserved. For t-channel production, to p quarks are decayed using M adSpin [38]. T he p a rto n shower, fragm entation, and th e underlying event are sim ulated using

6.428 [39] w ith th e CTEQ 6L1 [40]

P D F sets and th e corresponding P eru g ia 2012 tu n e [41]. T he to p m ass is set to 172.5 GeV.

T he E v tG en v1.2.0 program is used for properties of th e b o tto m and charm hadron decays.

M ultiple pp in teractions in th e sam e or neighbouring bunch crossings superim posed on th e h ard physics process (referred to as pile-up) are sim ulated w ith th e soft QCD

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

processes of

8.186 using th e A2 tu n e [42] and th e M STW 2008LO P D F set [43].

T he events are rew eighted to accu rately reproduce th e average num ber of interactions per bunch crossing in d a ta .

All sim ulated sam ples are processed w ith a full ATLAS d etecto r sim ulation [44] based on

[45]. T he sim ulated events are recon stru cted and analysed w ith th e sam e analysis chain as for th e d a ta , using th e sam e trigger and event selection c riteria discussed in section 5.

4 E v en t r e c o n str u c tio n

P h o to n s are reco n stru cted from clusters of energy deposits in th e electrom agnetic calorim e­

te r m easured in projective towers. C lusters w ith o u t m atching track s are classified as un­

converted p h oton candid ates. A p h oton is considered as a converted photon can d id ate if it is m atched to a pair of track s th a t pass a requirem ent on T R T -h its [46] and form a vertex in th e ID which is consistent w ith originating from a massless particle, or if it is m atched to a single tra c k passing a T R T -h its requirem ent and has a first hit after th e innerm ost layer of th e pixel d etecto r. T he p h oton energy is corrected by applying th e energy scales m easured w ith Z ^ e+ e- decays [47]. T he tra je c to ry of th e p hoton is reco n structed using th e longi­

tu d in a l (shower d ep th ) segm entation of th e calorim eters and a co n strain t from th e average collision point of th e p ro to n beam s. For converted photons, th e position of th e conversion v ertex is also used if tracks from th e conversion have h its in th e silicon detectors. Id en ­ tification requirem ents are applied in order to reduce th e co n tam in atio n from n

or o th er n e u tra l hadrons decaying to two photons. T he p h o ton identification is based on th e profile of th e energy deposits in th e first and second layers of th e electrom agnetic calorim eter.

C an d id ate photons are required to have p T > 10 GeV, to satisfy th e “loose” identification c riteria defined in ref. [48] and to be w ithin |n| < 2.37. P h o to n s used in th e event selection m ust additionally satisfy th e “tig h t” identification c riteria [48] and be isolated as follows.

T he energy in th e calorim eters in a cone of size A R = ^ / (A n )

+ ( A ^ )

= 0.4 around th e cluster bary centre excluding th e energy associated w ith th e ph oton cluster is required to be less th a n 2.45 GeV + 0.022pT, w here pT is th e p T of th e p h oto n candid ate. T his cone energy is corrected for th e leakage of th e p hoton energy from th e cen tral core and for th e effects of pile-up [47].

E lectrons are reco nstru cted from clusters in th e electrom agnetic calorim eter m atched to a tra c k in th e ID. T he criteria for th e ir identification, and th e calibration steps, are sim i­

lar to those used for photons. E lectron can d idates m ust satisfy th e “m edium ” identification requirem ent of ref. [47]. M uons are identified eith er as a com bined tra c k in th e MS and ID system s, or as an ID tra c k th a t, once ex tra p o la ted to th e MS, is associated w ith a t least one tra c k segm ent in th e MS. M uon candidates m ust satisfy th e “m edium ” identification requirem ent [49]. T he significance of th e transverse im pact p aram eter, defined as th e tra n s ­ verse im pact p a ra m ete r d

divided by its estim ated uncertainty, a do, of track s w ith respect to th e p rim ary vertex

is required to satisfy |d

| /

d

|/ a do <

3T he prim ary vertex is defined as th e vertex w ith th e highest sum of th e squared transverse m om enta of its associated tracks. It is reconstructed from a t least two associated tracks w ith pT > 0.4 GeV.

- 6 -

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

3.0 for m uons. T he longitudinal im pact p a ra m ete r z0 m ust be |z0| s in 0 < 0.5 m m for b o th electrons and m uons. E lectrons are required to have p T > 7 GeV and |n| < 2.47, while m uons are required to have p T > 6 GeV and |n| < 2.7. If any selected electron shares its inner d e te c to r tra c k w ith a selected m uon, th e electron is removed and th e m uon is kept, in order to remove electron can d id ates com ing from m uon b rem sstrahlu ng followed by photon conversion.

Je ts are recon stru cted using th e an ti-k t algorithm [50, 51] w ith a radius p a ra m ete r R = 0.4 from clusters of energy deposits at th e electrom agnetic scale in th e calorim eters.

A correction used to calib rate th e je t energy to th e scale of its co n stitu en t particles [52, 53] is th e n applied. In addition, je ts are corrected for co ntribu tion s from pile-up in teractions [52].

C an d id ate jets are required to have p T > 20 GeV. To suppress pile-up jets, which are m ainly a t low

, a je t v ertex tagg er [54], based on tracking and vertexing inform ation, is applied in je ts w ith p T < 50 GeV and |n| < 2.4. Je ts used in th e event selection are required to have p T > 30 GeV and |n| < 4.5. H adronically decaying

leptons are considered as jets as in th e R u n 1 analysis [7].

To resolve am biguities which can hap p en in object reconstruction, an overlap removal procedure is perform ed in th e following order. If an electron lies w ithin A R < 0.2 of a can d id a te je t, th e je t is removed from th e event, while if an electron lies w ithin 0.2 < A R <

0.4 of a je t, th e electron is rem oved. M uons lying w ithin A R < 0.4 w ith respect to the rem aining c an d id ate je ts are removed, except if th e num ber of tracks w ith pT > 0.5 GeV associated w ith th e je t is less th a n three. In th e la tte r case, th e je t is discarded and th e m uon kept. F inally if a c an d id ate p hoton lies w ithin A R < 0.4 of a je t, th e je t is removed.

T he m om entum im balance in th e transverse plane is o b tain ed from th e negative vector sum of th e reco nstru cted and calib rated physics objects, selected as described above, and is referred to as m issing transverse m om entum , E m iss. T he symbol Em iss is used to denote its m agnitude. C alorim eter energy deposits and tracks are associated w ith a recon stru cted and identified high-pT ob ject in a specific order: electrons w ith p T > 7 GeV, photons w ith p T > 10 GeV, and je ts w ith p T > 20 GeV [55]. Tracks from th e p rim ary vertex not associated w ith any such objects (”soft te rm ” ) are also tak en into account in th e E m iss reco nstructio n [56]. This track -b ased soft term is m ore robu st against pile-up and provides a b e tte r E m iss m easurem ent in term s of resolution and scale th a n th e calorim eter-based soft term used in ref. [7].

C orrections are applied to th e objects in th e sim ulated sam ples to account for differ­

ences com pared to d a ta in object reconstruction, identification and isolation efficiencies for b o th th e selected leptons and photons and for th e vetoed leptons.

5 E v en t se le c tio n

T he d a ta were collected in pp collisions a t yfs = 13 TeV durin g 2015. T h e events for th e analysis are recorded using a trig ger requiring a t least one p hoton can d id a te w ith an online p T threshold of 120 GeV passing “loose” identification requirem ents based on th e shower shapes in th e EM calorim eter as well as on th e energy leaking into th e hadronic calorim eter from th e EM calorim eter [57]. Only d a ta satisfying beam , d e te c to r and d a ta

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

q u ality c riteria are considered. T he d a ta used for th e analysis correspond to an integ rated lum inosity of 3.2 fb-

. T he u n certain ty in th e in tegrated lum inosity is ± 5% . It is derived following a m ethodology sim ilar to th a t detailed in ref. [58], from a prelim inary calibration of th e lum inosity scale using x -y beam -separation scans perform ed in A ugust 2015.

Q uality requirem ents are applied to p h oton can didates in order to reject events con­

tain in g photons arising from in stru m en tal problem s or from non-collision background [46].

B eam -induced background is highly suppressed by applying th e c riteria described in sec­

tio n 6.5. In addition, q u ality requirem ents are applied to remove events containing candi­

d a te je ts arising from d e te c to r noise and out-of-tim e energy deposits in th e calorim eter from cosmic rays or o th er non-collision sources [59]. Events are required to have a reconstructed p rim ary vertex.

T he c riteria for selecting events in th e signal region (SR) are optim ized considering th e discovery p o ten tial for th e simplified d a rk m a tte r model. This SR also provides good sensitivity to th e o th er m odels described in section 1. E vents in th e SR are required to have Em iss > 150 GeV and th e leading p hoton has to satisfy th e “tig h t” identification criteria, to have pT > 150 GeV, |n| < 2.37, excluding th e calorim eter b a rre l/e n d -c a p tra n sitio n region 1.37 < |n| < 1.52, and to be isolated. W ith respect to th e R u n 1 analysis, a re­

o p tim ization was perform ed th a t leaded to th e following changes: a higher threshold for pT (150 GeV instead of 125 GeV) and a larger |n| region (|n| < 2.37 instead of 1.37) are used for th e leading photon. It is required th a t th e p ho ton and E m iss do not overlap in th e azim uth: A

(

, E m iss) > 0.4. E vents w ith m ore th a n one je t or w ith a je t w ith A 0 (je t, E m iss) < 0.4 are rejected. T he rem aining events w ith one je t are retained to increase th e signal acceptance and reduce system atic un certain ties related to th e m odelling of in itia l-state rad iatio n. E vents are required to have no electrons or m uons passing th e requirem ents described in section 4 . T he lepton veto m ainly rejects W /Z events w ith charged leptons in th e final sta te . For events satisfying these criteria, th e efficiency of th e trig ger used in th e analysis is 0.997+0o°8, as determ ined using a control sam ple of events selected w ith a ETpiss trigger w ith a thresh old of 70 GeV.

T he final d a ta sam ple contains 264 events, of which 80 have a converted photon, and 170 and 94 events have zero and one je t, respectively.

T he to ta l num ber of events observed in th e SR in d a ta is com pared w ith th e estim ated to ta l num ber of events in th e SR from SM backgrounds. T he la tte r is ob tain ed from a sim ultaneous fit to various control regions (CR) defined in th e following. Single-bin SR and CRs are considered in th e fit: no shape inform ation w ithin these regions is used.

6 B a ck g ro u n d e stim a tio n

T he SM background to th e

+ ETpiss final s ta te is dom inated by th e Z ( ^ v v )

process, w here th e p h oton is due to in itia l-state radiation. Secondary contribu tion s come from W

and Z

pro d u ctio n w ith unidentified electrons, m uons or w ith hadronically decaying t leptons. T here is also a co n trib u tio n from W /Z p ro d u ctio n w here a lepton or an associated rad ia te d je t is m isidentified as a photon. In addition, th ere are sm aller contribu tion s from to p -q u a rk pair, diboson,

+ je ts and m ulti-jet production.

- 8 -

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

All background estim ations are e x tra p o la ted from orthogonal d a ta sam ples. Control regions, built to be enriched in a specific background, are used to co nstrain th e norm aliza­

tio n of W / Z

and Y + jets backgrounds. T he norm alization is obtain ed via a sim ultaneous likelihood fit [60] to th e observed yields in all single-bin CRs. Poisson likelihood functions are used to m odel th e expected event yields in all regions. T he system atic uncertainties described in section

are tre a te d as G au ssian-distribu ted nuisance p aram eters in th e like­

lihood function. T he fit in th e CRs is perform ed to o b tain th e n orm alization factors for th e W y , Z y and Y + jets processes, which are th e n used to co n strain background estim ates in th e SR. T h e sam e n orm alization factor is used for b o th Z ( ^ v v )y and Z decaying to charged leptons in SR events.

T he backgrounds due to fake photons from th e m isidentification of electrons or jets in W /Z + je ts, top, diboson and m ulti-jet events are estim ated using data-d riv en techniques based on studies of electrons and je ts faking photons (see sections 6.3 and 6.4) .

6 .1 Zy a n d W y b a c k g r o u n d s

For th e estim atio n of th e W /Z y background, th re e control regions are defined by selecting events w ith th e sam e c riteria used for th e SR b u t inverting th e lepton vetoes. In th e first control region (1m uC R) th e W y co n trib u tio n is enhanced by requiring th e presence of a m uon. T he second and th ird control regions enhance th e ZY background by requiring th e presence of a p air of m uons (2m uC R) or electrons (2eleCR). In b o th 1m uC R and 2m uCR, to ensure th a t th e ETpiss sp ectrum is sim ilar to th e one in th e SR, m uons are tre a te d as non-in teracting particles in th e ETpiss reconstruction. T he sam e procedure is followed for electrons in th e 2eleCR. In each case, th e CR lepton selection follows th e sam e requirem ents as th e SR lepton veto, w ith th e ad d itio n th a t th e leptons m ust be isolated w ith “loose”

c riteria [49]. In b o th th e ZY-enriched control regions, th e d ilepton invariant m ass m u is required to be g reater th a n 20 GeV. T he norm alization of th e d om inant Z y background process is largely co n strained by th e event yields in th e 2m uC R and th e 2eleCR. T he signal co n tam in atio n in all CRs is negligible. T he expected fraction of signal events in th e 1m uCR is a t th e level of 0.15%. In th e 2m uC R and 2eleCR th e co n tam in atio n is zero due to th e requirem ent of two leptons.

6 .2 y + j e t s b a c k g r o u n d

T he Y + jets background in th e signal region consists of events w here th e je t is poorly reconstru cted and p a rtially lost, creating fake ETpiss. T his background is suppressed by th e large ETpiss and th e large je t- E m iss azim uthal sep aration requirem ents. It is estim ated from sim ulated Y + jets events corrected w ith a norm alization factor th a t is determ ined in a specific control region (P h Je tC R ), enriched in y + je ts events. This C R is defined w ith th e sam e c riteria as used for th e SR, b u t requiring 85 GeV < ETpiss < 110 GeV and azim uthal sep aratio n betw een th e p ho ton and E m iss, A

(y, E miss), to be sm aller th a n 3, to m inim ize th e co n tam inatio n from signal events. T he u p p er lim it on th e expected fraction of signal events in th e P h J e tC R has been estim ated to be a t th e level of 3%. T he e x tra p o la tio n in ETpiss of th e g a m m a + je ts background from th e C R to th e SR was checked in a validation region defined w ith higher ETpiss (125 < ETpiss < 250 GeV) and requiring A 0 (y , E m iss) < 3.0; no evidence of m ism odeling was found.

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

6 .3 Fake p h o to n s fro m m is id e n tifie d e le c tr o n s

C ontrib utions from processes in which an electron is m isidentified as a p h oton are estim ated by scaling yields from a sam ple of e+ETpiss events by an electron-to-photon m isidentification factor. T his factor is m easured w ith m u tually exclusive sam ples of e + e - and

+ e events in d a ta . To establish a pure sam ple of electrons, m ee and m eY are b o th required to be consis­

te n t w ith th e Z boson m ass to w ithin 10 GeV, and th e ETpiss is required to be sm aller th a n 40 GeV. T he m isidentification factor, calculated as th e ratio of th e num ber of

+ e to th e num ber of e + e - events, is p aram eterized as a function of pT and p seudorapidity and it varies betw een 0.8% and 2.6%. System atic uncertainties from th ree different sources are added in q u a d ra tu re : th e difference betw een m isidentification factors m easured in d a ta in two differ­

ent windows aro und th e Z m ass (5 GeV and 10 GeV), th e difference w hen m easured in Z ( ^ ee) MC events w ith th e sam e m ethod as used in d a ta com pared to using generator-level inform ation, and th e difference w hen m easured in Z ( ^ ee) and W ( ^ ev ) MC events using generator-level inform ation. Sim ilar estim ates are m ade for th e th re e control regions w ith leptons, by applying th e m isidentification factor to events selected using th e sam e c riteria as used for these control regions b u t requiring an electron instead of a photon. T he estim ated c o n trib u tio n of th is background in th e SR and th e associated erro r are rep o rted in section 7.

6 .4 Fake p h o to n s fro m m is id e n tifie d j e t s

B ackground con trib ution s from events in which a je t is m isidentified as a photon are es­

tim a te d using a sideband counting m ethod [61]. This m eth od relies on counting photon can didates in four regions of a tw o-dim ensional space, defined by th e transverse isolation energy and by th e quality of th e identification criteria. A signal region (region A) is de­

fined by p hoton cand id ates th a t are isolated w ith tig h t identification. T hree background regions are defined, consisting of p h oton can did ates which are eith er tig h t and non-isolated (region B), non -tig h t and isolated (region C) or non-tight and non-isolated (region D).

T he m eth od relies on th e fact th a t signal co n tam in atio n in th e th re e background regions is small and th a t th e isolation profile in th e non-tigh t region is th e sam e as th a t of th e background in th e tig h t region. T he num ber of background can did ates in th e signal region (N a ) is calculated by tak in g th e ratio of th e two non-tight regions (N c / N d ) m ultiplied by th e num ber of c an d idates in th e tig h t, non-isolated region (N b ). T his m eth o d is applied in all analysis regions: th e SR and th e four CRs. T he system atic u n certain ty of th e m eth o d is evaluated by varying th e criteria of tightness and isolation used to define th e four regions.

T his estim ate also accounts for th e co n trib u tio n from m ulti-jet events, which can mimic th e

+ ETpiss sign ature if one je t is m isreconstructed as a p hoton and one or m ore of th e o th er je ts are poorly reconstructed, resulting in large ETpiss. T he estim ated con trib u tio n of th is background in th e SR and th e associated erro r are rep orted in section 7.

6 .5 B e a m -in d u c e d b a c k g r o u n d

M uons from beam background can leave significant energy deposits in th e calorim eters, m ainly in th e region a t large |n|, and hence can lead to recon stru cted fake photons. These beam -induced fakes do not point back to th e p rim ary vertex, and th e p hoton tra je c to ry

- 10 -

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

F i g u r e 3. D istrib u tio n of E m iss, rec o n stru c te d tre a tin g m uons as n o n -in terac tin g p articles, in th e d a ta an d for th e b ackground in th e 1m uC R (left) an d in th e 2m uC R (right). T he to ta l background e x p e c ta tio n is norm alized to th e p ost-fit resu lt in each control region. Overflows are included in th e final bin. T he erro r b ars are statistic a l, a n d th e d ash ed b a n d includes s ta tistic a l an d sy stem atic u n c e rtain ties d ete rm in e d by a bin-by-bin fit. T he lower panel shows th e ra tio of d a ta to expected b ackground event yields.

provides a powerful rejection criterion. T he |z| position of th e intersection of th e e x tra p o ­ lated ph oton tra je c to ry w ith th e beam axis is required to be sm aller th a n 0.25 m, which rejects 98.5% of these fake photons. T he residual beam background after th e final event selection is found to be negligible, a b o u t

.

%.

6 .6 F in a l b a c k g r o u n d e s tim a t io n

B ackground estim ates in th e S R are derived from a sim ultaneous fit to th e four single­

bin control regions (1m uC R ,

m uC R ,

eleC R and P h J e tC R ) in order to assess w hether th e observed S R yield is consistent w ith th e background model. For each C R , th e inputs to th e fit are: th e num ber of events seen in th e d a ta , th e num ber of events expected from MC sim ulation for th e W /Z

and

+ je ts backgrounds, whose norm alizations are free p aram eters, and th e num ber of fake-photon events obtain ed from th e d ata-d riv en techniques. T he fitted values of th e norm alization factors for W

and Z

are kWY = 1.50 ± 0.26 and k z Y = 1.19 ± 0.21, while th e norm alization factor for th e

+ je ts background is k7+jets = 0.98 ± 0.28. T he uncertainties include those from th e various sources described in section

. T h e factor k w Y is large owing to th e d ata-M C norm alization difference in th e 1m uCR, which can p o tentially be reduced using higher-order corrections for th e V

cross sections [62], which are not available for th e selection c riteria used here.

P ost-fit distrib u tio n s of E™ ss in th e th re e lepton CRs and in th e P h J e tC R are shown in figure 3 and figure 4. T hese d istrib u tio n s illu strate th e kinem atics of th e selected events.

T h eir shape is not used in th e sim ultaneous fit, which is perform ed on th e single-bin CRs.

7 R e su lts

Table 1 presents th e observed num ber of events and th e SM background predictions in th e SR, obtain ed from th e sim ultaneous fit to th e single-bin CRs. T he sam e num bers are also

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

F i g u r e 4. D istrib u tio n of E m iss in th e d a ta and for th e b ackground in th e 2eleCR, w here E m iss is re c o n stru c te d tre a tin g electrons as n o n -in tera ctin g p articles (left) an d in th e P h J e tC R (right). T he to ta l background ex p e c ta tio n is norm alized to th e post-fit resu lt in each control region. Overflows are included in th e final b in for th e left figure. T he erro r b ars are sta tistic a l, a n d th e d ash ed b a n d includes s ta tistic a l an d sy stem atic u n certain ties d eterm in ed by a bin-by-bin fit. T h e lower panel shows th e ra tio of d a ta to exp ected b ackground event yields.

SR 1m uC R 2m uC R 2eleCR P h J e tC R

O bserved events 264 145 29 2 0 214

F itte d B ackground 295±34 145±12 27 ± 4 23 ± 3 214±15

Z ( ^ v v) 7 171±29 0.15± 0.03 0.0 0±0 . 0 0 0.0 0±0 . 0 0 8 .6± 1.4

W ( ^ ^ ) 7 58± 9 119±17 0.14±0.04 0.11± 0.03 22 ± 4

Z ( ^ ^ ) 7 3 .3± 0.6 7 .9± 1.3 26 ± 4 20 ± 3 1.2±0 . 2

7 + je ts 15±4 0.7 ± 0 .5 0.0 0±0 . 0 0 0.03±0.03 166±17

Fake p h o to n s from electrons 22± 18 1.7±1.5 0.05±0.05 0.0 0±0 . 0 0 5.8±5.1 Fake p h o to n s from je ts 26± 12 16±11 1.1±0 . 8 2.5± 1.3 9.9±3.1

P re-fit background 249±29 105±14 23 ± 2 19±2 209±50

T a b le 1. O bserved event yields in 3.2 fb- 1 com pared to expected yields from SM backgrounds in th e signal region (SR) an d in th e four control regions (C R s), as p red icted from th e sim ultaneous fit to all single-bin C Rs. T h e M C yields before th e fit are also shown. T he u n c e rta in ty includes b o th th e s ta tistic a l an d sy stem atic u n c e rtain ties described in section 8. T he individual u n ce rtain ties can be co rrelated a n d do n o t necessarily ad d in q u a d ra tu re to equal th e to ta l b ackground u ncertainty.

shown in th e th re e lepton CRs and in th e P h J e tC R . T he con trib u tio n from W /Z

w ith W /Z decaying to t includes b o th th e leptonic and th e hadronic t decays, considered in this search as jets. T h e fraction of W ( ^ t v ) and Z ( ^ t t ) w ith respect to th e to ta l background corresponds to ab o u t 12% and 0.8%, respectively. T he post-fit ETpiss d istrib u tio n and the p h o to n p T d istrib u tio n in th e SR are shown in figure 5.

8 S y s te m a tic u n c e r ta in tie s

System atic uncertain ties in th e background predictions in th e SR are presented as p ercent­

ages of th e to ta l background prediction. This prediction is obtain ed from th e sim ultaneous

- ¹² -

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

Figure 5. D istrib u tio n of E m iss (left) an d p h o to n p T (right) in th e signal region for d a ta an d for th e b ackground p red icted from th e fit in th e C Rs. Overflows are included in th e final bin. T he erro r bars are statistic a l, an d th e d ash ed b a n d includes s ta tistic a l an d sy stem atic u n c e rtain ties d eterm in ed by a bin-by-bin fit. T he exp ected yield of events from th e sim plified m odel w ith m x = 150 GeV an d m med = 500 GeV is stacked on to p of th e b ackground prediction. T he lower panel shows th e ra tio of d a ta to exp ected b ackground event yields.

fit to all single-bin CRs, which provides co n strain ts on m any sources of system atic uncer­

tainty, as th e norm alizations of th e d om inant background processes are fitted param eters.

T he dom inant system atic un certainties are sum m arised in tab le 2 .

T he to ta l background prediction uncertainty, including system atic and sta tistic a l con­

trib u tio n s, is approxim ately 11%, dom inated by th e sta tistic a l u n certain ty in th e control regions, which am ounts to approxim ately 9%. T he largest relative system atic un certain ty of 5.8% is due to th e electron fake rate. T his is m ainly driven by th e small num ber of events available for th e estim atio n of th e electron -to-ph oto n m isidentification factor yield­

ing a precision of 30-100% , depending on p T and n. P D F uncertain ties have an im pact on th e V

sam ples in each region b u t th e effect on norm alization is largely absorbed in th e fit.

T hey are evaluated following th e prescriptions of th e P D F group recom m endations [63] and using a rew eighting procedure im plem ented in th e L H A P D F Tool [64]. These uncertainties co n trib u te 2.8% to th e background p rediction u n certain ty affecting m ainly th e Z

)

background. T he un certain ty on th e je t fake ra te co ntributes a relative u n certain ty of 2.4%

and affects m ainly th e n orm alization of W ( ^ ^

)

background, while th e u n certain ty on th e m uon reco nstru ctio n and isolation efficiency gives a relative u n certain ty of 1.5% and m ainly affects th e Z ( ^ ££)

background. F inally th e u ncertain ty on th e je t energy reso­

lution accounts for 1.2% of th e u n certain ty and th e m ost affected background is

+ jets.

A fter th e fit, th e u n certain ty on th e lum inosity [58] is found to have a negligible im pact on th e background estim ation.

For th e signal-related system atics, th e P D F uncertainties are evaluated in th e sam e way described above for th e background sam ples, while QCD scale uncertainties are evaluated by varying th e renorm alization and facto rization scales by factors 2.0 and 0.5 w ith respect to th e nom inal values used in th e MC generation. T he uncertain ties due to th e choice of underlying-event tu n e used w ith

8 .186 are com puted by generating MC sam ples w ith th e a ltern ativ e underlying-event tu nes described in ref. [26].

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

Total background

T otal background u n certain ty

295

%

E lectro n fake rate 5.8%

P D F uncertainties

.

%

Je t fake rate 2.4%

M uons rec o n stru ctio n /iso latio n efficiency 1.5%

E lectrons rec o n stru c tio n /id e n tific a tio n /iso la tio n efficiency 1.3%

J e t energy resolution [65]

.

%

P h o to n energy scale

.

%

E m iss soft term scale and resolution 0.4%

P h o to n energy resolution

.

%

J e t energy scale [53]

.

%

T a b le 2. B reakdow n of th e d o m in an t sy stem atic u n certain ties in th e background estim ates.

T he u n certain ties are given relativ e to th e exp ected to ta l background yield. T h e individual u n c e rtain ties can be co rrelated and do n o t necessarily add in q u a d ra tu re to equal th e to ta l b ackground u ncertainty.

9 In te r p r e ta tio n o f r esu lts

T he 264 events observed in d a ta are consistent w ith th e prediction of 295 ± 34 events from SM backgrounds. T he results are therefore in terp reted in term s of exclusion lim its in m odels th a t would produce an excess of

+ E™ ss events. U p per bounds are calculated using a one-sided profile likelihood ratio and th e C L S technique [

, 67], evaluated using th e asy m p to tic appro x im ation [

]. T he likelihood fit includes b o th th e SR and th e CRs.

L im its on th e fiducial cross section of a p o ten tial signal beyond th e SM, defined as th e p ro d u ct of th e cross section tim es th e fiducial acceptance A, are provided. These lim its can be ex tra p o la ted w ithin some approxim ations to m odels producing

+ E™ ss events once A is known. T he value of A for a p a rticu la r m odel is com puted by applying th e sam e selection c riteria as in th e SR b u t at th e particle level; in this co m p u tatio n E ™ ss is given by th e vector sum of th e tran sv erse m om enta of all invisible particles. T he value of A is 0.43­

0.56 (0.4) for th e DM (ADD) sam ples generated for this search following th e specifications given in section 3 . T he lim it is com puted by dividing th e lim it on th e visible cross section

be 78%, corresponding to th e lowest efficiency found in th e ADD and DM m odels studied here, for which th e efficiency ranges from 78% to 91%. T he observed (expected) up p er lim its on th e fiducial cross section a x A for th e pro du ctio n of

+ E™ ss events are 17.8 (25.5) fb a t 95% confidence level (CL) and 14.6 (21.7) fb a t 90% CL. T he observed u pp er lim it at 95% CL would be 15.3 fb using th e largest efficiency value of 91%.

W hen placing lim its on specific models, th e signal-related system atic u ncertainties calculated as described in section

affecting A x e (P D F , scales, initial- and final-state

- 14 -

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

Figure 6. T he observed an d ex p ected 95% CL exclusion lim it for a sim plified m odel of d a rk m a tte r p ro d u c tio n involving an axial-vector o p erato r, D irac DM an d couplings gq = 0.25 an d gx = 1 as a function of th e d a rk m a tte r m ass m x an d th e axial-m ed iato r m ass m med. T h e plane u n d e r th e lim it curves is excluded. T he region on th e left is excluded by th e p e rtu rb a tiv e lim it. T he relic d ensity curve [70] is also shown.

rad iatio n) are included in th e sta tistic a l analysis, while th e uncertain ties affecting th e cross section (P D F , scales) are indicated as bands arou nd th e observed lim its and w ritte n as otheo.

Simplified m odels w ith explicit m ediators are robust for all values of th e m om entum tra n sfe r Q tr [14]. For th e simplified m odel w ith an axial-vector m ediator, figure 6 shows th e observed and expected contours corresponding to a 95% C L exclusion as a function of m med and m

for gq = 0.25 and gx = 1 . T he region of th e plane under th e lim it curves is excluded. T h e region not allowed due to p e rtu rb a tiv e u n ita rity violation is to th e left of th e line defined by m

= y / n /2 m med [69]. T he line corresponding to th e DM th erm al relic ab un dan ce [70] is also indicated in th e figure. T he search excludes m ediato r masses below 710 GeV for x m asses below 150 GeV.

F igure 7 shows th e contour corresponding to a 90% C L exclusion tra n sla te d to th e X -proton scatterin g cross section vs. m

plane. B ounds on th e x -p ro to n cross section are o b tained following th e procedure described in ref. [71], assum ing th a t th e axial-vector m ediator w ith couplings gq = 0.25 and gx = 1.0 is solely responsible for b o th collider x pair pro ductio n and for x-nucleon scattering . In th is plane a com parison w ith th e result from direct DM searches [72- 74] is possible. T he search provides strin gen t lim its on th e scatterin g cross section at th e o rder of 10-41cm 2 up to m

masses of ab o u t 150 GeV. T he lim it placed in this search extends to arb itra rily low values of m

, as th e acceptance at lower m ass values is th e sam e as th e one at th e lowest m

value shown here.

In th e case of th e m odel of YYXX interactions, lower lim its are placed on th e effective m ass scale M* as a function of m

, as shown in figure 8 . T he E F T is not always valid, so a tru n c a tio n procedure is applied [75]. In this procedure, th e scale at which th e E F T description becomes invalid (M cut) is assum ed to be related to M* th ro u g h M cut = g*M*, w here g* is th e E F T coupling. Events having a centre-of-m ass energy larger th a n M cut are removed and th e lim it is recom puted. T he effect of th e tru n c a tio n for various representative

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

F i g u r e 7. T h e 90% CL exclusion lim it on th e x -p ro to n sc a tte rin g cross section in a sim plifed m odel of d a rk m a tte r p ro d u c tio n involving an axial-vector o p erato r, D irac DM a n d couplings gq

= 0.25 an d gx = 1 as a function of th e d a rk m a tte r m ass m x . Also show n are resu lts from th ree d irect d a rk m a tte r search experim ents [72- 74].

Figure 8. T he observed an d ex p ected 95% CL lim its on M* for a dim ension-7 o p e ra to r E F T m odel w ith a c o n ta c t in te ra c tio n of ty p e YYXX as a function of d a rk m a tte r m ass m x . R esults w here E F T tru n c a tio n is applied are also shown, assum ing rep resen tativ e coupling values of 2, 4, 8 an d 4n.

values of g* is shown in figure 8 : for th e m axim al coupling value of 4n, th e tru n c a tio n has alm ost no effect; for lower coupling values, th e exclusion lim its are confined to a sm aller area of th e p a ra m ete r space, a n d no lim it can be set for a coupling value of unity. For very low values of M*, m ost events would fail th e centre-of-m ass energy tru n c a tio n requirem ent, therefore, th e tru n c a te d lim its are not able to exclude very low M * values. T h e search excludes m odel values of M* u p to 570 GeV and effects of tru n c a tio n for various coupling values are shown in th e figure.

In th e ADD m odel of LED , th e observed and expected 95% CL lower lim its on th e fu ndam en tal P lanck m ass M

for various values of n are shown in figure 9 . T he values of M D excluded at 95% CL are larger for larger n values: th is is explained by th e increase of th e cross section a t th e centre-of-m ass energy of 13 TeV w ith increasing n, which is an

- ¹⁶ -

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

Figure 9. T he observed an d ex p ected 95% CL lower lim its on th e m ass scale M D in th e ADD m odels of large e x tra dim ensions, for several values of th e n u m b er of e x tra dim ensions. T he u n tru n c a te d lim its from th e search of 8 TeV ATLAS d a ta [7] are show n for com parison. T he lim it w ith tru n c a tio n is also shown.

expected behaviour for values of M D which are not large w ith respect to th e centre-of- m ass energy. R esults in corp o ratin g tru n c a tio n in th e phase-space region w here th e m odel im plem entation is not valid are also shown. T h is consists in suppressing th e graviton p ro d u ctio n cross section by a facto r M D /s 2 in events w ith centre-of-m ass energy y/s > M D.

T he procedure is rep eated iteratively w ith th e new tru n c a te d lim it un til it converges, i.e., un til th e difference betw een th e new tru n c a te d lim it and th e one obtained in th e previous ite ratio n differ by less th a n 0.1a. It results in a decrease of th e 95% CL lim it on M D . T he search sets lim its th a t are m ore strin gen t th a n those from LHC R u n 1, excluding M d up to ab o u t 2.3 TeV for n = 2 and up to 2.8 TeV for n = 6; th e lim it values are reduced by 20 to 40% depending on n when applying a tru n c a tio n procedure.

10 C o n c lu sio n

R esults are rep o rted on a search for new phenom ena in events w ith a high-px p h oton and large m issing tran sv erse m om entum in pp collisions at yfs = 13 TeV a t th e LHC, using d a ta collected by th e ATLAS experim ent corresponding to an in teg rated lum inosity of 3.2 fb - 1 . T he observed d a ta are consistent w ith th e S ta n d a rd M odel expectations. T he observed (expected) u p p er lim its on th e fiducial cross section for th e p ro d uctio n of events w ith a p h oton and large m issing transverse m om entum are 17.8 (25.5) fb a t 95% CL and 14.6 (21.7) fb a t 90% CL. For th e simplified DM m odel considered, th e search excludes m ediator m asses below 710 GeV for x masses below 150 GeV. For th e E W -E F T m odel values of M* up to 570 GeV are excluded and th e effect of tru n c a tio n for various coupling values is rep o rted . For th e ADD m odel th e search sets lim its th a t are m ore stringent th a n in th e R u n 1 d a ta search, excluding M D up to a b o u t 2.3 TeV for n = 2 and up to 2.8 TeV for n = 6; th e lim it values are reduced by 20-40% depending on n w hen applying a tru n c a tio n procedure.

J H E P 0 6 ( 2 0 1 6 ) 0 5 9

A c k n o w le d g m e n ts

We th a n k C E R N for th e very successful op eratio n of th e LHC, as well as th e su p p o rt staff from our in stitu tio n s w ith o u t whom ATLAS could not be o p erated efficiently.

We acknowledge th e su p p o rt of A N P C yT , A rgentina; Y erPhI, A rm enia; ARC, A us­

tralia; B M W F W and F W F , A ustria; ANAS, A zerbaijan; SSTC, Belarus; C N P q and FA PESP, Brazil; NSERC , NRC and C FI, C anada; CERN; C O N IC Y T , Chile; CAS, M O ST and NSFC, C hina; C O LC IEN C IA S, Colombia; M SM T CR, M PO C R and VSC CR, Czech Republic; D N R F and DN SRC, D enm ark; IN 2P3-C N R S, C E A -D S M /IR F U , France;

G N SF, Georgia; B M B F, H G F, and M PG , G erm any; GSRT, Greece; R G C , H ong K ong SAR, China; ISF, I-C O R E and Benoziyo C enter, Israel; IN FN , Italy; M E X T and JSPS , Jap an ; C N R ST, Morocco; FO M and N W O , N etherlands; RCN, Norway; M NiSW and NCN, Poland; F C T , P ortu gal; M N E /IF A , R om ania; M ES of R ussia and NRC KI, R ussian Fed­

eration; JIN R ; M ESTD , Serbia; MSSR, Slovakia; A RRS and MIZS, Slovenia; D S T /N R F , South Africa; M IN EC O , Spain; SRC and W allenberg F oundation, Sweden; SERI, SNSF and C antons of B ern and Geneva, Sw itzerland; M O ST, Taiwan; TA EK , Turkey; STFC, U nited K ingdom ; D O E and N SF, U nited S tates of A m erica. In addition, individual groups and m em bers have received su p p o rt from B C K D F, th e C an ad a Council, CA N A RIE, CRC, C om pute C anada, F Q R N T , and th e O n tario Innovation T ru st, C anada; E P L A N E T , ER C, F P 7 , Horizon 2020 and M arie Skłodowska-Curie A ctions, E u ro p ean Union; Investissem ents d ’Avenir L abex and Idex, ANR, Region A uvergne and F on datio n P a rta g e r le Savoir, France;

D FG and AvH F oundation, Germ any; H erakleitos, T hales and A risteia program m es co­

financed by E U -E S F and th e G reek N SRF; BSF, G IF and M inerva, Israel; B R F , Norway;

G en eralitat de C atalunya, G en eralitat V alenciana, Spain; th e Royal Society and Lever- hulm e T rust, U nited Kingdom .

T he crucial com puting su p p o rt from all W LC G p a rtn e rs is acknowledged gratefully, in p a rticu la r from C E R N and th e ATLAS Tier-1 facilities at T R IU M F (C anad a), N D G F (D enm ark, Norway, Sweden), CC -IN 2P3 (France), K IT /G rid K A (G erm any), IN FN -C N A F (Italy ), N L-T1 (N etherlands), P IC (Spain), ASGC (Taiw an), RAL (U.K.) and BNL (U.S.A.) and in th e Tier-2 facilities worldwide.

O p e n A c c e s s .

This article is d istrib u ted under th e term s of th e C reative Com m ons A ttrib u tio n License (CC -B Y 4.0) , which perm its any use, d istrib u tio n and reprodu ction in any m edium , provided th e original au th o r(s) and source are credited.

R e fe r e n c e s

[1] C D F collaboration, T. A alto n en et al., Search fo r large extra dim ensions in fin a l states containing one p h o to n or je t and large m issing transverse energy produced in pp collisions at VS = 1.96 TeV, Phys. Rev. Lett. 101 (2008) 181602 [a r X iv :0 8 0 7 .3 1 3 2 ] [^{i n}SPIRE].

[2] D 0 collaboration, V.M . A bazov e t al., Search fo r large extra dim ensions via single p h oton plus m issing energy fin a l states at \ f s = 1.96 TeV, P hys. Rev. Lett. 101 (2008) 011601

[a r X iv :0 8 0 3 .2 1 3 7 ] [^{i n}SPIRE].

- 18 -

J H E P 0 6 ( 2 0 1 6 ) 0 5 9