arX

iv:0

712.

1147

v2 [

astr

o-ph

] 2

9 Ja

n 20

08

Upper Limit on the Cosmi -Ray Photon FluxAbove 1019 eV Using the Surfa e Dete tor of thePierre Auger ObservatoryThe Pierre Auger Collaboration: J. Abraham14, P. Abreu69,M. Aglietta55, C. Aguirre17, D. Allard33, I. Allekotte7, J. Allen89,P. Allison91, J. Alvarez-Muñiz76, M. Ambrosio58,L. An hordoqui103, 90, S. Andringa69, A. Anzalone54, C. Aramo58,S. Argirò52, K. Arisaka94, E. Armengaud33, F. Arneodo56,F. Arqueros73, T. As h39, H. Asorey5, P. Assis69,B.S. Atulugama92, J. Aublin35, M. Ave95, G. Avila13, T. Bä ker43,D. Badagnani10, A.F. Barbosa19, D. Barnhill94, S.L.C. Barroso25,P. Bauleo83, J.J. Beatty91, T. Beau33, B.R. Be ker100,K.H. Be ker37, J.A. Bellido92, S. BenZvi102, C. Berat36,T. Bergmann42, P. Bernardini48, X. Bertou5, P.L. Biermann40,P. Billoir35, O. Blan h-Bigas35, F. Blan o73, P. Blasi86, 46, 57,C. Bleve79, H. Blümer42, 38, M. Bohá£ová31, C. Bonifazi35, 19,R. Bonino55, M. Boratav35, J. Bra k83, 96, P. Brogueira69,W.C. Brown84, P. Bu hholz43, A. Bueno75, R.E. Burton81,N.G. Bus a33, K.S. Caballero-Mora42, B. Cai98, D.V. Camin47,L. Caramete40, R. Caruso51, W. Carvalho21, A. Castellina55,O. Catalano54, G. Cataldi48, L. Cazon95, R. Cester52,J. Chauvin36, A. Chiavassa55, J.A. Chinellato23, A. Chou89, 86,J. Chye88, P.D.J. Clark78, R.W. Clay16, E. Colombo2,R. Con eição69, B. Connolly100, F. Contreras12, J. Coppens63, 65,A. Cordier34, U. Cotti61, S. Coutu92, C.E. Covault81,A. Creusot71, A. Criss92, J. Cronin95, A. Curutiu40,S. Dagoret-Campagne34, K. Daumiller38, B.R. Dawson16,R.M. de Almeida23, C. De Donato47, S.J. de Jong63,G. De La Vega15, W.J.M. de Mello Junior23,J.R.T. de Mello Neto95, 28, I. De Mitri48, V. de Souza42,L. del Peral74, O. Deligny32, A. Della Selva49, C. Delle Fratte50,H. Dembinski41, C. Di Giulio50, J.C. Diaz88, C. Dobrigkeit 23,Preprint submitted to Elsevier S ien e 2 February 2008

J.C. D'Olivo62, D. Dorni 32, A. Dorofeev87, J.C. dos Anjos19,M.T. Dova10, D. D'Urso49, I. Dutan40, M.A. DuVernois97, 98,R. Engel38, L. Epele10, M. Erdmann41, C.O. Es obar23,A. Et hegoyen3, P. Fa al San Luis76, H. Fal ke63, 66, G. Farrar89,A.C. Fauth23, N. Fazzini86, F. Ferrer81, S. Ferry71, B. Fi k88,A. Filevi h2, A. Filip£i£70, I. Fle k43, R. Fonte51,C.E. Fra hiolla20, W. Fulgione55, B. Gar ía14,D. Gar ía Gámez75, D. Gar ia-Pinto73, X. Garrido34,H. Geenen37, G. Gelmini94, H. Gemmeke39, P.L. Ghia32, 55,M. Giller68, H. Glass86, M.S. Gold100, G. Golup6,F. Gomez Albarra in10, M. Gómez Berisso6, R. Gómez Herrero74,P. Gonçalves69, M. Gonçalves do Amaral29, D. Gonzalez42,J.G. Gonzalez87, M. González60, D. Góra42, 67, A. Gorgi55,P. Gou�on21, V. Grassi47, A.F. Grillo56, C. Grunfeld10,Y. Guardin erri8, F. Guarino49, G.P. Guedes24, J. Gutiérrez74,J.D. Hague100, J.C. Hamilton33, P. Hansen76, D. Harari6,S. Harmsma64, J.L. Harton32, 83, A. Haungs38, T. Haus hildt55,M.D. Healy94, T. Hebbeker41, G. Hebrero74, D. He k38,C. Hojvat86, V.C. Holmes16, P. Homola67, J. Hörandel63,A. Horne�er63, M. Horvat71, M. Hrabovský31, T. Huege38,M. Hussain71, M. Iarlori46, A. Insolia51, F. Ionita95, A. Italiano51,M. Kadu ak86, K.H. Kampert37, T. Karova31, B. Kégl34,B. Keilhauer42, E. Kemp23, R.M. Kie khafer88, H.O. Klages38,M. Kleifges39, J. Kleinfeller38, R. Knapik83, J. Knapp79,D.-H. Koang36, A. Krieger2, O. Krömer39, D. Kuempel37,N. Kunka39, A. Kusenko94, G. La Rosa54, C. La haud33,B.L. Lago28, D. Lebrun36, P. LeBrun86, J. Lee94,M.A. Leigui de Oliveira27, A. Letessier-Selvon35, M. Leuthold41,I. Lhenry-Yvon32, R. López59, A. Lopez Agüera76,J. Lozano Bahilo75, R. Luna Gar ía60, M.C. Ma arone54,C. Ma olino46, S. Maldera55, G. Man arella48, M.E. Man eñido10,D. Mandat31, P. Mants h86, A.G. Mariazzi10, I.C. Maris42,H.R. Marquez Fal on61, D. Martello48, J. Martínez60,O. Martínez Bravo59, H.J. Mathes38, J. Matthews87, 93,J.A.J. Matthews100, G. Matthiae50, D. Maurizio52, P.O. Mazur86,2

T. M Cauley90, M. M Ewen74, 87, R.R. M Neil87, M.C. Medina3,G. Medina-Tan o62, A. Meli40, D. Melo2, E. Meni hetti52,A. Mens hikov39, Chr. Meurer38, R. Meyhandan64,M.I. Mi heletti3, G. Miele49, W. Miller100, S. Mollera h6,M. Monasor73, 74, D. Monnier Ragaigne34, F. Montanet36,B. Morales62, C. Morello55, J.C. Moreno10, C. Morris91,M. Mostafá101, M.A. Muller23, R. Mussa52, G. Navarra55,J.L. Navarro75, S. Navas75, P. Ne esal31, L. Nellen62,C. Newman-Holmes86, D. Newton79, 76, T. Nguyen Thi104,N. Nierstenhoefer37, D. Nitz88, D. Nosek30, L. Noºka31,J. Oehls hläger38, T. Ohnuki94, A. Olinto33, 95,V.M. Olmos-Gilbaja76, M. Ortiz73, F. Ortolani50,S. Ostap henko42, L. Otero14, N. Pa he o74, D. Pakk Selmi-Dei23,M. Palatka31, J. Pallotta1, G. Parente76, E. Parizot33, S. Parlati56,S. Pastor72, M. Patel79, T. Paul90, V. Pavlidou95, K. Payet36,M. Pe h31, J. P�ekala67, R. Pelayo60, I.M. Pepe26, L. Perrone53,S. Petrera46, P. Petrin a50, Y. Petrov83, Diep Pham Ngo 104,Dong Pham Ngo 104, T.N. Pham Thi104, A. Pi hel11, R. Piegaia8,T. Pierog38, M. Pimenta69, T. Pinto72, V. Pirronello51,O. Pisanti49, M. Platino2, J. Po hon5, P. Privitera50, M. Prouza31,E.J. Quel1, J. Rautenberg37, A. Redondo74, S. Reu roft90,B. Revenu33, F.A.S. Rezende19, J. Ridky31, S. Riggi51, M. Risse37,C. Rivière36, V. Rizi46, M. Roberts92, C. Robledo59,G. Rodriguez76, D. Rodríguez Frías74, J. Rodriguez Martino51,J. Rodriguez Rojo12, I. Rodriguez-Cabo76, G. Ros73, 74,J. Rosado73, M. Roth38, C. Rou elle35, B. Rouillé-d'Orfeuil33,E. Roulet6, A.C. Rovero11, F. Salamida46, H. Salazar59,G. Salina50, F. Sán hez62, M. Santander12, C.E. Santo69,E.M. Santos35, 19, F. Sarazin82, S. Sarkar77, R. Sato12,V. S herini37, H. S hieler38, A. S hmidt39, F. S hmidt95,T. S hmidt42, O. S holten64, P. S hovánek31, F. S hüssler38,S.J. S iutto10, M. S uderi51, A. Segreto54, D. Semikoz33,M. Settimo48, R.C. Shellard19, 20, I. Sidelnik3, B.B. Si�ert28,G. Sigl33, N. Smetniansky De Grande2, A. Smiaªkowski68,R. �mída31, A.G.K. Smith16, B.E. Smith79, G.R. Snow99,3

P. Sokolsky101, P. Sommers92, J. Sorokin16, H. Spinka80, 86,R. Squartini12, E. Strazzeri50, A. Stutz36, F. Suarez55,T. Suomijärvi32, A.D. Supanitsky62, M.S. Sutherland91,J. Swain90, Z. Szadkowski68, J. Takahashi23, A. Tamashiro11,A. Tamburro42, O. Ta³  u37, R. T a iu 43, D. Thomas101,R. Ti ona18, J. Ti�enberg8, C. Timmermans65, 63, W. Tka zyk68,C.J. Todero Peixoto23, B. Tomé69, A. Tona hini52, I. Torres59,D. Torresi54, P. Travni ek31, A. Tripathi94, G. Tristram33,D. Ts herniakhovski39, M. Tueros9, V. Tunni li�e78, R. Ulri h38,M. Unger38, M. Urban34, J.F. Valdés Gali ia62, I. Valiño76,L. Valore49, A.M. van den Berg64, V. van Elewy k32,R.A. Vázquez76, D. Veberi£71, A. Veiga10, A. Velarde18,T. Venters95, 33, V. Verzi50, M. Videla15, L. Villaseñor61,S. Vorobiov71, L. Voyvodi 86, H. Wahlberg10, O. Wainberg4,P. Walker78, D. Warner83, A.A. Watson79, S. Westerho�102,G. Wie zorek68, L. Wien ke82, B. Wil zy«ska67, H. Wil zy«ski67,C. Wileman79, M.G. Winni k16, H. Wu34, B. Wundheiler2,T. Yamamoto95, P. Younk101, E. Zas76, D. Zavrtanik71,M. Zavrtanik70, A. Ze h35, A. Zepeda60, M. Ziolkowski43

1 Centro de Investiga iones en Láseres y Apli a iones, CITEFA and CONICET,Argentina2 Centro Atómi o Constituyentes, CNEA, Buenos Aires, Argentina

3 Centro Atómi o Constituyentes, Comisión Na ional de Energía Atómi a andCONICET, Argentina4 Centro Atómi o Constituyentes, Comisión Na ional de Energía Atómi a andUTN-FRBA, Argentina

5 Centro Atómi o Barilo he, Comisión Na ional de Energía Atómi a, San Carlosde Barilo he, Argentina6 Departamento de Físi a, Centro Atómi o Barilo he, Comisión Na ional deEnergía Atómi a and CONICET, Argentina

7 Centro Atómi o Barilo he, Comision Na ional de Energía Atómi a and InstitutoBalseiro (CNEA-UNC), San Carlos de Barilo he, Argentina8 Departamento de Físi a, FCEyN, Universidad de Buenos Aires y CONICET,Argentina

9 Departamento de Físi a, Universidad Na ional de La Plata and Funda iónUniversidad Te nológi a Na ional, Argentina10 IFLP, Universidad Na ional de La Plata and CONICET, La Plata, Argentina

11 Instituto de Astronomía y Físi a del Espa io (CONICET), Buenos Aires,Argentina12 Pierre Auger Southern Observatory, Malargüe, Argentina4

13 Pierre Auger Southern Observatory and Comisión Na ional de EnergíaAtómi a, Malargüe, Argentina14 Universidad Te nológi a Na ional, FR-Mendoza, Argentina

15 Universidad Te nológi a Na ional, FR-Mendoza and Funda ión UniversidadTe nológi a Na ional, Argentina16 University of Adelaide, Adelaide, S.A., Australia17 Universidad Catoli a de Bolivia, La Paz, Bolivia

18 Universidad Mayor de San Andrés, Bolivia19 Centro Brasileiro de Pesquisas Fisi as, Rio de Janeiro, RJ, Brazil

20 Pontifí ia Universidade Católi a, Rio de Janeiro, RJ, Brazil21 Universidade de Sao Paulo, Inst. de Fisi a, Sao Paulo, SP, Brazil23 Universidade Estadual de Campinas, IFGW, Campinas, SP, Brazil

24 Univ. Estadual de Feira de Santana, Brazil25 Universidade Estadual do Sudoeste da Bahia, Vitoria da Conquista, BA, Brazil

26 Universidade Federal da Bahia, Salvador, BA, Brazil27 Universidade Federal do ABC, Santo André, SP, Brazil

28 Univ. Federal do Rio de Janeiro, Instituto de Físi a, Rio de Janeiro, RJ, Brazil29 Univ. Federal Fluminense, Inst. de Fisi a, Niterói, RJ, Brazil

30 Charles University, Institute of Parti le & Nu lear Physi s, Prague, Cze hRepubli 31 Institute of Physi s of the A ademy of S ien es of the Cze h Republi , Prague,Cze h Republi

32 Institut de Physique Nu léaire, Université Paris-Sud, IN2P3/CNRS, Orsay,Fran e33 Laboratoire AstroParti ule et Cosmologie, Université Paris 7, IN2P3/CNRS,Paris, Fran e34 Laboratoire de l'A élérateur Linéaire, Université Paris-Sud, IN2P3/CNRS,Orsay, Fran e

35 Laboratoire de Physique Nu léaire et de Hautes Energies, Universités Paris 6 &7 and IN2P3/CNRS, Paris Cedex 05, Fran e36 Laboratoire de Physique Subatomique et de Cosmologie, IN2P3/CNRS,Université Grenoble 1 et INPG, Grenoble, Fran e

37 Bergis he Universität Wuppertal, Wuppertal, Germany38 Fors hungszentrum Karlsruhe, Institut für Kernphysik, Karlsruhe, Germany

39 Fors hungszentrum Karlsruhe, Institut für Prozessdatenverarbeitung undElektronik, Germany40 Max-Plan k-Institut für Radioastronomie, Bonn, Germany

41 RWTH Aa hen University, III. Physikalis hes Institut A, Aa hen, Germany42 Universität Karlsruhe (TH), Institut für Experimentelle Kernphysik (IEKP),Karlsruhe, Germany

43 Universität Siegen, Siegen, Germany46 Università de l'Aquila and Sezione INFN, Aquila, Italy47 Università di Milano and Sezione INFN, Milan, Italy48 Università del Salento and Sezione INFN, Le e, Italy

49 Università di Napoli "Federi o II" and Sezione INFN, Napoli, Italy50 Università di Roma II "Tor Vergata" and Sezione INFN, Roma, Italy

51 Università di Catania and Sezione INFN, Catania, Italy5

52 Università di Torino and Sezione INFN, Torino, Italy53 Università del Salento and Sezione INFN, Le e, Italy

54 Istituto di Astro�si a Spaziale e Fisi a Cosmi a di Palermo (INAF), Palermo,Italy55 Istituto di Fisi a dello Spazio Interplanetario (INAF), Università di Torino andSezione INFN, Torino, Italy

56 INFN, Laboratori Nazionali del Gran Sasso, Assergi (L'Aquila), Italy57 Osservatorio Astro�si o di Ar etri, Floren e, Italy

58 Sezione INFN di Napoli, Napoli, Italy59 Benemérita Universidad Autónoma de Puebla, Puebla, Mexi o

60 Centro de Investiga ión y de Estudios Avanzados del IPN (CINVESTAV),Méxi o, D.F., Mexi o61 Universidad Mi hoa ana de San Ni olas de Hidalgo, Morelia, Mi hoa an, Mexi o

62 Universidad Na ional Autonoma de Mexi o, Mexi o, D.F., Mexi o63 IMAPP, Radboud University, Nijmegen, Netherlands

64 Kernfysis h Versneller Instituut, University of Groningen, Groningen,Netherlands65 NIKHEF, Amsterdam, Netherlands66 ASTRON, Dwingeloo, Netherlands

67 Institute of Nu lear Physi s PAN, Krakow, Poland68 University of �ód¹, �ódz, Poland

69 LIP and Instituto Superior Té ni o, Lisboa, Portugal70 J. Stefan Institute, Ljubljana, Slovenia

71 Laboratory for Astroparti le Physi s, University of Nova Gori a, Slovenia72 Instituto de Físi a Corpus ular, CSIC-Universitat de Valèn ia, Valen ia, Spain

73 Universidad Complutense de Madrid, Madrid, Spain74 Universidad de Al alá, Al alá de Henares (Madrid), Spain75 Universidad de Granada & C.A.F.P.E., Granada, Spain

76 Universidad de Santiago de Compostela, Spain77 Rudolf Peierls Centre for Theoreti al Physi s, University of Oxford, Oxford,United Kingdom

78 Institute of Integrated Information Systems, University of Leeds, UnitedKingdom79 S hool of Physi s and Astronomy, University of Leeds, United Kingdom

80 Argonne National Laboratory, Argonne, IL, USA81 Case Western Reserve University, Cleveland, OH, USA

82 Colorado S hool of Mines, Golden, CO, USA83 Colorado State University, Fort Collins, CO, USA

84 Colorado State University, Pueblo, CO, USA86 Fermilab, Batavia, IL, USA

87 Louisiana State University, Baton Rouge, LA, USA88 Mi higan Te hnologi al University, Houghton, MI, USA

89 New York University, New York, NY, USA90 Northeastern University, Boston, MA, USA91 Ohio State University, Columbus, OH, USA

92 Pennsylvania State University, University Park, PA, USA93 Southern University, Baton Rouge, LA, USA6

94 University of California, Los Angeles, CA, USA95 University of Chi ago, Enri o Fermi Institute, Chi ago, IL, USA

96 University of Colorado, Boulder, CO, USA97 University of Hawaii, Honolulu, HI, USA

98 University of Minnesota, Minneapolis, MN, USA99 University of Nebraska, Lin oln, NE, USA

100 University of New Mexi o, Albuquerque, NM, USA100 University of Pennsylvania, Philadelphia, PA, USA

101 University of Utah, Salt Lake City, UT, USA102 University of Wis onsin, Madison, WI, USA

103 University of Wis onsin, Milwaukee, WI, USA104 Institute for Nu lear S ien e and Te hnology, Hanoi, Vietnam

Abstra tA method is developed to sear h for air showers initiated by photons using datare orded by the surfa e dete tor of the Auger Observatory. The approa h is basedon observables sensitive to the longitudinal shower development, the signal risetimeand the urvature of the shower front. Applying this method to the data, upper limitson the �ux of photons of 3.8×10−3, 2.5×10−3, and 2.2×10−3 km−2 sr−1 yr−1 above1019 eV, 2 × 1019 eV, and 4 × 1019 eV are derived, with orresponding limits on thefra tion of photons being 2.0%, 5.1%, and 31% (all limits at 95% .l.). These photonlimits disfavor ertain exoti models of sour es of osmi rays. The results also showthat the approa h adopted by the Auger Observatory to alibrate the shower energyis not strongly biased by a ontamination from photons.1 Introdu tionThe sear h for photons in the ultra-high energy (UHE) osmi -ray �ux hasbeen stimulated by the observation of osmi rays with energies ex eedingEGZK ∼ 6×1019 eV [1,2,3,4,5,6℄. If these parti les are due to osmologi allydistant sour es, the �ux spe trum is expe ted to steepen above this energy. In-triguingly, a �ux spe trum with no apparent steepening above EGZK has beenreported by the AGASA Collaboration [7℄. To a ount for this observation andto ir umvent the theoreti al hallenge of explaining parti le a eleration tosu h energies, models involving new physi s have been proposed in whi h the osmi rays are reated at the observed energies at relatively lose distan esfrom the Earth. These �top-down� models [8,9℄ may involve super heavy darkmatter (SHDM) [10,11,12℄, topologi al defe ts [13℄, or neutrino intera tionswith the reli neutrino ba kground (Z-bursts) [14℄. A ommon feature of thesemodels is the predi tion of a substantial photon �ux at highest energies.7

The Auger Collaboration has re ently reported a measurement of the osmi -ray spe trum from the Auger South site showing a �ux suppression aboveEGZK [15℄. The Auger method is based on a large surfa e array to olle t therequired statisti s and a �uores en e dete tor to alibrate the energy s ale. Us-ing this �hybrid� approa h, the energy re onstru tion is largely independent ofhadroni intera tion parameters and, in ase of nu lear primaries, of the pri-mary mass omposition. However, as explained later, the energy assignmentfrom surfa e arrays an be substantially altered in the ase of primary pho-tons. This would a�e t the re onstru ted primary spe trum if a non-negligiblenumber of the highest-energy events, where data from the �uores en e tele-s opes are sparse due to their ∼10% duty yl e, was a tually due to photons(see also [16℄). It is worthwhile to note that the a eptan e of �uores en edete tors (as also applied in the HiRes experiment [5℄) an be altered in the ase of photon primaries [17,18,19℄.UHE photons an also a t as tra ers of the GZK (Greisen-Zatsepin-Kuzmin)pro ess [20℄ of resonant photopion produ tion of nu leons o� the osmi mi- rowave ba kground. The orresponding photon �uxes are sensitive to sour efeatures (type of primary, inje tion spe trum, distan e to sour es ...) andto propagation parameters (extragala ti radio ba kgrounds and magneti �elds) [9,21,22,23,24℄.Thus, the sear h for primary photons remains an important subje t for variousreasons [25℄, parti ularly• to set signi� ant limits to the possible ontribution of top-down me hanismsto the primary osmi -ray �ux;• to sear h for GZK photons, to prove the GZK e�e t and onstrain sour eand propagation models;• to establish the maximum photon fra tion in the primary �ux, for whi hthe energy estimate in the surfa e array dete tor would be altered;• to obtain input to fundamental physi s, for instan e, to probe quantumgravity e�e ts in the ele tromagneti se tor [26℄.Showers initiated by UHE photons develop di�erently from showers indu ed bynu lear primaries. Parti ularly, observables related to the development stageor �age� of a shower (su h as the depth of shower maximum Xmax) and tothe ontent of shower muons provide good sensitivity to identify primary pho-tons. Photon showers are expe ted to develop deeper in the atmosphere (largerXmax). This is onne ted to the smaller multipli ity in ele tromagneti inter-a tions ompared to hadroni ones, su h that a larger number of intera tionsis required to degrade the energy to the riti al energy where the as adingpro ess stops. Additionally, the LPM e�e t [27℄ results in a suppression ofthe pair produ tion and bremsstrahlung ross-se tions. Photon showers also ontain fewer se ondary muons, sin e photoprodu tion and dire t muon pair8

produ tion are expe ted to play only a sub-dominant role.Sear hes for photons were previously ondu ted based on surfa e arrays [28,29,30,31,32℄,and limits to the fra tion of photons were reported (see [25℄ for a review). Thederivation of limits to the photon fra tion using surfa e array data alone isan experimental and on eptual hallenge (see also Se tion 2.3). Firstly, for on lusions on the fra tion, the energy s ales for photon and nu lear primariesare needed. These energy s ales may di�er from ea h other for surfa e arrays,and the di�eren e between the s ales may depend in a non-trivial way onprimary parameters su h as the shower zenith angle. Se ondly, the energy re- onstru tion of nu lear primaries su�ers from substantial un ertainties due toour limited knowledge of high-energy hadron dynami s.Both issues an be resolved using the �uores en e te hnique, whi h is near- alorimetri and largely independent of simulating hadron intera tions. A or-responding approa h has been developed and applied re ently to obtain a�rst bound on the fra tion of photons from data taken at the Auger Observa-tory [19℄.In this work, using the larger number of events re orded by the surfa e array,we derive for the �rst time a dire t limit to the �ux of photons by sear hing forphoton andidates and relating their number to the well-known exposure of thesurfa e array. This avoids the need of simulating events inititated by nu learprimaries; only the photon energy s ale is needed whi h an be simulatedwith mu h higher on�den e. Two observables of the surfa e dete tors are hosen whi h have signi� antly di�erent behavior for nu lear primaries when ompared to photons: the risetime of the re orded shower signal and the radiusof urvature of the shower front.We also derive a limit to the fra tion of photons. While the hallenge of usingtwo energy s ales remains for this part of the analysis, hadron simulations anstill be avoided by using the hybrid alibration [15℄ to re onstru t the energiesof the observed events.The plan of the paper is as follows. In Se tion 2, the observables used inthe analysis and their relationship with the omposition of osmi rays areexplained. In Se tion 3, the simulation of UHE photons is onsidered. Themethod developed to distinguish events whi h are photon andidates usingobservables of the surfa e dete tor is detailed in Se tion 4. In Se tion 5, theresults are presented. The on lusions are given in Se tion 6.9

2 ObservablesThe analysis in this paper is based on data taken during 21,400 hours ofoperation of the surfa e dete tor re orded in the period 1 January 2004 to 31De ember 2006. The surfa e dete tor, when ompleted, will have 1600 waterCherenkov dete tors spa ed 1.5 km apart and overing ∼3000 km2 [33,34℄.Ea h water Cherenkov dete tor, or station, is a ylinder 1.2 m in height and3.6 m in diameter. Ea h dete tor is lined with a re�e tive ontainer that holds12 tonnes of puri�ed water and is �tted with three nine-in h photomultipliertubes (PMTs) looking down into the water.When a relativisti parti le passes through a station, Cherenkov radiationis emitted. The radiated photons then propagate through the water, beingre�e ted at the station walls, and are either eventually absorbed or dete tedby a PMT. The signals from the PMTs are digitised by a �ash analog todigital onverter (FADC) whi h samples the signal every 25 ns. These digitisedsignals are then transmitted to a entral data a quisition system where eventtriggers are built. Ea h event, then, has a detailed time pro�le si(ri, t) ofthe energy deposited in ea h station i at distan e ri in the shower plane.The fun tion s(r, t) depends in a omplex way both on the parameters ofthe primary parti le (energy, type, dire tion) and on the dete tor responseto di�erent se ondary parti les (parti ularly the ele tromagneti and muoni shower omponents).In this work, we extra t two relatively simple but robust observables from thesedata, noting that the wealth of information ontained in the time pro�les anfurther be exploited in future work. The observables, the radius of urvatureof the shower front and the risetime at 1000 m ore distan e, were foundto provide good dis rimination between photon and nu lear primaries (seee.g. also Ref. [35℄). In addition to the quantitative studies of these observablesby means of the simulation-re onstru tion hain, we will also sket h (in asimpli�ed way) why these observables are indeed expe ted to di�er betweennu lear and photon primaries.2.1 Radius of CurvatureDue to geometri al reasons, the arrival of the �rst parti les at lateral distan er from the axis is expe ted to be delayed with respe t to an (imaginary) planarshower front (see also Fig. 1, left plot). For a parti le that is due to an earlierintera tion at height H along the shower axis and observed at r, the delay10

from the longer path length an be approximated ast =

1

c(√

H2 + r2 − H) ∝ r2

H(r ≪ H). (1)The delay in reases (for r ≪ H about quadrati ally) with r. Importantly, thedelay de reases with in reasing height H . Air showers with the �rst groundparti les oming from relatively large heights will have smaller delays t at �xeddistan e r ompared to showers where the registered parti les originated fromsmaller heights. Compared to primary photons, showers from nu lear primariesdevelop higher in the atmosphere (smaller Xmax). Additionally, shower muons(mu h more abundant in showers from nu lear primaries) an rea h the groundfrom still higher altitudes further redu ing the time delay. Thus, for nu learprimaries smaller delays are expe ted ompared to photon primaries.We make use of this relation by �tting a shower front (abstra t surfa e with onvex urvature de�ned by the fastest shower parti les) to the measuredtrigger times ti(ri) of the �rst parti les registered at distan es ri. In the presentstudy, the shape of the shower front is approximated using a spheri al model(in a ord with Eq. (1)), and the radius of urvature R of the shower front isobtained by minimizing χ2 in the fun tion

χ2 =∑

i

[c(ti − t0) − |R~a − ~xi|]2c2σ2

t

(2)where ti is the trigger time for station i as de�ned in [36℄, t0 is the time of theshower in the enter of urvature, ~a is the unit ve tor along the shower axis,~xi is the lo ation of the station on the ground relative to the shower ore, andσt is the un ertainty in the shower arrival time [37℄. In the determination ofti, a software �lter is applied to redu e ontributions from spurious signals notrelated to the a tual shower.2.2 RisetimeAlso the spread in time of the signal si(ri, t) registered at distan e ri, whi h orresponds to the thi kness of the lo al shower disk, an be extra ted. UsingEq. (1), the di�eren e of arrival times of parti les originating from a heightinterval [H1, H1 − ∆H ℄ follows as

∆t(H1, ∆H)∝ r2

(

1

H1 − ∆H− 1

H1

)

=r2∆H

H1(H1 − ∆H)

< ∆t(H2, ∆H) for H2 < H1. (3)11

Fig. 1. Illustration of geometri al e�e ts on radius of urvature and risetime of theshower front. (Left) With respe t to an imaginary planar shower front, parti lesarrive more delayed at distan e r when originating from a smaller height H2 < H1.Correspondingly, the radius of urvature of the a tual shower front is smaller in ase of the deep developing photon primaries. (Right) The spread of arrival times ofparti les produ ed over a pathlength ∆H and arriving at distan e r in reases for asmaller produ tion height H2 < H1. Correspondingly, the risetime of the shower isin reased in ase of the deep developing photon primaries.The spread of arrival times of these parti les at �xed ore distan e in reasesfor smaller produ tion heights (see also Fig. 1, right plot). A ordingly, alarger spread is expe ted in ase of the deep developing photon primaries(larger Xmax). We note that in general, the situation is more omplex. Thetime spread may depend on details of the previous shower development, par-ti ularly also on the ompetition between the signals from the ele tromagneti and muoni shower omponents whi h will be ommented on below. Still, geo-metri al e�e ts are essential in the relation between time spread and primary omposition.In this study, we use the risetime t1/2(1000) of the shower signal re onstru tedfor 1000 m distan e and lo ated along the line given by the proje tion of theshower axis onto the ground. First, the risetime tmeas1/2 (ri) of a single stationis de�ned as the time it takes to in rease from 10% to 50% of the total sig-nal deposited in that station. A ording to Eq. (3), for non-verti al showersa (moderate) dependen e of tmeas

1/2 (ri) on the internal azimuth angle of thestations within the shower plane is expe ted. This is be ause the height Hmeasured along the shower axis is larger for those stations on the exterior side12

of the shower ompared to those on the interior side of the shower. To a ountfor this, the observed tmeas1/2 (ri) are orre ted depending on the internal azimuthangle ζ of that station:

tcor1/2(ri) = tmeas1/2 (ri) − g · cos ζ (4)

g =−66.61 + 95.13 · sec θ − 30.73 · sec2 θ + [0.001993 · sec θ

−0.001259 · sec2 θ + 0.0002546 · sec3 θ − 0.0009721] · r2iwhere the parameter g depends on distan e r and primary zenith angle θand is parameterised from the data, and ζ is the lo kwise angle between theproje tion of the shower axis on the ground and the line onne ting the showerimpa t point and the station.It is also expe ted from Eq. (3) that the values tcor1/2(ri) depend on the distan e

ri of the stations. We obtain the �nal risetime t1/2(1000) of the shower byperforming a �t to tcor1/2(ri) using the fun tiont1/2(r) = (40 + ar + br2) ns . (5)The parameters a and b are determined for ea h event by �tting the stationdata (typi al values are 50 ns km−1 and 100 ns km−2 respe tively). The fun -tion is an hored at 40 ns at r=0 as that is the mean single parti le responsein the water Cherenkov dete tors.While geometri al e�e ts onne ted to the di�erent shower developments fromnu lear and photon primaries are a main reason for the risetime di�eren e(larger t1/2(1000) in photon showers), again this sensitivity to omposition isfurther strengthened by shower mouns whi h are more abundant in the aseof nu lear primaries and an dominate the registered signal at larger zenithangles. As muons tend to arrive within a shorter time window ompared tothe ele tromagneti omponent whi h su�ers from multiple s attering, thisfurther redu es the risetime t1/2(1000) for nu lear primaries.2.3 EnergyAs an energy estimator, the time-integrated energy deposit S(1000) at 1000 m ore distan e is used [38℄. However, for the same initial energy and dire tionthe average S(1000) from primary photons an be a fa tor ≥2 below thatfrom nu lear primaries [39,40℄. Reasons are the (typi ally fa tor ∼4) smallernumber of muons and, due to the later development, the steeper ground lateraldistribution in primary photon showers. For a limit to the fra tion of primaryphotons, the energy s ales (transformation from S(1000) to primary energy)13

for both photon and nu lear primaries are required, while the determinationof a limit to the �ux an rely on the photon energy s ale alone.The energy s ale for nu lear primaries is based on the �uores en e te hniqueby using events that are dete ted with both the surfa e dete tor and the �uo-res en e teles opes [41℄. The energy s ale for photon primaries (whi h indu ealmost purely ele tromagneti as ades) is taken from simulations. Thus, bothapproa hes are largely independent from assumptions about hadron intera -tions at high energy.Using a dire t relationship between S(1000) and primary energy for the photonenergy s ale results in a (relatively poor) resolution of about 40%. To improvethis, a unique energy onversion for photons is applied that is des ribed indetail in Ref. [40℄. It is based on the universality of shower development [42℄,i.e. the ele tromagneti part of the shower is expe ted to develop in a well-predi table manner for depths ex eeding Xmax. In brief, for given values ofS(1000) and Xmax, the primary energy is estimated by

S(1000)

Eγ= 1.4(1 +

∆X − 100

1000)[1 + (

∆X − 100

340)2]−1 (6)with ∆X = Xground − Xmax ,where S(1000) is measured in units of verti al equivalent muons (VEM) [36℄,the photon energy Eγ is in EeV, and ∆X is in g m−2. Sin e Xmax is not dire tlymeasured by the surfa e dete tor alone, an iterative approa h using Eq. (6) istaken to estimate the energy. After an initial guess of the photon energy using

S(1000) alone, the typi al Xmax of the photon showers at this energy is takenfrom simulations. With this estimate of Xmax, a new estimate of the photonenergy is obtained using Eq. (6), and the pro edure is repeated. The energyestimate is found stable after few iterations and an energy resolution of ∼25%is a hieved [40℄. We use this improved estimation of the photon energy, butnote that the main on lusions remain valid also when using a dire t energyestimation.3 Monte Carlo SimulationsThe QED pro esses of LPM e�e t [27℄ and geomagneti as ading ([43,35℄ andreferen es therein) need to be onsidered for photon showers at highest energy.As mentioned before, the LPM e�e t leads to a suppression of the pair produ -tion and bremsstrahlung ross-se tions and, thus, additionally in reases theseparation of photon and nu lear primaries in terms ofXmax (for a review of the14

LPM e�e t and experimental observations of the LPM suppression, see [44℄). 1In ase of geomagneti as ading of UHE photons, the initial onversion of theUHE photon into an ele tron-positron pair an indu e a �preshower� (mostlysyn hrotron photons plus ele tron-positron pair(s)) outside the atmosphere.The subsequent air showers from su h � onverted� photons develop higher inthe atmosphere (smaller Xmax) than air showers dire tly initiated by UHEphotons do. As geomagneti as ading be omes important at energies above∼50 EeV at the southern site of the Auger Observatory, this pro ess is ofminor relevan e for the bulk of data used in this analysis.The shower simulations were generated with the Aires simulation pa kage(v2.8), whi h in ludes the LPM e�e t and geomagneti as ading [45℄. QGS-JET 01 [46℄ was used as the hadroni intera tion model. The simulation ofthe water Cherenkov dete tors uses the GEANT4 [47℄ simulation pa kagealong with spe i� ode that handles PMT response and data a quisitionele troni s. The result is that the output of a simulated event is in a formatthat is identi al to the data format re orded with the Auger Observatory. Theshower re onstru tion pro edure used is the same for real events as it is forsimulated events to avoid systemati di�eren es at the re onstru tion stage.4 MethodIn brief, the limit to the photon �ux is obtained as follows. Sele tion uts areapplied to the data (and simulations) to ensure events of good re onstru tionquality and a high a eptan e of the dete tor to photons. Based on S(1000),showers above a minimum primary energy are sele ted. This data set is thensear hed for photon andidates using t1/2(1000) and R (see Se tion 2 for de�ni-tions). Simulations assuming photons are used to determine the orrespondingsele tion e� ien ies. From the number of photon andidates, the e� ien ieswith respe t to photons, and the experimental exposure (obtained from thegeometri al a eptan e known from dete tor monitoring), the upper limit tothe photon �ux is derived.The riteria to sele t events of good quality are:• the station with the largest signal is surrounded by 6 a tive stations;• ≥5 stations used in the �tting of the lateral distribution fun tion [48℄ outof whi h ≥4 stations have a non-saturated signal of ≥10 VEM (verti alequivalent muons) [36℄;1 Even when arti� ially swit hing o� the LPM e�e t, photon showers still havea signi� antly larger Xmax than nu lear primaries (di�eren es >150 g m2 above1019 eV) and a smaller number of muons.15

Fig. 2. Photon dete tion and re onstru tion e� ien y (right hand s ale) as a fun tionof the energy (in EeV) and zenith angle of the primary photon. The analysis isrestri ted to a minimum energy of 10 EeV and zenith angles greater than 30◦ andless than 60◦ (0.866 > cos θ > 0.5).• redu ed χ2 < 10 (χ2 from Eq. (2)).The �rst ut restri ts the analysis to well- ontained events, eliminating inparti ular events near the border of the array. It a�e ts the geometri al a - eptan e only. The multipli ity riterion in the se ond ut is important alsoto ensure a good re onstru tion of t1/2(1000) and R. As the multipli ity isrelated to primary energy, this ut also a�e ts the energy-dependent a ep-tan e of the array to photons. The third ut reje ts the extreme tail of the χ2distribution when re onstru ting R, removing ∼4% of data. As noted before,the assumption of a spheri al model used in Eq. (2) is a simpli� ation and,thus, not expe ted to provide a perfe t des ription of the omplex features ofthe shower front. This ut restri ts the analysis to events where a single valueof R an be reasonably extra ted. It has been he ked with simulations thatno bias to photons is introdu ed this way.As an be seen from Fig. 2, the resulting photon e� ien y drops to smallvalues below ∼10 EeV. At higher energy, near-verti al photons an also failthe station multipli ity ut due to their deep development. Therefore, theanalysis is restri ted to• primary energies ≥10 EeV;• primary zenith angles of 30−60◦. 16

)θsec(1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2

Rad

ius

of

Cu

rvat

ure

[km

]

2

4

6

8

10

12

14

16

18

20

22

24Photon

Data

Sibyll

QGSJet

)θsec(1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2

[n

s] a

t 1

km1/

2t

100

200

300

400

500

600Photon

Data

Sibyll

QGSJet

Fig. 3. Parameterization of the mean behavior of R and t1/2 for 20 EeV primaryphotons as a fun tion of the zenith angle using QGSJET 01 [46℄ or SIBYLL 2.1 [49℄.The rms values are indi ated for the ase of QGSJET 01. An in rease (a de rease)of R (of t1/2) with zenith angle is qualitatively expe ted from Eqs. (1) and (3) dueto the generally longer path lengths to ground in ase of larger in lination. Realevents of 19�21 EeV (photon energy s ale) are added. The signi� ant deviation ofthe observed values from those expe ted for primary photons is visible.Events with zenith angles below 60◦ are sele ted here sin e in lined show-ers require dedi ated algorithms for an optimum re onstru tion [50℄ (this utmight be relaxed in the future).The sear h for photon andidates makes use of t1/2(1000) and R and onsistsof the following steps. Firstly, the deviation ∆x of the observable x (withx = t1/2 or R referring to risetime or radius of urvature, respe tively) fromthe mean value x̄γ predi ted for photons is derived in units of the spread σx,γof the observable x,

∆x =x − x̄γ(S(1000), θ)

σx,γ(S(1000), θ). (7)where x̄γ(S(1000), θ) and σx,γ(S(1000), θ) are parameterized from simulationsusing primary photons. In Fig. 3, examples are shown for these parameteriza-tions of the observables along with distributions of real events.Se ondly, we ombine the information ontained in the quantities ∆t1/2

and∆R by performing a prin ipal omponent analysis [51℄, leaving a more sophisti- ated statisti al analysis for the future. To determine the prin ipal omponent(de�ned as the axis with the largest varian e), 5% of the real events are usedtogether with results from photon simulations, see Fig. 4. For the simulations,a power law spe trum of index -2.0 has been assumed (see below for otherindi es). The remaining 95% of the data are then proje ted onto the prin ipalaxis along with the simulated photons.This pro edure allows the a priori de�nition of a simple ut in the proje ted17

(1000)1/2Deviation of t-10 -8 -6 -4 -2 0 2 4 6 8 10

Dev

iati

on o

f R

-10

-8

-6

-4

-2

0

2

4

6

8

10

MC Photons

Data

Fig. 4. The deviation from a photon predi tion for 5% of the data ( losed squares)and simulated photon events ( rosses). The solid grey line is the prin ipal omponentaxis identi�ed using the limited set of real showers while the dashed line is theaxis perpendi ular to the prin ipal omponent. The minimum energy is 10 EeV(Eγ > 10 EeV).distribution to �nally obtain photon andidate events. The ut was hosenat the mean of the distribution for photons, su h that the e� ien y of this ut is f = 0.5 by onstru tion. Any real event falling above this ut willbe onsidered a photon andidate. We note that su h photon andidates, ifo uring, an not yet be onsidered as being photons, as they a tually might bedue to ba kground events from nu lear primaries. A presen e of ba kgroundevents would result in weaker upper limits (larger numeri al values) in theanalysis approa h adopted here.Finally, an upper limit on the number of photonsN CLγ at on�den e level CL is al ulated from the number of photon andidate events Nγ above a minimumenergy, Emin. The upper limit on the �ux or fra tion of photons above a givenenergy is based on N CL

γ along with the integrated e� ien y ε of a eptingphotons, the photon sele tion ut e� ien y (f = 0.5), and either the exposureA of the dete tor for the �ux limit:

ΦCL(E > Emin) =N CL

γ (Eγ > Emin) × 1f× 1

ε

0.95A, (8)18

or the number of non-photon andidate events Nnon−γ in the data set for thefra tion limit:FCL(E > Emin) =

N CLγ (Eγ > Emin) × 1

f× 1

ε

Nγ(Eγ > Emin) + Nnon−γ(Enon−γ > Emin). (9)In Eq. (8), the fa tor 0.95 is from the fa t that only 95% of the data are usedto determine the number of photon andidate events. The energy is labeledas either the energy a ording to the photon energy re onstru tion, Eγ , or(required in Eq. (9)) the energy a ording to the non-photon energy re on-stru tion, Enon−γ .Experimentally, the limit ΦCL to the �ux is more robust than the limit FCLto the fra tion due to the di�erent denominators of Eqs. (8) and (9). For FCL,two energy s ales are required; also, with in reasing energy, the statisti alun ertainty of the quantity (Nγ +Nnon−γ) be omes large. For ΦCL, in ontrast,the aperture is known to good (∼3%) a ura y.Though the present work does not aim at extra ting a omposition of nu learprimaries, it is interesting to he k whether the prin ipal omponent axis foundfrom real data and the separation along it re�e ts what would be expe tedif the bulk of the real data is due to nu lear primaries. In Fig. 5, the samesimulated photon events are used as in Fig. 4 but the 5% of real data arerepla ed with a set of ∼750 Monte Carlo proton and iron showers with anenergy of 10 EeV. The separation observed in real data is both in the samedire tion and of a similar magnitude as that expe ted from simulated nu learprimaries.5 ResultsThe data from 2004�2006 are analysed as des ribed in the pre eding se tion.The integrated aperture of the Observatory is 3130 km2 sr yr for the angular overage regarded in this analysis. Above 10, 20, and 40 EeV, for the energys ale of photons (in bra kets for nu lear primaries), the data set onsists of2761 (570), 1329 (145), and 372 (21) events. The measured values of t1/2(1000)and R are used to determine the proje tion on the prin ipal axis. A s atterplot of this quantity vs. the primary energy is shown in Fig. 6, while in Fig. 7the orresponding distributions are plotted for the three threshold energies.No event passes the photon andidate ut. The upper limits on the photon�ux above 10, 20, and 40 EeV are then 3.8 × 10−3, 2.5 × 10−3, and 2.2 ×

10−3 km−2 sr−1 yr−1 (at 95% CL). The limits on the photon fra tion are 2.0%,5.1%, and 31% (at 95% CL) above 10, 20, and 40 EeV. In Tab. 1, all relevantquantities (number of events, e� ien ies, resulting limits) are summarized.19

(1000)1/2Deviation of t-10 -8 -6 -4 -2 0 2 4 6 8 10

Dev

iati

on o

f R

-10

-8

-6

-4

-2

0

2

4

6

8

10

MC Photons

MC Hadrons

Fig. 5. The bla k rosses are simulated photon showers while the squares are amixture of Monte Carlo proton and iron with an energy of 10 EeV. For omparison,the lines shown in Fig. 4 (prin ipal omponent axes) are added. The distribution ofsimulated nu lear primaries is similar to the distribution of real data seen in Fig. 4.Emin N(Eγ > Emin) Nγ N 0.95

γ Nnon−γ ε Φ0.95 F0.9510 2761 0 3.0 570 0.53 3.8 × 10−3 2.0%20 1329 0 3.0 145 0.81 2.5 × 10−3 5.1%40 372 0 3.0 21 0.92 2.2 × 10−3 31%Table 1Results of the analysis sear hing for photon andidate events. The fra tion and�ux limits are integral limits above Emin (EeV), ε is the e� ien y of dete tion andre onstru tion, Φ0.95 is in units of km−2 sr−1 yr−1, and all results are 95% on�den elevel.From Fig. 6 it an also be seen that the separation of data and photon pri-maries in reases with energy. In parti ular at highest energies above EGZKfor the photon energy s ale, there is no indi ation for photon-initiated events.Thus, the absen e of photons, within the improved limits pla ed in this work,shows that the method applied by the Auger Observatory to alibrate theshower energy is not strongly biased by a photon � ontamination�.We studied potential sour es of systemati e�e ts in the analysis. To determinethe e� ien y to photons and to establish the photon andidate ut, a primary20

Log(Energy/EeV)1 1.2 1.4 1.6 1.8 2 2.2

Pri

ncip

al C

ompo

nent

Dev

iati

on

-10

-8

-6

-4

-2

0

2

4

6

8

10MC Photons

Data

Fig. 6. The deviation of data (bla k rosses) and photons (open red ir les) fromthe prin ipal omponent as a fun tion of the primary energy (photon energy s ale).Data lying above the dashed line, whi h indi ates the mean of the distribution forphotons, are taken as photon andidates. No event meets this requirement.photon spe trum of power law index -2.0 has been used in the simulations,motivated by predi tions from top-down models in (e.g. in Ref. [10℄). The e�e tof hanging the power law index to -1.7, -2.5, and -3.0 has been investigated.The number of events whi h are photon andidates is un hanged (along withthe number of non-photon andidate events), but the orre tion for the photone� ien y hanges. Spe i� ally, for a steeper input spe trum (in reased fra tionof lower-energy photons), the e� ien y de reases. The summary of the results an be seen in Table 2. For 10 EeV threshold energy, limits hange from(3.8→5.5)×10−3 km−2 sr−1 yr−1 for the �ux and from (2.0→2.9)% for thefra tion. The di�eren es get smaller with in reased threshold energy.The photonu lear ross-se tion used in the simulation is based on the Parti leData Group (PDG) extrapolation [52℄. For an in reased ross-se tion, moreenergy would be transfered to the hadron (and muon) omponent whi h oulddiminish the separation power between data and primary photons [53℄. Fromunitarity onstraints, the ross-se tion is not expe ted to ex eed the PDG ex-trapolation by more than∼75% at 10 EeV [54℄; at 1015 eV, where the di�eren ein ross-se tion would have a greater impa t on the shower development, themaximum di�eren e is ∼20%. From simulations with modi�ed ross-se tionsit was veri�ed that this leads to a negligible variation of the average values of21

-10 -8 -6 -4 -2 0 2 4 6 8 100

0.2

0.4

0.6

0.8

1

MC Photons

Data

-10 -8 -6 -4 -2 0 2 4 6 8 100

0.2

0.4

0.6

0.8

1

MC Photons

Data

-10 -8 -6 -4 -2 0 2 4 6 8 100

0.2

0.4

0.6

0.8

1

MC Photons

Data

Fig. 7. Distribution of real events ( losed squares) along with simulated photonevents (open ir les) for the proje tion on the prin ipal omponent axis. The photon andidate ut is set at the mean of the distribution for photons and is shown as thedotted line. The plots are made requiring a minimum energy (a ording to the photonenergy onverter) of 10 EeV (top-left), 20 EeV (top-right), and 40 EeV (bottom).Distributions are normalised to unity at maximum.the dis riminating variables used in the urrent analysis.The simulations have been performed with AIRES using the QGSJET had-roni intera tion model. As a ross- he k, al ulations with CORSIKA [55℄ /QGSJET and AIRES / SIBYLL were ondu ted, both of whi h show reason-able agreement to the AIRES / QGSJET ase. As the as ade initiated byprimary photons has an almost pure ele tromagneti nature, indeed no sig-ni� ant e�e t is expe ted when hanging to another intera tion model. Thisminor dependen e of the results on the details of hadron intera tions, whi hare largely un ertain at high energy, is an important advantage of sear hes forprimary photons.The new limits are ompared to previous results and to theoreti al predi tions22

Emin 10 20 40 10 20 40 10 20 40α E� ien y (ε) Flux (×10−3) Fra tion [%℄1.7 0.60 0.83 0.93 3.3 2.4 2.2 1.8 5.0 312.0 0.53 0.81 0.92 3.8 2.5 2.2 2.0 5.1 312.5 0.43 0.76 0.91 4.7 2.6 2.2 2.5 5.4 313.0 0.36 0.71 0.90 5.5 2.8 2.2 2.9 5.9 32Table 2Results when hanging the exponent (α) in the power law of the simulated spe trum.The default value is 2.0. The e� ien y of dete tion and re onstru tion is on the left,the resulting limit on the fra tion of photons is on the right, and the limit on theintegrated �ux is listed in the middle in units of km−2 sr−1 yr−1 (95% CL).in Fig. 8 for the photon �ux and in Fig. 9 for the photon fra tion. We pla ed the�rst dire t limit to the �ux of UHE photons (an earlier bound from AGASA,about an order of magnitude weaker than the urrent bounds, was derivedindire tly via a limit to the fra tion and the �ux spe trum [29℄). In terms ofthe photon fra tion, the urrent bound at 10 EeV approa hes the 10−2 levelwhile previous bounds were at the 10−1 level.A dis overy of a substantial photon �ux ould have been interpreted as a signa-ture of top-down models. In turn, the experimental limits now put strong on-straints on these models. For instan e, ertain SHDM or TD models dis ussedin the literature (SHDM and TD from Ref. [21℄ based on the fragmentation al ulations of Ref. [11℄, SHDM' from Ref. [12℄ 2 ) predi t �uxes that ex eedthe limits by a fa tor ∼10. It should be noted that a simple res aling of the�ux predi tions from top-down models, whi h were motivated by and based onthe energy spe trum observed by AGASA, would redu e the predi ted photon�ux by only a fa tor ∼2 whi h would still overshoot our experimental limitby a fa tor ∼5 at 1019 eV. While a minor ontribution from top-down modelsto the observed UHE osmi -ray �ux might still be allowed within the limitsderived in this work, urrent top-down models do not appear to provide anadequate explanation of the origin of the highest-energy osmi rays (see alsoRef. [56℄ for a omparison of photon �ux predi tions to the Auger limits fordi�erent top-down model parameters).In a eleration models, photon �uxes are usually expe ted to be a fa tor 2 ormore below the urrent bounds ( f. the GZK photon predi tions in the Figs. 8and 9 from Ref. [21℄). Su h �uxes an be tested with future data taken at theAuger Observatory (see also Ref. [25℄). After �ve years of operation with the omplete surfa e dete tor, sensitivities at the level of ∼4×10−4 km−2 sr−1 yr−1for the integrated �ux and ∼0.7% for the fra tion of photons above 20 EeV

2 Two others of the eight photon �ux spe tra al ulated in Ref. [12℄ from ryptonde ays may still be ompatible with our limits within a fa tor ∼2.23

[eV]0E1910 2010

]-1

yr-1

sr-2

[km

0P

hoto

n F

lux

for

E>E

-310

-210

-110

SHDMSHDM’TDZ BurstGZK Photons

)0

Limit (E>EA

limits at 95% CL

Fig. 8. The upper limits on the integral �ux of photons derived in this work (bla karrows) along with predi tions from top-down models (SHDM, TD and ZB fromRef. [21℄, SHDM' from Ref. [12℄) and with predi tions of the GZK photon �ux [21℄.A �ux limit derived indire tly by AGASA (�A�) is shown for omparison [29℄.(95% CL) ould be rea hed.6 Con lusionsUsing data from the surfa e dete tor we obtained 95% .l. upper limits on thephoton �ux of 3.8 × 10−3, 2.5 × 10−3, and 2.2 × 10−3 km−2 sr−1 yr−1 above1019 eV, 2× 1019 eV, and 4× 1019 eV. These are the �rst dire t bounds on the�ux of UHE photons. For the photon fra tion, limits of 2.0%, 5.1%, and 31%were pla ed.These limit improve signi� antly upon bounds from previous experiments andput strong onstraints on ertain models of the origin of osmi rays. Currenttop-down models su h as the super-heavy dark matter s enario do not appearto provide an adequate explanation of the UHE osmi rays. In bottom-upmodels of a eleration of nu lear primaries in astrophysi al sour es, the ex-pe ted photon �uxes are typi ally well below the urrent bounds. An astro-physi al origin of UHE osmi rays is also suggested by the re ent dis overy24

[eV]0E1910 2010

[%

]0

Pho

ton

Fra

ctio

n fo

r E

>E

1

10

100

SHDMSHDM’TDZ BurstGZK Photons

)0

Limit (E>E

A

AA2

HP HPAY

Y

Y

FD

limits at 95% CL

Fig. 9. The upper limits on the fra tion of photons in the integral osmi -ray �ux de-rived in this work (bla k arrows) along with previous experimental limits (HP: Hav-erah Park [28℄; A1, A2: AGASA [29,30℄; AY: AGASA-Yakutsk [31℄; Y: Yakutsk [32℄;FD: Auger hybrid limit [19℄). Also shown are predi tions from top-down models(SHDM, TD and ZB from Ref. [21℄, SHDM' from Ref. [12℄) and predi tions of theGZK photon fra tion [21℄.of a orrelation of UHE osmi rays with the dire tions of nearby AGNs [57℄.Con erning the method of energy alibration as applied by the Auger Obser-vatory, the photon bounds derived in this work show that there is no strongbias due to a ontamination from UHE photons.With the data a umulating over the next years, and parti ularly when om-plementing the Auger southern site by an extended northern one, the �uxlevels expe ted for GZK photons may be in rea h.A knowledgements:The su essful installation and ommissioning of the Pierre Auger Observatorywould not have been possible without the strong ommitment and e�ort fromthe te hni al and administrative sta� in Malargüe.We are very grateful to the following agen ies and organizations for �nan- ial support: Comisión Na ional de Energía Atómi a, Funda ión Antor has,Gobierno De La Provin ia de Mendoza, Muni ipalidad de Malargüe, NDM25

Holdings and Valle Las Leñas, in gratitude for their ontinuing ooperationover land a ess, Argentina; the Australian Resear h Coun il; Conselho Na- ional de Desenvolvimento Cientí� o e Te nológi o (CNPq), Finan iadora deEstudos e Projetos (FINEP), Fundação de Amparo à Pesquisa do Estado deRio de Janeiro (FAPERJ), Fundação de Amparo à Pesquisa do Estado de SãoPaulo (FAPESP), Ministério de Ciên ia e Te nologia (MCT), Brazil; Min-istry of Edu ation, Youth and Sports of the Cze h Republi ; Centre de Cal ulIN2P3/CNRS, Centre National de la Re her he S ienti�que (CNRS), Con-seil Régional Ile-de-Fran e, Département Physique Nu léaire et Corpus ulaire(PNC-IN2P3/CNRS), Département S ien es de l'Univers (SDU-INSU/CNRS),Fran e; Bundesministerium für Bildung und Fors hung (BMBF), Deuts heFors hungsgemeins haft (DFG), FinanzministeriumBaden-Württemberg, Helmholtz-Gemeins haft Deuts her Fors hungszentren (HGF), Ministerium fürWissens haftund Fors hung, Nordrhein-Westfalen, Ministerium fürWissens haft, Fors hungund Kunst, Baden-Württemberg, Germany; Istituto Nazionale di Fisi a Nu le-are (INFN), Ministero dell'Istruzione, dell'Università e della Ri er a (MIUR),Italy; Consejo Na ional de Cien ia y Te nología (CONACYT), Mexi o; Min-isterie van Onderwijs, Cultuur en Wetens hap, Nederlandse Organisatie voorWetens happelijk Onderzoek (NWO), Sti hting voor Fundamenteel Onder-zoek der Materie (FOM), Netherlands; Ministry of S ien e and Higher Edu- ation, Grant Nos. 1 P03 D 014 30, N202 090 31/0623, and PAP/218/2006,Poland; Fundação para a Ciên ia e a Te nologia, Portugal; Ministry for HigherEdu ation, S ien e, and Te hnology, Slovenian Resear h Agen y, Slovenia; Co-munidad de Madrid, Consejería de Edu a ión de la Comunidad de Castilla LaMan ha, FEDER funds, Ministerio de Edu a ión y Cien ia, Xunta de Gali- ia, Spain; S ien e and Te hnology Fa ilities Coun il, United Kingdom; De-partment of Energy, Contra t No. DE-AC02-07CH11359, National S ien eFoundation, Grant No. 0450696, The Grainger Foundation USA; ALFA-EC/ HELEN, European Union 6th Framework Program, Grant No. MEIF-CT-2005-025057, and UNESCO.Referen es[1℄ J. Linsley, Phys. Rev. Lett. 10, 146 (1963).[2℄ M.A. Lawren e, R.J.O. Reid, A.A. Watson, J. Phys. G17, 733 (1991).[3℄ D.J. Bird et al., Phys. Rev. Lett. 71, 3401 (1993).[4℄ N. Hayashida et al., Phys. Rev. Lett. 73, 3491 (1994).[5℄ R.U. Abbasi et al., Phys. Lett. B 619, 271 (2005).[6℄ Pierre Auger Collaboration, Pro . 29th Intern. Cosmi Ray Conf., Pune, 7, 283(2005). 26

[7℄ M. Takeda et al., Astropart. Phys. 19, 447 (2003).[8℄ P. Bhatta harjee, G. Sigl, Phys. Rep. 327, 109 (2000).[9℄ S. Sarkar, A ta Phys. Polon. B35, 351 (2004).[10℄ V. Berezinsky, M. Ka helrieÿ, A. Vilenkin, Phys. Rev. Lett. 79, 4302 (1997);M. Birkel, S. Sarkar, Astropart. Phys. 9, 297 (1998); S. Sarkar, R. Toldra, Nu l.Phys. B 621, 495 (2002); C. Barbot, M. Drees, Astropart. Phys. 20, 5 (2003);R. Aloisio, V. Berezinsky, M. Ka helriess, Phys. Rev. D 74, 023516 (2006).[11℄ R. Aloisio, V. Berezinsky, M. Ka helriess, Phys. Rev. D 69, 094023 (2004).[12℄ J. Ellis, V. Mayes, D.V. Nanopoulos, Phys. Rev. D 74, 115003 (2006).[13℄ C.T. Hill, Nu l. Phys. B224, 469 (1983); M.B. Hindmarsh, T.W.B. Kibble,Rep. Prog. Phys. 58, 477 (1995).[14℄ T.J. Weiler, Phys. Rev. Lett. 49, 234 (1982); T.J. Weiler, Astropart. Phys. 11,303 (1999); D. Fargion, B. Mele, A. Salis, Astrophys. J. 517, 725 (1999).[15℄ Pierre Auger Collaboration, Pro . 30th Intern. Cosmi Ray Conf., Merida (2007);[arXiv:astro-ph/0706.2096℄.[16℄ N. Bus a, D. Hooper, E.W. Kolb, Phys. Rev. D 73, 123001 (2006).[17℄ V. de Souza, G. Medina-Tan o, J.A. Ortiz, Phys. Rev. D 72, 103009 (2005).[18℄ A.S. Chou, Phys. Rev. D 74, 103001 (2006).[19℄ Pierre Auger Collaboration, Astropart. Phys. 27, 155 (2007); PierreAuger Collaboration, Pro . 30th Intern. Cosmi Ray Conf., Merida (2007),arXiv:0710.0025 [astro-ph℄.[20℄ K. Greisen, Phys. Rev. Lett. 16, 748 (1966); G.T. Zatsepin, V.A. Kuzmin, JETPLett. 4, 78 (1966).[21℄ G. Gelmini, O.E. Kalashev, D.V. Semikoz, [arXiv:astro-ph/0506128℄.[22℄ S. Lee, A.V. Olinto, G. Sigl, Astrophys. J. 455, L21 (1995).[23℄ G. Sigl, Phys. Rev. D 75, 103001 (2007) [arXiv:astro-ph/0703403℄.[24℄ G. Gelmini, O. Kalashev, D.V. Semikoz, [arXiv:astro-ph/0706.2181℄.[25℄ M. Risse, P. Homola, Mod. Phys. Lett. A 22, 749 (2007).[26℄ M. Galaverni, G. Sigl, to appear in Phys. Rev. Lett., [arXiv:astro-ph/0708.1737℄;R. Aloisio, P. Blasi, P. L. Ghia, A. F. Grillo, Phys. Rev. D 62, 053010 (2000)[arXiv:astro-ph/0001258℄.[27℄ L.D. Landau, I.Ya. Pomeran huk, Dokl. Akad. Nauk SSSR 92, 535 & 735 (1953);A.B. Migdal, Phys. Rev. 103, 1811 (1956).[28℄ M. Ave et al., Phys. Rev. Lett. 85, 2244 (2000); Phys. Rev. D65, 063007 (2002).27

[29℄ K. Shinozaki et al., Astrophys. J. 571, L117 (2002).[30℄ M. Risse et al., Phys. Rev. Lett. 95, 171102 (2005).[31℄ G.I. Rubtsov et al., Phys. Rev. D 73, 063009 (2006).[32℄ A.V. Glushkov et al., JETP Lett. 85, 163 (2007).[33℄ Pierre Auger Collaboration, Nu l. Instrum. Meth. A 523, 50 (2004).[34℄ Pierre Auger Collaboration, Pro . 30th Intern. Cosmi Ray Conf., Merida (2007);arXiv:0709.1823 [astro-ph℄.[35℄ X. Bertou, P. Billoir, S. Dagoret-Campagne, Astropart. Phys. 14, 121 (2000).[36℄ X. Bertou et al., P. Auger Collaboration, Nu l. Instrum. Meth. A 568, 839(2006).[37℄ Pierre Auger Collaboration, Pro . 29th Intern. Cosmi Ray Conf., Pune, 7, 17(2005).[38℄ D. Newton, J. Knapp, A.A. Watson, Astropart. Phys. 26, 414 (2007).[39℄ Pierre Auger Collaboration, Pro . 29th Intern. Cosmi Ray Conf., Pune, 7, 387(2005), [arXiv:astro-ph/0507150℄.[40℄ P. Billoir, C. Rou elle, J.C. Hamilton, [arXiv:astro-ph/0701583℄.[41℄ Pierre Auger Collaboration, Pro . 30th Intern. Cosmi Ray Conf., Merida,arXiv:0706.1105 [astro-ph℄.[42℄ A.S. Chou et al., Pro . 29th Intern. Cosmi Ray Conf., Pune, 7, 319 (2005).[43℄ T. Erber, Rev. Mod. Phys. 38, 626 (1966); B. M Breen, C.J. Lambert,Phys. Rev. D 24, 2536 (1981); T. Stanev, H.P. Vankov, Phys. Rev. D 55, 1365(1997); P. Homola et al., Comput. Phys. Commun. 173, 71 (2005); P. Homolaet al., Astropart. Phys. 27, 174 (2007).[44℄ S. Klein, Rev. Mod. Phys. 71, 1501 (1999); S. Klein, Rad. Phys. Chem. 75, 696(2006).[45℄ S.J. S iutto, �AIRES: A system for air shower simulations (version 2.2.0),�arXiv:astro-ph/9911331; available at http://www.�si a.unlp.edu.ar/auger/aires[46℄ N.N. Kalmykov, S.S. Ostap henko, A.I. Pavlov, Nu l. Phys. B (Pro . Suppl.)52B, 17 (1997).[47℄ S. Agostinelli et al., Nu l. Instrum. Meth. A 506, 250 (2003). J. Allisonet al., IEEE Transa tions on Nu lear S ien e, 53 270 (2006). See alsohttp://geant4. ern. h/.[48℄ Pierre Auger Collaboration, Pro . 29th Intern. Cosmi Ray Conf., Pune, 7, 291(2005).[49℄ R. Engel, T.K. Gaisser, P. Lipari, T. Stanev, Pro . 26th Intern. Cosmi RayConf., Salt Lake City, 1, 415 (1999).28

[50℄ Pierre Auger Collaboration, Pro . 30th Intern. Cosmi Ray Conf., Merida (2007);[arXiv:astro-ph/0706.2096℄.[51℄ E. Oja, International Journal of Neural Systems, 1, 61 (1989).[52℄ S. Eidelmann et al., Parti le Data Group, Phys. Lett. B592, 1 (2004).[53℄ M. Risse et al., Cze h. J. Phys. 56, A327 (2006).[54℄ T.C. Rogers, M.I. Strikman, J. Phys. G: Nu l. Part. Phys. 32, 2041 (2006).[55℄ D. He k et al., Reports FZKA 6019 & 6097, Fors hungszentrum Karlsruhe(1998).[56℄ Pierre Auger Collaboration, Pro . 30th Intern. Cosmi Ray Conf., Merida (2007);arXiv:0706.2960 [astro-ph℄.[57℄ Pierre Auger Collaboration, S ien e 318, 939 (2007).

29