Forward Particle Measurements at Colliders and Air Shower Development

Forward particle measurements at colliders and air shower development

Takashi Sako High Energy Cosmic Ray Research Division, ICRR

2018/4/20 ICRR seminar 1 What I was doing before joining ICRR Forward particle measurements at colliders and air shower development

What I am studying at ICRR Takashi Sako High Energy Cosmic Ray Research Division, ICRR

2 Outline

• Air shower observation and hadron interaction • Forward particle measurements at collider • LHCf and RHICf experiments • Introduction to the LHCf experiment • Forward �0 and neutron results • √s dependence • RHICf • Model tuning to the LHCf result and impact on air shower simulation • Summary and future

3 High-energy cosmic ray observations

• High-energy particles from outer space, cosmic rays, are observed • Majority of CRs is fully ionized nuclei

Space/balloon including protons observations • 1020eV(=16J) CRs are observed Air shower observations • LHC can accelerate up to 7x1012eV

Where is the origin of CRs? How are they accelerated?

• Very low flux at high energy Ex) >1020eV 1 ptcl/1km2/century • Direct observation is not available

=> Air shower technique

4 5 High-energy cosmic ray observations

• High-energy particles from outer space, cosmic rays, are observed • Majority of CRs is fully ionized nuclei

Space/balloon including protons observations • 1020eV(=16J) CRs are observed Air shower observations • LHC can accelerate up to 7x1012eV

Where is the origin of CRs? How are they accelerated?

• Very low flux at high energy Ex) >1020eV 1 ptcl/1km2/century • Direct observation is not available

=> Air shower technique

6 Air shower technique • km2~1000km2 detection area is achieved using a sparse array of the ground detectors • Identification of primary particle (nuclei, gamma, etc…) is possible by measuring the difference in the air shower development

• We can (want to) extract • Energy • Direction • Type (mass number, �, e-, �) of the primary particles

=> Analyses strongly rely on the MC Spread over 100m – few km simulation of air shower development, especially fundamental hadronic interaction is essential 7 Model dependent mass interpretation QGSJET1 QGSJETII

SIBYLL EPOS

(Kampert and Unger, Astropart. Phys., 2012) • Is difference between proton and Helium small? => Factor 4 in mass number!! • Is the truth really between the existing models? => Nobody knows!! Models must be tested by accelerator data 8 Hadronic interaction in air shower proton CR proton • km2~1000km2 detection area is neutron

=> Analyses strongly rely on the MC Spread over 100m – few km simulation of air shower development, especially fundamental hadronic interaction is essential 9 Hadronic interaction in air shower proton CR proton • km2~1000km2 detection area is neutron

�± achieved using a sparse array of �0 the ground detectors photon (�) Atmospheric • Identification of primary particle muon nucleus neutrino (nuclei, gamma, etc…) is possible by measuring the difference in the air shower development Electromagnetic cascade • We can (want to) extract SUPER simplified view: • Energy incident particle [E] • Direction • Type -> a leading baryon [kelaEof the primary particles] + multi mesons [(1-Kela)E/Nmulti] => Analyses strongly rely on the MC Spread over 100m – few km simulation of air shower development, especially fundamental hadronic interaction is essential 10 Cosmic-ray spectrum and collision energy （D’Enterria et al., APP, 35,98-113, 2011 ）

End of galactic CR and transition to extra-gal CR

Ankle (GZK) cutoff: end of CR spectrum Knee: end of galactic proton CR

LHC beam energy

Indirect observation through air shower

11 Cosmic-ray spectrum and collision energy （D’Enterria et al., APP, 35,98-113, 2011 ）

RHIC LHC FCC

End of galactic CR and transition to extra-gal CR

Ankle (GZK) cutoff: end of CR spectrum Knee: end of galactic proton CR

7x1012eV+7x1012eV

1017eV + (at rest) Indirect observation through air shower

12 17 Though the particle energy is 7x10 eV at LHC, the collision energy corresponds to ECR=1012 eV Detectors @ Colliders

Central detector (ATLAS, CMS, ALICE, STAR, …) Beam particle (black solid) Neutral particles

collision

Elastic scattering Dipole （black dashed） Beam pipe

ü Main physics at colliders are achieved using the “Central detectors” ü But…

13 Angular distribution at colliders multiplicity and energy flux at LHC 14TeV collisions � =0 �=1 ∘ �>1 pseudo-rapidity; η= -ln(tan(θ/2)) �<0 (�=45 ) Multiplicity Energy Flux All particles

neutral

ü Most of the particles are produced in the central region ü Most of the energy flows into very forward = relevant to CR air shower • forward = soft interaction; theoretical difficulty 14 • experimental difficulty Angular distribution at colliders multiplicity and energy flux at LHC 14TeV collisions

pseudo-rapidity; η= -ln(tan(θ/2)) Energy Flux 1017eV proton showerElectron Profile Muon Density ] Electron Profile Muon Density 2 |η|<1 � in the 1st interaction 1<|η|<3.5 3.5<|η|<5 ] 2 Number |η|<1 10-1 5<|η|<6.6 7 6.6<|η|<8 10 1<|η|<3.5 η>8

3.5<|η|<5 Density [1/m -2 Number 10-1 5<|η|<6.6 10 6.6<| |<8 7 6 η 10 10 η>8 -3 Density [1/m 10 10-2 105 10-4 106

-3 0 200 400 600 800 1000 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 10 2 Depth [g/cm ] Distance [m] Fig. 6.1: Fraction of the air shower development for one proton induced air shower at 1017eV primary 105 ∼ 10-4 energy, which is determined� by=8 => the hadronic� 1mrad (CMS) particle production in the initial inelastic p-air collision in differentü Most of the particles are produced in the central region acceptance regions for electrons in longitudinal profile (left hand-side and muons in lateral 0 200 400 600 800 1000 0 500distribution1000ü 1500 at2000 ground2500 (right3000 hand-side).3500 4000 The4500 acceptance5000 is calculated in the center-of-mass frame of the 2 Most of the energy flows into very forward = relevant to CR air shower Depth [g/cm ] collision, and• theforward = soft interaction; theoretical difficulty shown values are relatedDistance to typical [m] LHC detectors. The major part of the air shower is determined by particle production in the forward region. 15 Fig. 6.1: Fraction of the air shower development for one proton induced• experimental difficulty air shower at 1017eV primary energy, which is determined by the hadronic particle production in the initial inelastic p-air collision The LHC data on total, elastic and diffractive cross sections and other measurements in different acceptance regions for electrons in longitudinal profilerelated to (left soft diffraction hand-side (rapidity and muons gaps, energy in lateral loss, ...) are examples of the first category, while distribution at ground (right hand-side). The acceptance is calculatedmean particle in the multiplicities, center-of-mass multiplicity frame distributions, of the jet cross sections at low p , particle ? collision, and the shown values are related to typical LHC detectors.spectra and The correlations major part between of the observables air shower belong is to the second one. determined by particle production in the forward region. 6.2.1 LHC data and hadronic interaction models For instance, measurement of the pseudorapidity dependence of the transverse energy flow and charged particle multiplicity distributions in proton-proton collisions are sensitive to the mod- The LHC data on total, elastic and diffractive crosseling of sections soft fragmentation and other effects, measurements MPI and diffractive interactions. As well as allowing for a deeper understanding of these effects in their own right, the tuning of MC models yields more related to soft diffraction (rapidity gaps, energy loss, ...)accurate are examples simulations of of the the first “underlying category, event” while - comprising MPI and additional soft interac- mean particle multiplicities, multiplicity distributions,tions jet between cross sections the primary at partons low inp events, particle with a hard perturbative scatter. The dynamics of soft interactions are also important? to understand at the LHC due to the large number of spectra and correlations between observables belong tosoft the interactions second one. (pile-up) which occur during every event. An example of how models can be retuned using these data is shown on Fig. 6.2. On the left-hand side, predictions of pre-LHC models used for air shower simulations (EPOS 1.99 [18,19] (solid line), QGSJETII-03 [21,22] 6.2.1 LHC data and hadronic interaction models (dashed line), QGSJET01 [23, 24] (dash-dotted line) and SIBYLL 2.1 [25–27] (dotted line)) For instance, measurement of the pseudorapidity dependenceare compared of the to transverse ALICE data [28], energy while flow on the and right-hand side results are presented for the two models (EPOS LHC [29] (solid line) and QGSJETII-04 [30] (dashed line)) which where charged particle multiplicity distributions in proton-protonretuned collisions using first LHC are data. sensitive to the modeling of soft fragmentation effects, MPI and diffractive interactions.By requiring As a forward well proton as allowing to be tagged for in a a LHC Roman pot based detector, a subset deeper understanding of these effects in their own right,of the inelastic tuning interactions of MC are models probed which yields will allow more diffraction to be investigated in more detail. This in turn will lead to more accurate tunes and possibly highlight areas of tension where the accurate simulations of the “underlying event” - comprisingcurrent phenomenological MPI and additional models are soft unable interac- to describe the data and would therefore need tions between the primary partons in events with a hardrevisiting. perturbative Such samples scatter. are especially The dynamics sensitive to the modeling of the forward regions and of soft interactions are also important to understand atwill the be of LHC use to due constrain to the cosmic-ray large air number shower physics. of The CASTOR (CMS) calorimeter provides the unique possibility to minimize the gap in soft interactions (pile-up) which occur during every event.the forward An example coverage of of detectors how modelsat LHC. While can other be forward charged particle detectors reach up to h < 5, this is extended by CASTOR up to 6.6. For the physics in extensive air showers retuned using these data is shown on Fig. 6.2. On the left-hand| | side, predictions of pre-LHC models used for air shower simulations (EPOS 1.99 [18,19] (solid line), QGSJETII-03 [21,22] (dashed line), QGSJET01 [23, 24] (dash-dotted line) and SIBYLL 2.1 [25–27] (dotted179 line)) are compared to ALICE data [28], while on the right-hand side results are presented for the two models (EPOS LHC [29] (solid line) and QGSJETII-04 [30] (dashed line)) which where retuned using first LHC data. By requiring a forward proton to be tagged in a LHC Roman pot based detector, a subset of inelastic interactions are probed which will allow diffraction to be investigated in more detail. This in turn will lead to more accurate tunes and possibly highlight areas of tension where the current phenomenological models are unable to describe the data and would therefore need revisiting. Such samples are especially sensitive to the modeling of the forward regions and will be of use to constrain cosmic-ray air shower physics. The CASTOR (CMS) calorimeter provides the unique possibility to minimize the gap in the forward coverage of detectors at LHC. While other forward charged particle detectors reach up to h < 5, this is extended by CASTOR up to 6.6. For the physics in extensive air showers | |

179 Forward detectors @ Colliders

Central detector (ATLAS, CMS, ALICE, STAR, …) Beam particle (black solid) Neutral ZDC/LHCf/ particles ZDC/LHCf/ RHICf RHICf collision Zero Degree Calorimeter (ZDC@ATLAS/CMS/ALICE/STAR/ PHENIX, LHCf, RHICf) Roman Elastic scattering Dipole Pot （black dashed） Forward detector (CASTOR, TOTEM T1, T2) Beam pipe Roman Pot (TOTEM RP, ATLAS ALFA, Minimum bias counter pp2pp, UA7) (MBTS, LUCID, BBC)

ü CMS CASTOR and TOTEM T1/T2 cover most forward at CMS ü TOTEM/ALFA roman pots are powerful for total cross section measurements ü ZDC/LHCf/RHICf cover neutral particles at zero degree (zero degree measurement is possible only in p-p, but not in p-pbar) 16 Forward detectors @ Colliders

140m Two independent detectors at either side of IP1 (Arm#1, Arm#2 ) Beam Beam pipe Charged particles (+) LHCf Arm#2 Neutral 96mm particles Charged particles (-)

ü All charged particles are swept by dipole magnet ü Neutral particles (photons and neutrons) arrive at LHCf

ü η>8.4 (to infinity) is covered 18 LHCf Detectors ü Imaging sampling shower calorimeters ü Two calorimeter towers in each of Arm1 and Arm2 ü Each tower has 44 r.l. (1.6�) of Tungsten,16 sampling scintillator and 4 position sensitive layers ü Plastic scintillators => GSO scintillators, SciFi => GSO bars in Run2 LHCf Arm#1 Detector 20mmx20mm+40mmx40mm 4 XY SciFi+MAPMT

LHCf Arm#2 Detector 25mmx25mm+32mmx32mm

4 XY Silicon strip detectors 19 LHCf Detectors ü Imaging sampling shower calorimeters ü Two calorimeter towers in each of Arm1 and Arm2 ü Each tower has 44 r.l. of Tungsten,16 sampling scintillator and 4 position sensitive layers ü Plastic scintillators => GSO scintillators, SciFi => GSO bars in Run2 LHCf Arm#1 Detector 20mmx20mm+40mmx40mm 4 XY SciFi+MAPMT Technique of CALET

LHCf Arm#2 Detector 25mmx25mm+32mmx32mm

4 XY Silicon strip detectors 20 Technique of PAMELA Detector construction in Japan and Italy

• All of design, construction, calibration of the detectors, data acquisition and analyses are led by the Nagoya U group 290mm 90mm Largest and Smallest TeV gamma-ray detectors

23 Sampling calorimeter

n, gamma

• Incident particles develop showers in Tungsten • Deposited energy is sampled by scintillators interleaved (3% for EM showers) • Four strip detector layers record lateral distribution of showers Event example @ LHC

Longitudinal development

Energy Particle ID

Lateral distribution Incident position Nb. of incident Silicon X

Silicon Y Event categories of LHCf Responsible for air shower core (elasticity) LHCf calorimeters Leading baryon (neutron) Single hadron event

Multi meson production

π0 photon Single photon event

Responsible for air shower EM particles (inelasticity)

π0 photon Pi-zero event (photon pair)

26 MEASUREMENTS OF LONGITUDINAL AND TRANSVERSE … PHYSICAL REVIEW D 94, 032007 (2016) V. ANALYSIS FRAMEWORK are determined according to the peak height and width of the corresponding distinct peaks in the lateral-shower- A. π0 event reconstruction and selection impact distribution. The standard reconstruction algorithms consist of four steps: hit position reconstruction, energy reconstruction, particle identification, and π0 event selection. 3. Particle identification Particle identification (PID) is applied in order to 1. Position reconstruction efficiently select pure electromagnetic showers and to Hit position reconstruction starts with a search for reduce hadron (predominantly neutron) contamination. multihit and single-hit events. A multihit event is defined PID in the study of this paper depends only on the to have more than one photon registered in a single parameter L90%. L90% is defined as the longitudinal dis- calorimeter. A single-hit event is defined to have a single tance, in units of radiation length (X0), measured from the hit in each of the two calorimeters in a given detector, Arm1 first tungsten layer of the calorimeter to the position where or Arm2. the energy deposition integral reaches 90% of the total Therefore, multihit event candidates should have two or shower energy deposition. Events with an electromagnetic more distinct peaks in the lateral-shower-impact distribu- shower generally have a L90% value smaller than 20 X0, tion of a given calorimeter and are then identified using the while events with a hadronic shower generally have L90% TSpectrum algorithm [42] implemented in ROOT [43]. larger than 20 X0. The threshold L90% value as a function of 0 TSpectrum provided the basic functionality for peak find- the photon energy is defined in order to keep the π ing in a spectrum with a continuous background and selection efficiency at 90% over the entire energy range of statistical fluctuations. the individual photons. PID criteria are determined by MC The MC simulation estimated efficiencies for identifying simulations for each calorimeter. multihit events are larger than 70% and 90% for Arm1 and Arm2, respectively [25]. Given the list of shower peak 4. π0 event selection position candidates that have been obtained above, the 0 lateral distributions are fit to a Lorenzian function [44] to The π are then identified by their decay into two photons, leading to the distinct peak in the invariant mass obtain more precise estimates of the shower peak positions, 0 heights, and widths. In the case of multihit events, two distribution around the π rest mass. The invariant mass of the two photons is calculated using the reconstructed peaks are fit using superimposed Lorenzian functions. 0 Multihit events with three or more peaks are rejected from photon energies and incident positions. The π events usedParticle ID (PID) in the analysis of this paper are classified into two the analysis. Conversely, single-hit events, not having two 0 0 or more identifiable peaks in a single calorimeter but categories: Type-I π and Type-II π events. A Type-I event having a single hit in each calorimeter are correctly selected is defined as having a single photon in each of the two Neutron event with an efficiency better than 98% for true single-photon calorimetersPhoton event of Arm1 or Arm2 (the left panel of Fig. 1). A events with energy greater than 100 GeV for both Arm1 Type-II event is defined as having two photons in the same and Arm2. calorimeter (the right panel of Fig. 1). Note that Type-II events were not used in the previous analyses [18,19] and 2. Energy reconstruction thus are taken into account for the first time in this paper. As detailed in Sec. VB, the phase spaces covered by Type-I The photon energy is reconstructed using the measured and Type-II events are complementary. In particular, the energy deposited in the LHCf calorimeters. The charge 1.6 interaction lengths inclusion of Type-II events extends the pT upper limit for information in each scintillation layer is first converted to a analysis from 0.6 GeV in44 the previousr.l. analyses to 1.0 GeV. deposited energy by using the calibration coefficients obtained from the electron test beam data taken at the The LHCf Collaboration / Physics Letters B 780 (2018) 233–239 235 Super Proton Synchrotron (SPS) below 200 GeV [26]. The sum of the energy deposited in the 2nd to 13th scintillation layers is then converted to the primary photon energy using � = � �� imately 5% and 10% for Arm1 and Arm2, respectively. In the an empirical function. The coefficients of the function are high-energy range, most of the multi-hit events are caused by determined from the response of the calorimeters to single photon pairs from π 0 decay. In these events, the separation photons using MC simulations. Corrections for shower leakage effects and the light-yield collection efficiency of O. ADRIANI et al. PHYSICAL REVIEW D 86, 092001 (2012) between photons is kinematically limited above 5.8 mm. This the scintillation layers are carried out during the energy makes the identification of multi-hits simpler. 500 signal window can be estimated using the likelihood func- reconstruction process [20]. In the case of multihit events, 0 FIG. 1. Observation of π decay by a LHCf detector. Left: -1 About 4% of the total triggered events were identified as multi- LHCf-Arm1 s=7TeV, Ldt=2.53nb tion [L y; p ;m ] characterized by the best-fit third- the reconstructed energy based on the measured energy Type-I π0 event having one photon entering each∫ calorimeter. BGð T Þ 0 hit events. Two corrections were applied to the measured deposited is split into two energies, primary and secondary. Right: Type-II π 400event9.0 having < y < two 9.2 photons entering one order Chebyshev polynomial function. For simplicity, Fractions of the energy for the primary and secondary hits calorimeter, here entering the small calorimeter. L y; p ;m is shortened as L in the following cross-section: BGð T Þ BG text. Thus the signal distribution with background sub- 1. ‘Multi-hit performance’ correction: 300 tracted is given by The contamination of multi-hit events misidentified as 032007-5 Sig Sig BG BG single-hits and the loss of single-hit events misidentified f y;pT f y;pT þ R y;pT;m;^ l;u f y;pT ; 200 ð Þ ¼ ð Þ À ð Þ ð Þ as multi-hits are corrected with an energy-dependent fac- (4)

Events / (1 MeV) tor based on the MC dataset of QGSJET II-04. This correction factor depends mostly on the detector performance, while 100 where R y; pT; m;^ l; u is the normalization for the background distributionð and writtenÞ as it depends weakly on the model chosen to generate the dataset. m^ 3u 0 þ LBGdm 2. ‘Multi-hit cut’ correction: 80 100 120 140 160 180 m^ 3l R y;pT;m;^ l;u À : m^ 3l Shower depth parameterm^ 6u Reconstructed m [MeV] ð Þ¼ À LBGRdm þ LBGdm As the single-photon cross-section is measured by the de- γγ Fig. 1. L m^ distribution6l in Arm1m^ 3 foru the events with the reconstructed energy 90% À þ þ tector, another correction factor based on the same MC between 1.1R and 1.2 TeV. TheR black points represent(5) the experimental data with Invariant mass of photon pairFIG. 4 (color online). Reconstructed invariant mass distribu- (Adriani et al., PLB, 2018) statistical error bars. The red and blue colored lines correspond to the template dis- dataset was applied to correct for the multi-hit cut and re- tion within the rapidity range from 9.0 to 9.2. Solid curve shows 0 tributions obtained from the MC simulation for photons and hadrons, respectively.27 cover the inclusive production cross-section. This correction Peak the best-fit@ 135MeV composite physicsfrom model� decay events to the invariant mass distribution. Dashed curve indicates the background component. (AdrianiThe blacket al., D. Unfolding line representsPRD, 2012) of spectra the total of the template distributions. These distributions factor ranged within 50%, which was the largest contribu- were normalized by the results of the template fitting. (For interpretation of the ± Solid and dashed curves indicate the signal and background The raw rapidity–p distributions must be corrected for tion among the corrections and was strongly dependent on windows, respectively. colors in theT figure(s), the reader is referred to the web version of this article.) unavoidable reconstruction inefficiency and for the smear- the choice of event-generation model in the MC simulation. ing caused by finite position and energy resolutions. An This is because the multi-hit rate is related to the cross- C. Background subtraction 0 iterative Bayesianthe methodPID estimator, [39,40] is L used90%, defined to simulta- as the longitudinal depth, in section of high-energy π production, as discussed above. Background contamination of two-photon 0 events by neously correct for both effects. The advantages of an units of radiation length (X0), at which the integral of the Both multi-hit corrections were performed inside the unfold- hadron events and the accidental coincidence of two pho- iterative Bayesian method with respect to other unfolding 0 energy deposition in a calorimeter reached 90% of the total. ing algorithm, which is described below. tons not coming from the decay of a single are sub- algorithms are discussed in another report [ 39]. The un- tracted using the so-called ‘‘sideband’’ method. folding procedureAs for a thecriterion data is organizedof the selection as follows. of the photon component, we Unfolding: 0 • Figure 4 shows an example of the reconstructed two- First, the responseset an of energy-dependent the LHCf detectors to criterion single L90%,thr, which defines the We corrected for detector biases (as energy resolution and photon invariant mass distribution of the experimental data events is simulatedL90% byvalue toy MC to calculations. maintain Ina the90% toy effi MCciency of photon selection in multi-hit effects) in the obtained cross-section by perform- of Arm1 in the rapidity range from 9.0 to 9.2. The energy simulations, two photons from the decay of 0s and low the MC simulations. Fig. 1 presents the L90% distribution of ing an unfolding technique based on the iterative Bayesian scale correction discussed in the previous section has been energy background particles such as those originating in a applied. The sharp peak around 135 MeV is due to 0 Arm1-Region A for the reconstructed energy range between 1.1 method [25]provided by the RooUnfold package [26]. The MC prompt photon event or a beam pipe interaction are traced 8 events. The solid curve represents the best fit of a compos- through the detectorand and1.2 thenTeV. reconstructed The red and with blue the lines event in Fig. 1, obtained from the simulation dataset with 10 inelastic collisions generated by ite physics model to the invariant mass distribution of the reconstruction algorithmMC simulation introduced dataset above. Noteof QGSJET that the II-04, indicate the template the QGSJET II-04 model was used as a training sample. data. The model consists of an asymmetric Gaussian dis- single 0 kinematicsdistributions that are simulated for the within pure the photon allowed and pure hadron samples, Decay correction: • tribution (also known as a bifurcated Gaussian distribution) phase space arerespectively. independent of These the particular distributions interaction were produced with normal- The photons detected by the LHCf experiment mainly come for the signal component and a third-order Chebyshev model that is being used. The background particles are 0 ization obtained from the template-fit result. According to the from the decay of short-lifetime particles such as π and η polynomial function for the background component. The simulated by a hadronic interaction model, which is dis- mesons, which decay near the interaction point. Particles with dashed curve indicates the background component. cussed later, sincetemplate-fit the amount ofresults, background the hadron particles contamination, is typically 10%, a longer lifetime (such as K 0, K and #) can decay along the Using the expected mean (m^ ) and 1 deviations (l for not directly measuredcan be by theestimated LHCf detector. as a function of energy and it is corrected ± 0 lower side and u for upper side) of the signal component, The detector responsetogether to with events the depends90% effi onciency rapidity in the analysis. beam pipe between the interaction point and detector and can the signal window is defined as the invariant mass region and pT, since theMulti-hit performance correction of the particle identification contribute to the photon yield. In order to remove the con- within the two solid arrows shown in Fig. 4, where the algorithm and• the selection efficiency of events with a Because the mis-reconstruction of multi-hit events as single- tribution of long-lifetime particles, an energy-dependent cor- lower and upper limits are given by m^ 3 and m^ 3 , l u single-photon hit in both calorimeters depend upon the rection was estimated with MC simulations by comparing the respectively. The background windowÀ is constructedþ energy and the incidenthit events position makes of a the particle. measured The recon- spectra more complex, multi- from the two sideband regions, m^ 6 ; m^ 3 and structed rapidity—hitp eventsdistributions were for rejected given true from rapidity— the analysis. In order to iden- photon production cross-section at the interaction point with ½ À l À l T m^ 3u; m^ 6u , that are defined as the invariant mass p distributions thentify leadmulti-hit to the calculation events, a of lateral the response shower profile measured by the that after transportation along the beam pipe to the detector ½ þ þ T regions within the dashed arrows in Fig. 4. function. Then theposition-sensitive reconstructed rapidity layers and waspT spectra fitted by an empirical function. (i.e. after step ‘2’ described in Sec. 3). The correction reaches a The rapidity and pT distributions of the signal are corrected with the response function that is equivalent maximum of about 15% in the lowest-energy bin and becomes Sig The difference in the goodness-of-fit between the single and [f y; pT ] are then obtained by subtracting the back- to the likelihood function in Bayes’s theorem. The correc- less than 1% above 2 TeV. ð Þ BG double peak assumptions, the distance between two peaks, ground distribution [f y; pT ], estimated by the back- tions are carried out iteratively whereby the starting point ground window, fromð theÞ signal-rich distribution of the current iterationand the is theratio ending between point of two the previous peak heights were used to iden- Sig BG 5. Systematic uncertainties [f y; p þ ] selected from the signal window. The iteration. Statisticaltify uncertaintymulti-hit events. is also propagated These criteria from were adjusted to achieve a ð TÞ fraction of the background component included in the the first iterationhigh to theeffi last.ciency Iteration of multi-hit is stopped detection at or while maintaining a rea- We considered the following contributions as systematic uncer- sonably low incidence of single-hit-event mis-reconstructions tainties of the measured production cross-section. Fig. 2 shows the as multi-hit events. estimated systematic uncertainties for each detector and each re- 092001-6 The consistency of the multi-hit identification efficiencies ex- gion as a function of photon energy. hibited by the data and MC simulation was tested using ‘artificial’ multi-hit event sets. These artificial multi-hit events 5.1. Energy scale were created by merging two independent single-hit events. The combinations of single-hit events were selected to repre- Energy scale errors are attributable to a) the absolute gain cal- sent the distributions of photon-pair energies and hit-position ibration of each sampling layer, b) uniformity, c) relative gain cali- distances in the true multi-hit events of QGSJET II-04. The bration of the photomultiplier tubes (PMTs) used for the readout of same procedure was performed for the MC simulation also. scintillator lights, and d) the Landau–Pomeranchuk–Migdal (LPM) The multi-hit detection efficiency exceeds 85% across the full effect [27,28]. The first two contributions were studied in beam energy range and reaches nearly 100% above 2 TeV, while in- tests and are described in Ref. [17]. The third source of errors is re- consistencies between the data and MC are less than approx- lated to the differences in the high-voltage configurations of PMTs Dynamic range and linearity test using UV pulse laser

結果:各HVにおける光量とADC値の関係 N2レーザー（KEN-1020,300psec）波⻑:377nm (0.025pC) ADC

部分反射フィルター Laser intensity（MIP eq.）

N2 laser（KEN-1020） • 337.1nm • 0.3ns pulse

Scintillators with �=9.6ns (EJ-260) and �=30ns (GSO) are used to avoid saturation Non-uniformity calibration at HIMAC

Ø Calorimeters WITHOUT tungsten layers were scanned over the ion beam Ø Signal intensity as a function of the position was extracted 29 Detector performance @ CERN SPS Y.Makino et al., JINST 12 P03023 (2017) Energy resolution Position resolution 5 5 m] m] Data µ Data Data X µ Data X

E/E [%] E/E [%] 250 Data Y 250 Data Y ∆ Simulation ∆ Simulation 4 4 MC X MC X MC Y MC Y 200 200 Resolution [ Resolution [

3 3 150 150

100 100 2 2 2017 JINST 12 P03023

3000 2nd layer 1600 3rd layer 2500 4th layer Events Events Events χ2/DOF=1.07 1400 χ2/DOF=0.90 χ2/DOF=0.36 2017 JINST 12 P03023 2500 2000 1200 50 50 2000 1000 1500 1500 800

600 1000 1000 1 1 400 500 500 200

0 0 0 0 0 0 100 200 300 400 500 600 700 800 600 800 1000 1200 1400 1600 1800 800 900 1000 1100 1200 1300 1400 1500 1600 1700 ADC count ADC count ADC count 0 50 100 150 200 250 0 5050 100100 150150 200200 250 250 50 100 150 200 250 Electron beam energy [GeV] Electron beam energy [GeV] 4000 Electron beam energy [GeV] Electron beam energy [GeV] 3500 5th layer 6th layer 7th layer 5000 Events 2/DOF=1.32 Events 2/DOF=0.90 Events 3500 2/DOF=0.77 3000 χ χ χ 4000 3000 20 mm 1st layer 20 mm 2nd layer 2500 (a) 20 mm (a)(b) 40 mm (b) 2500 2000 3000 2000 1500 2000 1500 2017 JINST 12 P03023 1000 1000

1000 m] m] 500 Energy linearity 500 Figure 12. Energy dependence of the energy resolution of the Arm1 detector for data (ﬁlled circles) and MC µ µ 0 0 0 800 900 1000 1100 1200 1300 1400 1500 1600 200 300 400 500 600 700 800 900 1000 200 300 400 500 600 700 800 900 1000 Data X Data X ADC count ADC count ADC count (open circles, shifted horizontally by 5 GeV). The events in a250 4(8)8 mm 4(8) mm square aroundData the Y center 250 Data Y 7000 ⇥ 8th layer 9000 9th layer 10th layer 6000 12000

Events 2 Events 8000 2 Events 2 MC X MC X χ /DOF=0.79 χ /DOF=0.86 χ /DOF=1.12 of the 20(40) mm calorimeter tower were selected. 5000 7000 10000 8 6000 MC Y MC Y 4000 8000 5000 6 3000 4000 6000 200 200 Resolution [ Resolution [

2000 3000 4000 2000 6 1000 2000 1000 0 0 0 5 5 0 100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800 ADC count ADC count ADC count 1504 150 4 Figure 5. The energy deposit distributions of the experimental data (black points) and simulation (blue line) Data Data E/E [%] E/E [%]

of the Arm1 40 mm tower which have the minimum-chi-square parameters set from the 2nd layer to the 10th ∆ ∆ Simulation 2 Simulation layer. Injected• particlesBeam test using 100 were 200 GeV electrons. The horizontal axis represents the-250GeV electron and proton beams ADC counts. The area 4 4 100 1002 of each MC distribution is normalized to the area of the correspondent data distribution. Total collected energy [GeV] Total collected energy [GeV]

3500 4500 0 0 2nd layer 3500 3rd layer 4th layer 4000 50 50

Events • Events Events 3000 Charge to energy calibration factors2 3000 2 2 /DOF=0.52 /DOF=0.81 /DOF=0.77 χ χ 3500 χ Electron beam energy [GeV] Electron beam energy [GeV] 2017 JINST 12 P03023 2500 2500 3000 3 3 1 1 2000 2000 2500

1500 1500 2000 1500 0 0 1000 1000 0 0 • ∑dE to E function1000 50 100 150 200 250 50 100 150 200 250 500 500 500

0 0 0 100 200 300 400 500 600 700 800 900 400 600 800 1000 1200 1400 1200 1400 1600 1800 2000 2200 2400 2600 2800 Residual [%] −1 Electron beam energy [GeV] Residual [%] −1 Electron beam energy [GeV] ADC count ADC count ADC count 2 2 0 50 100 150 200 250 0 50 100 150 200 250 4000 4000 5th layer 3500 6th layer 7th layer (c) 40 mm 1st layer (d) 40 mm 2nd layer 3500 Electron beam energy [GeV] Electron beam energy [GeV] Events 3500• 2 Events 2 Events 2 30 Energy resolution, linearity, position resolutions χ /DOF=0.65 3000 χ /DOF=0.91 χ /DOF=0.35 3000 3000 2500 2500 2500 (a) Arm1: 20 mm tower. (b) Arm1: 40 mm tower. 2000 2000 2000 1500 Figure 18. Position resolution of the GSO-bar hodoscope layers depending on the incident electron beam 1500 1500 1000 1 1 1000 1000 500 500 500 energies. Black and white markers represent data and simulation results, respectively. Events within a 0 0 0 900 1000 1100 1200 1300 1400 1500 500 600 700 800 900 1000 600 800 1000 1200 1400 8 ADC count ADC count ADC count 0 50 100 150 200 250 2(4) mm0 square50 around100 the150 center200 of the250 calorimeter were selected8 for the 20(40) mm calorimeter tower. 6000 4500 8th layer 5000 9th layer 10th layer Electron beam energy [GeV] Electron beam energy [GeV] 4000 5000 Events χ2/DOF=1.22 Events χ2/DOF=0.68 Events χ2/DOF=0.25 3500 4000

3000 4000 6 3000 (a) 25 mm (b) 32 mm 6 2500 3000 2000 2000 1500 2000 dependence of the calorimeter was reduced below the level of 1% after using the correction maps 1000 1000 1000 500 Figure 13. Energy dependence of the energy resolution of the4 Arm2 detector for data (ﬁlled circles) and 4 0 0 0 100 200 300 400 500 600 100 200 300 400 500 600 700 0 100 200 300 400 500 600 generated from MC simulations. The linearity of the detector response to the beam energy has been ADC count ADC count ADC count MC (open circles, shifted horizontally by 5 GeV). The events in a 5(10) mm 5(10) mm square around the measured to be better⇥ than 0.5% for both Arm1 and Arm2 in the energy range between 100 and Figure 6. The energy deposit distributions of the experimental data (black points) and simulation (blue line) center of the 25(32) mm calorimeter tower were selected. 2 2 of the Arm2 25 mm tower which have the minimum-chi-square parameters set from the 2nd layer to the 10th 250 GeV.Total collected energy [GeV] After the calibration, we conﬁrm that the detectorsTotal collected energy [GeV] meet all the requirements of the LHCf layer. Injected particles were 200 GeV electrons. The horizontal axis represents the ADC counts. The area of each MC distribution is normalized to the area of the correspondent data distribution. experiment0 for proton-proton collisions at 13 TeV. 0 Electron beam energy [GeV] Electron beam energy [GeV] The correction was tested by checking the position dependence1 of S for each calorimeter. Data 1 with 150 and 200 GeV electron beams were usedAcknowledgments for this0 study of Arm1 and Arm2, respectively. 0

–9– The uniformity of calorimeter responses before and afterResidual [%] −1 correction is demonstrated in figure 14 Residual [%] −1 0 50 100 150 200 250 0 50 100 150 200 250 and 15 for Arm1 and Arm2, respectively. We greatly appreciateElectron the beam support energy [GeV] of the CERN sta. This work wasElectron supported beam energy [GeV] by JSPS KAK- ENHI (23244050,(c) 23340076,Arm2: 25 mm 23740183, tower. 25287076, and 26247037)(d) Arm2: and Istituto 32 mm Nazionaletower. di Fisica Nucleare (INFN) in Italy. Figure 16. The relation between the energy of the incident particle and S for both Arm1 and Arm2 detectors. The black points represent measured data while the fit function is drawn as a dashed line. In each plot, the – 14 – bottom panel shows the residuals from the fitting results.

– 19 – dependence of the signal is too large and it cannot be properly corrected using this method. The same cut is applied also in the physics analysis of the LHC data.

5.5 Linearity of the calorimeter response

The linearity of the calorimeter response was checked by fitting with a linear function of the relation between S and the energy of the incident particle (S = ↵ E + , where E is the incident energy and · ↵, are parameters) for all the electron beam energies. The energy vs S relation and fit function are shown in the top of figure 16 for both Arm1 and Arm2. The deviations from linearity shown in the bottom of figure 16 were less than 0.5% in the whole energy range between 100 GeV and 250 GeV for both Arm1 and Arm2.

– 16 – Installation into the LHC tunnel LHCf/RHICf History ü2004 LOI submitted to CERN ü2006 TDR approved by CERN ü2009 First data taking at √s=900GeV p-p collision ü2010 √s=7TeV p-p collision

ü2013 √s=2.76TeV p-p & √sNN=5TeV p-Pb collisions ü2015 √s=13TeV p-p collision

ü2016 √sNN=8.1TeV p-Pb collision ü2017 √s=510GeV polarized p-p collision as RHICf

32 physics results

Publications performance results

Neutron Photon π0 (limited π0 (full (hadron Performance (EM shower) acceptance) acceptance) shower) NIM, A671 JINST, 9 (2014) Beam test (2012) 129- P03016 136 PLB, 715 0.9TeV p-p (2012) 298- 303 IJMPA, 28 PLB, 703 PLB, 750 (2013) 1330036 PRD, 86 7TeV p-p (2011) 128- (2015) 360- (2012) 092001 134 366 PRD, 94 (2016) 032007 2.76TeV p-p PRC, 89 (2014) 5.02TeV p-Pb 065209 PLB, 780 13TeV p-p (2018) 233- Analysis in progress 239

33 physics results

Publications performance results

34 0 π pz spectra in 7TeV p-p collisions (PRD, 94 (2016) 032007) 11

(a) 0.0 < p [GeV] < 0.2 (b) 0.2 < p [GeV] < 0.4 (c) 0.4 < p [GeV] < 0.6 ] T ] T ] T

-2 1 -2 -2 10−1 [GeV [GeV [GeV 3 3 3 10−2 /dp /dp /dp σ σ σ 3 10−3 LHCf s=7TeV 3 3

Ed -1 Ed Ed 10−4 Ldt=2.64+2.85nb inel ∫ inel inel σ −5 σ σ 1/ 10 1/ 1/

(d) 0.6 < p [GeV] < 0.8 (e) 0.8 < p [GeV] < 1.0 ] p [GeV]T ] p [GeV]T p [GeV]

-2 1 z -2 z z 10−1 LHCf (stat.+syst.) [GeV [GeV 3 3 10−2 DPMJET 3.06 /dp /dp

σ σ QGSJET II-04 3 10−3 3 SIBYLL 2.1 Ed −4 Ed

inel 10 inel PYTHIA 8.185 σ σ 5

1/ − 1/ 10 EPOS LHC 1000 2000 3000 1000 2000 3000 p [GeV] p [GeV] z z ü DPMJET3 and PYTHIA8 overestimate over all E-pT range FIG. 5: (color online). Experimental combined pz spectra of the LHCf detector (ﬁlled circles) in p + p collisions at ps =7TeV. Shadedü EPOS rectangles-LHC indicateand theQGSJET II total statistical-04 and, standard in air shower MC, are not bad systematic uncertainties. The predictions of hadronic interaction models35 are shown for comparison (see text for details.)

] (a) 8.8 < y < 9.0 ] (b) 9.0 < y < 9.2] (c) 9.2 < y < 9.4] (d) 9.4 < y < 9.6 -2 1 -2 -2 -2 10−1 [GeV [GeV [GeV [GeV 3 3 3 3 10−2 /dp /dp /dp /dp σ σ σ σ 3 10−3 3 3 3 Ed Ed Ed Ed 10−4 LHCf s=2.76TeV inel inel inel inel

σ -1 σ σ σ −5 Ldt=2.36nb 1/ 10 ∫ 1/ 1/ 1/

] (e)p 9.6 [GeV] < y < 9.8 p [GeV] p [GeV] p [GeV]

-2 1 T T T T 10−1 LHCf (stat.+syst.) [GeV 3 10−2 DPMJET 3.06 /dp

σ QGSJET II-04 3 10−3 SIBYLL 2.1 Ed −4

inel 10 PYTHIA 8.185 σ 5

1/ − 10 EPOS LHC 0 0.1 0.2 0.3 0.4 p [GeV] T

FIG. 6: (color online). Experimental pT spectra of the LHCf detector (ﬁlled circles) in p + p collisions at ps =2.76 TeV. Shaded rectangles indicate the total statistical and systematic uncertainties. The predictions from hadronic interaction models are shown for comparison (see text for details.) �0 in 7TeV p-p collision LHCf and models

36 �0 in 7TeV p-p collision LHCf and models (ratio to data) EPOS-LHC/LHCf data PT ratio 1GeV

2.0

1.5 0.5GeV 1.0

0.5

0GeV 0 1TeV 2TeV 3TeV Pz QGSJET II-04/LHCf data SIBYLL 2.1/LHCf data

37 Neutrons in 7TeV p-p collision (√s=7TeV p-p；PLB 750 (2015) 360-366)

Zero degree

(∼10% of other neutral hadrons at 140m are included both in data and MC)

ü Only QGSJET II explains the characteristic peak near zero degree ü DPM and PYTHIA under production at zero degree ü DPM and PYTHIA not bad at off-zero degree. DPM is best. 38 Neutron, p-p √s=13TeV Unfolded Spectra

8.99 < η < 9.22 8.81 < η < 8.99 η > 10.76 8.99η > η< >10.76 10.76η < 9.22 −3 8.99 < η < 9.22 8.81 −<3 8.81 η < 8.99 < η < 8.99 −3 ×10 ×10 −3 −3 ×10 −3 LHCf p-p s = 13 TeV ×10 ×10 1.2 ×10 QGSJET II-04 0.8 LHCf p-p s = 13 TeV EPOS-LHC 1 1.2 QGSJETDPMJET 3.06 II-04 0.7 1 PYTHIA 8.212 0.6 EPOS-LHC SIBYLL 2.3 0.6 1 0.8 0.8 DPMJET 3.06 1 0.5 PYTHIA 8.212 0.5 0.6 0.4 0.6 /dE [mb/GeV] /dE [mb/GeV] SIBYLL 2.3 /dE [mb/GeV] n n n σ σ σ d 0.8 d d 0.3 0.4 0.8 0.4 /dE [mb/GeV] /dE

0.4 n 0.2

σ 0.2 0.2

d 0.1 0.6 0.6 0 0 0 0.3 2 2 /dE [mb/GeV] /dE [mb/GeV] 2 /dE [mb/GeV] n n n

σ σ 1.5 1.5 σ 1.5 d d d 0.4 1 1 1 Ratio

0.4 Ratio 0.2 Ratio

Ratio 0.5 0.5 0.5

0 0 0 0 1000 2000 3000 4000 5000 6000 0 1000 2000 3000 4000 5000 6000 0 1000 2000 3000 4000 5000 6000 0.2 2000 4000 Energy6000 [GeV] 2000 0.24000 Energy6000 [GeV] 2000 4000 Energy 6000 [GeV] 0.1 Energy [GeV] Energy [GeV] Energy [GeV] • In η > 10.76, data shows a strong increasing of neutron production in the 0 0 0 2 2 high energy region. This behavior is not2 predicted by all models.

1.5 1.5 • EPOS-LHC and SIBYLL 2.3 have the best1.5 agreement in 8.99 < η < 9.22, 8.81 < η < 8.99, respectively. 1 1 28-Feb.-2018 133rd LHCC1 Meeting 8 Ratio Ratio Ratio 0.5 0.5 0.5

0 0 • High-energy peak @ 0 degree is also confirmed with the 13TeV data0 0 1000 2000 3000 4000 5000 6000 0 1000 2000 3000 4000 5000 6000 0 1000 2000 3000 4000 5000 6000 Energy [GeV] Energy [GeV] Energy [GeV] • NOTE: pT coverage is different from the 7TeV analysis 39

Figure 4: Unfolded neutrons energy spectra for p-p collisions at ps =13TeVmeasuredby the LHCf Arm2 detector. Black markers are experimental data with statistical uncertainty, whereas gray bands represent the quadrature sum of statistical and systematic uncertainty. Histograms refer to models spectra at the generator level. Top are energy distributions ex- pressed as dn/dE and bottom are the ratios of these distributions to the experimental data.

8.99, respectively. In particular, they are compatible with data in the region between 1.5 and 2 TeV, where neutron production is maximum, but they are softer or harder otherwise. The other models underestimate (QGSJET II-04)

395 or overestimate (DPMJET 3.06, PYTHIA 8.212) the di↵erential cross section in all the energy range. The general trend of experimental data is similar to what observed at ps= 7 TeV [13]. Direct comparison of models can not be done because the version used here is di↵erent respect to the one employed in [13]: in particular, QGSJET II-04,

400 EPOS-LHC and SIBYLL 2.3 were tuned using LHC Run I results. Comparing the pre-LHC and post-LHC version of SIBYLL, we can observe a signiﬁcant increase of the neutron production in all the pseudorapidity regions, fact that improves the agreement of the model with experimental measurements. Di↵er- ently, QGSJET and EPOS are not a↵ected by relevant changes. Whereas no

405 strong variation is found also in PYTHIA, DPMJET exhibits a very di↵erent neutron production in the two cases. Because no signiﬁcant changes in di↵erential cross section are expected between ps= 7 and 13 TeV, this variation is

18 √s scaling ; π0 collision energy (in)dependence

ü Scaling is essential to extrapolate beyond LHC ü (630GeV −) 2.76TeV – 7TeV good scaling within uncertainties

ü Wider coverage in y and pT with 13TeV data ü Wider √s coverage with RHICf experiment in 2017 at √s=510GeV

Feynman x; xF = 2pz/√s

40 A. ADARE et al. PHYSICAL REVIEW D 88, 032006 (2013)

TABLE I. The expected pT for r<2 cm, mean pT value with TABLE III. The result of the differential cross section the experimental cut, and the efficiency for the experimental cut d=dxF mb for neutron production in p p collisions at ps estimated by the simulation (Fig. 12). The errors were derived 200 GeVð. TheÞ first uncertainty is statistical,þ after the unfolding,¼ ffiffiffi considering the uncertainty in the parameter a xF in the and the second is the systematic uncertainty. The absolute Gaussian form evaluated by HERA. ð Þ normalization error, 9.7%, is not included.

Neutron x Mean p (GeV=c) Efficiency x Exponential p form Gaussian p form F T h Fi T T 0.45–0.60 0.072 0:779 0:014 1:8% 0.53 0:243 0:024 0:043 0:194 0:021 0:037 0.60–0.75 0.085 0:750 Æ 0:009ð1:2%Þ 0.68 0:491 Æ 0:039 Æ 0:052 0:455 Æ 0:036 Æ 0:085 0.75–0.90 0.096 0:723 Æ 0:006ð0:8%Þ 0.83 0:680 Æ 0:044 Æ 0:094 0:612 Æ 0:044 Æ 0:096 0.90–1.00 0.104 0:680 Æ 0:016ð2:3%Þ 0.93 0:334 Æ 0:035 Æ 0:111 0:319 Æ 0:037 Æ 0:123 Æ ð Þ Æ Æ Æ Æ are the efficiency for the experimental cuts and are listed in B. Result Table I. The errors were derived considering the The differential cross section, d=dx , for forward uncertainty in the parameter a xF in the Gaussian form F evaluated by HERA. There is noð significantÞ difference in neutron production in p p collisions at ps 200 GeV was determined using twoþ p distributions:¼ a Gaussian the result in case of using the ISR (exponential) pT T ffiffiffi distribution. form, as used in HERA analysis, and an exponential form, used for ISR data analysis. The results are listed in The mean values of the simulated pT distributions in each energy region are also listed in Table I. The cross Table III and plotted in Fig. 13. We show the results for xF section was obtained after the correction of the energy above 0.45 since the data below 0.45 are significantly unfolding and the cut efficiency. affected by the energy cutoff before the unfolding. The Table II summarizes all systematic uncertainties eval- pT range in each xF bin is 0

0.5 [mb] F In the cross section analysis, we evaluated the beam (mb) 0.5 s=62.7 GeV : ISR dx / center shift described in Appendix A as a systematic 0.4 /dx

σ σ d uncertainty. For the evaluation, cross sections were calcu- d 0.4 lated in the different acceptances according to the result of 0.3 0.3 the beam center shift while requiring r<2 cm, and the variations were applied as a systematic uncertainty. 0.2 0.2

TABLE II. Systematic uncertainties for the cross section mea- 0.1 0.1 surement. The absolute normalization error is not included in these errors. The absolute normalization uncertainty was esti- 0 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 mated by BBC counts to be 9.7% (22:9 2:2 mb for the BBC x xF F trigger cross section). Æ FIG. 13PHENIX, PRD, 88, 032006 (2013) (color online). The cross section results for forward LHCf, K.Kawade, PhD thesis, CERN-THESIS-2014-315 Exponential p Gaussian p T T neutron productionp < 0.11 in p xp GeV/ccollisions at ps Figure200 GeV 6.8:are xF distribution of neutrons atNarrower angle than pT range 0

109 √s scaling, or breaking? LHCf 2.76TeV and 7TeV data shows √s scaling of forward �0 A. ADARE et al. PHYSICAL REVIEW D 88, 032006 (2013)

TABLE I. The expected pT for r<2 cm, mean pT value with TABLE III. The result of the differential cross section the experimental cut, and the efficiency for the experimental cut d=dxF mb for neutron production in p p collisions at ps estimated by the simulation (Fig. 12). The errors were derived 200 GeVð. TheÞ first uncertainty is statistical,þ after the unfolding,¼ ffiffiffi considering the uncertainty in the parameter a xF in theRHICfand the second is the systematic uncertainty. The absolute Gaussian form evaluated by HERA. ð Þ normalization error, 9.7%, is not included.

Neutron x Mean p (GeV=c) Efficiency x Exponential p form Gaussian p form F T h Fi T T 0.45–0.60 0.072 0:779 0:014 1:8% 0.53 0:243 0:024 0:043 0:194 0:021 0:037 Æ ð Þ Æ Æ Æ Æ 0 0.60–0.75 0.085 0:750 0:009 1:2% 0.68 0:491 0:039 0:052 0:455 0:036 0:085 0.75–0.90 0.096 0:723 Æ 0:006ð0:8%Þ 0.83 0:680 Æ 0:044 Æ 0:094 0:612 Æ 0:044 Æ 0:096 0.90–1.00 0.104 0:680 Æ 0:016ð2:3%Þ 0.93 0:334 Æ 0:035 Æ 0:111 0:319 Æ 0:037 Æ 0:123 Æ ð Þ Æ Æ Æ Æ � are the efficiency for the experimental cuts and are listed in B. Result Table I. The errors were derived considering the The differentialLHCf cross section, d=dx , for forward uncertainty in the parameter a xF in the Gaussian form F evaluated by HERA. There is noð significantÞ difference in neutron production in p p collisions at ps 200 GeV was determined using twoþ p distributions:¼ a Gaussian the result in case of using the ISR (exponential) pT T ffiffiffi distribution. form, as used in HERA analysis, and an exponential form, used for ISR data analysis. The results are listed in The mean values of the simulated pT distributions in each energy region are also listed in Table I. The cross Table III and plotted in Fig. 13. We show the results for xF section was obtained after the correction of the energy above 0.45 since the data below 0.45 are significantly unfolding and the cut efficiency. affected by the energy cutoff before the unfolding. The Table II summarizes all systematic uncertainties eval- pT range in each xF bin is 0

was estimated to be 7% with the r<2 cm cut. s=52.8 GeV : ISR [mb] 0.5 F 0.5

In the cross section analysis, we evaluated the beam (mb) s=62.7 GeV : ISR /dx dx σ center shift described in Appendix A as a systematic /

0.4 d 0.4 σ

uncertainty. For the evaluation, cross sections were calcu- d lated in the different acceptances according to the result of 0.3 0.3 the beam center shift while requiring r<2 cm, and the 0.2 variations were applied as a systematic uncertainty. 0.2 neutron 0.1 TABLE II. Systematic uncertainties for the cross section mea- 0.1 surement. The absolute normalization error is not included in 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 these errors. The absolute normalization uncertainty was esti- 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 xF mated by BBC counts to be 9.7% (22:9 2:2 mb for the BBC xF trigger cross section). Æ FIG. 13 (color online). The cross section results for forwardFigure 6.8: xF distribution of neutrons at pT range 0

109 RHICf is a kind of Zero degree calorimeters @STAR interaction point

RHICf detector (traditional) Zero • Former LHCf Arm1 detector used at LHC Degree Calorimeter • Two compact sampling calorimeters behind RHICf • 44 r.l. (1.7 hadron interaction lengths) • <5% and 40% energy resolutions for EM and hadronic showers, respectively • <0.2mm and <1mm position resolutions for EM and hadronic showers, respectively

43 Collision rates in RHICf week

TL center Top TS center TS center

• Higher �* (=8m) than usual RHIC operation • Radial polarization (usually vertical) to maximize the single-spin asymmetry in vertical • Luminosity∼1031 cm-2s-1 44 Quick look（statistics）

RHICf DAQ rate • Max rate was limited ~1kHz • High rate events were prescaled • Low rate events were enhanced with special triggers • Prescale factors were optimized from time to time

Total : 110M events RHICf+STAR RHICf (shower event) Total acquisition time 1659min = 27.7 hours RHICf (High-energy EM trigger)

RHICf (Type-I �0 trigger)

45 Quick look（polarization & spectrum） Though physics of polarized beam collision is not covered in this talk…

• Polarization angle is 0 in usual RHIC operation (vertical pol) • Radial polarization (90°) was required for RHICf operation • Stable radial pol and asymmetry was observed by ZDC

Enhanced by special triggers

• Energy spectrum of EM-like showers in a 30 minutes run ∼250GeV • High-energy EM showers and �0 were selectively triggered to compensate the limited DAQ speed.

46 Quick look（basic performance）

Hit maps of >200GeV hadron-like events at different detector positions => Determination of “zero degree”

Invariant mass of photon pairs => 135MeV peak by �0

Correction factors considering the final alignment and RHIC energy range are in study. 47 Quick look（common run with STAR）

East West

Neutral particles ZDC RHICf ZDC

collision

Roman Pot

• Hadron-like (deep penetrating) showers were selected • Anticorrelation between the RHICf raw (folded) energy and ZDC measured energy (in ADC unit) is confirmed

• (Anti)correlation only with West ZDC as expected => correct event matching 48 Model tuning to the LHCf resultsSIBYLL2.3c-Diffの光子スペクトル Tuned SIBYLL2.3 η>10.94 8.81<η<8.99 −5 −5 ] 10 ] 10 -1 -1

s = 13 TeV, 0.191 nb-1 s = 13 TeV, 0.191 nb-1 [GeV [GeV γ γ 10−6 η>10.94, ∆φ=180° 10−6 8.81<η<8.99, ∆φ=20°

/dE /dE Data LHCf-Arm1, arXiv:1703.07678 γ γ

Data Nch=0, ATLAS-CONF-2017-075

dN dN Syst. + Stat. Unc. LHCf-Arm1

inel −7 inel −7 Syst. + Stat. Unc. N =0 10 10 ch 1/N 1/N

−8 −8 10 Original SIBYLL2.310 Data LHCf-Arm1, arXiv:1703.07678 EPOS-LHC Data N =0, ATLAS-CONF-2017-075 −9 ch −9 SIBYLL2.3c 10 Syst. + Stat. Unc. LHCf-Arm1 10 SIBYLL2.3 Syst. + Stat. Unc. N =0 ch SIBYLL2.3c-Diff

EPOS-LHC EPOS-LHC Nch=0 EPOS-LHC Nch=0 −10 SIBYLL2.3c SIBYLL2.3c Nch=0 −10 10 10 SIBYLL2.3c Nch=0 SIBYLL2.3 SIBYLL2.3 Nch=0 SIBYLL2.3 Nch=0 SIBYLL2.3c-Diff SIBYLL2.3c-Diff Nch=0 SIBYLL2.3c-Diff Nch=0 10−11 10−11 1000 2000 3000 4000 5000 6000 1000 2000 3000 4000 5000 6000 Photon Energy [GeV] Photon Energy [GeV]

• LHCf photon events are categorized in ‘diffractive’ and SIBYLL2.3c-Diffは連動実験の結果をよく再現できるようになった。 ‘non-diffractive’ using the ATLAS information X

• SIBYLL2.3c–diff tuned to LHCf diffractive photons 49 SIBYLL2.3c-DiffProton-proton equivalentによる cm. energyXmaxps (TeV)の予測

) 10 100

2 Effect of new tune to air showerCredit to Felix Reihn SIBYLL2.3c-DiffによるXmaxの予測 ) 2 Credit to Felix Reihn 2 max 10 Small but visible impact (gcm Auger 2014 i HiRes / MIA

max 800 q = 67 X h (gcm Auger 2014Sibyll 2.3c-diffFly’s Eye Epos-LHC i Proton Iron 6 10 3 HiRes /Sibyll MIA700 2.3c QGSjetII-04 max 800 600

0 Sibyll 2.3c-diff Epos-LHC

⇥ X Sibyll 2.3c QGSjetII-04 h E Fly’s Eye

/ 3 16 17 18 19 20

Avg. depth of shower max. 10 10 10 10 10 i 4 10 Primary energy E (eV) Proton-proton equivalent cm. energy ps (TeV) 0 10 100 ) Iron 2 X 0 ⇥ 10 3 q = 67 Sibyll 2.3c-diff Epos-LHC

E 700 Proton 3 Iron 3 10 N� 6 10 Sibyll 2.3c QGSjetII-04 0 ( ⇥ E

/ 3 i 4 10 ) µ ⇥ Iron

0 ⇥ 3 E 3 10 (

µ ⇥ N

N 3 3 h 2 10 h 2 10 ⇥ Proton 2 Eµ > 1GeV 2240g/cm ⇥ 3 Proton 10 600 1016 1017 1018 1019 1020 50 SibyllSignificant impact 2.3c-diff Epos-LHCPrimary energy E0 (eV) 2 Eµ > 1GeV 2240g/cm 3 Sibyll 2.3c QGSjetII-04 10 16 17 18 19 20 10 16 10 17 10 18 10 19 10 20

Avg. depth of shower max. 10 10 10 10 10 Primary energy E0 (eV) Primary energy E0 (eV)

X Summary

• Air shower analyses are influenced by the uncertainty of hadronic interaction • Current and future colliders cover the collision energies appropriate to the air shower observations • Dedicated forward measurements are important • LHCf and RHICf observed forward particle production from 14 17 ECR=10 eV to 10 eV • Current popular models have reasonable agreement with the LHCf results and they are updating • Next keys are • Combined analyses with the central detector (LHCf-ATLAS, RHICf-STAR) • Collision energy (in)dependence of particle production (ISR-PHENIX- RHICf-LHCf) • Light ion collisions like p-O, O-O at LHC • Feedback from the air shower observations

51 52 proton CR proton • km2~1000km2 detection neutron

�± area is achieved using a

�0 sparse array of the ground photon (�) Atmospheric detectors nucleus muon • Identification of primary neutrino particle (nuclei, gamma, etc…) is possible by measuring the difference in

Electromagnetic the air shower development cascade => Analyses strongly rely on the MC simulation of air shower development, especially fundamental hadronic Spread over 100m – few km interaction is essential 53 粒⼦⽣成素過程と空気シャワー発達 ① Inelastic cross section

If large s rapid development ④ 2ndary interactions If small s nucleon, p deep penetrating ② Forward energy spectrum If softer shallow development If harder deep penetrating

③ Inelasticity k= 1-plead/pbeam If large k (π0s carry more energy) rapid development If small k ( baryons carry more energy) deep penetrating

54 DEPTH OF MAXIMUM OF … . I. MEASUREMENTS AT … PHYSICAL REVIEW D 90, 122005 (2014) 19 2 E 7.9 0.3 × 10 eV and Xmax 762 2 g=cm , point in Fig. 13 corresponds to a center-of-mass energy that respectively,¼ð Æ whereÞ the uncertainties are¼ statisticalÆ only. is only four times larger than the one currently available at The Xmax distributions after event selection are shown the LHC (ps 8 TeV). Therefore, unless the models have rec ¼ in Fig. 12. These are the “raw” distributions [fobs Xmax in deficiencies in phase-space regions that are not covered Eq. (4)] that still include effects of the detector resolutionð Þ well by LHCffiffiffi measurements, the uncertainties due to the and the acceptance. Electronically readable tables of the extrapolation of hadronic interactions to the lower energy distributions, as well as the parameters of the resolution and threshold of this analysis should be small. On the other acceptance, are available at [89]. A thorough discussion of hand, the last energy bin at lg E=eV 19.62 corre- h ð Þi ¼ the distributions can be found in an accompanying paper sponds to a center-of-mass energy that is a factor of about [94], where a fit of the data with simulated templates for 40 higher than the LHC energies and the model predictions different primary masses is presented. have to be treated more carefully. In this paper we will concentrate on the discussion of Comparing the energy evolution of Xmax for data h i the first two moments of the X distribution, X and and simulations in Fig. 13 it can be seen that the slope max h maxi σ Xmax , which are listed in Table IV together with their of the data is different than what would be expected for statisticalð Þ and systematic uncertainties. The statistical either a pure-proton or pure-iron composition. The change uncertainties are calculated with the parametric bootstrap of Xmax with the logarithm of energy is usually referred h i 58 W.D. Apel et al. / Astroparticlemethod. Physics For 47 this (2013) purpose, 54–66 the data are fitted with Eq. (4) to as elongation rate [17–19], assuming the functional form suggested in [76] as f Xmax . ð Þ d Xmax mixed composition,Given this parametric model of the truemixedX maxcomposition,distribution, D10 h i : 9 6.6

the datanumber of simulated showers is shown in Fig. 13. The simulations have been

1 performed1 using the three contemporary hadronic inter- d ln A D Dˆ 1 − h i : 10 action models that were either tuned to recent LHC data 10 ¼ 10 d ln E=eV ð Þ -0.3 -0.2 -0.1 0 0.1 0.2 0.3(QGSJetII-04-0.2[95,96] -0.1, Epos-LHC 0[97,98] 0.1) or 0.2 found in 0.3 good ð Þ deviation after correction log (Nrec)-Cbias-log (Ntru) deviation after correction log (Nrec)-Cbias-log (Ntru) CR mass estimation from air shower10 ch ch 10 chagreement with these measurements10 µ (Sibyll2.1µ 10 µ[81], see A single linear fit of Xmax as a function of lg E does h 2i ð Þ Fig. 4. The distribution of the deviations between the reconstructed (bias corrected)[99]). and It is true worth shower sizes noting in case that of charged the energy particles (left, of exemplarily the firstfor data the intervalnot describe our data well (χ =ndf 138.4=16). Allowing 6:6 < log Ntru < 6:7) and of muons (right, exemplarily for the interval 5:7 < log Ntru < 5:8), as well as the determined parametrization (curves). To increase the available ¼ 10ð ch Þ 10ð l Þ simulation statistics, a mixed composition is used. KASCADE Grande, Astropart. Phys., 47 (2013) 54-66 PAO, PRD, 90, 122005 (2014)

ations as well as that of the properties of the experiment. Shown ) KASCADE-Grande data (full efficiency) rec rec rec ch 9 0°-18° are some isolines representing the cells log10 Nch ; log10 Nl (N ð Þ ð Þ i

10 13 c. 78000 events 103 of the data plane with constant probability (from the inner to 4 8.5 the outermost isoline: 0:1; 0:05 and 10À probability density). For isolines according to 0.01% reasons of clarity, only the results for two exemplary primaries are response matrix for: 5% p 10% 9.22-9.36 illustrated: protons and iron nuclei. The isolines, which correspond 8 rec rec Fe 102 to the log N -log N combinations with a probability of 8.80-8.94 10ð ch Þ 10ð l Þ 4 7.5 10À , represent the smallest probability value just considered in 8.38-8.52 the response matrix after its conditioning. As can be seen, these outer isolines cover almost all measured data; hence, the minimal prob-

7 0.01% number of showers 7.96-8.10 10 ability is not set too large. intervals of true energy 5% 6.5 in log (E/GeV): 7.54-7.68 10 4. Error propagation

rec. number of charged particles log 7.12-7.26 6 1 The determination of the elemental energy spectra will be sub- 5 5.5 6 6.5 7 7.5 jected to inﬂuences of different error sources. They can roughly be reconstructed number of muons log (Nrec) 10 µ classiﬁed in four categories (cf. [17] for details):

Fig. 5.Response (color contours) was calculated using A comparison between the measured shower size distribution (grey histogram) and some isolines with log Nrec -log Nrec combinations of constant (i) Statistical uncertainties due to the limited measurement time: QGSJET II-02 + FLUKA 2002.4 10ð l Þ 10ð ch Þ probability according to the parametrizations (from the inner to the outermost line Due to the limited exposure, the measured data sample will ü10%, 5% and 0.01%). This is illustrated exemplarily for protons as well as iron nuclei,FIG. 13. Energysuffer evolution from unpreventable of the first two statistical central moments uncertainties, of the whichX distribution compared to air-shower simulations for proton and and inTechniques to estimate mass from air shower observations case of six energy bins (labelled below each isoline set). are expected to be Poisson distributed. These uncertaintiesmax iron primaries [80,81,95 98]. will be propagated– through the applied unfolding algorithm In order• toKnee to Ankle : N parametrize the dependencee of– theNmu resolution of the experiment on the true sizes, a possible bias in the charged particle and are usually amplified thereby. The statistical uncertain- and muon• numberAnkle to Cutoff : < reconstruction must firstXmax be corrected>, Nmu by using ties can be determined by means of a frequentist approach: appropriate correction functions Cbias and Cbias, respectively, deter- The measured two-dimensional shower size plane is consid- ü Interpretations rely on the MC predictions with an ch l assumed hadronic 122005-19 mined based on the simulations. The correction is typically in the ered as probability distribution. Based on a random55 genera- orderinteraction model of less than 10%. The distributions of the remaining devia- tor, a couple of artificial data sets are generated, which are tions between the reconstructed (and bias corrected) and true unfolded individually. The spread of the solutions represents shower sizes are depicted in Fig. 4 for the charged particle number a good estimate for the statistical uncertainty due to the lim- (left panel) and for the muon number (right panel), in case of dis- ited measurement time. Systematic bias induced by the unfolding method: crete exemplary true shower size intervals (corresponding to about (ii) In the con- 30 PeV to 40 PeV primary energy). Since the resolution does not text of the convergence properties of the iterative unfolding differ significantly between different primaries, in order to increase algorithms, small numbers of iteration steps will on the one statistics, the simulations for the five primary particles can be com- hand reduce the amplification of the statistical uncertainties bined to a mixed composition set serving for the parametrization. of the data sample, and on the other hand will result in a In Fig. 5, the measured shower size plane is compared to the solution that is deviating from the exact one. In case of the probabilities given by the final response matrix taking into account regularized techniques it is similar, since the regularization the entire parametrizations, i.e. that of the intrinsic shower fluctu- damps oscillations, but, conversely, results in a biased solution. In this work, the number of iteration steps, respectively the regularization parameter, is chosen such that an optimal 13 In case of smaller energies, the widths of the probability distributions are as large balance between the statistical uncertainties and the sys- that there are no individual probabilities larger than 0.1 or even 0.05, such that the inner isolines are missing in these cases. tematic bias is achieved. The bias can be estimated based What is difficulty in hadronic interaction? • Hard scattering (large Q2); Perturbative QCD 2->2 scattering of individual partons (quarks and gluons) • Parton-parton cross section • Parton distribution function (PDF) • Hadronization (fragmentation function) (Collins and Martin, 1984) are relatively well understood separately • Soft scattering (small Q2); non-perturbative exchange of group of partons（mesons, Reggeon, Pomeron） difficult in first principle calculation of cross sections even for the elastic scattering phenomenological modeling is necessary u Nature does not have a boarder between soft and hard…

• Nuclear effect； not so simple like “Glauber effect” collective effect in dense matter depends on the impact parameter and issue of modeling Models must be tested by accelerator data 56 Setup of SPS beam test

Detector (Arm#1 or Arm#2)

Silicon Tracker (ADAMO)

Calorimeter

Moving Table Trigger Scintillator

Ø Calorimeters were exposed to the SPS electron (100-200GeV), proton (350GeV) and muon (150GeV) beams Ø Calorimeters were scanned over the beams

Ø Impact points were measured using a silicon strip tracker 57 Channel to channel calibration Beam test results vs. MC simulation

180GeV electron

Ø Charge => energy factors are determined by scaling the MC results

to the experimental data 58 Cosmic-ray flux

End of galactic nuclear CR and transition to extra-gal CR

Ankle Cutoff: Knee: end of end of CR spectrum galactic proton CR

• Maximum energy up to 1020 eV (LHC beam energy is 7x1012 eV) • Structures in the energy spectrum suggest their origins • Knee @ 1015eV … end of the galactic protons ?? • Ankle @ 1018eV … transition from galactic to ex-galactic origins ?? • Cutoff @ 1020eV … end of the ex-galactic origin ?? 59 0 π SIBYLL 2.1 11

(a) 0.0 < p [GeV] < 0.2 (b) 0.2 < p [GeV] < 0.4 (c) 0.4 < p [GeV] < 0.6 ] T ] T ] T

-2 1 -2 -2 10−1 [GeV [GeV [GeV 3 3 3 10−2 /dp /dp /dp σ σ σ 3 10−3 LHCf s=7TeV 3 3

Ed -1 Ed Ed 10−4 Ldt=2.64+2.85nb inel ∫ inel inel σ −5 σ σ 1/ 10 1/ 1/

(d) 0.6 < p [GeV] < 0.8 (e) 0.8 < p [GeV] < 1.0 ] p [GeV]T ] p [GeV]T p [GeV]

-2 1 z -2 z z 10−1 LHCf (stat.+syst.) [GeV [GeV 3 3 10−2 DPMJET 3.06 /dp /dp

σ σ QGSJET II-04 3 10−3 3 SIBYLL 2.1 Ed −4 Ed

inel 10 inel PYTHIA 8.185 σ σ 5

1/ − 1/ 10 EPOS LHC 1000 2000 3000 1000 2000 3000 p [GeV] p [GeV] z z ü FIG. 5: (color online).Underestimate in low Experimental combined pz spectrapT, overestimate in high of the LHCf detector (ﬁlled circles)pT in p + p collisions at ps =7TeV. Shaded rectangles indicate the total statistical and systematic uncertainties. The predictions of hadronic interaction models ü 60 are shown for comparisonTotally overestimate because of larger phase space in high (see text for details.) pT

σ -1 σ σ σ −5 Ldt=2.36nb 1/ 10 ∫ 1/ 1/ 1/

] (e)p 9.6 [GeV] < y < 9.8 p [GeV] p [GeV] p [GeV]

-2 1 T T T T 10−1 LHCf (stat.+syst.) [GeV 3 10−2 DPMJET 3.06 /dp

σ QGSJET II-04 3 10−3 SIBYLL 2.1 Ed −4

inel 10 PYTHIA 8.185 σ 5

1/ − 10 EPOS LHC 0 0.1 0.2 0.3 0.4 p [GeV] T

(a) 0.0 < p [GeV] < 0.2 (b) 0.2 < p [GeV] < 0.4 (c) 0.4 < p [GeV] < 0.6 ] T ] T ] T

-2 1 -2 -2 10−1 [GeV [GeV [GeV 3 3 3 10−2 /dp /dp /dp σ σ σ 3 10−3 LHCf s=7TeV 3 3

Ed -1 Ed Ed 10−4 Ldt=2.64+2.85nb inel ∫ inel inel σ −5 σ σ 1/ 10 1/ 1/

(d) 0.6 < p [GeV] < 0.8 (e) 0.8 < p [GeV] < 1.0 ] p [GeV]T ] p [GeV]T p [GeV]

-2 1 z -2 z z 10−1 LHCf (stat.+syst.) [GeV [GeV 3 3 10−2 DPMJET 3.06 /dp /dp

σ σ QGSJET II-04 3 10−3 3 SIBYLL 2.1 Ed −4 Ed

inel 10 inel PYTHIA 8.185 σ σ 5

1/ − 1/ 10 EPOS LHC 1000 2000 3000 1000 2000 3000 p [GeV] p [GeV] z z

FIG. 5: (colorü online).Very good agreement at mid Experimental combined pz spectra of- theenergy (large cross section range) LHCf detector (ﬁlled circles) in p + p collisions at ps =7TeV. Shaded rectangles indicate the total statistical and systematic uncertainties. The predictions of hadronic interaction models 61 are shownü for comparisonSlightly overestimate in high energy (small cross section range) (see text for details.)

σ -1 σ σ σ −5 Ldt=2.36nb 1/ 10 ∫ 1/ 1/ 1/

] (e)p 9.6 [GeV] < y < 9.8 p [GeV] p [GeV] p [GeV]

-2 1 T T T T 10−1 LHCf (stat.+syst.) [GeV 3 10−2 DPMJET 3.06 /dp

σ QGSJET II-04 3 10−3 SIBYLL 2.1 Ed −4

inel 10 PYTHIA 8.185 σ 5

1/ − 10 EPOS LHC 0 0.1 0.2 0.3 0.4 p [GeV] T

(a) 0.0 < p [GeV] < 0.2 (b) 0.2 < p [GeV] < 0.4 (c) 0.4 < p [GeV] < 0.6 ] T ] T ] T

-2 1 -2 -2 10−1 [GeV [GeV [GeV 3 3 3 10−2 /dp /dp /dp σ σ σ 3 10−3 LHCf s=7TeV 3 3

Ed -1 Ed Ed 10−4 Ldt=2.64+2.85nb inel ∫ inel inel σ −5 σ σ 1/ 10 1/ 1/

(d) 0.6 < p [GeV] < 0.8 (e) 0.8 < p [GeV] < 1.0 ] p [GeV]T ] p [GeV]T p [GeV]

-2 1 z -2 z z 10−1 LHCf (stat.+syst.) [GeV [GeV 3 3 10−2 DPMJET 3.06 /dp /dp

σ σ QGSJET II-04 3 10−3 3 SIBYLL 2.1 Ed −4 Ed

inel 10 inel PYTHIA 8.185 σ σ 5

1/ − 1/ 10 EPOS LHC 1000 2000 3000 1000 2000 3000 p [GeV] p [GeV] z z

FIG. 5: (colorü Perfect in shape, slightly underestimate in higher online). Experimental combined pz spectra of the LHCf detector (ﬁlled circles)p inT p + p collisions at ps =7TeV. Shaded rectangles indicate the total statistical and systematic uncertainties. The predictions of hadronic interaction models 62 are shownü forTotally slightly underestimate comparison (see text for details.)

σ -1 σ σ σ −5 Ldt=2.36nb 1/ 10 ∫ 1/ 1/ 1/

] (e)p 9.6 [GeV] < y < 9.8 p [GeV] p [GeV] p [GeV]

-2 1 T T T T 10−1 LHCf (stat.+syst.) [GeV 3 10−2 DPMJET 3.06 /dp

σ QGSJET II-04 3 10−3 SIBYLL 2.1 Ed −4

inel 10 PYTHIA 8.185 σ 5

1/ − 10 EPOS LHC 0 0.1 0.2 0.3 0.4 p [GeV] T

ü Lowest neutron yield, especially at zero degree

63 Neutron QGSJET II-03

ü Qualitatively nice agreement, only model, at zero degree ü Lower yield at non-zero angle

64 Neutron EPOS 1.99

ü Generally low yield

65 Detector performance

Arm2

ΔE/E ≈ 40%

ΔE/E < 5%

66 モデルのdiffraction事象の生成質量分布 Energy flow Δη low-mass Δη üPost-LHC models (EPOS-LHC and QGSJET II-04) well explain 0 high-mass P the π1 results, but not for neutronsX ∆η 18 16 14 12 10 8 6 4 2 0 ü P DPMJET3 explains the neutron results, but recently it is not a Rapidity gap )] X )] ξ

popular model in CR simulations X (

ξ 1.4 EPOS-LHC

6 ( 10 P P 10 QGSJET-II-04 üLHCf 13 2 TeV data allows to access smaller rapidity than 7TeV 3 1.2 1 SIBYLL2.3 )[mb/log X

ξ 0.8 PYTHIA8212-SS ( 10 0 0.6 Neutron PYTHIA8212-DL )[mb/log 4

MX : Diffractiveπ 質量, X /dlog −8(data contains all neutral hadrons −7 −6 SD ξ σ ( at 140m while MC is only neutrons)log (ξ ) d 10 X X系統の不変質量。 10 2 /dlog SD

LHCfM σ X 2 −Δη d ξX = (DPM3) ! e 0 QGS IIs -04 −8 −6 −4 −2 0 EPOS-LHC log (ξ ) 10 Low-mass X 非弾性度 rapidity (y) pseudo-rapidity (η) to be covered in 13TeV 小大 DiffractionBlack solid circle : LHCfの生成質量分布はモデルによって違う、特にdata (π0 , LHCf 2012) low-mass 0 Dotted lines : π energy flow distribution of each model 67 diffractionThick horizontal line : Energy flow calculation after の取り扱い。 pT cut X low-mass diffraction ∆η 事象の感度 18 16 14 12 10 8 6 4 2 0 )]

X ❖

)] Central-veto low-mass ξ は X (

ξ 1.4 EPOS-LHC

6 ( 10 10 1.2 QGSJET-II-04 1 SIBYLL2.3 diffraction (log10(ξx) < -6.0)に対し )[mb/log X

ξ 0.8 PYTHIA8212-SS ( 10 0.6 PYTHIA8212-DL )[mb/log 4 100% X /dlog −8 −7 −6 て約の検出効率がある。 SD ξ σ ( log (ξ ) d 10 X 10 ❖ ATLAS-LHCf連動実験によるこれ 2 /dlog

SD まで測定例のないlow-mass σ ? d 0 diffraction事象の選別方法を確立。 −8 −6 −4 −2 0 log (ξ ) Low-mass 10 X )]

X 1.0 ξ

( QGSJET-II-04 p-p, s = 13 TeV

10 4 0.8

SD(pp -> pX) Efficiency 3 SD w/ central-veto

)[mb/log 0.6 X ξ ( 10 2 0.4 /dlog

SD 1 0.2 σ d 0.0 0 −8 −6 −4 −2 0 −8 −6 −4 −2 0 log68 (ξ ) log (ξ ) 10 X 10 X X 結果１：光子スペクトル η>10.94 8.81<η<8.99 −5 −5 ] 10 ] 10 -1 -1 ATLAS-LHCf Preliminary ATLAS-LHCf Preliminary s = 13 TeV, 0.191 nb-1 s = 13 TeV, 0.191 nb-1 [GeV [GeV γ LHCfのみの結果 γ 10−6 η>10.94, ∆φ=180° 10−6 8.81<η<8.99, ∆φ=20°

/dE /dE Data LHCf-Arm1, arXiv:1703.07678

γ (Diff.+non-diff.) γ

Data Nch=0, ATLAS-CONF-2017-075

dN dN Syst. + Stat. Unc. LHCf-Arm1

inel −7 inel −7 Syst. + Stat. Unc. N =0 10 10 ch 1/N 1/N

10−8 10−8 連動実験の結果 Data LHCf-Arm1, arXiv:1703.07678 EPOS-LHC Data N =0, ATLAS-CONF-2017-075 Diff. only) −9 ch （ −9 QGSJET-II-04 10 Syst. + Stat. Unc. LHCf-Arm1 10 SIBYLL2.3 Syst. + Stat. Unc. N =0 ch PYTHIA8212DL

EPOS-LHC EPOS-LHC Nch=0 EPOS-LHC Nch=0 −10 QGSJET-II-04 QGSJET-II-04 Nch=0 −10 10 10 QGSJET-II-04 Nch=0 SIBYLL2.3 SIBYLL2.3 Nch=0 SIBYLL2.3 Nch=0 PYTHIA8212DL PYTHIA8212DL Nch=0 PYTHIA8212DL Nch=0 10−11 10−11 1000 2000 3000 4000 5000 6000 1000 2000 3000 4000 5000 6000 Photon Energy [GeV] Photon Energy [GeV]

✦ Nch = 0 の光子スベクトルのデータとモデル予測の比較： ‣ η > 10.94：EPOS-LHCはデータをよく再現した。

‣ 8.81 < η < 8.99：PYTHIA8212DLはデータをよく再現した。 69 X Diffractionの生成質量分布の改良

P 1 X MX : Diffractive mass, X系統の不変質量 P Rapidity gap Δη M X 2 −Δη ξX = ( ) ! e :陽子が失った運動量の比 P P s 2 3 SD dσ 2 1−2α P (t ) 2 α P (0)−1 = fP (ξX ,t)σ (M X ) ∝ξX ((M X ) ) dξX dt pomeron ﬂux pomeron cross-section

α P (t) = α P (0) +α P′t = 1+ Δ +α ′t : Pomeron 軌道 ∆η 18 16 14 12 10 8 6 4 2 0

)] SIBYLL2.3に実装した X ξ

)] 2 (

X EPOS-LHC

6 ξ ( 10 10 QGSJET-II-04 1.5 pomeron ﬂux SIBYLL2.3 元の

)[mb/log SIBYLL2.3c

X 1 ξ ( SIBYLL2.3c-Diff 10

)[mb/log 4 SIBYLL2.3 0.5 の作者である X

/dlog −8 −7 −6 ξ SD (

σ log (ξ )

d 10 X 10 Felix Reihnは連動解析の結 2 /dlog 果に基づいて、SIBYLL2.3c SD σ d のdiffractionの生成質量分布 0 −8 −6 −4 −2 0 を改良した。 70 log (ξ ) 10 X X SIBYLL2.3c-Diffの光子スペクトル Tuned SIBYLL2.3 η>10.94 8.81<η<8.99 −5 −5 ] 10 ] 10 -1 -1

s = 13 TeV, 0.191 nb-1 s = 13 TeV, 0.191 nb-1 [GeV [GeV γ γ 10−6 η>10.94, ∆φ=180° 10−6 8.81<η<8.99, ∆φ=20°

/dE /dE Data LHCf-Arm1, arXiv:1703.07678 γ γ

Data Nch=0, ATLAS-CONF-2017-075

dN dN Syst. + Stat. Unc. LHCf-Arm1

inel −7 inel −7 Syst. + Stat. Unc. N =0 10 10 ch 1/N 1/N

SIBYLL2.3c-Diffは連動実験の結果をよく再現できるようになった。71 X