UNIVERSITAT AUTONOMA DE BARCELONAFACULTAD DE CIENCIASP H D T H E S I Sby the Universitat Autonoma de Bar elonaSpe ialty : Astrophysi sSome observational andtheoreti al aspe ts of osmi -raydiusionDefended byElsa de Cea del PozoThesis Advisor: Diego F. Torres

prepared at Institut de Cien ies de l'Espai, IEEC - CSICdefended on July 2011

To those who believed in me, despite myself.

iiiA knowledgmentsEsta tesis nun a habría sido posible sin mi jefe. Diego me ha abierto las puertas almundo de la investiga ión, enseñándome el ompromiso que impli a, tanto a nivelpersonal omo profesional. Por su dedi a ión y esfuerzo, por a ompañarme durantemis primeros pasos en la dura arrera ientí a, quiero darle las gra ias. No sólome ha aportado su experien ia y ono imientos ientí os, sino que también meha brindado la oportunidad de trabajar on estupendos olaboradores, de los queespero seguir aprendiendo en el futuro.I want to a knowledge my great group-mates: those who are, and those whowere. They have helped me through the hardest parts, being patient many manytimes, and always trying to tea h me something. They even onsent to go with meoutside the working world, and shared a few laughs! Agni, Ana Y.: I miss you girls,thanks for putting up with me in my lueless rst years. Nanda, Andrea, Ana andGio: I have nothing but smiles and huge thanks to you all, I'll see you on the road,I'm sure. Also, I sent my gratitude to those whom I have the pleasure to work withand that have taught me a great deal of useful things: Stefano and Olaf, thanks alot. I have met wonderful people inside the MAGIC Collaboration, and I would liketo thank them, too. Working with you has been a really interesting experien e.Moreover, I had a really ni e time with many of you outside the work, both in LaPalma and in any ity we had a meeting or a s hool. I hope to be in lose onta tin the future, no matter where I end up being.I almost wish I hadn'tgone down that rabbit-hole and yet and yet it's rather urious, you know,this sort of life!Lewis Carroll, Ali e's Adventures in WonderlandHe tenido la suerte de ono er a gente estupenda en estos uatro años enBar elona. Sin ellos, mi vida aquí habría sido insoportable, y es a ellos a los quemás e haré de menos uando me haya ido. Siempre había pensado que onseguiríaes ribir algo espe ial para ada uno de ellos, pero dado que soy un po o desastrey que, bueno, no me estoy muriendo, sólo me voy un po o más lejos, pues tendránque onformarse on lo que hay.Empezaré por mis ompis en desdi has: los iberianos. Esa extraña pobla iónde estudiantes apaz de trabajar en un lugar sin puertas ni ventanas, y aún asíreír y animarse mutuamente. Por orden alfabéti o: Ane, Antonio, Carlos, Dani,Diego, Felipe, Ja obo, Jonatan, Jorge(s), Jose, Juan Carlos, Nataly, Pris, Santi. Ser iberiano se lleva en el orazón, re ordadlo. Y si me he dejado a alguien, no me

ivpeguéis muy fuerte en la abeza, que estoy estudiando. Gra ias por ha er mi día adía menos gris, y mis nes de semana más ajetreados.En esta ategoría entran también un buen puñado de gente del IFAE, `los delinstituto de al lado'. A los seniors, gra ias por aguantar estoi amente mis in esantesdudas, ruegos y preguntas, en espe ial a Stefan, Daniel y Abelardo. De los niños nome olvido, por supuesto, también les doy las gra ias: Manel (por darme ollejas yanimarme, todo a la vez: volveré a visitarte y es una amenaza), Ignasi (por ha ermetemer más que antes los petardos y adorar más que antes la montaña), Roberta (mi` riatura' y ompañera en desgra ias, la tesis se a aba pero nosotras no), sin olvidara las re ientes in orpora iones, Ali ia (a.k.a. la piraña más di hara hera, ojalá noshubiéramos en ontrado antes, niña) y Adiv ( ompi de piso durante 6 meses, quetiene su mérito). Y ya que estoy, también aprove ho a mandar un abrazo enorme aesa omunidad de italianos que me han he ho ono er más lugares de esta iudadque ualquier persona lo al: me alegro de que nos hayáis invadido.Y antes de que deje de hablar del trabajo, quiero re ono er aquí la labor deIsabel, Del y Josep. Desde la administra ión, la gestión o la informáti a, se hano upado de mí y me han ayudado on la mejor de las inten iones siempre que hanpodido, y más allá de lo que les to aba. Y a Alina, que no se me olvida, gra ias pornuestras harlas diarias aguantando tormentas y tempestades, y por ser la primeraen leerse esta tesis y orregir mi inglés para que fuera más fa ilmente legible. Loserrores que queden son míos.Llegados a este punto, tengo que parar de ha er listas. En los on ursos yentregas de premios se llama a esta ategoría `men iones espe iales', y omo taldeben onsiderarse. Ellos me han ambiado para bien, espero, y gra ias a ellos hellegado a donde estoy. Es ribir esta tesis no ha sido sólo a umular ono imientosen astrofísi a. Mi salud mental se ha visto laramente resentida, y de no ser porellos, llevaría ya una bonita amisa blan a, on mangas atadas a la espalda. Gra iaspor estar a mi lado todos estos años, y por uidar, omo bonus extra, de mi saludemo ional. A Carlos, le doy las gra ias por nuestras ina abables onversa ionessobre libros, pelí ulas, omi s y demás, pero sobre todo por re ordarme que valgomás de lo que re onoz o, y por tratar de enseñarme a quererme un po o más.Tengo tus palabras en la mente, apli arlas me llevará tiempo, pero que sepas quetendrás una gran parte de ulpa uando lo onsiga. A Del, por ser mi amigay ompartir onmigo el tro ito de su vida que he tenido la suerte de presen iar,peluquera in luida. Suya es la ulpa de que en estos agrade imientos in luya unade las frases on más sinsentido de la historia: gra ias por ayudarme a apre iar elinestimable uso de una bola de demoli ión, por mostrarme la verdad que es ondela palabra perro omo deni ión de jornada laboral, y por ha erme entender que lavida sería mu ho más triste sin pingüinos ni ovejas. Re uerda lo que dijo el telardel destino. . . A Daniela, por onvertirse en mi maligna aprendiz y sufrir uando hesufrido y reir uando yo he reído. Pink girl, lo a-lo a-lo a, no defraudes al LOC (quete ayude Ali ia, es su deber pirañil). Volveré y tendremos esta, salsa y elefantitosde olores.Cambio de iudad y me vuelvo a Madrid, aunque sea sólo on la mente: mis

vamigos no me han olvidado en este tiempo, y omo en ada separa ión me he dado uenta de la suerte que tengo ono iendo gente. Con ada regreso, he podido robarosminutos, risas y mu hos ánimos. Desde aquí, os doy las gra ias: a los del master, porser los afe tados que mejor saben salir de ena y bares (Cris, Álvaro, Igna io, Pablo yJuan-ito), a los del barrio, por a ogerme on tanto ariño desde el primer día (Jewi,Riki, Guille, Fer, et al.), y a los de la arrera, porque ya sois mis amigos de verdady uento on vosotros (Arturo, Carmen, Ele-rizos, Fer, Juan, Marta, Sol, Super-Ele,váis por orden alfabéti o, ea). Y ya para terminar, agrade er de todo orazón, y meda igual que suene ursi, a la piedra angular en mi vida que onstituyen las tres:Ana, Isa y Elena. Cada una guardáis una parte diferente de mí, sois las mejoresamigas que podría haber soñado, y ualquier osa que diga sonará a po o. Gra ias,de verdad.Everyone seems quite relieved, though, onsidering they all knew I'd get o.J. K. Rowling, Harry Potter and the Order of the PhoenixPor último, y por tanto más importante, quiero darles las gra ias a mi familia.Tanto a los que me a ogieron a mi llegada a Bar elona omo a una más, omo alos que me han seguido animando desde Madrid. Tengo mu ha suerte de teneros,y está laro que seguiremos unidos sin importar dónde esté. De todos ellos quierodesta ar a mi padre, a mi madre y a mis hermanos. Mis palabras no van a sersu ientes, pero lo voy a intentar. Mis padres han sido y seguirán siendo mi lugarseguro en el mundo. Desde el prin ipio me han apoyado en mi ele ión, aunque ellosupusiera mandarme a 600 km de asa que era 600 km más lejos de lo que elloshubieran deseado. Pero siempre han estado ahí para mí, y siempre los he sentido er a. Estoy orgullosa de la forma en la que me habéis riado, y espero que vosotroslo estéis de mí. A mi hermano Pablo le doy las gra ias por seguir onando en mí,aunque me fuera lejos. . . justo después de embar arle en la lo ura de la Físi a. Y amis enanitos, Jorge y Luis, que después de tanto tiempo ya no son tan pequeños,también les quiero dar las gra ias por llenarme de abrazos y besos on los que llenarmis largas ausen ias. Os quiero mu ho, a todos. Gra ias.There's no pla e I an besin e I found serenity.You an't take the sky from me.Joss Wheddon, Firey

Contents1 Introdu tion 11.1 Conne ting CR and gamma rays . . . . . . . . . . . . . . . . . . . . 11.2 Gamma-ray Astronomy, There and Ba k Again . . . . . . . . . . . . 41.2.1 Observing photons of GeV energies . . . . . . . . . . . . . . . 41.2.2 Observing photons of TeV energies . . . . . . . . . . . . . . . 71.3 Re ent s ienti impa t . . . . . . . . . . . . . . . . . . . . . . . . . 81.3.1 A brief summary of gala ti highlights . . . . . . . . . . . . . 81.3.2 A brief summary of extragala ti highlights . . . . . . . . . . 111.4 Multimessenger astronomy . . . . . . . . . . . . . . . . . . . . . . . . 131.5 On this Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 Diusion of osmi -rays 192.1 The role of diusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 202.2 Astrophysi al s enarios in the Pre-Fermi era . . . . . . . . . . . . . . 222.2.1 Dis ussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252.3 Mole ular louds illuminated by CRs from SNRs . . . . . . . . . . . 283 Pre-Fermi study on the environment of SNR IC443 313.1 IC443 pla ed into ontext . . . . . . . . . . . . . . . . . . . . . . . . 313.2 MAGIC and EGRET observations of the region . . . . . . . . . . . . 343.3 A model for MAGIC J0616+225 . . . . . . . . . . . . . . . . . . . . 353.3.1 Results of the model . . . . . . . . . . . . . . . . . . . . . . . 353.4 Dis ussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404 The GeV to TeV onne tion in SNR IC 443 414.1 New high and very high-energy observations . . . . . . . . . . . . . . 414.1.1 Relative lo alization of sour es . . . . . . . . . . . . . . . . . 424.1.2 Possible relationship between gamma-ray emission and the PWN 434.2 Comparison with nominal model . . . . . . . . . . . . . . . . . . . . 434.3 Using Fermi LAT data to onstrain model parameters . . . . . . . . 454.4 Cosmi -ray distributions and their ee ts . . . . . . . . . . . . . . . 464.5 Degenera ies and un ertainties . . . . . . . . . . . . . . . . . . . . . 494.5.1 Inuen e of the δ-parameter . . . . . . . . . . . . . . . . . . . 504.5.2 Un ertainties due to the ross se tion parameterization . . . . 514.6 Computation of se ondaries other than photons . . . . . . . . . . . . 544.7 Con luding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

viii Contents5 Starburst galaxies 575.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 575.2 Theoreti al model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585.3 M 82 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 625.3.1 Comparison with previous studies . . . . . . . . . . . . . . . . 635.3.2 Results and Dis ussion . . . . . . . . . . . . . . . . . . . . . . 645.4 Dis overy of HE and VHE emission from starbursts . . . . . . . . . . 745.4.1 Gamma-ray emission dete ted from M82 . . . . . . . . . . . . 745.4.2 NGC 253, onfronted with the model . . . . . . . . . . . . . . 745.5 Con luding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . 776 Analysis of MAGIC data 836.1 Cherenkov te hnique and teles opes . . . . . . . . . . . . . . . . . . . 836.1.1 Cherenkov light . . . . . . . . . . . . . . . . . . . . . . . . . . 846.1.2 Hadroni and ele tromagneti showers . . . . . . . . . . . . . 856.1.3 Imaging Air Cherenkov Te hnique . . . . . . . . . . . . . . . 866.2 The MAGIC teles opes . . . . . . . . . . . . . . . . . . . . . . . . . . 886.2.1 Stru ture and ree tor . . . . . . . . . . . . . . . . . . . . . . 886.2.2 Camera . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 896.2.3 Readout, trigger and data adquisition . . . . . . . . . . . . . 906.2.4 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 916.2.5 Observation modes . . . . . . . . . . . . . . . . . . . . . . . . 916.3 Analysis method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 916.3.1 Mono observations . . . . . . . . . . . . . . . . . . . . . . . . 966.3.2 Stereo observations . . . . . . . . . . . . . . . . . . . . . . . . 1007 MAGIC upper limits in the region of SNR G65.1 1017.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1017.2 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1037.3 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1047.4 Upper limits on the gamma-ray ux . . . . . . . . . . . . . . . . . . 1067.5 Interpretation and dis ussion . . . . . . . . . . . . . . . . . . . . . . 1097.6 Con lusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1108 Simulations of CTA response to parti ular s ien e ases 1118.1 Sorting out dierent layouts and ongurations for CTA . . . . . . . 1118.2 A brief look at starting-up tools . . . . . . . . . . . . . . . . . . . . . 1138.3 Spe tral studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1158.3.1 Mole ular louds illuminated by CR from nearby SNR . . . . 1168.3.2 Starburst galaxies M82 & NGC 253 . . . . . . . . . . . . . . 1188.4 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

Contents ix9 IC 443 in MAGIC stereo and prospe ts with CTA 1259.1 Proposal and observations with MAGIC stereo . . . . . . . . . . . . 1289.2 Analysis and results . . . . . . . . . . . . . . . . . . . . . . . . . . . 1299.3 IC443 as seen in CTA . . . . . . . . . . . . . . . . . . . . . . . . . . 13110 Con lusions and future work 13710.1 Final remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13710.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138Bibliography 141List of Figures 169List of Tables 179

Chapter 1Introdu tionContents1.1 Conne ting CR and gamma rays . . . . . . . . . . . . . . . . 11.2 Gamma-ray Astronomy, There and Ba k Again . . . . . . . 41.2.1 Observing photons of GeV energies . . . . . . . . . . . . . . . 41.2.2 Observing photons of TeV energies . . . . . . . . . . . . . . . 71.3 Re ent s ienti impa t . . . . . . . . . . . . . . . . . . . . . . 81.3.1 A brief summary of gala ti highlights . . . . . . . . . . . . . 81.3.2 A brief summary of extragala ti highlights . . . . . . . . . . 111.4 Multimessenger astronomy . . . . . . . . . . . . . . . . . . . . 131.5 On this Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 151.1 Conne ting CR and gamma raysGamma-ray astronomy appeared as a mean to obtain an answer for the long-standingproblem of the origin of osmi rays (CR). Those CR parti les were dis overed byVi tor Hess in 1912, who observed an ionizing radiation impa ting on nu lei in theEarth's atmosphere [Hess 1912. The CR spe trum, shown in Figure 1.1, extendsover an energy range of 13 orders of magnitude.Below E = 1GeV, the CR ux is ae ted by the solar wind, via the 11 yrsmodulation of its magneti eld. At higher energies the dierential spe trum anbe des ribed by a power law dN/dE ∝ E−Γ with a spe tral index Γ ≈ 2.7 upto E = 1015.5 eV and Γ ≈ 3.0 above that energy. The transition region is alled`knee'. The spe trum hardens again at about E = 1018 eV (the so- alled `ankle'). Itis believed that CRs below the `knee' are produ ed at gala ti sites like supernovaremnants, pulsars, or binary systems, while CRs with higher energies are more likelyextragala ti . Due to gala ti and intergala ti magneti elds, harged osmi raysup to 1019 eV are isotropized. Therefore, their arrival dire tion at Earth does notpoint ba k to their origin. Only at even higher energies, in the regime of the ultrahigh energy osmi rays, the rigidity R = E/(Ze) of the parti les is so high thatthey are not dee ted by extragala ti or gala ti magneti elds (rgyro = R/B).The osmi -originated parti les an be ele tri ally harged (ele trons, protons,heavy nu lei, . . . ), but there is also photons. Sin e photons do not have harge,they are not dee ted by the interestellar magneti elds, thus providing dire t

2 Chapter 1. Introdu tion

Figure 1.1: All-parti le osmi ray spe trum. From [Be ker 2008.information on the lo ation of the original sour e. These energeti photons areprodu ed through several non-thermal pro esses and, depending on the nature ofthe parental osmi -ray parti le, they an be either leptoni (mostly ele trons) orhadroni (mostly protons and nu lei) pro esses.The main physi al pro esses that ontribute to gamma-ray reation by ele tronsand other parti les alike are: syn hrotron radiation, uvature radiation, inverseCompton intera tions, relativisti bremsstrahlung and ele tron-positron annihila-tion. Whereas for proton-originated gamma rays there are: photomeson produ tionand neutral pion de ay from proton-proton (pp) intera tions. For a omprehensivestudy on the details of these pro esses see [Ginzburg & Syrovatskii 1964, a generalsket h an be seen in Figure 1.2. Essentially the pro esses an be briey des ribedas follows.Syn hrotron radiationExtremely energeti parti les moving in a strong magneti eld will emitgamma-ray photons within an angle θ ∼ mc2/E of its dire tion of motion.Syn hrotron radiation usually generates seed photons for inverse Comptons attering (see below). Moreover, ultra-high energy CRs an emit syn hrotronphotons dire tly in the gamma-ray energy domain.Curvature radiationInside strong magneti elds harged parti les move along the eld lines andare a elerated (be ause the urvature radius of the eld is small) and thenradiate.Inverse Compton intera tionRelativisti ele trons are s attered on soft photons, transfering part of theirenergy, thus being able to produ e gamma rays. Two regimes an be onsidered

1.1. Conne ting CR and gamma rays 3

Figure 1.2: Main physi al pro esses that an generate gamma-ray photons throughdierent intera tionsm, both hadroni (b) and leptoni ( , d, e). Annihilation antake pla e among leptons or hadrons.for the ross-se tion of the intera tion: the Thompson (for EeEγ ≪ m2ec

4) andthe KleinNishina (for EeEγ ≫ m2ec

4) approximation. For EeEγ ≈ m2ec

4, theexa t Klein-Nishina formula should be used, see (Weinberg 1995, pp.362).σT = 3

8πr2e (1.1)

σKN = πr2e1

ε(ln(2ε) + 0.5) , ε =

Eγ

mec2(1.2)Photomeson produ tionIntera tion of a very relativisti proton with a photon produ e pions. Thispro ess is responsible for the GZK ut-o at distan es above 100Mp .Relativisti bremsstrahlungA relativisti ele tron is a elerated in the ele trostati eld of a nu leus or a harged parti le, radiating energeti photons. The emitted photon spe trum an be expressed with a power law with the same index as the a eleratedele tron.Hadroni gamma-ray emissionThe ollision of two protons (or proton antiproton) produ e neutral and harged parti les. The neutral pions de ay inmediately (∼ 10−16) in twogamma-ray photons. The harged pions also de ay fast (∼ 10−8), generatingmuons and neutrinos. The former se ondaries, the muons, also de ay produ -ing ele trons (or positrons, depending on the harge of the original muon and

4 Chapter 1. Introdu tionpion) and neutrinos (or antineutrinos). This neutrino signature is hara ter-isti and unique for hadroni intera tions.π0 → 2γ

π± → µ± + ν

µ± → e± + ν + νEle tron-positron annihilationTwo gamma rays are produ ed when an ele tron and a positron ollide andannihilate.1.2 Gamma-ray Astronomy, There and Ba k Again1In the early years of gamma-ray astronomy, the su ess of a mission over its pre e-dents was quantied by the number of photons, rather than the sour es dete ted.The rst experiments devoted to observe energeti photons in the MeVGeV rangewere pla ed in balloon- and spa e-borne dete tors. However, gamma rays abovehundreds of GeV ould not be dete ted by these dete tors mainly due to their small olle tion area. Ground-based teles opes over this range of energies, for instan e,proting from the imaging Cherenkov te hnique, or dete ting other se ondary par-ti les that arry information of the primary osmi ray.1.2.1 Observing photons of GeV energiesAlthough several gamma-ray dete tors were pla ed into spa e in the 1950's and1960's, the rst su esful s ienti mission dedi ated to the study of gammarays was laun hed in 1972. The se ond Small Astronomy Satellite (SAS-2) onlyoperated for 7 months, but in that short time it dete ted photons above 30MeVfrom the Crab and Vela pulsars, and onrmed the Gala ti enter as a sour e ofgamma rays. Shortly afterwards, in 1975, the COS-B satellite started taking data atenergies above 50MeV. For the rst time, the COS-B mission ompleted a skymap ofthe Gala ti plane between 100MeV and 6GeV, allowing the study of the large-s alegamma-ray diuse emission. This diuse emission was supposed to be originatedfrom osmi rays intera ting with the interestellar medium and, in fa t, a orrelationwith gas distribution (both in HI and CO maps) was found. Emission ould beresolved from loud omplexes as Ophiu hus and Orion-Mono eros. Moreover, theCrab and Vela pulsars were onrmed as gamma-ray sour es, as well as the yet-unidentied Geminga sour e and the Cygnus region [Hermsen 1990. Moreover, therst extragala ti sour e at these high energies was dete ted: the quasar 3C 273.However, the large positional un ertainty from 0.4 to 1.5 degrees, depending onthe energy prevented further identi ation of point-like sour es.1The Hobbit, or There and Ba k Again, better known by its abbreviated title The Hobbit, is afantasy novel by J. R. R. Tolkien, published in 1937.

1.2. Gamma-ray Astronomy, There and Ba k Again 5

Figure 1.3: Skymap above 100MeV from our Galaxy by the entire EGRET mission(phases 1 to 4), with the main sour es dete ted in gamma rays. Credit: EGRETTeam/NASA.Following these en ouraging results, other satellites were laun hed during thenext de ade. The next gamma-ray experiment, the Energeti Gamma-Ray Ex-periment Teles ope (EGRET), on board of the Compton Gamma-Ray Obser-vatory (CGRO), was laun hed in 1991. After nine years of observations be-tween 30MeV and 30GeV, EGRET left behind a atalog ontaining 271 sour es[Hartman et al. 1999. Blazars were established as the more numerous gamma-rayemitters. Pulsars were also established as a sour e lass by adding 5 new mem-bers, in luding the identi ation of Geminga, a radio-quiet pulsar. The skymapof the Galaxy is shown in Figure 1.3 and presents some of the prin ipal identiedsour es. Among them, the Large Magellani Cloud (LMC) is the rst galaxy de-te ted in gamma rays without having an a tive gala ti nu lei (AGN). Moreover,the radio galaxy Centaurus A (Cen A) is also dete ted. Nonetheless, 60% of thesour es in the Third EGRET Catalog (∼ 170) are unidentied. Studies on thegala ti diuse emission were arried on thanks to the performed all-sky survey.Together with this, the teles ope allowed determining the isotropi nature of theextragala ti diuse emission. Another s ienti ontribution provided by EGRETwas the prolonged GeV emission (or high energy tail) dete ted from gamma-raybursts. Regarding aveats of this instrument, the de rease of dete tion e ien y athigh energies and the large error boxes that prevented from nding ounterparts forsome of the unidentied sour es are among the most importants.Almost a de ade after the end of the CGRO mission in 2000, AGILE and Fermi

6 Chapter 1. Introdu tion

Figure 1.4: All-sky map above 300MeV during the rst year of the Fermi LATteles ope. Credit: NASA/DOE/Fermi LAT Collaboration.Table 1.1: Mission parameters of the three latest spa e teles opes in the MeV GeV range: EGRET, AGILE and Fermi LAT. The sensitivity above 100MeV is onsidered for a 2-year survey at high latitudes.EGRET AGILE Fermi LATEnergy range 30MeV 30GeV 30MeV 50GeV 20MeV 300GeVEnergy resolution 20 25% ∆E/E = 1 18 6%Ee tive area (peak) 1500 m2 700 m2 10000 m2Field of view 0.5 sr 3 sr 2.4 srAngular resolution 5.5 at 100MeV 4.7 at 100MeV 3.5 at 100MeV0.5 at 10GeV 0.2 at 10GeV 0.15 at 10GeVSensitivity > 100MeV 10−7 m−2 s−1 5× 10−8 m−2 s−1 2× 10−9 m−2 s−1Mass 1800 kg 60 kg 3000 kghave been laun hed in 2007 and 2008, respe tively. AGILE (Astrorivelatore Gammaa Immagini LEggero) ontains an X-ray monitor observing from 18 to 60 keV and agamma-ray teles ope that overs the sky in the energy range from 30MeV to 50GeV.The two instruments on board of Fermi are a Gamma-ray Burst Monitor (GBM),sensitive to hard X-rays and soft gamma-rays, and a Large Area Teles ope (LAT)that observes in sky-survey mode from 20MeV to 300GeV. Table 1.2.1 omparesthe parameters of the latest three satellite teles opes. Already in the rst year ofobservation (Figure 1.4), Fermi LAT dete ted around 1451 sour es, in luding newtypes of gamma-ray emitters, i.e., pulsar wind nebulae, supernova remnants, X-raybinaries, starburst galaxies and globular lusters. The in reased angular resolutionhas nally allowed to arry on morphologi al and spe tral studies of individualsour es. Key results provided by these latest experiments are dis ussed in the nextse tion.

1.2. Gamma-ray Astronomy, There and Ba k Again 7The dete tion te hnique ommon to SAS-2, COS-B, EGRET, AGILE and FermiLAT onsist in pair-produ tion teles opes with a tra ker, a alorimeter and an an-ti oin iden e dete tor. The ele tron and positron generated by the original gammaray hits the sili on-trip dete tor, and the paths are tra ked through the dierentlayers. The pair onversion signature is also used to distinguish the signal from themore abundant ba kground of osmi rays. The energy of the parti le is measuredin the alorimeter, when it is totally absorbed. The ultimate defense against the CRba kground is the anti oin iden e dete tor that overs the tra ker. Flashes of lightare produ ed whenever a harged parti le hits on it. All the information olle tedfrom the previous omponents is handled by the data a quisition system (DAQ),whi h also tells apart CRs from gamma rays for the rst time.1.2.2 Observing photons of TeV energiesAbove 100GeV, the ux of gamma rays is too low for the small olle tion area ofthe spa e-borne teles opes to dete t them. Nonetheless, the Earth's atmosphere,that prevents the primary gamma-ray photons to arrive at ground level, also a tsas a alorimeter where a as ade of parti les reated by energeti osmi parti les develops. The se ondary parti les that are generated in this way an be dete tedfrom ground-based teles opes and provide information from the primary parti le,like the arrival dire tion and the primary energy. One of the existing te hniquesprots from the Cherenkov radiation: when the gamma-ray photon enters the at-mosphere, it generates an ele tromagneti as ade (ele tron-positron pair), then,the se ondary parti les propagate at a faster-than-light speed in the medium, gen-erating a ash of Cherenkov light that rea hes ground level (see Chapter 6 for moredetails in Cherenkov te hnique and dete tion).Gamma rays and harged osmi rays an both produ e parti le as ades inthe atmosphere, being the latter several orders of magnitude more numerous andthus produ ing an overall ba kground. The imaging te hnique tries to dis rimi-nate among both kind of parti les through geometri onsiderations on the shapeof the as ade when it arrives at the dete tor. In 1989, the 10-meter Whippleteles ope pioneered this te hnique and dete ted the Crab Nebula at TeV energies[Weekes et al. 1989. The se ond generation of Cherenkov teles opes in the 90's wasleaded by the High Energy Gamma Ray Astronomy (HEGRA) and Cherenkov Ar-ray at Th'emis (CAT) experiments. They were responsible of dete ting the rstextragala ti sour e at TeV energies, the blazar Mkn 421 [Pun h et al. 1992, and afew other gala ti and extragala ti sour es (like, for instan e, the SNR Cas A andthe blazar Mkn 501).The Imaging Air Cherenkov (IAC) teles opes improved the ba kground reje tionthrough stereos opi observations. The introdu ed progress relied on the betterdetermination of the point of the atmosphere where the primary photon impa t, thedire tion where it ame from and the original energy. At present, the three mainIAC teles opes are:• High Energy Stereos opi System (H.E.S.S.) operating sin e 2004 with 4 tele-

8 Chapter 1. Introdu tionTable 1.2: Performan e of the three main Cherenkov experiments in the GeV TeVrange: H.E.S.S., MAGIC and VERITAS. In the title: ♯ Tels. stands for numberof teles opes, Tels. Area is the area of ea h teles ope, f.o.v. is the eld of view,Tot. Area is the total area of the array of telse opes, Eth is the energy threshold,Ang. res. means angular resolution, and Sensitivity 50 h onveys the sour e uxin 50 hours of observation respe t to the Crab Nebula ux with a signi an e of 5sigma. From [De Angelis 2011.Instrument ♯ Tels. Tels. Area f.o.v. Tot. Area Eth Ang. res. Sens. 50 h(m2) () (m2) (TeV) () (% Crab)H.E.S.S. 4 107 5 428 0.1 0.06 0.7MAGIC 2 236 3.5 472 0.05 0.07 0.8(0.025)VERITAS 4 106 4 424 0.1 0.07 0.7s opes, of 13m diameter ea h, in Namibia (South Afri a).• Major Atmospheri Gamma-ray Imaging Cherenkov Teles ope (MAGIC) on-sisted on a single-dish teles ope by 2004, and then be ome a stereos opi system by adding a se ond teles ope in 2009. Both teles opes have a diameterof 17m and are lo ated in La Palma (Canary Islands, Spain).• Very Energeti Radiation Imaging Teles ope Array System (VERITAS) ob-serving with 4 teles opes, 12m diameter ea h, sin e 2007. The re ent re-arrangement of the array improved signi atively its sensitivity.Further information on their perfoman e are displayed in Table 1.2.2.In addition to these experiments, there are other ways to observe at TeV energiesfrom ground level: for instan e, dete ting the se ondary parti les generated in thementioned atmospheri as ades. These experiments have higher energy thresholdsand longer duty y les than IAC teles opes (whi h annot observe during full Moon).Tibet and Milagro are the best examples of air shower dete tors.1.3 Re ent s ienti impa tIn the past few years, the latest instruments mentioned in the previous se tion haveprovided the s ienti ommunity with key results from the gamma-ray astronomy,from GeV to TeV energies.1.3.1 A brief summary of gala ti highlightsSupernova remnants (SNRs) have always been the most favoured andidates fora elerating osmi rays, and thus an expe ted sour e of gamma rays at high andvery high energies (HE and VHE, respe tively). In fa t, su h emission has been

1.3. Re ent s ienti impa t 9

Figure 1.5: Left: Skymap RX J1713.7-3946 by H.E.S.S., bla k ontours over-plotted show the X-ray brightness that ASCA dete tes from 1 to 3 keV. From[Aharonian et al. 2004b. Right: Fermi LAT spe tral energy distribution (SED) ofthe SNR W44, ea h urve orresponds to π0 de ay (solid), ele tron bremsstrahlung(dashed), inverse Compton s attering (dotted) and bremsstrahlung from se ondaryele trons and positrons (thin dashed). From [Abdo et al. 2010freported, although the nature of the parental osmi -ray parti les that are a el-erated (protons or ele trons) still remains a subje t of debate. The biggest steptowards the disentanglement of hadroni /leptoni s enarios has been the spatialresolution of shell-like SNR, reported both at GeV and TeV energies. RX J1713.7-3946 [Aharonian et al. 2004b and W44 [Abdo et al. 2010f onstituted the pioneer-ing sour es in ea h energy regime.The rst onrmed shell-like SNR, RX J1713 in short, was dete ted and spatiallyresolved by H.E.S.S. [Aharonian et al. 2004b. The unpre edently well-resolved shellmorphology, a quired thanks to the stereos opi imaging te hnique, oin ides withthe X-ray morphology, as an be seen in Figure 1.5 (left). The observed TeV spe -trum (without signs of a uto up to several TeVs) indi ates that parti les are beinga elerated in the shell of the SNR to energies up to 100TeV. To further distinguishbetween protons or ele trons as the parent parti les produ ing gamma rays, obser-vations of SNRs on the GeV energy range are needed. The Fermi LAT dete torresolved the shell morphology of the SNR W44 [Abdo et al. 2010f, and presentedits spe trum with a low energy ut-o (around 2GeV), as seen in Figure 1.5 (right).Subsequent dete tions of both young and middle-aged SNRs ontinue to provideinformation on their emission me hanism.Not only expe ted gamma-ray emitters have been dete ted and studied, butalso other type of obje ts have surprised the s ienti omunity by showing gamma-ray emission. This has been the ase of the nova that was found in the symbioti binary V407 Cygni (a binary system onsisting on a pulsating red giant and awhite dwarf ompanion). The Fermi LAT reported HE emission oin iding withthe maximun opti al emission from a nova outburst [Abdo et al. 2010e. The GeVemission lasted for 2 weeks, by that time, the X-ray emission reported by Swiftstarted rising. The gamma-ray spe trum suggest for the emission to be produ ed

10 Chapter 1. Introdu tioneither by neutral pion de ay or by inverse Compton s attering of infrared photonsfrom the red giant. A predi tion for novas in the gamma-ray sky was previously madeby [Tatis he & Hernanz 2007, Tatis he & Hernanz 2008, although the positionon the sky of the studied sou e prevented observations of gamma rays with ground-based dete tors. Notwithstanding, these gamma-ray novae are thought to be rareevents.Another lass of long expe ted gamma-ray generators are the X-ray binaries(XRBs). The rst one ever dete ted was Cygnus X-3 (Cyg X-3), by SAS-2. But itwas only re ently that it has been onrmed as a gamma-ray sour e between 100MeVand 100GeV by the Fermi satellite [Abdo et al. 2009g. Furthermore, the rstXRB dete ted at TeV energies was PSR B1259−63 [Aharonian et al. 2005a, very losely followed by the dete tion of LS 5039 by H.E.S.S. [Aharonian et al. 2005 .The next year, another TeV obje t was added to this lass when LS I +61 303[Albert et al. 2006 was dete ted by the MAGIC teles ope. In ea h ase, the gamma-ray emission appear to be modulated by the orbit of the system. The a elerationsite of the parental osmi ray parti les that produ e the gamma-ray emission seems to be geometry dependent and onned inside the binary system. Nonethe-less, one of the open dis ussions regarding XRBs is the nature of the ompa t obje t:wheter it is a bla k hole or a neutron star. In the former s enario, the binary systemis alled mi roquasar and it displays relativisti radio jets. Hen e, the VHE gammarays may be produ ed in the jet via inverse Compton s attering. In the ase ofthe pulsar binary system, the VHE gamma rays are expe ted to be generated inthe wind-wind ollisions between the pulsar and the massive ompanion star. Themultiwavelength approa h to solve the puzzle would be either dete ting the jet inradio, or pulsations from the ompa t obje t. Up to date, 7 XRBs have reportedeither at GeV or TeV energies, adding PSR B1259−63 [Abdo et al. 2010i, CygX-3 [Abdo et al. 2009g, Cyg X-1 [Albert et al. 2007a, Sabatini et al. 2010, 1FGLJ1018.6-5856 [Corbet et al. 2011 and HESS J0632+057 [Ong 2011, Mariotti 2011to the already mentioned binary systems. It is worth noting that, of these sour es,only LS I +61 303, LS 5039 and PSR B12594−63 have been dete ted at both GeVand TeV energies.Diuse emission from the Gala ti Center has been observed sin e the beginingof the spa e-borne missions. This emission was asso iated with intera tions of CRwith the interestellar medium. Thanks to the s an of the Gala ti Plane performedby H.E.S.S. [Aharonian et al. 2005b, eight new sour es were dis overed. In deeperand longer observations, the aorementioned asso iation spe i ally, to densemole ular louds ould be su essfully proved [Aharonian et al. 2006e. Therefore,hadroni pro esses are favoured as the origin of the gamma-ray emission in the enterof our Galaxy.The latest ares measured in the Crab Nebula [Abdo et al. 2011, defy the deni-tion of this obje t as the 'standard andle' both in X-ray and gamma-ray astronomy.In X-rays, variation in the ux of the Crab Nebula have been reported for the lastyears. The gamma-ray ares dete ted by the Fermi and AGILE satellites, however,are not orrelated with the emission in the keV energy regime.

1.3. Re ent s ienti impa t 11Pulsars, the rst established sour e- lass in gamma-rays, are starting to be stud-ied as a population in gamma rays. Fermi LAT dete ted more than 60 of theseobje ts, from whi h 8 were millise ond pulsars (MSPs) [Abdo et al. 2009e, and 16new gamma-ray pulsars were found through blind2 sear hes [Abdo et al. 2009a.The starting point of these dis overies was starred by the pulsar found inside theSNR CTA [Abdo et al. 2008. CTA-1 belongs to a ertain type of radio-quiet, butgamma-ray-loud pulsars, like Geminga.The MSPs are usually found in binary systems and thougth to be spun up by thetorque resulting from a retion of mass from their ompanion. They are signi antlymore stable than younger pulsars, although their basi emission me hanism seemsto be the same. The globular luster 47 Tu anae [Abdo et al. 2009f belongs toanother lass of predi ted sour e of gamma rays, but it had eluded dete tion untilvery re ently. Globular lusters ontains a large amount of MSPs, and in fa t theirGeV emission is largely explained by the umulative gamma-ray emission from theseyoung pulsars.Furthermore, in 2008, MAGIC opened the window of the dete tion of pul-sars from ground-based teles opes: the Crab pulsar was seen above 25GeV[Aliu et al. 2008b, thanks to a spe ial trigger setup. This re-opened a dis ussionon models that were trying to explain the emission and me hanism of pulsars. Veryre ently, the Crab pulsar as been dete ted by VERITAS at the Fermi Symposium(2010) at energies as high as 100GeV, thus onstraining even more the model forthe me hanism of gamma-ray produ tion.1.3.2 A brief summary of extragala ti highlightsBlazars are the most numerous population of gamma-ray sour es, and are in ludedin the ategory of a tive gala ti nu lei (AGN) obje ts. They allow studies on theextragala ti ba kground light (EBL), test general relativity, and shed some lighton the jet pro esses o uring next to their entral bla k holes.The starlight emitted by galaxies and a umulated over time is largely assumedto be the main ontributor for EBL. Another possible sour e of this diuse extra-gala ti emission was thought to be the emission oming from the rst stars, thatwere formed in the early Universe: metal-free massive stars, known as population III.Given that dire t measurements are not straightforward, observing distant obje ts spe i ally, their absorbed spe tra like blazars seems a better approa h. Those ab-sorption features are a onsequen e of photon-photon ollision and pair produ tion.The original spe trum of the sour e (named intrinsin spe trum) is thus modiedand depends on the spe tral energy distribution (SED) of the EBL. This spe tral hange (steepening above 1TeV) be omes more pronoun ed at larger redshifts. Forinstan e, the H.E.S.S. experiment dis overed gamma-ray emission from the distantblazars H 2356-309 and 1ES 1101-232, at redshifts z = 0.165 and 0.186, respe -tively [Aharonian et al. 2006f. By assuming an intrinsi and reasonable spe trumfor both blazars, an upper limit on the EBL ould be derived. The EBL ux was2None of them were dete ted previously in any other lower frequen y

12 Chapter 1. Introdu tion onstrained to lower values, onsistent with the limit provided by the integratedlight of resolved galaxies, and it ex luded a major ontribution from the rst stars.Up to date, ontinued observations of these obje ts is further onstraining the shapeof the EBL.Blazars owe their large number of dete tions to the fa t of having their jetspointing towards us. Their TeV emission is due to the photons emitted in the jetbeing boosted by relativisti ee ts. Besides blazars, there have also been gamma-ray emission reported from other AGNs, like the radio galaxies M87 and Cen A andfar-away radio quasar 3C 279.In the GeV regime, Cen A was dete ted by the EGRET satellite[Hartman et al. 1999, and later on the Fermi LAT teles ope onrmed and im-aged its giant radio lobes [Abdo et al. 2010d. Cen A is the nearest and one of thebrightest radio galaxies, and has also been monitored at radio wavenlenghts.M87 is one of the best studied radio galaxies in almost every wavelength,with a resolved jet (in radio, opti al and X-ray) in lined 30 from the our lineof sight. It is one of the best andidates to study the jet onne tion to the observedgamma-ray emission. A rst hint of dete tion was laimed by the HEGRA experi-ment [Aharonian et al. 2003 and was nally onrmed above 730GeV by H.E.S.S.[A iari et al. 2009 . The fast TeV variability in a daily s ale reported in 2005already pointed to a small emission region for the VHE radiation, in the imme-diate vi inity of the entral supermassive bla k hole. Another rapid and strongoutburst in 2008 was reported by the MAGIC teles ope, subsequently followed bythe VERITAS and H.E.S.S., a response that allowed a dense sampling of the event[Aharonian et al. 2006d. All the previous VHE gamma-ray dete tions, togetherwith the simultaneous high-frequen y radio overage provided by the Very LongBaseline Interferometry (VLBI), further onstrained the VHE emission site to takepla e within the jet ollimation, in a region small even for the highly resolved radioimages.The more distant extragala ti obje t dete ted in gamma rays is the ra-dio quasar 3C 279. Its dete tion in aring state by the MAGIC experiment im-plied not only the rst report at VHE of a quasar, but also that the Universewas more transparent to gamma-ray radiation than what was previously thougth[Albert et al. 2008a.The year 2009 brought to the gamma-ray astronomy the addition of a new lassof sour e: starburst galaxies were found to emit both at GeV and TeV energies(see below). These long predi ted gamma-ray emitters are lo ated at intermediatedistan es between the already dete ted LMC lose to our Galaxy and the distantblazars, and are supposed to have intermediate gamma-ray luminosities. The pro-du ion of gamma rays in these galaxies was predi ted to ome from the a elerationof CRs in SNRs and subsequent ollision with interestellar gas. An ex ess in gammarays is expe ted at the entral region of starburst galaxies due to the hara teristi enhan ed SN explosion rate, enhan ed star formation rate and ri h mole ular gasregions. More dis ussion on this topi is provided in Chapter 5.The two losest starburst galaxies that have been dete ted are M82, by

1.4. Multimessenger astronomy 13

Right Ascension

-12.729 -11.059

Dec

linat

ion

[d

eg]

-26

-25.5

-25

-80-60-40-20020406080100120140

PSF

00h46m00h48m00h50m

NGC 253

H.E.S.S.

Figure 1.6: Left: Skymap NGC 253 by HESS. Right: Spe trum M 82 by VERITAS.VERITAS [A iari et al. 2009a, and NGC 253, by H.E.S.S. [A ero et al. 2009 inthe TeV range, and both of them were dete ted by Fermi in the GeV range[Abdo et al. 2010 . The VERITAS Collaboration presented a spe trum above700GeV, see Figure 1.6 (left). H.E.S.S. showed that the gamma-ray emission omesfrom the entral part of the starburst galaxy, see Figure 1.6 (right).Gamma-Ray Bursts (GRBs) were established as extragala ti transient sour es,due to its isotropi distribution on the sky by BATSE, on board of CGRO. Theirspe tra helps onstraining EBL [Abdo et al. 2009h and, re ently, quantum gravity[Abdo et al. 2009b. The GRB 090510 at a redshift z = 0.903 was dete ted byFermi over a very broad energy range (from 8 keV to 31GeV). Despite its distan e,the rising edge time of the burst varied less than 1 s over the entire energy range.Many theories on quantum gravity predi t Lorenz invarian e violation (LIV) thatshould be dete ted as a delay: an energy-dependent variation in the rising edge timeof the burst. This delay was not dete ted in this GRB 090510 and set the strongest onstrain on the Lorenz invarian e up to now: less than the Plan k length dividedby 1.2 at 99% onden e level.1.4 Multimessenger astronomyNeutrinos are the ideal astronomi al messenger. Sin e they intera t weakly withmatter, they travel long distan es without being dee ted by magneti elds and arrying almost unbiased information of the sour e that produ es them. However,their dete tion on Earth is not so easy, basi ally for the same reason: their la k ofele tri harge results in very s ar e intera tions. Therefore, the neutrino de te torsneed to be inmense enough to olle t su ient statisti s.One of the strongest motivations to promote neutrino astronomy from thegamma-ray astronomer perspe tive is that dete ting neutrinos from an astrophys-i al sour e would provide univo al eviden e for the hadroni produ tion of gamma-

14 Chapter 1. Introdu tionrays (see se tion 1.1). The gamma-ray and neutrino astronomy shared a ouple of ommon aims:• the study of the type of sour es and me hanisms responsible of the osmi -raya eleration,• and the nature and distribution of dark matter.The most prominent km3 neutrino experiments are about to open this window.The I eCube Neutrino Observatory, in the Antarti a, has been ompleted re ently[Halzen & Klein 2010, thanks to the su ess of its prede essor, Antar ti Muonand Neutrino Dete tor Array (AMANDA). On the other side of the Earth, the Cu-bi Kilometer Neutrino Teles ope (KM3NeT) will soon ontinue the eorts madeby the Astronomy with a Neutrino Teles ope and Abyss Environmental Resear h(ANTARES) experiment in the Mediterranean sea. Being pla ed in dierent hemi-spheres, they will be both omplementary. As has been mentioned, the inmenseparti le dete tors I eCube and KM3NeT have their performan e previously testedin smaller dete tors, (namely AMANDA and ANTARES respe tively).The neutrino dete tion te hnique is based on apturing Cherenkov light, whi h isprodu ed when these parti les intera t with a nu leus in the i e or in the water. Thefew intera tions that take pla e reate muons, ele trons and hadrons in a as adeof parti les. The harged se ondary parti les are the ones that radiate Cherenkovlight and penetrate deep into the i e/water. The dire tion of the neutrino an bederived from the light pattern, that is re orded by photomultipliers (PMTs). Thetime of arrival and the digitized waveforms of the light ontains the informationto re onstru t the energy of the neutrino events, as well as their arrival dire tions.The largest sour e of ba kground for osmi neutrinos are the so- alled atmospheri neutrinos: they ome from the de ay of pions and kaons that were produ ed in ppintera tions in the atmosphere. The undesired radiation is present up to 1000TeV,but its ux an be al ulated. The good angular resolution is provided thanks tothe big size of the dete tors (in the i e, the mean free path for muons an rea k10 km).I eCube ontains a total of 5160 digital opti al modules, deployed on 86 verti alstrings, with 60 digital opti al modules (DOMs) atta hed at depths around 2000m.The on eptual design is ilustrated in Figure 1.7. The bulk of I eCube is sensitiveto neutrinos with energies above 100GeV; the DeepCore inll array may observeneutrinos with energies as low as 10GeV. The I eTop surfa e array, lo ated onthe i e above I eCube, onsists of 160 i e-lled tanks, ea h instrumented with twoDOMs. It observes osmi -ray air showers with a threshold of about 300GeV. Theearly data from I eCube are very promising, and the dete tor is observing over 10000 neutrino events per year.Also proting from the Cherenkov radiation and ontaining water dete tors isthe Auger Proje t, named after Pierre Auger, the Fren h s ientist who rst inves-tigated air showers. A tually, Auger is a hybrid experiment that ombines parti leand uores en e dete tors. The parti le dete tor part of the array onsists of a total

1.5. On this Thesis 15

Figure 1.7: A tual design of the I eCube neutrino dete tor with 5160 opti al sensorsviewing a kilometer ubed of natural i e. The signals dete ted by ea h sensor aretransmitted to the surfa e over the 86 strings to whi h the sensors are atta hed.I eCube en loses its smaller prede essor, AMANDA. From I eCube S ien e Team -Fran is Halzen.of 1,600 tanks of water, of large apa ity and sealed from external light ontamina-tion. It is dotted with PMTs to dete t the passage of air shower parti les throughtheir emission of Cherenkov light. The tanks are spa ed in a huge grid as to allowdete tion of a single air shower by ve to ten dete tor tanks. There is a set of fourarrays of uores en e dete tor teles opes interspersed among the water tank grid.There is one teles ope array in the enter of the parti le dete tor grid, with theother three eyes around the rim of the grid.1.5 On this ThesisThis Thesis ontains some studies on aspe ts related to osmi -ray diusion.It is presented in two parts, one des ribing models on the phenomenology ofCR diusion (Chapters 2 to 5), and another that shows observations using theMAGIC experiment and simulations of the future Cherenkov Teles ope Array, CTA(Chapters 6 to 9). In the rst part, the general a epted theory on CR diusionis introdu ed in Chapter 2. From this starting point, a model is presented forthe environment of the SNR IC 443 on Chapters 3 and 4 in order to explainthe high-energy phenomenology, and it is ontrasted with urrent observationsof the sour e. The Chapter 5 ontains a multi-messenger model for the diuseemission of the starburst galaxy M82. The gamma-ray predi tions are omparedwith the re ent dete tions in the GeV and TeV energy range. In the se ond part,

16 Chapter 1. Introdu tionthe Cherenkov te hnique and the MAGIC experiments is des ribed. The upperlimits obtained with the MAGIC-I teles ope from two Milagro-dete ted BrigthFermi sour es in the region of the SNR G65.1+0.6 are presented on Chapter 7. TheChapter 8 ontains simulations of CTA and initial spe tral studies on parti ulars ien e ases. Observations with MAGIC stereo on IC 443, together with preliminarstudies performed with CTA an be found in Chapter 9. Finally, the Chapter 10 ontains on lusions and advan es some future work.Part of the work in this Thesis has already been published in refereed journals:Diusion of osmi -rays and the Gamma-ray Large Area Teles ope: Phe-nomenology at the 1-100 GeV regime, [Rodríguez Marrero et al. 2008.MAGIC J0616+225 as delayed TeV emission of osmi -rays diusing from SNRIC 443, [Torres et al. 2008.The GeV to TeV view of SNR IC443: predi tions for Fermi,[Rodríguez Marrero et al. 2009.The GeV to TeV onne tion in the environment of SNR IC 443,[Torres et al. 2010.Multi-messenger model for the starburst galaxy M82,[de Cea del Pozo et al. 2009b.MAGIC Upper Limits for two Milagro-dete ted, Bright Fermi Sour es in the Re-gion of SNR G65.1+0.6 (as one of the orresponding authors), [Aleksi¢ et al. 2010.Moreover oral and poster ontributions have also been presented by the PhDstudent:Diusion of osmi -rays and gamma-ray sour es, oral talk, VIII Reunión Cien-tí a de la So iedad Española de Astronomía, 2008, Santander (Spain).Model analysis of the very high energy dete tions of the starburst galaxiesM82 and NGC 253, poster, Fermi Symposium, 2009, Washington DC (USA)[de Cea del Pozo et al. 2009a.First exploration of the spe tral CTA response on mole ular louds nearsupernova remnants, oral talk, General CTA Meeting, 2010, Zeuthen (Germany).More ontributions displaying several simulations with CTA an be found in thefollowing address:http://www. ta-observatory.org/ tawp wiki/index.php/PHYS-TLs-work.

PART I:Studies on the phenomenology of osmi -ray diusion

Chapter 2Diusion of osmi -raysContents2.1 The role of diusion . . . . . . . . . . . . . . . . . . . . . . . . 202.2 Astrophysi al s enarios in the Pre-Fermi era . . . . . . . . . 222.2.1 Dis ussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252.3 Mole ular louds illuminated by CRs from SNRs . . . . . . 28It is ommonly a epted that supernova remnants (SNR) are one of the mostprobable s enarios of leptoni and hadroni osmi -ray (CR) a eleration. The par-ti le a eleration me hanism in individual SNRs is usually assumed to be diusivesho k a eleration, whi h naturally leads to a power-law population of relativisti parti les. In the standard version of this me hanism, e.g. [Bell 1978, parti les ares attered by magnetohydrodynami waves repeatedly through the sho k front. Ele -trons suer syn hrotron losses, produ ing the non-thermal emission from radio toX-rays usually seen in shell-type SNRs.The maximum energy a hieved depends on the sho k speed and SNR age aswell as on any ompeting loss pro esses. In young SNRs, ele trons an eas-ily rea h energies in ex ess of 1TeV, and they produ e X-rays. Non-thermalX-ray emission asso iated with sho k a eleration has been learly observed inmany SNRs. But in order to have an observational onrmation of protons andother nu lei being a elerated, parti ularly, in order to be able to distinguishthis from leptoni emission, one should try and isolate the multi-messenger ef-fe ts of the se ondary parti les produ ed when the a elerated hadrons intera tin nearby mole ular louds through pp ollisions. These ideas go ba k, for instan e,to the works by [Dogel & Sharov 1990, Naito & Takahara 1994, Drury et al. 1994,Sturner et al. 1997, Gaisser et al. 1998, Baring et al. 1999, among others. In fa t,more than 30 years ago, [Montmerle 1979 suggested that SNRs within OB stel-lar asso iations, i.e. star forming regions with plenty of mole ular gas, ouldgenerate observable gamma-ray sour es. A mole ular loud being illuminatedby parti les that es aped from a nearby SNR ould then a t as a target forpp intera tions, greatly enhan ing the gamma-ray emission, see, e.g., the re- ent works by [Gabi i & Aharonian 2007, Gabi i et al. 2009, Casanova et al. 2010,Rodríguez Marrero et al. 2008. Furthermore, observing gamma rays from loudsnearby SNRs, an provide feedba k on our knowledge of the diusion hara teris-ti s of the environment.

20 Chapter 2. Diusion of osmi -raysAs investigated by [Aharonian & Atoyan 1996, the observed gamma rays anhave a signi antly dierent spe trum from that expe ted from the primary parti lepopulation at the immediate vi inity of sour e (the SNR sho k). For instan e, astandard diusion oe ient in the range δ ∼ 0.3 − 0.6 an explain gamma-rayspe tra as steep as Γ ∼ 2.3−2.6 in sour es with parti les a elerated to a power-lawJp(Ep) ∝ E−2 if the target that is illuminated by the π0-de ays is su iently faraway from the a elerator. Measuring gamma-ray emission around SNRs would thenallow to a quire knowledge of the diusion environment in whi h the CRs propagateat several kp from Earth.EGRET was unsu essful in performing detailed studies of the gamma-ray skyabove 10GeV, partly due to ba k-splash of se ondary parti les produ ed by high-energy gamma rays ausing a self-veto in the monolithi anti- oin iden e dete torused to reje t harged parti les, and partly due to a non- alibrated dete tor response.Fermi is not strongly ae ted by this ee t sin e the anti- oin iden e shield wasdesigned in a segmented fashion [Moiseev et al. 2007. The ee tive area of Fermiis roughly an order of magnitude larger than that of EGRET leading to an in reasedsensitivity, see gure 1 of [Funk et al. 2008.This Chapter presents the general theory that up to day tries to explain thediusion of osmi rays in the interestellar medium, ISM. In se tion 2.2, possibleastrophysi al s enarios that were predi ted to be dete ted at the high-energy endof the observations with Fermi have been analyzed. Those s enarios have beenpublished in [Rodríguez Marrero et al. 2008, previous to the Fermi laun h, and arethought to o ur as result of CR diusion in the ISM.2.1 The role of diusionWhen CR protons intera t with ambient nu lei, several types of parti les are pro-du ed. One of these produ ts are neutral pions, π0, whi h naturally de ay in twogamma-ray photons. Above hundreds of MeVs, the π0-de ay gamma-ray emis-sion dominates over bremsstrahlung and inverse Compton (IC) in the gala ti plane [Fi htel et al. 1976, Berts h et al. 1993. The π0-de ay gamma-ray ux froma sour e of proton-density np is

F (Eγ) = 2

∫ ∞

Eminπ

Fπ(Eπ)√

E2π −m2

π

dEπ, (2.1)whereFπ(Eπ) = 4πnp

∫ Emaxp

Eminp

Jp(E)dσπ(Eπ, Ep)

dEπdEp, (2.2)and dσπ(Eπ, Ep)/dEπ is the dierential ross-se tion for the produ tion of π0, e.g.,[Domingo-Santamaría & Torres 2005, Kelner et al. 2006. The limits of integrationin the last expression are obtained by kinemati onsiderations. Any possible gradi-ent of CR or gas number density in the target has been impli itly negle ted. The CR

2.1. The role of diusion 21spe trum, whi h is essentially mimi ked by π0-de ay gamma rays at high energies,is given by:Jp(E, r, t) =

[

cβ

4π

]

f, (2.3)where f(E, r, t) is the distribution fun tion of protons at an instant t and a distan er from the sour e.The distribution fun tion f satises the radial-temporal-energy dependent dif-fusion equation, [Ginzburg & Syrovatskii 1964, whi h in the spheri ally symmetri ase has the form:

∂f

∂t=

D(E)

r2∂

∂rr2

∂f

∂r+

∂

∂E(Pf) +Q, (2.4)where P = −dE/dt is the energy loss rate of the parti les, Q = Q(E, r, t)is the sour e fun tion, and D(E) is the diusion oe ient, for whi h a depen-den e only on the parti le's energy is assumed. The energy loss rate are dueto ionization and nu lear intera tions, with the latter dominating over the for-mer for energies larger than 1GeV. The nu lear loss rate is Pnuc = E/τpp, with

τpp = (np c κσpp)−1 being the times ale for the orresponding nu lear loss, κ ∼ 0.45being the inelasti ity of the intera tion, and σpp being the ross se tion (Gaisser1990). [Aharonian & Atoyan 1996 presented a solution for the diusion equationwith an arbitrary diusion oe ient, and an impulsive inje tion spe trum finj(E),su h that Q(E, r, t) = N0finj(E)δrδ(t). For the parti ular ase in whi h D(E) ∝ Eδand finj ∝ E−α, above 1 10GeV, where the ross-se tion to pp intera tions is aweak fun tion of E, it readsf(E, r, t) ∼

N0E−α

π3/2R3dif

exp

[

−(α− 1)t

τpp−

(

R

Rdif

)2]

, (2.5)whereRdif = 2

(

D(E)texp( tδ

τpp)− 1

tδ/τpp

)1/2 (2.6)stands for the radius of the sphere up to whi h the parti les of energy E have timeto propagate after their inje tion. In ase of ontinuous inje tion of a eleratedparti les, Q(E, t) = Q0E−αT (t), the previous solution needs to be onvolved withthe fun tion T (t− t′) in the time interval 0 ≤ t′ ≤ t [Atoyan et al. 1995.

f(E, r, t) ∼N0E

−α

4πD(E)Rerf [ R

Rdif(E, t)

]

, (2.7)where erf (z) = 2

π

∫

∞

zexp(−x2) dx (2.8)is the error fun tion.In the following se tion, typi al values, α = 2.2 and δ = 0.5, will be assumed.

22 Chapter 2. Diusion of osmi -rays

Figure 2.1: SEDs generated by CR propagation in ISM with dierent properties.Fluxes orrespond to a loud with M5/d2kpc = 0.5. Curve for D10 = 1026, 1027,and 1028 m2 s−1 are shown with solid, dotted, and dashed lines respe tively. Sen-sitivities of EGRET (red) and Fermi (blue) (both for dierent dire tions in thesky with dierent ba kground ontribution), H.E.S.S. (magenta) (survey mode andpointed observations with typi al integrations), and MAGIC (yellow), are shown for omparison purposes (see gure 1 of [Funk et al. 2008 for details on sensitivities).2.2 Astrophysi al s enarios in the Pre-Fermi eraIn the ase of energy-dependent propagation of CRs, a large variety of γ-rayspe tra is expe ted, e.g., [Aharonian & Atoyan 1996, Gabi i & Aharonian 2007,Torres et al. 2008. This study presents systemati ally and numeri ally produ ed more than 2000 E2F distributions, and their dependen es with the involved pa-rameters. Table 2.1 summarizes the results both for an impulsive and a ontinuousa elerator. The most surprising dependen es are related with the age of the a - elerator and the diusion oe ient, see Figure 2.1. These parameters have adire t impa t on the CR distribution. As the diusion in reases in speed, highenergy CR over larger distan es. The sour e ( loud) is parameterized in units of

M5 = MCl/105M⊙ and dkpc = d/ 1 kp . The ttransition parameter, dened in the ase of an impulsive a elerator, is the age for whi h the times ale for the orrespond-ing nu lear loss be omes omparable to the age of the a elerator itself. Dtransitionis the value of the diusion oe ient for whi h the SEDs stop displa ing in energykeeping approximately the same ux, as inferred from Figure 2.1.Setting, as an example, reasonable parameters for the energy inje ted by the

2.2. Astrophysi al s enarios in the Pre-Fermi era 23

Table 2.1: Dependen e of the SED (E2F vs. E) on various parameters. Imp.( ont.) stands for the impulsive ( ontinuous) a elerator ase. Dependen es upon loud parameters su h as density (nCl), mass (MCl), and radius (RCl) are obviousand related.Parameter symbol and meaning Ee t on the E2F distributions versus EA eleratorWp: total energy inje ted as CRs imp.: overall s aling, small ee ts in the rangeif in the typi al range 10501051 ergLp: energy inje ted per unit time ont.: overall s aling, small ee ts in the rangeif in the typi al range 10371038 erg s−1in reasing t: age of the a elerator imp.: peak displa es to smaller energies for axed distan e, until t > ttransition, and the peakdispla es to smaller uxes ont.: peak displa es to smaller energies andlarger uxes, for a xed distan eInterstellar mediumn: density negligible ee ts in the typi al range 0.510 m−3,sin e τpp ≫ t.in reasing D10: for a xed age: displa ement to smaller energiesdiusion oe ient of the medium until D10 > Dtransition where peaks generated(at 10GeV) by louds at large separation, R, displa e upand peaks generated by louds at smaller Rdispla e down in the SEDfor a xed distan e: displa ement to smallerenergies until D10 > Dtransition where peaksgenerated by older a elerators (larger t)displa es down and peaks generated byyounger a elerators (smaller t) displa es up

24 Chapter 2. Diusion of osmi -rays

Figure 2.2: Examples of the model predi tions for a hadroni maxima in the 1 −

100GeV regime. The left top (bottom) panel shows the predi tions for a louds aled at M5/d2kpc = 0.025 (0.04), lo ated at 20 (30) p from an a elerator of 104(3 × 104) yr, diusing with D10 = 1027 m2 s−1. The right top (bottom) panel urve shows the predi tions for a loud s aled at M5/d

2kpc = 0.08 (0.06) lo ated at10 (20) p from an a elerator of 103 (104) yr, diusing with D10 = 1028 m2 s−1.In reasing the ratio M5/d

2kpc, the urves move up maintaining all other features.a elerator into osmi -rays (e.g., Wp = 5×1049 erg for an impulsive sour e and Lp =

5× 1037 erg s−1 for a ontinuous one) and for the interstellar medium density (e.g.,n = 1 m−3), several s enarios for the appearan e of hadroni maxima produ ed bydiusion are found. Some examples are shown in Figure 2.2, for the two types ofa elerators. Two kinds of peaks at this energy regime are possible: those that arenot to be dete ted by an instrument with the sensitivity of EGRET or MAGIC,and those that are not to be dete ted by an instrument like H.E.S.S. or VERITAS.The impulsive a elerator produ es a more narrow peak maxima. A maxima in theSED, hadroni ally produ ed as an ee t of diusion of CRs, is possible and notun ommon at the high-energy end, where they produ e a level of ux dete table byFermi LAT.Figure 2.3 displays, as ontour plots, the energy at whi h the maximum of theSED is found for the ases of impulsive a eleration of osmi rays, at dierent dis-tan es, ages of the a elerator, and diusion oe ients. The impa t of the diusionis learly shown by this gure, whi h hanges are due to pure energy-dependent prop-agation ee ts. The diusion radius, for t ≪ τpp, is Rdif(E) = 2

√

D(E)t, so that ata xed age and distan e, only parti les of higher energy will be able to ompensate a

2.2. Astrophysi al s enarios in the Pre-Fermi era 25smaller D10, produ ing SED maxima at higher E-values. The smaller values of D10are expe ted in dense regions of ISM, e.g., [Ormes et al. 1988, Torres et al. 2008.It is interesting to note that for many, albeit not for all, of the SEDs studied, themaximum is found at energies beyond the Fermi a eptan e. On the other hand,Figure 2.3 interpretes a Fermi observational dis overy of a 1− 100GeV maximum,and provides interesting lues about the nature of the astrophysi al system thatgenerates the gamma rays. First, these SEDs are found in ases where the s enariodoes not predi t dete table emission at the EGRET sensitivity, so that they willrepresent new phenomenology. Se ond, the range of a elerator-target separationsand ages of the a elerator that would produ e su h a 1 − 100 GeV maximum israther limited (see in Figure 2.3 the narrow ontours for maxima at su h energies).This would lead to a dire t identi ation of the sour e.Another interesting possibility is the ase of two unresolved sour es. Two sepa-rate a elerator- loud omplexes are onsidered lose to the line of sight, su h thatthey would be observed as a single sour e. This kind of s enarios would produ ean inverted spe trum. Figure 2.4 shows four possible inverted spe tra. The twogures in the top (bottom) panel are generated by an impulsive ( ontinuous) a el-erator. The SED reated by the oldest (youngest) a elerator is shown by dashed(dot-dashed) lines in ea h of the s enarios. For the left ases, EGRET should havebeen able to weakly dete t the sour e produ ing uxes at smallest energies. In any ase, EGRET ould not on lusively relate it to su h phenomenon due to its largelow-energy PSF. The ounterpart at higher energies is a bright sour e potentiallydete table by ground-based teles opes. Due to ontinuous energy overage, Fermi isa prime instrument to tra k this phenomenology, although none has been reportedup to now. The right panel ases show parti ular examples in whi h the dete tionof the sour e by an instrument with the sensitivity of EGRET is not possible at all.The inverted spe trum is less deep in these s enarios. Less pronoun ed V-shapedspe tra an be obtained with on omitantly lower uxes at TeV energies.2.2.1 Dis ussionCompton peaks (whi h rst example ould have been found already,[Aharonian et al. 2006 ) are not the only way to generate a maximum in a SED. Alarge variety of parameters representing physi al onditions in the vi inity of a CRa elerator ould produ e a rather similar ee t. Distinguishing between these aseswould require multiwavelength information, sear h for ounterparts, and modelling.If su h a maximum is interpreted hadroni ally, as a result of diusion of CR in theISM and their subsequent intera tion with a nearby target, the results presentedherein onstrain, given the energy at whi h the maximum of the SED is rea hed,the hara teristi s of the putative a elerator, helping to the identi ation pro ess.Indeed, one of the most distinguishing aspe ts of this study is the realization thatthese signatures (in parti ular, peaks at the 1−100GeV energy region) is indi ativefor an identi ation of the underlying me hanism produ ing the gamma rays thato urs in nature: whi h a elerator (age and relative position to the target loud)

26 Chapter 2. Diusion of osmi -rays

Figure 2.3: For ea h ombination of age and a elerator-target separation, forwhi h more than two thousand spe tra where numeri ally produ ed, the energy ofthe maximum of su h spe tra are shown in a ontour plot. The olor of the dierent ontours orresponds to the range of energy where the maximum is found a ordingto the olor bar above ea h gure. From top to bottom, plots are reated for the ase of an impulsive sour e inje ting protons in a medium with D10=1026 m2 s−1,1027 m2 s−1 and 1028 m2 s−1.

2.2. Astrophysi al s enarios in the Pre-Fermi era 27

Figure 2.4: The parameters for the plots are as follows, (top left) the dashed urveon the left: t = 4×105 yr, R = 5p , M5/d2kpc = 0.01; the dashed urve on the right:

t = 104 yr, R = 20p , M5/d2kpc = 0.1; (top right) the dashed urve on the left: t =2×106 yr, R=100 p , M5/d

2kpc = 3; the dashed urve on the right: t = 4×103 yr, R= 15p ,M5/d

2kpc = 0.1; (bottom left) the dashed urve on the left: t = 2×106 yr,

R=15p , M5/d2kpc = 0.004; the dashed urve on the right: t = 103 yr, R = 5p ,

M5/d2kpc = 1; ( bottom right) the dashed urve on the left: t = 2×106 yr, R = 40p ,

M5/d2kpc = 0.017; the dashed urve on the right: t = 6×104 yr, R = 30p ,M5/d

2kpc= 2.5. D10 is set to 1026 m2 s−1. R is the a elerator- loud separation.

28 Chapter 2. Diusion of osmi -raysand under whi h diusion properties CR propagate, as it is exemplied in Figure2.3. In a survey mode su h as the one Fermi LAT performs, it is also possible to ndrather unexpe ted, tell-tailing SEDs, like those V-shaped presented here, if observedwith instruments having a limited PSF, predi tably leaving many Gala ti sour esunresolved.2.3 Mole ular louds illuminated by CRs from SNRsNon-thermal emission is expe ted to ome from mole ular louds, due to in-tera tions of CR that penetrate the loud. This emission would be enhan edwhenever the mole ular loud is in the proximity of a SNR [Gabi i et al. 2009,[Rodríguez Marrero et al. 2009. The mole ular louds provide a dense target for theCR that es ape and subsequently diuse away from the a elerator (i.e., the SNR).When both sour es of CR are lose with respe t to the line of sight, a on ave spe -trum appears in gamma rays, ree ting the shape of the underlying CRs. This al-ternative V-shaped spe trum of CRs was thouroghly studied in [Gabi i et al. 2009.A hadroni s enario, where gamma rays mainly appear as a result of neutralpion de ay produ ed in pp intera tions, would be favoured if spatial orrelationwere found between TeV gamma rays and dense gas. A mole ular loud, sit-uated lose to a SNR, would appear illuminated by the CR es aping the SNRand produ e gamma rays. This asso iation was already presented as the possiblesour e of unidentied TeV sour es dete ted by ground based Cherenkov teles opes[Gabi i & Aharonian 2007. The model explored in [Gabi i et al. 2009 in ludes notonly the hadroni ally produ ed photons, but also the se ondary ele trons generatedin the loud, together with the bremsstrahlung and IC emission. The diusion o-e ient inside the loud is assumed to be similar to the Gala ti one, in order tohave a free penetration of the CRs.The total CR spe trum onsidered therein has two ontributions: one omingfrom the Gala ti ba kground and the other from the CRs that es ape the a el-erator. The rst ontribution is hara terized by a steep spe trum that peaks inthe GeV energy region, and it is not supposed to hange with time. On the otherhand, the runaway CRs will be variable in time and their spe trum is hard. At thehighest energies, in the TeV range, this se ond peak de reases and moves to lowerenergies. This ee t omes from the fa t that CRs diuse earlier and faster awayfrom the a elerator at PeV energies than at lower energies (GeVs, TeVs). Takinginto a ount both ontributions, the CR spe trum a quires a on ave shape, whi his ree ted in gamma rays. The aoredmentioned evolution in time is proportionalto the square of the separation between the mole ular loud and the SNR. The CRspe trum presents a ut-o, whi h position depends on the last parti les with energyenough that have time to rea h the loud.The dete tion of su h a shape ould prove the presen e of a CR a elerator lose to mole ular louds, and maybe larify the nature of up-to-now unidentiedTeV sour es. One possibility to a hieve this goal would be overing the whole

2.3. Mole ular louds illuminated by CRs from SNRs 29energy range with joint observations of Fermi and the next generation of groundbased Cherenkov teles opes (CTA). An initial study on this dire tion is explored inChapter 8.

Chapter 3Appli ations to the environmentof SNR IC443: Pre-Fermi studyContents3.1 IC443 pla ed into ontext . . . . . . . . . . . . . . . . . . . . 313.2 MAGIC and EGRET observations of the region . . . . . . . 343.3 A model for MAGIC J0616+225 . . . . . . . . . . . . . . . . 353.3.1 Results of the model . . . . . . . . . . . . . . . . . . . . . . . 353.4 Dis ussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40Of all supernova remnants (SNRs) that were found to be positionally oin identwith gamma-ray sour es in the MeV range in the EGRET era, IC 443 was one ofthe most appealing for subsequent observations with higher sensitivity instruments(see the ase-by- ase study by [Torres et al. 2003). It was, together with W28, theonly ase in whi h the mole ular environment - as mapped for instan e with COobservations - showed a peak in density lose by, but separated in sky proje tion,from the SNR enter, as shown in Figure 3.1. This would allow distinguishingpossible osmi -ray (CR) diusion ee ts, in ase the gamma-ray emission observedwould be hadroni ally produ ed. Several observations of the IC 443 environmenthave been made at the highest energies, and in this Chapter, these are used in thesetting of a theoreti al model in whi h CRs from the SNR IC 443 are diusing awayfrom it and intera ting with louds nearby. This model has been originally presentedin [Torres et al. 2008.3.1 IC443 pla ed into ontextIC 443 is an asymmetri shell-type SNR with a diameter of ∼45' (e.g.,[Fesen & Kirshner 1980). Two half shells appear in opti al and radio images,e.g, [Braun & Strom 1986, Leahy 2004, Lasker et al. 1990. The intera tion re-gion, with eviden e for multiple dense lumps, is also seen in 2MASS im-ages, e.g. [Rho et al. 2001. In radio, IC 443 has a spe tral index of 0.36,and a ux density of 160 Jy at 1GHz [Green 2004. [Claussen et al. 1997 re-ported the presen e of maser emission at 1720MHz at (l, b) ∼ (−171.0, 2.9).Later on, [Hewitt et al. 2006 onrmed these measurements and dis overed

32 Chapter 3. Pre-Fermi study on the environment of SNR IC443

Figure 3.1: CO distribution around the remnant IC 443 (G189.1+3.0). The 3EGgamma-ray sour e J0617+2238 is plotted with white ontours. The opti al boundaryof the SNR is superimposed as a bla k ontour [Lasker et al. 1990. The opti alemission seems to fade in regions where CO emission in reases. This indi atesthat the mole ular material is likely lo ated on the foreground side of the remnant,absorbing the opti al radiation. Plot taken from [Torres et al. 2003, gure 10.

3.1. IC443 pla ed into ontext 33weaker maser sour es in the region of intera tion. IC 443 is a promi-nent X-ray sour e, observed with Rosat [Asaoka & As henba h 1994, ASCA[Keohane et al. 1997, XMM [Bo hino & Bykov 2000, Bo hino & Bykov 2001,Bo hino & Bykov 2003, Bykov et al. 2005, Troja et al. 2006, and Chandra[Olbert et al. 2001, Gaensler et al. 2006. The works by [Troja et al. 2006 and[Bykov et al. 2008 summarize these observations. Some additional features ofIC 443 are presented next, due to their relevan e for the model.Age A small (∼ 1000 yr) age was determined by [Wang et al. 1992 and se ondedby [Asaoka & As henba h 1994 and [Keohane et al. 1997, although IC 443is now agreed to have a middle-age of about 3 × 104 yrs. This age has beeninitially advo ated by [Lozinskaya 1981 and was later onsistently obtainedas a result of the SNR evolution model [Chevalier 1999. Observations by[Bykov et al. 2008 onrm that there are a few X-ray-emitting eje ta frag-ments, a number mu h smaller than that expe ted for a younger SNR.Distan eKinemati al distan es from opti al systemi velo ities span from 0.7 to 1.5 kp ,e.g., [Lozinskaya 1981. The assumption that the SNR is asso iated witha nearby HII region, S249, implies a distan e of ∼ 1.5 − 2.0 kp . Sev-eral authors laimed that the photometri distan e is more reliable, e.g,[Rosado et al. 2007, and on urrently with all other works on IC 443, a dis-tan e of 1.5 kp is adopted here (thus, 1 ar min orresponds to 0.44 p ).Energy of the explosionThere is no lear indi ator for E51, a term dened as the energy of the explosionin units of 1051 erg. [Chevalier 1999 obtains a lower limit of 4 × 1050 erg,whereas lower estimations are provided by [Di kman et al. 1992, based on[Mufson et al. 1986, albeit the latter assumed an age of ∼5000 yr. La kinga strong reason for other numeri al assumptions, in the following model itwill be assumed E51 = 1, although to be onservative, it will subsequentlybe assumed that only 5% of this energy is onverted into relativisti CRs.Reasonable dieren es in this assumed value of E51 are not expe ted to haveany impa t on the model.The mole ular environment[Cornett et al. 1977 and [De Noyer & Frerking 1981 were among the rstto present detailed observations of mole ular lines towards IC 443.Later on, [Di kman et al. 1992, [Seta et al. 1998, [Butt et al. 2003 and[Torres et al. 2003 among others, presented further analysis. These works onform the urrent pi ture for the environment of IC 443: a total massof ∼ 1.1 × 104 M⊙ mainly lo ated in a quies ent loud in front of theremnant (with linear s ales of a few parse s and densities of a few hun-dred parti les m−3) that is absorbing opti al and X-ray radiation, e.g.,

34 Chapter 3. Pre-Fermi study on the environment of SNR IC443[Lasker et al. 1990, Troja et al. 2006, a s enario already put forward by[Cornett et al. 1977. [Di kman et al. 1992 estimated that 500− 2000M⊙ aredire tly perturbed by the sho k in the northern region of intera tion, near theSNR itself. [Huang et al. 1986 found several lumps of mole ular materialalong this intera ting shell, with subparse linear s ales. [Rosado et al. 2007found inhomogeneities down to 0.007 p . As a rst approximation to the prob-lem, these latter inhomogeneities are negle ted when onsidering the propaga-tion of CRs in the ISM, i.e. it is assumed an homogeneous medium of typi alISM density where CRs diuse. Then, the mole ular mass s enario is a maingiant loud in front of the SNR ontaining most of the quies ent mole ularmaterial found in the region, and smaller loud(s) totalizing the remainingmass lo ated loser to the SNR.3.2 MAGIC and EGRET observations of the regionThe rst gamma-ray emission oming from the SNR IC 443 was labeled asthe EGRET sour e 3EG J0617+2238 [Hartman et al. 1999. The EGRET uxwas (51.4± 3.5) ×10−8 ph m−2 s−1, with a photon spe tral index of 2.01± 0.06[Hartman et al. 1999. The EGRET sour e was lassied as non-variable by[Torres et al. 2001a, Nolan et al. 2003. An independent analysis of GeV pho-tons measured by EGRET ended up being the sour e GeV J0617+2237[Lamb & Ma omb 1997, also at the same lo ation of 3EG J0617+2238 at the enterof the SNR.Later on, MAGIC observations towards IC 443 yielded the dete tion ofJ0616+225 nearby, but displa ed from the enter of the SNR IC 443, with en-troid lo ated at (RA,DEC)J2000=(06h16m43s, +2231' 48), ± 0.025stat ± 0.017sys[Albert et al. 2007b. The MAGIC Collaboration showed that the very high energy(VHE) sour e is lo ated at the position of a giant loud right in front of the SNR.A simple power law was tted to the measured spe tral points:dNγ

dAdtdE= (1.0 ± 0.2stat ± 0.35sys)× 10−11

(

E

0.4TeV

)−3.1±0.3stat±0.2sys

cm−2s−1TeV−1 (3.1)with quoted errors being statisti al. The systemati error was estimated to be 35%in ux and 0.2 in spe tral index. The integral ux of MAGIC J0616+225 above100GeV is about 6.5% of the Crab Nebula. No variability was found along theobservation time (over one year). No signi ant tails or extended stru ture wasfound at the MAGIC angular resolution. These results were onrmed by observa-tions with the VERITAS array [Humensky & the VERITAS Collaboration 2008.In addition, onsistent upper limits were reported by Whipple dNγ/(dAdt) <

6 × 10−12cm−2s−1 (0.11 Crab) above 500GeV [Holder et al. 2005 and by CATdNγ/(dAdt) < 9× 10−12cm−2s−1 above 250GeV [Kheli 2003.

3.3. A model for MAGIC J0616+225 35MAGIC J0616+225 is displa ed with respe t to the position of the lower energysour e 3EG J0617+2238. Indeed, the EGRET entral position is lo ated dire tlytowards the SNR, whereas the MAGIC sour e is south of it, lose to the 95% CL ontour of the EGRET dete tion. As [Albert et al. 2007b showed, the MAGICsour e is lo ated at the position of a giant loud in front of the SNR. The aim of themodel presented here is to show eviden e of them being related. Extrapolating thespe trum of the EGRET sour e into the VHE regime, a higher ux and a harderspe trum than the one observed for MAGIC J0616+225 would be obtained, sup-porting the view that a dire t extrapolation of this and other EGRET measurementsinto the VHE range is not valid [Funk et al. 2008.3.3 A model for MAGIC J0616+225A theoreti al model is presented in this Chapter and the previous one, explain-ing the high energy phenomenology of IC 443 (see [Aharonian & Atoyan 1996,Gabi i & Aharonian 2007), pla ing an emphasis on the displa ement betweenEGRET and MAGIC sour es. MAGIC J0616+225 is interpreted as a delayed TeVemission of CRs diusing from the SNR. This model is ompared with other ontem-porary studies, together with a dis ussion on how it an be tested using observationswith the Fermi-LAT instrument (at that time still to ome).The model [Torres et al. 2008 is based on the s enario of CR diusion, as de-s ribed in the previous Chapter. The parameters have been adjusted here forthis spe i sour e. CRs are a elerated at the SNR site and end up intera t-ing with the nu lei present in the interestellar medium (ISM), produ ing a set ofparti les, in luding both harged and neutral pions. Gamma rays are generatedthrough π0−de ay. The CR spe trum of the protons is given by a distributionfun tion that satises the radial-temporal-energy dependent diusion equation of[Ginzburg & Syrovatskii 1964. Details on this issue are given in Chapter 2. Twoassumptions are onsidered: a diusion oe ient dependent only on the energy,D(E)∝ E−δ, and a power-law for the distribution fun tion of the inje ted protonspe trum (∝ E−α). Both impulsive and ontinuous inje tion of CR protons from thea elerator are onsidered (equations 2.5 and 2.7) in the following results. Finally,α = 2.2 is assumed.3.3.1 Results of the modelFigure 3.2 shows the urrent CR spe trum generated by IC 443 at two dierentdistan es from the a elerator, 10 (solid) and 30 (dashed) p . The SNR is onsid-ered both as a ontinuous and an impulsive a elerator. In the ontinuous ase, arelativisti proton power of Lp = 5× 1037 erg s−1 is assumed, and the proton lumi-nosity is su h that the energy inje ted into relativisti CRs through the SNR age is5× 1049 erg. If an impulsive inje tor is onsidered, then it has the same total powerbut the inje tion of high energy parti les o ur in a mu h shorter times ale thanthe SNR age. The horizontal line in Figure 3.2 marks the CR spe trum near Earth,

36 Chapter 3. Pre-Fermi study on the environment of SNR IC443

Figure 3.2: CR spe trum generated by IC 443 at two dierent distan es, 10 (solid)and 30 (dashed) p , at the age of the SNR. Two types of a elerator are onsidered,one providing a ontinuous inje tion (bla k) and other providing a more impulsiveinje tion of CRs (red). The horizontal line marks the CR spe trum near the Earth.The Y-axis units have been hosen to emphasize the ex ess of CRs in the SNRenvironment.so that the ex ess of CRs in the SNR environment an be seen. For this ase, thediusion oe ient at 10 GeV, D10, was hosen to be 1026 m2 s−1, with δ = 0.5.CRs propagate through the ISM, whi h is assumed to have a typi al density, e.g.nISM = 0.5, 1, 5, and 10 m−3 . In the s ale of Figure 3.2, urves for this set ofvalues would be superimposed, so that nISM be omes an irrelevant parameter inthis range. The reason for this is that the times ale for nu lear loss τpp, obtainedwith the densities onsidered for the ISM, is orders of magnitude larger than the ageof the a elerator. Dieren es between the dierent kind of a elerators assumedare also minimal for the SNR parameters.Figure 3.3 shows the result for the gamma-ray emission oming from the loudlo ated at the position of the MAGIC sour e, when it is assumed that the sour elies at dierent distan es in front of IC 443. The giant loud mass is assumed( onsistently with observations) to be 8000M⊙. The a elerator properties andpower of IC 443 are those shown in Figure 3.2, for ea h ase. Fluxes are given foran ISM propagation in a medium of n = 1 m−3, although this is not a relevantparameter, as it has been dis ussed beforehand. Clouds lo ated from ∼20 to ∼30 p produ e an a eptable mat h to MAGIC data. In the ase of a more impulsivea elerator, the VHE predi ted spe tra is slightly steeper than that produ ed inthe ontinuous ase at the same distan e, providing a orrespondingly better t tothe MAGIC spe trum. Figure 3.3 also shows, apart from MAGIC data, EGRETmeasurements of the neighborhood of IC 443. As it has been stated previously, thesetwo sour es are not lo ated at the same pla e, highlighted with dierent symbols in

3.4. Dis ussion 37the plot. Figure 3.3 shows that there is plenty of room for a loud the size of theone dete ted in front of IC 443 to generate the MAGIC sour e and not a o-spatialEGRET dete tion.The existen e of a VHE sour e without a ounterpart at lower energies is theresult of diusion of the high-energy CRs from the SNR sho k, whi h is an energydependent pro ess leading to an in reasing de it of low energy protons as thedistan e from the a elerator be omes larger. To larify this assertion, and sin ethe solution to the diusion-loss equation is a fun tion of time, the evolution of theux along the age of the SNR is shown in Figure 3.4. The integrated photon ux oming from the position of the giant loud is presented as a fun tion of time above100MeV and 100GeV in the impulsive ase. Dierent qualities of the a elerator(impulsive or ontinuous) produ e a rather omparable pi ture. At the age of theSNR, Fermi is predi ted to see a sour e only for the losest separations. On the ontrary, the integrated photon uxes above 100GeV present minimal deviations,and a MAGIC sour e is always expe ted.Figure 3.3 also presents the results of this theoreti al model fo using in theenergy range of EGRET. At these low energies, the CR spe trum intera ting witha lo al-to-the-SNR loud is obtained assuming an average distan e of intera tionof 3 − 4p . A few hundred M⊙ lo ated at this distan e (∼700M⊙ in the ase ofan impulsive, and ∼300M⊙ for a ontinuous ase) produ e an ex ellent mat h tothe EGRET data, without generating a o-spatial MAGIC sour e. As defended in[Gaisser et al. 1998, the lowest energy data points in the EGRET range (below 100MeV) are produ ed by bremsstrahlung of a elerated ele trons.The present model imposes some onstraints provided by the observed phe-nomenology, like e.g., the mole ular environment and the position of the gamma-ray sour es. One of the onsequen es is that D10 should be low, of the order of1026 m2 s−1, whereas the hara teristi value on the Galaxy is ∼1028 m2 s−1. Byvarying the diusion oe ient, its inuen e in the results an be studied. If theseparation between the giant loud and the SNR is > 10 p , a slower diusion(D10 > 1026 m2 s−1) would not allow su ient high energy parti les to rea h thetarget material; thus, the MAGIC sour e would not be there. On the other hand, ifthe separation between the main loud and the SNR is < 10 p , the EGRET sour ewould have been dete ted at the position of the loud, whi h is not the ase. Thevalue of D10 established in this study, at 1.5 kp from Earth, has been dened by ombining MAGIC and EGRET observations. Su h low values of D10 are expe tedin dense regions of ISM [Ormes et al. 1988, Gabi i & Aharonian 2007.3.4 Dis ussionIn this se tion, the dis ussed model is pla ed in the ontext of others.[Bykov et al. 2000 suggested that the GeV emission seen towards IC 443 is mostlydue to relativisti bremsstrahlung. As noted by [Butt et al. 2003, who already favora hadroni emission of gamma-rays, the syn hrotron radio emission seen towards the

38 Chapter 3. Pre-Fermi study on the environment of SNR IC443

Figure 3.3: MAGIC and EGRET measurement of the neighborhood of IC 443 (starsand squares, respe tively) as ompared with model predi tions. The top (bottom)panel shows the results for an impulsive ( ontinuous) ase. At the MAGIC energyrange, the top panel urves show the predi tions for a loud of 8000M⊙ lo atedat 20 (1), 25 (2), and 30 (3) p , whereas they orrespond to 15 (1), 20 (2), 25(3), and 30 (4) p in the bottom panel. At lower energies, the urve shows thepredi tion for a few hundred M⊙ lo ated at 3 − 4p . The EGRET sensitivity urves (in red) are shown for the whole lifetime of the mission for the Gala ti anti- entre (solid), whi h re eived the largest exposure time and has a lower level ofdiuse gamma-ray emission, and for a typi al position in the Inner Galaxy (dashed),more dominated by diuse gamma-ray ba kground. The Fermi sensitivity urves(in blue) show the simulated 1-year sky-survey sensitivity for the Gala ti Northpole, whi h orresponds to a position with low diuse emission (solid), and fora typi al position in the Inner Galaxy (dashed). These urves were taken fromhttp://www-glast.sla .stanford.edu/software/IS/glast_latperforman e.html. From[Rodríguez Marrero et al. 2009.

3.4. Dis ussion 39

Figure 3.4: Integrated photon ux as a fun tion of time above 100MeV and 100GeV,solid (dashed) lines orrespond to the ase of the loud lo ated at 10 (30) p . Thehorizontal lines represent the values of integrated uxes in the ase that the CRspe trum intera ting with the loud is the one found near Earth. The verti alline stands for the SNR age. EGRET and Fermi initially predi ted integralsensitivity are shown, onsistent in value and olor oding with those in Figure 3.3.rim of the SNR and the entrally lo ated EGRET sour e must then be asso iated.Judging from the lo alization of the multiwavelength emissions, this does not seemto be the ase. An unavoidable (but subdominant at high energies) bremsstrahlung omponent oming from primary and se ondary ele trons an however play a role atthe lowest energies in the EGRET range, as already shown by [Gaisser et al. 1998and Figure 3.3.[Bo hino & Bykov 2001 suggested that the systemati errors in the EGRETlo ation ontours ould yield the pulsar CXOU J061705.3+222127 dis overed by[Olbert et al. 2001 and its nebula, as the sour e of the GeV emission. As dis ussedabove, this ontradi ts urrent observations. The same assumption (i.e., that thepositions of the measured EGRET and GeV sour es are wrong by half a degree)was taken by [Bartko & Bednarek 2008. The pulsar nebula is also displa ed fromthe MAGIC dete tion by 20 ar min, but these authors suggested that they may be onne ted if a pulsar with a velo ity of 250 km s−1 moves along the SNR age. Thisimplies that the pulsar CXOU J061705.3+222127 should have been born at the SNR enter and should have traveled to its urrent position while a elerating parti lesthat intera t with the loud, giving rise to the MAGIC sour e. This s enario doesnot seem to mat h the observed phenomenology: the EGRET sour e should be ontop of the urrent position of the pulsar and not where it a tually is, and physi ally,it should be the result of pulsed emission (like in Vela), although pulses were notreported during the EGRET era. The only argument supporting the latter assump-tion is that the ux and spe trum of 3EG J0617+2238 are similar to that of PSR

40 Chapter 3. Pre-Fermi study on the environment of SNR IC4431706-44, also observed by EGRET. This would apply to dozens of other EGRETsour es but an not be sustained as ir umstantial eviden e of physi al similar-ity, e.g., [Romero et al. 1999, Reimer 2001, Torres et al. 2001b, Torres et al. 2003.Still within the same s enario, the MAGIC sour e is onsidered to be generated byinverse Compton from ele trons a elerated at an initial phase of the pulsar andtravelling towards the loud. The lo alization and size of the MAGIC sour e arenot explained sin e the dieren e in target photon elds in the region surroundingthe loud should not be signi ant, and the target eld should even be larger at theposition of the intera ting sho k in the northeast.[Zhang & Fang 2008 presented an alternative model for IC 443, in whi h a fra -tion of the SNR shell evolves in the mole ular loud, and other in the ambientinterstellar environment, en ountering dierent matter densities. In this model, thegamma-rays observed by EGRET are mainly produ ed via pp intera tions with theambient matter in the louds, similar to that found in the MAGIC sour e. Althoughthis may sound similar to the present model, there is a key dieren e between them:in Zhang and Fang s enario, both EGRET and MAGIC sour es should be at thesame position. The reason for this lies in the fa t that, in their model, the radialdependen e of the CR spe trum is not onsidered. [Gabi i & Aharonian 2007 notedin general that an old SNR annot onne multi-TeV parti les in their shells, as ithas also been shown in the IC 443 results of this Chapter.3.5 SummaryIn this Chapter, MAGIC J0616+225 is found to be onsistent with the interpreta-tion of CR intera tions with a giant mole ular loud lying in front of the remnant,produ ing no ounterpart at lower energies. Moreover, the nearby EGRET sour e an be explained as produ ed by the same a elerator, and in this ase, a o-spatialMAGIC sour e is not expe ted. In the present model, the displa ement betweenEGRET and MAGIC sour es has a physi al origin. It is generated by the dierentproperties of the proton spe trum at dierent lo ations, in turn produ ed by thediusion of CRs from the a elerator (IC 443) to the target.At high energies, a morphologi al and spe tral hange should ome from theposition of the loud (i.e. the enter of MAGIC J0616+225) towards the enter ofIC 443. At a morphologi al level, the lower the energy, the more oin ident with theSNR the radiation will be dete ted. At a spe tral level, su ient statisti s shouldshow that the lower the gamma-ray energy the harder the spe trum is. In addition,Fermi LAT measurements are predi ted to be sensitive enough to dete t the same loud that shines at higher energy.All these predi tions will be dis ussed in the following Chapter, ombining theMAGIC stand-alone data with the new observations by Fermi and AGILE at HE,and by VERITAS at VHE. Related to this topi , preliminar observations made withthe MAGIC stereo system at VHE will be presented in Chapter 9, together withprospe ts on the forth oming Cherenkov Teles ope Array (CTA).

Chapter 4The GeV to TeV onne tion inSNR IC 443Contents4.1 New high and very high-energy observations . . . . . . . . . 414.1.1 Relative lo alization of sour es . . . . . . . . . . . . . . . . . 424.1.2 Possible relationship between gamma-ray emission and the PWN 434.2 Comparison with nominal model . . . . . . . . . . . . . . . . 434.3 Using Fermi LAT data to onstrain model parameters . . . 454.4 Cosmi -ray distributions and their ee ts . . . . . . . . . . . 464.5 Degenera ies and un ertainties . . . . . . . . . . . . . . . . . 494.5.1 Inuen e of the δ-parameter . . . . . . . . . . . . . . . . . . . 504.5.2 Un ertainties due to the ross se tion parameterization . . . 514.6 Computation of se ondaries other than photons . . . . . . . 544.7 Con luding remarks . . . . . . . . . . . . . . . . . . . . . . . . 55In the previous Chapter 3, the sour e MAGIC J0616+225 has been interpretedas a result of delayed TeV emission of osmi rays (CR) diusing from the super-nova remnant (SNR) IC 443 and intera ting with a loud in the foreground of theremnant. This model was used to make predi tions for observations with the Fermisatellite. Re ently, AGILE, Fermi LAT, and VERITAS released new results fromtheir observations of IC 443. In this Chapter, those results are ompared with thepredi tions of the model, exploring the GeV to TeV onne tion in this region. FermiLAT data is used to onsider the possibility of onstraining the osmi -ray diusionfeatures of the environment. Moreover, the osmi -ray distributions, their intera -tions, and a possible dete tion of the SNR environment in the neutrino hannel areanalyzed. The study presented here was published as [Torres et al. 2010.4.1 New high and very high-energy observationsRe ently, the Very Energeti Radiation Imaging Teles ope Array System (VER-ITAS) presented further observations towards IC 443 [A iari et al. 2009b. Re-garding the position of the entroid, it was found to be at (RA, DEC)J2000 =(06h 16m 51s, +22 30' 11), ± 0.03stat± 0.08sys thus, onsistent with that of MAGIC.

42 Chapter 4. The GeV to TeV onne tion in SNR IC 443Eviden e of a very-high-energy (VHE, E > 100GeV) extended gamma-ray emis-sion was also found. The extension derived was 0.16 ± 0.03stat ± 0.04sys. TheVHE spe trum was well t by a power law (dN/dE = N0 × (E/TeV)−Γ) witha photon index of 2.99 ± 0.38stat ± 0.3sys and an integral ux above 300GeV of(4.63 ± 0.90stat ± 0.93sys) × 10−12 m−2 s−1. Thus, the spe tral determination is onsistent with the MAGIC measurements; both present a steep slope, with VERI-TAS nding a slight overall in rease in the ux level. No variability of the gamma-rayemission was laimed by VERITAS either.Moreover, AGILE results on IC 443 has been re ently reported[Tavani et al. 2010. AGILE dis overed a distin t pattern of diuse emissionin the energy range 100MeV3GeV oming from the SNR, with a prominentmaximum lo alized in the Northeastern shell, displa ed (as it was the ase withEGRET) respe t to the MAGIC/VERITAS sour es. The latter are ∼0.4o apartfrom the maximum of the AGILE emission (whi h is also separated from theposition of the nearby pulsar wind nebula (PWN), as dis ussed below). Finally,Fermi has also re ently presented an analysis of its rst 11 months of observationstowards the region of SNR IC 443 [Abdo et al. 2010g. These results reinfor e thoseobtained by AGILE, given the better instrument sensitivity. Therefore, Fermi LATmeasurements will be preferably used when analyzing GeV results in this Chapter.The sour e has been dete ted in a broad range of energies, from 200MeV up to50GeV, with a SED that rolls over at about 3GeV to seemingly mat h in slopethat found at the highest energies. The spe trum an be represented, for instan e,with a broken power law with slopes of 1.93 ± 0.03 and 2.56 ± 0.11, with a breakat 3.25 ± 0.6GeV. This is one important dieren e with the previously dis ussedEGRET data. The spe tral energy distribution (SED) ould not unveil ba k thenneither that the emission would maintain a hard spe trum up to su h tens-of-GeVenergies, nor the existen e of a roll over in the spe trum at the energies found. Theux above 200MeV resulted in (28.5±0.7) ×10−8 ph m−2 s−1, thus allowing for avery signi ant dete tion in Fermi. The entroid of the emission is onsistent withthat of EGRET 3EG J0617+2238.4.1.1 Relative lo alization of sour es[Abdo et al. 2010g report that the entroid of the Fermi LAT emission is displa edmore than 5 × θerror68 (MAGIC error) from that of MAGIC (J0610+225), and morethan 1.5× θerror68 (VERITAS error) from that of the VERITAS sour e. These num-bers are obtained assuming that the systemati and statisti al errors in lo alizationadd up in quadrature, and onsidering the worse error of ea h of the pairs of mea-surements (FermiMAGIC, FermiVERITAS), whi h in both ases orrespond tothe imaging air Cherenkov teles opes (IACTs). The signi an e of the entroid sep-aration greatly improves when a) the best measured position is onsidered (i.e., theerror by Fermi LAT), for whi h both pairs of measurements are about 5σ away,and/or b) when only statisti al errors are onsidered for VERITAS (the systemati errors in this latter measurement is about a fa tor of 3 larger than the statisti s

4.2. Comparison with nominal model 43and signi antly dierent from all others; but of ourse, one an not ne essarily as-sume it to approa h the dete tion in the dire tion of the Fermi LAT sour e). Thus,albeit urrent measurements are not on lusive about energy dependent morphol-ogy, they are onsistent with it. [Abdo et al. 2010g report that the entroid of theFermi LAT emission moves towards that of the VERITAS sour e as the energy band hanges from 1 − 5GeV to 5 − 50GeV. However, the signi an e of this displa e-ment is low: only ∼ 1.5σ. It might be that the angular resolution and/or sensitivityand/or the separation of the real mole ular mass distribution on sky proje tion arenot enough to distinguish when su h nearby energy ranges are onsidered. Newmeasurements from MAGIC (using the just-obtained stereos opi apability) ouldprovide ontinuous overage from 50GeV up.4.1.2 Possible relationship between gamma-ray emission and thePWNIn all energy bands, the entroid of the orrespondingly dete ted sour es is in onsis-tent with the PWN (and the putative pulsar) CXOU J061705.3+222127, dis overedby [Olbert et al. 2001, and lying nearby. Both the 3EG and the GeV sour e in the atalogs of [Hartman et al. 1999 and [Lamb & Ma omb 1997, whi h are o-spatial,are in onsistent with the PWN lo ation. Similarly, the position of the PWN is sep-arated from Fermi LAT sour e by 0.26o, or ∼11σ away from the lo alization of theFermi LAT peak.Also at higher energies, the gamma-ray emission observed by VERITAS andMAGIC is oset from the lo ation of the PWN by 10− 20 ar min. This latter fa t ould be understood in ase the PWN is a gamma-ray emitter, like in HESS J1825-137 [Aharonian et al. 2006b or HESS J1908+063 [Aharonian et al. 2009 where sim-ilar osets were found, see also [Abdo et al. 2010h. The emission ould be onsistentwith a s enario in whi h the VHE emission arises from inverse Compton s atteringo ele trons a elerated early in the PWN's life.However, if one would assume that the PWN CXOU J061705.3+222127 is pro-du ing the emission (note that pulsed radiation from this obje t has not been foundat any frequen y), the highest energy TeV-band radiation should peak there: it ouldbe extended, but due to losses, as the energy in reases, the emission should be maxi-mum towards the PWN. Moreover, the GeV radiation should be an unresolved pulsaremission: it should also peak there and be pulsed, see [Bartko & Bednarek 2008.Therefore, Fermi/VERITAS data guarantee that the GeV and TeV emissions de-te ted do not originate in the PWN.4.2 Comparison with nominal modelFirst of all, the predi tions made in Chapter 3 are ompared to/with the mostre ent results obtained by VERITAS and Fermi. In the ase of VERITAS, giventhat their measured SED is ompatible with the earlier one obtained by MAGIC,no signi ant dieren e in the response of the models is expe ted. In the ase of

44 Chapter 4. The GeV to TeV onne tion in SNR IC 443

Figure 4.1: Earlier MAGIC and EGRET (stars and diamonds, respe tively), andre ent Fermi LAT and VERITAS (squares and upper trianges, respe tively) mea-surements of the neighborhood of IC 443 as ompared with model predi tions for animpulsive and a ontinuous a elerator, as onsidered in Chapter 3. The nominalvalues of parameters for these models are the same as in Figure 3.3, although herethe dierent ontributions are summed up. See the text for details.Fermi, the situation is dierent be ause Fermi LAT results have extended the energydomain of the SED mu h beyond what was possible for EGRET. Thus, the exploredmodels in Chapter 3 were un onstrained for that energy range.In the model presented in Chapter 3, IC 443 was onsidered both as a on-tinuous and an impulsive a elerator, as explained in Se tion 3.3.1. CRs were as-sumed to propagate with a diusion oe ient at 10GeV, e.g., D10 = 1026 m2 s−1,and δ = 0.5 in a medium of typi al density. In this medium, the times alefor nu lear loss τpp is orders of magnitude larger than the age of the a elera-tor. The nominal models also explored the possibility of having dierent dis-tan es between the SNR sho k and the intera ting louds, together with as-sumptions on the mole ular mass ae ted by the CRs. These assumptionswere based on the observations of mole ular lines towards the region IC 443made by, e.g., [Cornett et al. 1977, De Noyer & Frerking 1981, Di kman et al. 1992,Seta et al. 1998 and [Torres et al. 2003. From these studies, the overall pi tureemerges: a total mass of ∼ 1.1 × 104 M⊙ pla ed mostly at the foreground ofthe remnant, sin e the opti al and X-ray radiation appear absorbed, with smaller loud(s) lo ated loser to the SNR. However, there are several un ertainties at play:whether there is one or several foreground louds, the distan e between the fore-ground loud(s) and the SNR shell, the number and spe i lo ation(s) of the fore-ground loud(s), and their mass distribution if more than one loud is present.Fermi LAT data is expe ted to elu idate some of these parameters by a posteriori omparison with data, regarding not only the mole ular environment but also thediusion properties of the medium.Figure 4.1 shows the result of the nominal model predi tions (theoreti al urves are exa tly as in Figure 3.3, from Chapter 3, ex ept for the fa t that all ontributions are summed up), ompared with the newest data. The ele tronbremsstrahlung ontribution, visible only at the smallest energies, an hardly explain

4.3. Using Fermi LAT data to onstrain model parameters 45the whole of the observed IC 443 gamma-ray emission, a on lusion also rea hed by[Abdo et al. 2010g, and previously by [Butt et al. 2003 but with EGRET data. Inthe present model, the bremsstrahlung ontribution is onsidered for primary par-ti les with a proton to ele tron ratio of 150, following [Torres 2004 formulae. The ross se tion of bremsstrahlung and pion produ tion are similar at the Fermi LATrange. Hen e, the bremsstrahlung to pion ratio an be approximated with the ratioof CR ele tron and proton uxes (whi h is almost 0.01). The observed gamma-rayux is too high for bremsstrahlung to be the dominant pro ess, although it has likelya non-negligible ontribution below 200 MeV.The urves in Figure 4.1 are based on assuming 8000M⊙ at the dierent distan esenumerated in the plot and a few hundred M⊙ lo ated loser to the SNR (as forexample, ∼700M⊙ for the ase of an impulsive, and ∼300M⊙ for a ontinuous ase, lo ated at 3 − 4p ). What is striking to the eye is that the previous goodagreement between theory and the observations performed by EGRET and MAGIC(and obviously VERITAS), parti ularly in the ase of the impulsive a elerator,is now in disagreement with Fermi LAT data. The spe trum is harder than thatsuggested by EGRET, presenting an almost at SED up to 10GeV, with a roll-overin the spe trum between 10 and 100GeV. The models in Chapter 3 are unable toreprodu e the details of these trends. In fa t, the ase of ontinuous a elerationwas already disfavored in Chapter 3 due to both, the middle age of the remnant andthe behavior at the highest energies, whi h were produ ing a mu h harder SED thanthat observed. This ase is now denitely ruled out, and will not be onsidered anyfurther. In the ase of impulsive a eleration, it is at the earlier unexplored regionof energies, between 10 and 100GeV, where signi ant deviations between theoryand data are found. There is no model among the ones explored above whi h ana ommodate at the same time a SED that is both, su iently steep at VHEs to on ur with MAGIC/VERITAS observations and su iently at at lower energiesto on ur with Fermi LAT data.4.3 Using Fermi LAT data to onstrain model parame-tersAlthough the previous se tion seems to present a di ult-to-solve failure of thes enario, the a tual failure omes only from some numeri al values of parameters.In parti ular, dieren es in the lo ation and masses of the overtaken louds anmove the peaks of their orresponding ontributions, see [Aharonian & Atoyan 1996,Gabi i & Aharonian 2007, Rodríguez Marrero et al. 2009 and this Chapter for de-tailed analysis of the dependen es. Certainly, kinemati distan e estimations arenot a urate enough to obtain the exa t separation of the loud(s) from the SNRshell. Thus, Fermi observations hold the key to make some pre isions on the as-sumptions made in this sense, given that the unknowns an ae t the nal results onthe predi ted spe tra. Using Fermi LAT results, we nd that a loser less massivegiant loud (e.g., ∼5300 M⊙ at 10 p ) is being overtaken by CRs diusing away

46 Chapter 4. The GeV to TeV onne tion in SNR IC 443

Figure 4.2: As in Figure 4.1, summed results (right) are produ ed by two main omponents (left) oming from a giant loud in front of the SNR, whi h is at leastpartially overtaken by the diusing osmi rays (∼5300M⊙ at 10 p ) and a loser-to-the-shell loud (at 4 p , with 350M⊙), similar to the previous examples. Thedotted and dashed lines at the VHE range orresponds to dierent normalizations,whi h an also be understood as intera ting masses of ∼4000 and ∼3200M⊙ at thesame distan e. The diusion oe ient is as before, D10 = 1026 m2 s−1.from IC 443. In addition, a smaller amount of mole ular material in loud(s) loserto the SNR shell (e.g., at 4 p , with 350M⊙). The ombination of both produ ean ex ellent mat h to the whole range of observations, see Figure 4.2. A smalleramount of mass than the total quoted: 1.1 × 104M⊙ in the foreground giantmole ular loud(s) being overtaken by diusing CRs from IC 443 is perfe tly possi-ble, despite the various un ertainties in the absolute position of the loud, its realnumber, and the velo ity model used; and essentially, due to the fa t that the totalamount of mass orresponds to a larger proje ted sky area. Whereas this implies nosubstantial hange to the model, it allows a mat h with data at all high-energy fre-quen ies, as shown in Figure 4.2. In this Figure, the non-solid lines at the VHE range orresponds to dierent normalizations, whi h an also be interpreted as dierentintera ting masses, of ∼4000 and ∼3200M⊙ (dotted and dashed, respe tively), atthe same distan e. The diusion oe ient is as before, D10 = 1026 m2 s−1.4.4 Cosmi -ray distributions and their ee tsIn Figure 4.3, the distribution of CRs appears as generated by the impulsive a - elerator (IC 443) at the two dierent distan es onsidered for the mole ular massdistribution in the model of Figure 4.2 (solid bla k line). Moreover, the ratio of thesedistributions with respe t to Earth's osmi -ray distribution is also shown. The keyissue extra ted from these plots is that the osmi -ray energy density is greatlyenhan ed along the energy range of interest as ompared to that in our vi in-ity, des ribed with a spe trum of the form J⊙(E) ∼ 2.2E−2.75GeV m−2GeV−1 s−1 sr−1(e.g. [Dermer 1986. Another relevant aspe t is that signi ant deviations of the osmi -ray density are obtained when the diusion is slower (i.e., D10 is larger). Ata xed SNR age of 30 kyrs, in reasing D10 produ es a displa ement of the gamma-

4.4. Cosmi -ray distributions and their ee ts 47

Figure 4.3: Cosmi -ray spe trum generated by the impulsive a elerator (IC 443)at the two dierent loud distan es onsidered in Figure 4.2: 10 (solid) and 4 p (dashed), at the age of the SNR, as a fun tion of energy. Dierent olors showresults for dierent diusion oe ients (bla k, D10 = 1026 m2 s−1; and red, D10 =

1027 m2 s−1). The right panel shows the ratio between the osmi -ray spe tra ofthe left panel, and the osmi -ray spe trum near Earth, as a fun tion of energy.

Figure 4.4: Example of a model output with D10 = 1027 m2 s−1. The dierent urves represent results for the lo ation of the giant mole ular loud at 10, 15,20, 25, 30 p from the SNR shell, whereas the lose-to-the-SNR loud is at 4 p .Neither in this nor in any other of the studied models, the VHE sour e spe trum an be reprodu ed by varying the parameters with su h a diusion oe ient s ale.Furthermore, the resulting SED in the Fermi LAT range is not hard enough tomat h the data.

48 Chapter 4. The GeV to TeV onne tion in SNR IC 443

Figure 4.5: Contour plot depi ting the position of the peak of the SED generatedby a 30 kyrs old inje tion intera ting with louds at dierent distan es, for a rangeof diusion oe ient s ale, D10.ray emission predi tion to smaller energies, typi ally, up to D10 > Dtransition, wherepeaks generated by louds at large separation (e.g, 100 p ) shift up and peaks gen-erated by louds at smaller separation (e.g., 10 p ) shift down in the SED (e.g.,[Rodríguez Marrero et al. 2009 and referen es therein). This fa t implies that, forthe range of distan es onsidered for the giant and lose-to-the-SNR mole ular louds(10 30 p and 2 6 p , respe tively), there is no solution with large D10 able tot the whole range of data. This was already suggested in Chapter 3, where it waspossible to set a strong onstraint over the diusion times ale, using MAGIC dataonly. D10 should be of the order of 1026 m2 s−1. If the separation between the giant loud and the SNR is greater than 10 p , a slower diusion would not allow su ienthigh energy parti les to rea h the target material and it would be impossible to re-produ e the VHE data. On the other hand, the separation between the foreground loud(s) and the SNR shell an not be mu h smaller than 10 p , given that there isa displa ement between the entroid positions of EGRET/Fermi and VHE sour esand that mole ular material is absorbing lower frequen y emission from the rem-nant. The urrent Fermi LAT data emphasizes this on lusion. Figure 4.4 showsan example of a full range of models onstru ted with D10 = 1027 m2 s−1, and itsdisagreement with data. Note that even if the urves are res aled and assuming,for instan e, a mu h higher mole ular mass (whi h would itself be in oni t withmulti-frequen y observations), it is not possible to obtain a good t a ross the wholerange of observations.Figure 4.5 presents a ontour plot of the energy at whi h the maximum of theSED is found for the ases of impulsive a eleration of CRs. The age orresponds to aSNR like IC 443, and the CRs are intera ting with louds at dierent distan es, for arange of diusion oe ient s ale, D10, from 1025 to 1027 m2 s−1. This plot is useful

4.5. Degenera ies and un ertainties 49

Figure 4.6: Examples of solutions around the dis ussed main values, exploring thedegenera ies (or un ertainties) in determining the numeri al values of model param-eters mat hing the observational data. The order of the panels in this plot, top tobottom and left to right, orresponds with the parameters des ribed in Table 4.5, be-ing ea h olumn one of the three groups therein. Adapted from [Torres et al. 2010.to understand whi h is the favored solution and its degree of un ertainty/degenera y.In order to t the ombined MAGIC/VERITAS and Fermi LAT data, a giant loudis needed to produ e a peak at about the Fermi spe tral turnover. This implies that,by looking at the plot, there is either a very large separation between the loud andthe SNR shell for a high D10, or a smaller distan e with a faster diusion, i.e. alower D10. The rst option is dis arded be ause it is not possible to t the VHEdata in this onguration, whereas the latter is the favored solution in this study.4.5 Degenera ies and un ertaintiesFigure 4.6 explores the range of parameters around the solutions mat hing the ob-servational data. The aim is to show the degenera ies (or un ertainties) within

50 Chapter 4. The GeV to TeV onne tion in SNR IC 443whi h this model provides a reasonable agreement with observations. Given thedistan es to ea h of the louds, the values of masses and diusion oe ients usedin Figure 4.6 to obtain good data-mat hing are displayed in Table 4.5. Fits ould be onsidered good for distan es to the giant mole ular loud, DGMC , between 9 and11 p . For an average distan e of 10 p , the mass in the lose-to-the-SNR loud (or louds) de reases the farther the loud is, as shown in the se ond olumn of Figure4.6. For these ases, good solutions with DGMC between 9 and 11 p (i.e., 10 p ) analways be found, adjusting other parameters, an example of whi h appears in these ond olumn of Figure 4.6. The average model explored in Figure 4.2 orrespondsto the model 3 in Table 4.5 with: DGMC = 10p , dsnr = 4p , and the three urves onstru ted with ∼5300, ∼4000, ∼3200M⊙, and Msnr = 350M⊙.The results in the third olumn of Figure 4.6 and bottom group in Table 4.5 showthat the smaller the diusion oe ient, the worse is the t at VHEs, overpredi tingthe data. Corre ting this via a mass adjustment would in turn make for a poor tat lower energies. Thanks to this, a lower limit to the diusion oe ient D10 anbe imposed. On the other hand, when D10 in reases, the VHE spe tra is qui klyunderpredi ted, and again, orre ting this via a mass adjustment would not providea good t at lower energies.In summary, in order for this model to mat h the multi-frequen y observationaldata, the range of variation in the parameters be omes onstrained as 9 . DGMC .

11p ; 3 . dsnr . 6p , and D10 ∼ 1026 m2 s−1. Therefore, these parameters onstitute a dire t estimation of D10 and the mole ular environment in the IC 443vi inity, under the assumed validity of this model.As mentioned before, the gamma-ray emissivity was assumed to be onstantwithin the louds; i.e., there is no signi ant osmi -ray gradient in the target. Thisassumption is an approximation, and it improves whenever the size of the loud isless than the distan e to the a elerator and the diusion oe ients inside andoutside the loud are not signi antly dierent (or even if they are, the proton-proton times ale is larger than the time it takes for osmi rays to overtake thewhole loud). In the ase of IC 443, these onditions an be a ommodated for thesolutions in Table 4.5, ex ept perhaps for the very massive loud lo ated lose tothe SNR at dsnr=2p . For this massive lose-to-the-SNR loud, it would imply anaverage density higher than that usually found, although even this might also bepossible given the small s ale lumps found therein, see e.g., [Rosado et al. 2007.4.5.1 Inuen e of the δ-parameterIn addition, other parameters of the model are explored, parti ularly those inu-en ing the way in whi h the diusion oe ient varies with energy (the parameterδ), or the inje tion spe trum of osmi rays (referred to as α). Their orrespondingvalues appear to be rather onstrained. For instan e, the δ parameter is expe ted tobe around δ = 0.4 − 0.7, e.g., [Berezinskii et al. 1990, and a typi al value of 0.5 isusually assumed. Figure 4.7 gives a ount on how small variations in δ hange theslope of good-tting solutions to the high and very-high energy data. One an see

4.5. Degenera ies and un ertainties 51Table 4.1: Main model parameters for solutions shown in Figure 4.6, ex ept model3, whi h instead is shown in Figure 4.2. DGMC and dsnr are the distan e to theGMC and the loser-to-the-SNR mole ular louds. The three quoted f values deneMGMC = (1/f) 8000M⊙. The three groups explore dierent degenera ies: in theposition of the GMC, in the position of the smaller loud, and on the diusion oe ient. Model DGMC dsnr D10 f Msnrp p m2 s−1 . . . M⊙1 8 4 1026 3.0 4.0 5.0 3502 9 4 1026 1.8 2.5 3.2 3503 10 4 1026 1.5 2.0 2.5 3504 11 4 1026 1.0 1.4 1.8 3505 12 4 1026 0.7 1.0 1.3 3506 10 2 1026 1.5 2.0 2.5 17507 10 3 1026 1.5 2.0 2.5 5808 10 5 1026 1.5 2.0 2.5 2509 10 6 1026 1.5 2.0 2.5 19510 10 4 8 ×1025 1.5 2.0 2.5 35011 10 4 9 ×1025 1.5 2.0 2.5 35012 10 4 2 ×1026 1.5 2.0 2.5 35013 10 4 3 ×1026 1.5 2.0 2.5 350that for steeper δ-parameters, steeper gamma-ray spe tra are found. If the massesof the mole ular louds are maintained and δ is larger, in order to have a good tone would need an even lower D10 (lower than D10 = 1026 m2 s−1, see Figure 4.6),obtaining a less plausible solution.4.5.2 Un ertainties due to the ross se tion parameterizationAnother he k is performed to nd out whether hanges in the ross se tion pa-rameterization an produ e signi ant varian e in the results. In the appendixof [Domingo-Santamaría & Torres 2005, the dierent predi ted yields in gamma-rays obtained when using alternate ross se tion parameterizations (known ba kthen) were ompared among themselves and with the data. The onsidered pa-rameterizations therein were: the one introdu ed by [Kamae et al. 2005, the δ-fun tional form by [Aharonian & Atoyan 1996 that is used above, the one by[Stephens & Badhwar 1981, and [Blattnig et al. 2000a, Blattnig et al. 2000b. Itwas found that Kamae's and the δ-fun tional form mat hed quite well. Figure 11 ofthat study showed this by plotting the gamma-ray emissivities obtained with the or-responding use of ea h of the parameterizations of the ross se tion. In addition, itwas also found that neither the parameterizations from [Stephens & Badhwar 1981nor from [Blattnig et al. 2000a, Blattnig et al. 2000b were appropriate for their usein broad-band high-energy modeling su h as the one pursued here. More re ently,

52 Chapter 4. The GeV to TeV onne tion in SNR IC 443

Figure 4.7: Comparing gamma-ray yields with dierent δ parameters, from left toright δ = 0.4, 0.5, 0.6, and 0.7. The other parameters are as in Figure 4.2.[Kelner et al. 2006 presented a new approa h for obtaining the ross se tion in ppintera tions. These authors used two shapes for representing the ross se tion, sep-arated in energy. At low energies (up to 100GeV), Kelner et al. approa h uses aslightly modied but similarly-shaped δ-fun tional form. At higher energies, the ap-proa h is dierent, and presents an analyti al shape that ts the results of the simu-lations of the energy distribution of π mesons by the SYBILL ode. Figure 4.8 showsa omparison of the ross se tion parameterizations used in the previous gures, theδ-fun tional form by [Aharonian & Atoyan 1996, with that of [Kelner et al. 2006.Dieren es are within 20%, with the δ-fun tional approximation being larger.Hen e, the on omitant hange in ux predi tions, due only to dierent rossse tion parameterizations, an be reabsorbed as part of the un ertainty in the de-termination of the model. Figure 4.9 shows the impa t of this in two ways. Theleft panel shows several alternatives for the position of the large (TeV-produ ing) loud in the model, lo ated at 10, 15, and 20 p . Here, the mass of this loud ismaintained xed, and only the shape of the urves for dierent distan es is impor-tant. It is lear that the hange in ross se tion parameterization does not enableany of the previous models to be feasible. Also, it singles out a distan e of about10 p from the SNR shell to the giant TeV-produ ing loud for obtaining a good t.The right panel assumes this distan e of 10 p and explores the un ertainty in thedetermination of the giant loud mass. The parameters therein shown are 4 p and350M⊙ for the lose-to-the-remnant loud (the same as in the left panel), and 10 p and 7272, 5333, 4210M⊙ for the TeV-produ ing giant loud.

4.5. Degenera ies and un ertainties 53

Figure 4.8: Comparison (left) and ratio (right) of the ross se tion parameterizationsused in the previous Figures, the δ-fun tional form by [Aharonian & Atoyan 1996,with that of [Kelner et al. 2006.

Figure 4.9: Left: Gamma-ray ux results using the [Kelner et al. 2006 approxima-tion for dierent distan es from the shell to the TeV-produ ing loud, 10 (solid),15 (dotted) and 20 (dashed) p . The lose-to-the-remnant loud is xed at 4 p and ontains 350M⊙ hanges in this latter value do not improve the overall t.Right: Gamma-ray ux results using the parameterization from [Kelner et al. 2006for dierent values of the giant loud mass pla ed at a xed distan e of 10 p (seetext for details).

54 Chapter 4. The GeV to TeV onne tion in SNR IC 443

Figure 4.10: Ele trons (ele trons and positrons are shown together), photons andtwo avors of neutrinos produ ed within the louds onsidered nearby IC 443, usinga set of parameters shown in Figure 4.9 (right panel) with mass of the giant loudequal to 7272M⊙. The νµ and νe neutrino urves show both the parti le and theanti-parti le ux. Data should only be ompared with the photon urve.4.6 Computation of se ondaries other than photonsUsing the parameterization from [Kelner et al. 2006, se ondary parti les otherthan photons an be readily omputed, and this is shown in Figure 4.10.[Gabi i et al. 2009 showed, by testing a wide range of parameters, that se ondaryele trons produ ed within louds an es ape without being ae ted by signi antlosses. In other words, that the propagation time through the loud for osmi -rayele trons is shorter than the energy loss time for parti le energies between ∼ 100MeVand few hundreds of TeV. Thus, there would be little ee t of the se ondary ele -trons produ ed on the non-thermal emission from the loud. In addition, for typi aldensities of louds, in the several hundred to several thousand parti les per m3, thedominant energy loss from ∼ 100MeV and ∼ 10TeV would be bremsstrahlung andnot syn hrotron.Nonetheless, a on lusive proof of the hadroni nature of the gamma-ray emis-sion ould ome from the dete tion of neutrinos, ν. Neutrino teles opes sear h forup-going muons produ ed deep inside the Earth, and are mainly sensitive to thein oming ux of muoni neutrinos, νµ, and their antiparti les, νµ. The nishedI eCube experiment, for example, will onsist on 4800 photomultipliers, arrangedon 80 strings pla ed at depths between 1400 and 2400m underneath the South Polei e, e.g., [Halzen 2006. The strings will be lo ated in a regular spa e grid overinga surfa e area of 1 km2. Ea h string will have 60 opti al modules (OM) spa ed17m apart. The number of OMs whi h have seen at least one photon is alled the hannel multipli ity, Nch. Note that ea h photon omes from Cherenkov radiation

4.7. Con luding remarks 55that has been produ ed by the muon whi h, in turn, results from the intera tionof the in oming ν inside the earth and i e rust. The multipli ity threshold is setto Nch = 10, whi h orresponds to an energy threshold of 200GeV. The angularresolution of I eCube will be around ∼ 0.7.A rst estimation of the event rate of the atmospheri ν-ba kground that willbe dete ted in the sear h bin an be obtained, e.g., [An hordoqui et al. 2003, as:dN

dt

∣

∣

∣

∣

B

= Aeff

∫

dEνdΦB

dEνPν→µ(Eν) ∆Ω , (4.1)where Aeff is the ee tive area of the dete tor, ∆Ω ≈ 1.5 × 10−4 sr is the angularsize of the sear h bin, and dΦB/dEν . 0.2 (Eν/GeV)−3.21GeV−1 m−2 s−1 sr−1 is the

νµ+ νµ atmospheri ν-ux [Volkova 1980, Lipari 1993. Here, Pν→µ(Eν) denotes theprobability that a ν of energy Eν on a traje tory through the dete tor, produ es amuon. For Eν ∼ 1−103 GeV, this probability is ≈ 3.3×10−13 (Eν/GeV)2.2, whereasfor Eν > 1TeV, Pν→µ(Eν) ≈ 1.3 × 10−6 (Eν/TeV)0.8 [Gaisser et al. 1995. On theother hand, the ν-signal is similarly obtained as

dN

dt

∣

∣

∣

∣

S

= Aeff

∫

dEν (Fνµ + Fνµ)Pν→µ(Eν) , (4.2)where (Fνµ + Fνµ) is the in oming νµ-ux. In the previous integrals, both expres-sions for Pν→µ(Eν) are used a ording to the energy, and integrated from 200GeVup to 10TeV. The ee t of ν-os illations is taken into a ount following table 2 of[Cavasinni et al. 2006, where the os illation probability in the average va uum os- illation hypothesis is given. The inter- onversion probability between avours andbetween anti-avours is assumed to be the same.As an ee t of os illations, the avour omposition of all the expe ted uxesfor ea h avour are within 50% of ea h other. Using the former equations andthe se ondary omputation shown in Figure 4.10, the number of muon neutrinosignal events is found to be 0.6 per year of observation. This is still signi antlybelow the estimation of the number of ba kground events, whi h under the previousprovisions is 6.4 along the same period with the full I eCube array. If only eventsabove 1TeV are onsidered, the expe ted signal is 0.25 yr−1, and the omputedba kground is 1.92 yr−1. I eCube does not seem to be able to distinguish this signalin reasonable integration times, at least within the rea h of this simplied treatmentof the dete tor.4.7 Con luding remarksThe re ent observations of the IC 443 environment made by AGILE, Fermi LAT,and VERITAS at the GeV and TeV energies, are spe trally onsistent with theinterpretation of osmi -ray intera tions with a giant mole ular loud lying in frontof the remnant. This s enario would be produ ing no signi ant ounterpart at lowerenergies at that lo ation, and thus would be leading to a natural interpretation of

56 Chapter 4. The GeV to TeV onne tion in SNR IC 443the displa ement between the entroids of the dete tions at the dierent energybands. Using the latest data, the diusion hara teristi s in this environment wereestimated, within the assumed validity and framework of this model. The resultsshow that the diusion oe ient is lower and the osmi -ray density is higherthan the Earth-values of both magnitudes. Un ertainties in the amount and thelo alization of the target mole ular mass still remains. Just as well, the densityat whi h this mole ular material is found remains unknown. For instan e, theun ertainty in the osmi -ray-overtaken mass dis ussed in the previous se tions isabout 100% for mat hing models at the extremes of this parameter. But evenallowing for su h a wide range, the model ould a ommodate some but not allvariations in other parameters, with the values of D10 and distan es from the SNRshell seemingly being solid onstraints.

Chapter 5Starburst galaxiesContents5.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 575.2 Theoreti al model . . . . . . . . . . . . . . . . . . . . . . . . . 585.3 M 82 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 625.3.1 Comparison with previous studies . . . . . . . . . . . . . . . 635.3.2 Results and Dis ussion . . . . . . . . . . . . . . . . . . . . . . 645.4 Dis overy of HE and VHE emission from starbursts . . . . 745.4.1 Gamma-ray emission dete ted from M82 . . . . . . . . . . . . 745.4.2 NGC 253, onfronted with the model . . . . . . . . . . . . . . 745.5 Con luding remarks . . . . . . . . . . . . . . . . . . . . . . . . 775.1 Introdu tionStarburst galaxies have been anti ipated as gamma-raysour es [Paglione et al. 1996, Blom et al. 1999, Torres 2004, Torres et al. 2004,Domingo-Santamaría & Torres 2005, Persi et al. 2008, provided su ient in-strumental sensitivity. Su h galaxies have an enhan ement both in the starformation and supernova (SN) explosion rate, and dense (gas and dust enri hed)environments. These hara teristi s suggest that they should emit gamma rays,and their luminosity would be greater than normal galaxies. SN remnants (SNR)and sho k winds from massive stars are supposed to a elerate osmi rays (CR).Very energeti gamma rays are produ ed due to CRs ollisions with ambient nu leiand subsequent π0 de ay. Those gamma rays an be dete ted both with spa e-bornand ground-based imaging atmospheri Cherenkov teles opes.With the arrival of the Fermi satellite, and the se ond-generation of Cherenkovteles opes (H.E.S.S., VERITAS, MAGIC) at their full potential, it is importantto have the most detailed model for these astronomi al obje ts to obtain a properfeedba k from observations and improve our knowledge of the osmi -ray populationand the physi al environment of nearby starbursts. Furthermore, with the re enton-going dis ussions [Loeb & Waxman 2006, Ste ker 2007, An hordoqui et al. 2008about the starbursts ontribution to the neutrino ba kground, it is important tohave onsistent and detailed models of neutrino produ tion in spe i galaxies. The

58 Chapter 5. Starburst galaxiesassumptions made in previous studies are disse ted, while observational dete tabilitywith, e.g., km3-observatories is onsidered.Re ently, the dete tion of M82 was reported by the VERITAS ollaboration,while gamma rays above hundreds of GeVs oming from the fainter NGC 253 were laimed by the H.E.S.S. ollaboration. During the same epo h, the Fermi exper-iment presented the data olle ted at lower energies oming from the dire tion ofboth starburst galaxies [Abdo et al. 2010 . Minor and reasonable variations in theparameter spa e of already published models [Domingo-Santamaría & Torres 2005,de Cea del Pozo et al. 2009b an fully a ount for the high and very high energyemission oming from both galaxies, while agreeing with previous data dete ted fromradio to infrared (IR). This Chapter presents the model that, prior to the dete -tions, predi ted the gamma-ray emission from M82 [de Cea del Pozo et al. 2009b,and the possible parameter variations to better a knowledge those dete tions[de Cea del Pozo et al. 2009a.5.2 Theoreti al modelThe aim of this study is to obtain multi-frequen y/multi-messenger predi tions forthe photon and neutrino emission oming from the entral region of M82, where theinner starburst is lo ated. The model seeked is intended to be onsistent, with allthe dierent omponents of the emission from radio to TeV photons and neutrinos tra ked to one and the same original osmi -ray population. This population isa onsequen e of all ele tromagneti and hadroni hannels from the primary andsubsequently-produ ed se ondary parti les. With this model at hand, a range ofun ertainties in the parameters an be explored, orrespondingly shifting upwardsor downwards the high energy end of the spe trum.In order to perform su h study, the ode Q-diffuse [Torres 2004 has beenadapted in several ways. Q-diffuse solves the diusion-loss equation for ele tronsand protons and nds the steady state distribution for these parti les subje t toa omplete set of losses in the interstellar medium (ISM). Subsequently, it om-putes se ondary parti les from hadroni intera tions (neutral and harged pions)and Coulomb pro esses (ele trons), and gives a ount of the radiation or de ayprodu ts that these parti les produ e. Se ondaries (photons, muons, neutrinos, ele -trons, and positrons) that are in turn produ ed by pion de ay are also al ulated.Additional pie es of the ode provide the dust emissivity, and the IR-FIR photondensity, whi h is onsistently used both as target for inverse Compton s atteringand for modeling the radiation at lower frequen ies. For radio photons, syn hrotronradiation, free-free emission, and absorption are omputed using the steady distri-bution of ele trons. Finally, and using the radiation transport equation, opa itiesto γγ and γZ pro esses, as well as absorbed gamma-ray uxes, are al ulated. Theimplementation of Q-diffuse in ludes several upgrades. Some are stri tly te h-ni al, allowing for a more versatile and automati input-output intera tion. Someare physi al: neutrino-produ tion subroutines have been developed for all neutrino

5.2. Theoreti al model 59 hannels in the de ay of positively and negatively harged pions and a re ent param-eterization of pp intera tions by [Kelner et al. 2006 has been used. The ow of theQ-diffuse ode is shown in Figure 5.1. Several loops an be distinguished there:they are used in order to determine a self- onsistent set of parameters, su h that allpredi tions for the dierent bands of the ele tromagneti spe trum are onsistentwith the data.Previous studies of diuse high energy emission, and of ele tron and positronprodu tion, with dierent levels of detail and aims, go ba k to the early yearsof gamma-ray astronomy. A summary of these rst eorts is summarizedby [Fazio 1967, Ginzburg & Syrovatskii 1964 and [Ramaty & Lingenfelter 1966,then followed by [Maras hi et al. 1968, Ste ker 1977, among many others. Se -ondary parti le omputations have a similarly long history see, e.g., [Ste ker 1970,Orth & Bungton 1976. Relatively more re ent eorts in lude [Drury et al. 1994,Moskalenko & Strong 1998, Strong & Moskalenko 1998, Marko et al. 1999 and[Fatuzzo & Melia 2003. The general ideas followed in those works were alreadyused by [Paglione et al. 1996 and [Blom et al. 1999 when modeling nearby star-bursts galaxies. On the other hand, the present study losely tra ks ideas from[Brown & Mars her 1977 and [Mars her & Brown 1978, regarding their studies of lose mole ular louds.An important aspe t of this approa h is that it la ks detail in modeling singlesour es. Isolated SN explosions, or SNRs, or even the impli ation of those SNexploding within the wind bubbles of their progenitors and the on urrent losses inthese environments are not treated by themselves. On the ontrary, the emphasislies on average properties of the whole starburst region, seen as a CR inje tor.Mixing these two aims would require an eort that is beyond urrent observational apabilities of what is known in starburst galaxies, as well as beyond what is possiblefor this work. Essentially, to make quantitative estimates in su h small-s ale detail,one would need to understand the features of ea h and all the SN explosions inthese kind of galaxies, and model them individually and a urately. From thatpoint (i.e., from a sta king of a sour e-by-sour e modeling), an averaged s enario ould be built. This has not been a hieved even for our own Galaxy. Similarly,the present approa h to the modeling of the super-wind is also simple, in order toen ompass it with the rest of the omponents of the model. See [Everett et al. 2008for a more detailed analysis of the properties of the Gala ti super wind, and also[Gallagher & Smith 2005.The assumption of a uniform distribution of a elerators in the inner region ofthe starburst justies having a steady state distribution therein. Also, this dis-tribution an be omputed via solving the full diusion-loss equation. Parti leswith energies of 100TeV have a range omparable to or greater than the size of theGalaxy, whereas the distan es travelled by hadroni parti les inje ted with energiesof the order of tens of GeV do not ex eed 1 − 2 kp . For ele trons, the situation iseven more ompli ated (see later in this se tion), due to the more signi ant lossesthey suer. TeV-energy ele trons ool down so qui kly due to syn hrotron and in-verse Compton losses that they annot diuse beyond few hundred parse s from

60 Chapter 5. Starburst galaxies

EM

Dat

a fit

ting

initi

al in

ject

ion

prot

on s

lope

, Np/N

e

IC target photon field

photon field for IC losses

ionization, production, diffusion, convection

Dat

a fit

ting

e.g.

, mag

netic

fiel

d B

CR injection spectrum (SN rate, eff ...)

steady proton spectrum: Np(Ep)

primary protons: Qinjp(Ep) primary electrons: Qinj

p(Ep) Np/N

e

- Losses and confinement timescales from

steady electron/positron spectrum: Ne(Ee)

secondary electrons/positrons: Qsece(Ee)

DIFFUSION-LOSS Equation

inverse Compton, adiabatic, diffusion, convection- Losses and confinement timescales from ionization, synchrotron, bremsstrahlungDIFFUSION-LOSS Equation

+

Emission: Bremsstrahlung, IC Emission: Synchrotron + Absorption: Free-free

-ray spectrumradio spectrum

hadronic interactions with ISM: , , production knock-on: e- production

Emission: 0 decay +, decay: e+, e- production

Absorption processes

-ray spectrum-ray spectrum

Dust Emission

-ray spectrumIR spectrum

charged pion decay:

production

spectrum

Figure 5.1: Code ow of Q-diffuse, adapted from[Torres & Domingo-Santamaría 2005 to take into a ount new updates devel-oped for this work.their inje tion site, for typi al values of the diusion oe ient and of the inter-stellar magneti /radiation elds. Nonetheless, the limited size and ompa tness ofthe starburst region is ompatible with a steady state ele tron distribution in that entral region.The steady-state parti le distribution is omputed as the result of an inje tiondistribution being subje t to losses and se ondary produ tion in the ISM. At su- iently high energies, the inje tion proton emissivity is assumed to have the followingform as a fun tion of proton kineti energies:Qinj(Ep, kin) = K

(

Ep, kin

GeV

)−p

exp

(

−Ep, kin

Ep, cut

) (5.1)where p is a power-law index, K is a normalization onstant, Ep, cut is an energy uto in the a elerated parti les that are inje ted (assumed as 100 TeV, with arange explored below). Units are [Q]= GeV−1 m−3 s−1. To get a numeri al initialvalue for the normalization, K, its value is assumed to ome from the total powertransferred by supernovae into CRs kineti energy within a given volume. Thus, thesupernova rate is essential to x the level of the osmi -ray sea:∫ Ep, kin,max

Ep, kin,min

Qinj(Ep, kin)Ep, kindEp, kin ≡ηPR

V. (5.2)

R is the rate of supernova explosions, V orresponds to its volume, and η is thetransferred fra tion of the supernova explosion power (P ∼ 1051 erg) into CRs. The

5.2. Theoreti al model 61average rate of power transfer is assumed to be 10%. The assumption of the p valuein 5.1 is not a-priori, but rather an a posteriori hoi e. That hoi e is made inorder to get a slightly better t to all the multi-wavelength data, and a ts as anaverage des ription of the inje ted osmi -ray sea in the star forming region, wheremultiple nearby sho ks ould ontribute.As noted by [Domingo-Santamaría & Torres 2005, the distribution of osmi rays is probably atter at low energies, e.g., it would be given by equation (6) of[Bell 1978, orrespondingly normalized. Negle ting this dieren e at low energyhas been numeri ally veried to not produ e any important hange in the ompu-tation of se ondaries, and espe ially on gamma-rays at the energies of interest. Asin [Berezhko et al. 2006, ele trons are assumed to be inje ted into the a elerationpro ess at the sho k fronts. The ele tron inje tion rate is hosen su h that theele tron-proton ratio is a onstant to be determined from the syn hrotron obser-vations. The diusive transport equation takes are of the hanges produ ed afterinje tion onto this distribution.The diusion-loss equation (see, e.g.,[Ginzburg & Syrovatskii 1964 p. 296;[Longair 1994, p. 279) is given by:−D 2 N(E) +

N(E)

τ(E)−

d

dE[b(E)N(E)] −Q(E) = −

∂N(E)

∂t(5.3)In this equation, D is the s alar diusion oe ient, Q(E) represents the sour eterm appropriate to the produ tion of parti les with energy E, τ(E) stands for the onnement times ale, N(E) is the distribution of parti les with energies in therange E and E + dE per unit volume, and b(E) = − (dE/dt) is the rate of lossof energy. The fun tions b(E), τ(E), and Q(E) depend on the kind of parti les onsidered. In the steady state, ∂N(E)/∂t = 0, and when the spatial dependen e is onsidered to be irrelevant: D 2 N(E) = 0. Then, the diusion-loss equation anbe numeri ally solved by using the orresponding Green fun tion, G, su h that forany given sour e fun tion or emissivity, Q(E), the solution is:

N(E) =

∫ Emax

EdE′Q(E′)G(E,E′) (5.4)The onnement times ale will take into a ount that parti les an be diusingaway, arried away by the olle tive ee t of stellar winds and supernovae, or ae tedby pion produ tion (for CRs). Pion losses are atastrophi , sin e the inelasti ity ofthe ollision is about 50%. They produ e a loss times ale in the form of:

τ−1pp =

(dE/dt)pion

E(5.5)(see, e.g., [Mannheim & S hli keiser 1994. Thus, in general, for energies higherthan the pion produ tion threshold:

τ−1(E) = τ−1D + τc

−1 + τ−1pp . (5.6)

62 Chapter 5. Starburst galaxiesThe onve tive times ale, τc, is ∼ R/V , where V is the olle tive wind velo ity.[Stri kland et al. 1997 found a value of 600 km s−1 for the wind velo ity, and thisvalue is used below. However, the response of the model to an arbitrary doublingor halving of this value has been studied. The un ertainty introdu ed by varyingthe velo ity is smaller than that produ ed by other parameters. Finally, assuminga homogeneous distribution of a elerators in the entral hundreds of parse s of thestarburst galaxy, the hara teristi es ape time orresponds to the homogeneousdiusion model value ([Berezinskii et al. 1990, p. 50-52 and 78):τD =

R2

2D(E)=

τ0β(E/GeV)µ

(5.7)where β is the velo ity of the parti le in units of c, R is the spatial extent ofthe region from where parti les diuse away, and D(E) is the energy-dependentdiusion oe ient, whose dependen e is assumed ∝ Eµ with µ ∼ 0.5, and τ0 is the hara teristi diusive es ape time at ∼ 1GeV. The results ontained in this Chapteruse τ0 = 10Myr, but a range was explored to judge un ertainties. For ele trons, thetotal rate of energy loss onsidered is given by the sum of that involving ionization,inverse Compton s attering, adiabati , bremsstrahlung, and syn hrotron radiation.The Klein-Nishina ross se tion is used.5.3 M 82M82 is a near starburst galaxy. It an be seen nearly edge-on (77,[Mayya et al. 2005) and has a gas ontent mostly on entrated in the inner 2 kp .This galaxy presents a high luminosity both in the far infrared and X-ray domain:1044 erg s−1 and 1040 erg s−1 respe tively, e.g., [Ranalli et al. 2008.As part of the M81 group, M82 shows hints of an en ounter with some of its mem-bers 1Gyr ago, spe i ally with the dwarf galaxy NGC 3077 and the former M81,materialized in an intergala ti gas bridge of 20 kp . Up to 10 kp above the planeof M82, gala ti superwinds an be dete ted. Following HI streamers, the externalpart of the disk (>5 kp ) presents a warped form. On the other hand, the innerpart (300 p ) harbours a starburst. Around this region, a mole ular ring of 400 p radius and a near-IR bar of ∼ 1 kp length an be dete ted [Teles o et al. 1991.From the tips of this bar, two symmetri al arms emerge [Mayya et al. 2005. Thisled to a hange in the morphologi al lassi ation. At rst, an irregular shape wasassumed due to opti al appearan e: bright star-forming knots interspersed by dustylaments [O'Connell & Mangano 1978. Nowadays, it seems more likely that M82is a SB galaxy. [Freedman et al. 1994 established a distan e for M81 of 3.63 ±0.34Mp , thanks to the dis overy of new Cepheids by the Hubble Spa e Teles ope(HST) in this galaxy. A few years later, [Sakai & Madore 1999 found a distan eof 3.9 (±0.3)random(±0.3)systematic Mp for M82, based on the dete tion of the redgiant bran h stars using HST photometry. Both distan es are onsistent withinerrors, the latter is the one used in this work. The starburst region is lo ated in

5.3. M 82 63the inner part of the galaxy, with a radius of ∼300 p , height of ∼200 p , e.g., see[Völk et al. 1996 and referen es therein, and [Mayya et al. 2006. Meanwhile, therest of the emission extends to a thin disk up to 7 kp , see e.g., [Persi et al. 2008,and referen es therein.A signi ant amount of mole ular material has been found in the entral region,where most of the starburst a tivity is lo ated. A dis ussion on the supernova explo-sion rate an be found in the se tion 5.3.2. The ontent of the gas mass in the wholegalaxy range from 2.9×109M⊙ (HI) by [Crut her et al. 1978 and 2.9×108M⊙ (H2)by [Young & S oville 1984 to the more re ent results found by [Casasola et al. 20047×108M⊙ (HI) and 1.8×109M⊙ (H2), where dierent assumptions for distan e andthe normalization of luminosity were made. [Weiÿ et al. 2001 report for the star-burst region around 2× 108M⊙ (H2), using separate methods for the determinationof CO and H2 densities. This value essentially agrees with those estimated from 450µm dust ontinuum measurements [Smith et al. 1991 and from CO(2 → 1) inten-sities [Wild et al. 1992, and is, therefore, the one used in the determination of theuniform density for the model of this Chapter (∼ 180 m−3).5.3.1 Comparison with previous studiesA natural omparison an be made between this approa h and previous studies of the entral starburst region of M82, in parti ular, [Akyüz et al. 1991, Völk et al. 1996,Paglione et al. 1996, and [Persi et al. 2008.The earlier work made by [Akyüz et al. 1991 fo used on gamma-ray emissionat the GeV regime (integral ux above 100MeV) and basi ally provides a simple or-der of magnitude estimation with data available at that time. [Völk et al. 1996also did not provide a multi-frequen y model, but rather onsider only neutralpion de ay from protons to obtain an order of magnitude estimation of the uxat high and very-high gamma-rays. This approa h was also taken (with mi-nor dieren es) by [Pavlidou & Fields 2001 and [Torres et al. 2004. These stud-ies do not in lude a omputation of se ondaries and their emission at low ener-gies. Their aim is dierent than the one in the present work, whi h is provid-ing a onsistent model along the entire ele tromagneti spe trum. The work by[Paglione et al. 1996 is more omplex, and it was dis ussed in detail, in parti -ular, when ompared with results by Q-diffuse on Arp 220 and NGC 253, by[Torres 2004 and [Domingo-Santamaría & Torres 2005 respe tively. There are sev-eral dierent physi al and methodologi al onsiderations embedded in their model ompared to those assumed here. Moreover, high-energy (TeV) predi tions are notprovided. None of these works onsider neutrino emission.The most re ent study on M82, from [Persi et al. 2008, also diers from thepresent work in some key parameters and, most importantly, in the method andassumptions for the modeling. Among the physi al parameters, whi h values aregiven by Persi et al., the distan e (3.6Mp , whi h is a tually the one to M81), thedust emissivity index (σ = 1), the proton to ele tron primary ratio (Np/Ne = 200),the magneti eld (180µG) and the slope of primary inje tion spe trum (2.4) are

64 Chapter 5. Starburst galaxiesall dierent from the ones used or derived in the present Chapter. The dieren ebetween these (latter) values in [Persi et al. 2008 omes from the method usedto x them. For instan e, Persi et al. obtained the magneti eld by assumingequipartition with primary parti les, while the magneti eld in this work is found bya multi-frequen y analysis that also takes into a ount produ ed se ondary parti les.As Figure 5.2 (left panel) shows, the steady population of ele trons that would resultafter solving the diusion loss equation, inje ting only primary and only se ondary1parti les, are omparable. This fa t was also found in the ases of other highly-denseenvironments, and it implies that it is not a safe assumption to onsider that theprimary population dominates when omputing the equipartition eld.The path of the numeri al treatment by [Persi et al. 2008 is inverse to the onepresented in this Chapter. In the former study, the normalization of the protonspe trum is xed starting from an assumed Np/Ne fa tor from equipartition, withthe normalization for ele trons Ne in turn obtained from the radio spe trum. Inthe present ase, protons are the starting point, with the supernova explosions in-je ting primaries whi h are left to intera t through all possible pro esses, produ ingele trons whi h in turn generate syn hrotron emission. It is this non-negligible on-tribution from se ondary ele trons what is taken into a ount when omputing Band Np/Ne, in order for the data to onsistently mat h at all frequen ies (see Fig-ure 5.1 for further lari ation). Neither equipartition is a priori assumed, nor theprimary ele tron population is a priori xed from the radio spe trum in the presentapproa h. Finally, neutrino emission is s aled from gamma-ray uxes by Persi etal., whereas it is omputed using orrespondingly parameterized ross se tions up totertiary parti les intera tions in this work. The two approa hes the one herein pre-sented and that of [Persi et al. 2008 are in any ase omplementary and provide onsistent results.5.3.2 Results and Dis ussionThe steady population of protons and ele trons of the starburst galaxy M82 om-puted in the model is shown in Figure 5.2 (right), where a power-law index p = 2.1for the inje tion of primary relativisti hadrons is used. An exponential uto is setto 100TeV, although no signi ant dieren e is observed if it is hanged to half thisvalue, even at energies as high as 10TeV. Moreover, a range of values is studied forthe normalization fa tor for the hadroni inje tion spe trum, as it depends on othersensitive parameters subje t to un ertainties. For instan e, the rate of SN explosionper year was assumed to be 0.3 in earlier studies, but re ently it is more ommonlyreferred to as ∼0.1 SNyr−1. In the present work, both a high and a low value for theSN rate are studied, revealing a range of un ertainties. This has been representedin Figure 5.4, where other exponential utos are also onsidered.Indeed, the dis ussion about the supernovae explosion rate is on-going. Thehighest value of 0.3 SN yr−1 is used by, e.g., [Völk et al. 1996 and other authors1The se ondary parti les in this study are those ele trons and positrons that result from kno k-on and pion de ay pro esses in the inner region of M82.

5.3. M 82 65

Figure 5.2: Left: Comparison of the steady population of ele trons that would resultafter solving the diuse loss equation inje ting only primary (solid blue) and onlyse ondary (dashed blue) parti les from kno k-on and pion de ay in the innerregion of M82. The total steady ele tron population (solid bla k), resulting from theinje tion of both primary and se ondary ele trons, is also shown. Parameters usedin this Figure oin ide with those presented later within the same model (Figures5.3 5.4). Right: Steady proton (solid) and ele tron (dashed) distributions in theinnermost region of both M82 (bla k) and NGC 253 (red).

66 Chapter 5. Starburst galaxies

Figure 5.3: Multi-frequen y spe trum of M82 from radio to infrared. The observa-tional data points orrespond to: [Klein et al. 1988 (triangles), [Hughes et al. 1994( ir les) and [Förster et al. 2003 (squares), and referen es therein in ea h ase.The results from modelling orrespond to: syn hrotron plus free-free emision(dashed), dust emission (dotted) splitted in a ool (blue, Tc = 45K) and a warm(purple,Tw ≃ 200K) omponent, and the total emission from radio and IR emission(solid).thereafter, and it is based on [Kronberg et al. 1985. The latter study has ompileddierent estimations, starting from 0.1 SNyr−1 by [Kronberg & Wilkinson 1975,based on the total non-thermal emission; going up to 0.16 yr−1, based only onIR ex ess; then 0.2 0.3 by [Kronberg & Sramek 1985, based on dire t monitor-ing of variability of dis rete sour es; and still going higher up to 0.3 SNyr−1 in[Rieke et al. 1980, based on estimating the number of new radio sour es. But riti sto [Kronberg & Sramek 1985, and in general to high values of the supernova rem-nant rate, have arised mainly be ause the rate of dete tion of new radio sour es doesnot orrespond to those values, e.g., see de [de Grijs et al. 2001, M Leod et al. 1993,Bartel et al. 1987. Lower values for the rates (0.07 − 0.08SNyr−1) have also beenre ently favored by Fene h et al. (2008), who made an 8-days deep MERLIN radioimaging of SNR in the starburst and use two dierent methods to ompute theexplosion rate. On one side, they assumed that the more radio luminous SNRs inM82 are younger than Cas A. On the other, they assumed that the SNR are in freeexpansion and used the measured N(< D)−D plot.The primary ele tron population is taken to be proportional to the inje tionproton spe trum, s aled by the ratio between protons and ele trons, Np/Ne. The

5.3. M 82 67

Figure 5.4: Left: Energy distribution of the dierential gamma-ray uxes, exploringa range of un ertainties in supernova explosion rate and utos in the primary en-ergy, as it is explained in the text. The sensitivity urves for EGRET (red), Fermi(blue), MAGIC (purple), all from [Funk et al. 2008, and the intended one for theforth oming Cerenkov Teles ope Array (CTA, violet) are shown. Right: Dierentialneutrino ux predi tions from the inner region of M82, total and separated in dier-ent hannels. The neutrino predi tions make use of the same explored parametersalready presented in the left panel and explained in the text.

68 Chapter 5. Starburst galaxiesvalue is set later on, in a re ursive appli ation of this model, as a t to radio data.The same applies to other values, like the magneti eld. Together with the primaryele tron population, all the produ ed se ondary ele trons are taken into a ountwhen omputing the steady state. For omparison, in Figure 5.2 (right), the steadyparti le populations present in M82 are plotted together with the ones present inNGC 253, from [Domingo-Santamaría & Torres 2005. The urves for ele tron andproton distribution display a similar behavior and level than those orresponding toM82, emphasizing the similaritity of the initial hara teristi s of these two starburstsas possible high-energy sour es. With the steady ele tron population determined,a multi-frequen y spe trum (from radio up) an then be omputed and omparedwith experimental data.Multifrequen y spe trumFigure 5.3 shows the best mat h from this model: a multifrequen y spe -trum is overplotted to previous radio and IR data (see the aption for de-tails). At the lowest frequen ies, radio data is no longer redu ed to the entral starburst region. Due to angular resolution of the observing instru-ment, emission at this frequen y omes from the whole galaxy and annotbe separated from that oming only from the inner region. Therefore, thismodel ontribution is just a omponent of the total radiation below a few108 Hz. In the region of the spe trum where syn hrotron emission dominates,several parameters an be determined through repeated iterations: magneti eld, EM, slope of proton inje tion, and Np/Ne ratio. In the range of un- ertainty mentioned before, the lowest value for the magneti eld, 120µG, orresponds to the highest steady distribution (i.e., with the highest super-novae explosion rate assumed, and a value Np/Ne = 30). Inversely, a lowsteady distribution implies the need for greater losses by syn hrotron in or-der to mat h the data, thus pumping the magneti eld up to 290µG. Thegeneral t to observations is quite good in ea h ase. These values of mag-neti eld are in agreement with previous results in [Völk et al. 1989 and arealso similar to the ones found for the disk of Arp 220 [Torres 2004 or NGC253 [Domingo-Santamaría & Torres 2005, as well as ompatible with mea-surements in mole ular louds [Crut her 1988, Crut her 1994, Crut her 1999.On the other hand, the slope of the proton inje tion p is determined in thisstep by mat hing the orresponding slope for radio emission. The range ofun ertainties does not seem to ae t p signi antly.[Tabatabaei et al. 2007 onrm, in a study of M33, that non-thermal emissionis spatially asso iated with star forming regions and propose that the ontribu-tion of lo alized stru tures (SNRs) be ome more important as the (syn hrotronphoton) energy shifts to higher values. Although this may be true, the spatial orrelation by itself is not really a proof of it, sin e star forming regions alsohave an enhan ed osmi ray sea, and thus an in rease in the produ tion ofse ondary ele trons, whi h emit non-thermally. In fa t, [Bressan et al. 2002

5.3. M 82 69have shown that the ontribution of radio SNRs to the non-thermal emission annot be dominant in galaxies, on luding that, for our Galaxy in parti ular,it annot be more than about 6%. In any ase, the un ertainty in determiningthe magneti eld in this model must be arefully taken into a ount: most ofthe radio emission is onsidered to ome from osmi ray ele trons that havebeen either inje ted by SNRs (primaries) or other a elerators, or produ edin osmi -ray intera tions (se ondaries). If there is a ontribution to the to-tal non-thermal radio signal made by unresolved SNRs, this would diminishthe estimation of the average magneti eld in the starburst. However, thisdistin tion annot be made with the data on M82 so far.Regarding the far IR emission, starbursts data generally require a dustemissivity law, νσB(E,T ), where typi ally σ = 1.5 and B(E,T ) is the Plan kfun tion. This greybody peaks at ∼45K and has a luminosity of 4× 1010L⊙(from [Teles o & Harper 1980 orre ted by distan e). However, at higherfrequen ies, a simple greybody emission annot explain observations, andan ex ess appears at near IR wavelengths. [Lindner 2003 proposed a dust loud model in whi h an envelope of dust grains surrounds a luminous sour e.Ea h on entri shell de reases its density with in reasing radius and hasits own temperature and asso iated ux. As a simpler but still a urateapproximation, an addition of just one se ondary greybody is enough to tthe data well. The temperature of this greybody is warmer (T ≃ 200K) thanthe main one, and it has a smaller luminosity of 7 × 109L⊙. Small hangesin the warm greybody luminosity or maximum of the peak are possible withresults similar to the ones presented here, and they have been explored. Thet to the IR data is quite good, as an be seen in Figure 5.3.The model dis ussed in this work yields inverse Compton uxes of justa few per ent or less than the upper limits at X-ray energies, see e.g.,[Persi et al. 2008 and referen es therein. The diuse emission is overwhelmedby emission from ompa t obje ts.Gamma-ray predi tionsFigure 5.4 shows the gamma-ray dierential ux that results from this model oming from the entral region of M82, together with the sensitivity urves ofpossible instruments observing it. The goodness of this predi tion is that itis onsistent with the remaining range of the ele tromagneti spe trum: theparti les generating this emission, both hadrons and leptons, are the sameones generating the radiation seen at lower energies. The model onrms thatthere is not an expe tation for dete tion by EGRET, onsistent with EGRETupper limit by [Torres et al. 2004, and the sta king analysis of EGRET databy [Cillis et al. 2005. Nonetheless, it predi ts M82 as a sour e for Fermi. Thetotal ux estimations at high and very high energies are as follows:

70 Chapter 5. Starburst galaxies• for E > 100MeV, 2.6× 10−8 (8.3× 10−9) m−2 s−1,• for E > 100GeV, 8.8× 10−12 (2.8× 10−12) m−2 s−1,with the parenthesis representing the results obtained with the lower energy uto and lower supernova explosion rate. Separate ontributions are plottedfrom ea h gamma-ray hannel: neutral pion de ay, bremsstrahlung and inverseCompton (against CMB, far and near IR photon densities). Finally, Figure5.4 also presents neutrino uxes oming from the inner part of the starburstgalaxy. The separate ontribution of ea h neutrino pro ess an be seen, to-gether with the total ux. As in the previous ase, parenthesis represent theresults obtained with the lower energy uto and lower supernova explosionrate. The total ux estimations at high and very high energies are as follows:• for E > 100GeV, 1.2× 10−11 (3.9× 10−12) m−2 s−1,• for E > 1TeV 3.8× 10−13 (9.9× 10−14) m−2 s−1.Consequently, only if the highest end of the predi tions happens to be arealisti representation of the galaxy, the MAGIC teles ope ould dete t itabove 300GeV with 5σ in 50 h. Although the time required by MAGIC alonewould be unrealisti in order to obtain a dete tion for the lowest end of thepredi tions, with the up oming arrival of MAGIC II, the time to expendon this sour e ould be aordable. The a tual estimations for MAGIC IIsensitivity are a fa tor of 2 to 3 better than for MAGIC I, so the ux omingfrom the starburst galaxy would be dete ted within 50h. Similar estimationswould apply for the VERITAS array. But indeed, the instrument of hoi e todeeply study this sour e would be the forth oming Cerenkov Teles ope Array(CTA). Presenting an instrument a eptan e extending both to the lowerand higher energy ends ompared to the previous teles opes, with a 1TeVsensitivity one order of magnitude or more better than present Cherenkovexperiments, M82 should be a bright sour e for CTA, if this model is arealisti representation of this obje t. Wit h this observatory, studies on thepossible uto in the proton spe tra ould be made, although M82 would stilllook as a point like sour e.The dependen e of these results on the initial inje tion slope p has been ex-plored by enfor ing the latter to be dierent. This depende e is not found tobe relevant, within a reasonable range. In Figure 5.5, this is shown by ana-lyzing the predi tions of the present model for a −2.3 average inje tion slope.Both the multi-wavelength predi tions in radio-IR and the gamma-ray emis-sion are displayed. No relevant dieren es (within un ertainties) are found inthe parameters that give rise to these urves.Equipartition estimates

5.3. M 82 71

Figure 5.5: Comparison of the multi-wavelength predi tions for dierent initial spe -tral slope in the inje tion, see Figure 5.3 for further details. The bla k urve or-respond to the model with proton inje tion spe trum p = 2.1 and magneti eldB = 130µG, whereas the red urve orrespond to the results of modelling withp = 2.3 and B = 170µG. The gamma-ray emission from the −2.3 model in the aseof the highest SN explosion rate is shown against those obtained with the harderinje tion (the green shadow, oming from the un ertainties des ribed before). Maindieren es appear at high energies.It has already been mentioned that are should be exer ised with equipartitionestimates and even more after the work done by [Be k & Krause 2005, whereit was shown that the usual formula to ompute the equipartition eld2 maybe of limited pra ti al use. This formula is based on the ratio K of the to-tal energies of osmi ray protons and ele trons, whi h is generally omputed2The equipartition eld is sometimes referred as the minimum-energy estimate of total magneti elds strengths from radio syn hrotron intensities

72 Chapter 5. Starburst galaxies

Figure 5.6: Dierent ontributions to the radio emission (syn hrotron + free-free)by the steady primary-only (blue) and se ondary-only (yellow) ele tron population,also ompared to the total radio emission of the whole ele tron population (bla k).without a full model of the system. In addition to other non-trivial te hni alproblems mentioned by Be k & Krause, if energy losses of ele trons are impor-tant, the number density between these parti les in reases with parti le energyand the equipartition eld may be signi antly underestimated. The orre tvalue an be omputed only by onstru ting a model of gas density and os-mi ray propagation. [Be k & Krause 2005 already emphasize that starburstgalaxies and regions of high star formation rate in the entral regions andmassive spiral arms of galaxies have high gas densities and mostly at ra-dio spe tra, thus non-thermal bremsstrahlung and other losses are important.Sin e the loss rate in reases linearly with gas density n, the proton-to-ele tronratio also in reases with gas density. In addition, protons are also subje tto energy-dependent losses or es ape (e.g., diusion times ales dier). Thus,the ratio Np/Ne depends really on energy, and any equipartition formula willnot ope. Using the steady population of relativisti parti les that have beenfound as a result of this model, the energy ontent in su h omponent an bedetermined. From this point, equipartition is imposed in order to see how themagneti eld behaves. The obtained value is then ompared to the outputof the model. The result is 150µG, whi h is lose (but not the same) to theestimation by [Weaver et al. 2002, and of the same order as the magneti eldfrom the model, 120µG.Neutrino predi tions[Loeb & Waxman 2006 have suggested that hadroni pro esses in starbursts an produ e a large enough ba kground of diuse high energy neutrinos to be

5.3. M 82 73observable with the ICECUBE experiment. They arrive to this on lusion riti ized by [Ste ker 2007 starting, essentially, from two assumptions andmaking order-of-magnitude estimations thereafter:1. protons lose essentially all of their energy to pion produ tion, and2. a lower limit to the energy loss rate of the protons an be obtained fromthe syn hrotron radio ux, by assuming that all of the radiating ele trons(and positrons) ome from pion de ay.Even when photo-disso iation in starbursts is a minor ee t[An hordoqui et al. 2008, the 1st assumption is arguable, as [Ste ker 2007emphasized, be ause parti les are subje t to diusion and onve tion bywinds in addition to pion losses, i.e., the galaxy is not a omplete alorimeter,espe ially at the highest energies. The 2nd assumption is arguable be ausesyn hrotron radiating ele trons are not only se ondary parti les, but theprimaries too. Figure 5.6 quanties this ee t for M82 and ompares the ontribution to the syn hrotron radiation of the inner starburst if only-primary or only-se ondary ele trons are onsidered. From Ee − me ∼ 10−1to ∼ 100GeV, the se ondary population of ele trons slightly dominates,see Figure 5.2 (left), with the primary population ontributing signi antly.Therefore, it would not be very a urate to normalize the whole radioemission to the se ondaries alone (or primaries alone), and then use this tox the energy loss rate of protons, from where one an estimate the neutrinoemission. This study presents the rst full neutrino emission omputationfrom a starburst galaxy whi h is self- onsistent with the emission at allele tromagneti frequen ies in luded in the same model. Given the detailneeded for su h a des ription, a generalization of the present results for M82to all the starbursts is not re ommended. However, if this is done, and if it issupported by basi assumptions on the starbursts number density within thehorizon, the neutrino diuse emission would be below the Waxman-Bah alllimit3.The sky-average sensitivity of the urrent ICECUBE installation (IC-22, with 22 strings) at 90%C.L. to a generi E−2 ux of νµ is1.3(2.0) × 10−11 TeV−1 m−2 s−1 (E/TeV)−2 depending on the analysis[Bazo Alba et al. 2009. This is not enough to dete t M82 dire tly (see Figure5.4). In fa t, using the present estimation of the neutrino ux from M82 andfollowing the omments made by [An hordoqui et al. 2004, less than 2 eventsper year would be expe ted in the full ICECUBE fa ility. A more denitive3[Waxman & Bah all 1999 show that osmi -ray observations set a model-independent upperbound of E2

νΦν < 2×10−8 GeV m−2 s−1sr−1 to the intensity of high-energy neutrinos produ ed byphoto-meson (or pp) intera tions in sour es of size not mu h larger than the proton photo-meson(or pp) mean-free-path. The bound applies, in parti ular, to neutrino produ tion by either AGNjets or GRBs. This upper limit is two orders of magnitude below the ux predi ted in some popularAGN jet models, but is onsistent with their predi tions from GRB models. Their upper bound isstated in eq. (3) and is illustrated in g. 1 [Waxman & Bah all 1999.

74 Chapter 5. Starburst galaxiesassessment of its sensitivity to su h a signal will need to await further re-nement of angular and energy resolutions, via improved knowledge of thedete tor response.5.4 Dis overy of HE and VHE emission from starbursts5.4.1 Gamma-ray emission dete ted from M82Re ently, VHE gamma-ray emission oming from M82 was laimed by VERITAS[A iari et al. 2009a. Meanwhile, the HE regime was overed by the Fermi tele-s ope, also dete ting emission from the same galaxy in the rst year of observations[Abdo et al. 2010 . In Figure 5.7, this data is displayed, together with the spe -tral energy distribution (SED) of the model presented in this Chapter. A separate ontribution is shown oming from ea h gamma-ray generator: neutral pion (π0)de ay, bremsstrahlung and inverse Compton radiation. The latter was omputedhaving the osmi mi rowave ba kground (CMB), far and near infrared (IR) photondensities as targets altogether, see Figure 5.8. The predi tion (green shade region)is perfe tly ompatible with the observational data presented both at HE and VHE,within errors. The Fermi LAT and VERITAS data allow to further pre ise the orig-inal predi tions. Therefore, a range of parameters and some spe i outputs (redand bla k solid urves) are explored to better onstrain the model to the observedresults, see Table 5.1. As an be seen, π0 de ay ontribution dominates at VHEenergies.5.4.2 NGC 253, onfronted with the modelThe also near, barred-spiral starburst galaxy NGC 253 has been deeply stud-ied through the years. The distan e has been subje t of un ertainty, its valueranging from 2.5Mp [Turner & Ho 1985, Mauersberger et al. 1996, going up to3.3Mp [Mouh ine et al. 2005, 3.5Mp [Rekola et al. 2005, and even 3.9Mp , a - ording to the dete tion of the red giant bran h [Kara hentsev et al. 2003. Its ontinuum spe trum peaks in the far IR, at about 100µm, with a high lumi-nosity of 4 × 1010L⊙, [Teles o & Harper 1980, Ri e et al. 1988, Melo et al. 2002.Its inner (100 p ) region is hara terized, as well as M82, by starburst a tivity[Paglione et al. 1996, Domingo-Santamaría & Torres 2005.The integral ux published by H.E.S.S. onstrains previous predi tions for NGC253 at the VHE regime. Moreover, the rst results from the Fermi teles ope addmore information to the spe trum of the galaxy at lower energies. A set of urves forthe SED are spe i ally plotted in Figure 5.9 to a hieve this low ux, exploring theun ertainties in the distan e to this galaxy and subsequent ranges in the magneti eld and diusive times ale.Apart from diusing away during 106−7 yrs (see Table 5.2), parti les an es apethe inner starburst onve tively, arried away by winds (∼ 300 km s−1), and throughpion ollisions with ambient gas, in even shorter times ales of around a few 105 yrs.

5.4. Dis overy of HE and VHE emission from starbursts 75Table 5.1: Physi al parameters used in the multi-wavelength model of M82, pre-sented in the previous se tion and in [de Cea del Pozo et al. 2009b, together withspe i values that try to mat h the emission from the VERITAS dete tion. Inany ase, the (small) variations explored below are within the former predi tionsof the original model. The list of parameters is divided in se tions: observationalvalues, derived from observational values, obtained from modelling, and assumed.SB stands for starburst.Physi al parameters From 2009 model VERITAS-drivenmodelDistan e (3.9± 0.3stat ± 0.3sys)Mp . . .In lination (77± 3) . . .Radius SB 300 p . . .Radius Disk 7 kp . . .Height SB 200 p . . .Gas Mass SB 2× 108M⊙ (H2) . . .Gas Mass Disk 7× 108M⊙ (HI), . . .

1.8× 109M⊙ (H2)IR Luminosity 4× 1010 L⊙ . . .SN explosion rate 0.3 yr−1 (0.1 yr−1) 0.2 yr−1 | 0.3 yr−1SN explosion energy 1051 erg . . .SN energy transferred 10% 10% | 5%to CRConve tive velo ity 600 km s−1 . . .Dust temperature 45K . . .Ionized temperature 10000K . . .Uniform density SB ∼ 180 m−3 . . .Dust emissivity index 1.5 . . .Emission measure 5× 105 p m−6 . . .Magneti eld 120µG (270µG) 170µG | 210µGProton to ele tron primary 50 (30) . . .ratioSlope of primary inje tion 2.1 . . .spe trumEmax for primaries 106GeV . . .Diusion oe ient slope 0.5 . . .Diusive times ale 1− 10Myr . . .

76 Chapter 5. Starburst galaxies

Figure 5.7: Energy distribution of the dierential gamma-ray uxes of M82, ex-ploring a range of un ertainties in supernova explosion rate and e ien y to inje tenergy from SN to CR. The shaded green area orresponds to the original model pre-sented in this Chapter and [de Cea del Pozo et al. 2009b. Data points and upperlimit orrespond to both VERITAS (stars) and Fermi (diamonds) dete tions.

5.5. Con luding remarks 77

Figure 5.8: Multifrequen y spe trum of NGC 253 from radio to in-frared. The observational data points orrespond to: [Carilli 1996 (triangles),[Elias et al. 1978 ( ir les), [Rieke et al. 1973 (asterisks), [Ott et al. 2005 (dia-monds) and [Teles o & Harper 1980 (squares), and referen es therein. The resultsfrom modelling orrespond to: syn hrotron plus free-free emission (dashed), dustemission (dotted) splitted in a ool (blue, Tcold = 45K) and a warm (purple, Twarm

∼ 200K) omponent, and the total emission from radio and IR emission (solid).The diusion times ale of the parti le depends on the energy. As the times alede reases, the slope of the gamma-ray spe trum be omes steeper (the losses arehigher). The diusion oe ient asso iated is ∼ 1026−27 m−2 s−1 at 1 − 10GeV, ompared to the ∼ 1028 m−2 s−1 value in our Galaxy.5.5 Con luding remarksThe multi-wavelength model presented in this Chapter explains reasonably well boththe HE and VHE emission oming from the two losest starburst galaxies M82 andNGC 253, within a range of the explored parameters already mentioned. Every om-ponent of the emission an be tra ked to one and the same original CR population.Moreover, this CR population results as a onsequen e of all ele tromagneti andhadroni hannels from the primary and subsequently-produ ed se ondary parti les.CR enhan ement present in these starburst galaxies is ree ted in the high energydensity values that an be obtained from the steady proton population. Above aproton energy of ∼ 1500GeV ( orresponding to Eγ ∼ 250GeV), the energy densityis around 10 eV m−3 for M82 and similar value for NGC 253. Now that the VHE

78 Chapter 5. Starburst galaxies

Figure 5.9: Energy distribution of the dierential gamma-ray uxes of NGC253, exploring the un ertainty in distan e, a range of times ale diusion (τ0)and possible utos in the proton inje tion spe trum. The original model from[Domingo-Santamaría & Torres 2005 is also shown for omparison, as well as datapoints from Fermi dete tion (diamonds) and the integral ux from the H.E.S.S.dete tion (square), transformed in dierential ux (assuming a range of inje tionspe tra).

5.5. Con luding remarks 79Table 5.2: Physi al parameters used in the multiwavelength model of NGC 253,as presented both in the previous [Domingo-Santamaría & Torres 2005 study andin this se tion, but exploring some variations allowed within the model in order tomat h the emission from the H.E.S.S. dete tion. The list of parameters is dividedin se tions: observational values, derived from observational values, obtained frommodelling, and assumed. SB stands for starburst.Physi al parameters From 2005 model H.E.S.S.-drivenmodelDistan e 2.5Mp 2.6Mp | 3.9Mp In lination 78 . . .Radius SB 100 p . . .Radius Disk 1 kp . . .Height SB 70 p . . .Gas Mass SB 3× 107M⊙ . . .Gas Mass Disk 2.5× 108 M⊙IR Luminosity (2− 4)× 1010 L⊙ . . .SN explosion rate 0.08 yr−1 . . .SN explosion energy 1051 erg . . .Conve tive velo ity 300 km s−1 . . .Dust temperature 50K 45KIonized temperature 10000K . . .Uniform density SB ∼ 600 m−3 . . .Dust emissivity index 1.5 . . .Emission measure 5× 105 p m−6 . . .Magneti eld 300µG 200µG | 270µGProton to ele tron primary 50 30ratioSlope of primary inje tion 2.3 2.1spe trumEmax for primaries 106GeV . . .Diusion oe ient slope 0.5 . . .Diusive times ale 10Myr 1Myr | 10Myr

80 Chapter 5. Starburst galaxiesregime has been nally a hieved by ground-based teles opes, and Fermi has proventhat dete tions are possible from lower energies (∼ 100MeV), a full and lear pi tureof gamma-ray emission is begining to appear.

PART II:Observations using the MAGIC experiment,and simulations of future observations using CTA

Chapter 6Analysis of MAGIC dataContents6.1 Cherenkov te hnique and teles opes . . . . . . . . . . . . . . 836.1.1 Cherenkov light . . . . . . . . . . . . . . . . . . . . . . . . . . 846.1.2 Hadroni and ele tromagneti showers . . . . . . . . . . . . . 856.1.3 Imaging Air Cherenkov Te hnique . . . . . . . . . . . . . . . 866.2 The MAGIC teles opes . . . . . . . . . . . . . . . . . . . . . . 886.2.1 Stru ture and ree tor . . . . . . . . . . . . . . . . . . . . . . 886.2.2 Camera . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 896.2.3 Readout, trigger and data adquisition . . . . . . . . . . . . . 906.2.4 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 916.2.5 Observation modes . . . . . . . . . . . . . . . . . . . . . . . . 916.3 Analysis method . . . . . . . . . . . . . . . . . . . . . . . . . . 916.3.1 Mono observations . . . . . . . . . . . . . . . . . . . . . . . . 966.3.2 Stereo observations . . . . . . . . . . . . . . . . . . . . . . . . 100The atmosphere blo ks most of the radiation that arrives to Earth, whi h oth-erwise would be harmful for life. In Astronomy, this implies that most of the ob-servations above ultraviolet (UV) radiation, for instan e, need to be arried outfrom spa e. However, the atmosphere an also a t as a alorimeter and thus al-low the indire t dete tion of very energeti parti les, osmi and gamma rays. TheMAGIC (Major Atmospheri Gamma-ray Imaging Cherenkov) teles opes, amongother ground-based experiments, prots from this ir umstan e using the Cherenkovte hnique.6.1 Cherenkov te hnique and teles opesCosmi rays that arrive to Earth ollide with the nu lei present in the atmosphere.The se ondary parti les produ ed intera t again with other atmosperi nu lei, gen-erating a as ade of parti les, also dened as EAS, extended atmospheri showers,[Longair 1992. This ollision pro ess ontinues, losing energy in ea h step, until the riti al energy for reating parti les is rea hed. By the time the radiation arrives tothe ground level, it an be dete ted by imaging atmospheri Cherenkov teles opes(IACT) as Cherenkov light.

84 Chapter 6. Analysis of MAGIC data

Figure 6.1: Polarization of the medium by a harged parti le with (a) low velo ity,v < c/n, and (b) high velo ity, v > c/n. Huygens onstru tion of Cherenkovwavefront in ( )6.1.1 Cherenkov lightThe Cherenkov ee t ([Cherenkov 1934, [Tamm & Frank 1937) takes pla e whena harged parti le travels at a speed larger than in a transparent medium(v = βc, β > 1/n), disrupting the lo al ele tromagneti eld. The light emissionresults from the re-orientation of the instantaneous ele tri dipoles indu ed by theparti le in the medium. When the distortion travels faster than the generated pho-tons, the wavefronts along the traje tory an sum oherently (a ording to Huygens onstru tion, see Figure 6.1) and, if the medium is transparent, the light propagatesat an angle θc:

cosθc =1

βn. (6.1)The energy threshold for the harged parti les to emit Cherenkov light is:

Eth =m0c

2

√

1− β2min

=m0c

2

√

1− 1/n2(6.2)where m0 is the rest mass of the parti le.The number of Cherenkov photons emitted per tra k length by a harged parti leis given by [Yao et al. 2006:

d2N

dxdλ=

2παZ2

λ2

(

1−1

(βn)2

) (6.3)where α is the ne stru ture onstant. From this equation, it appears thatmost of the Cherenkov photons are emitted at short wavelenghts and their number

6.1. Cherenkov te hnique and teles opes 85

Figure 6.2: S heme of a γ-ray indu ed ele tromagneti shower (right) and a hadron-indu ed shower (left) developing in the atmospherede reases when going from ultraviolet to opti al light. The main sour e of atten-uation (between 15 and 2 km above sea level) is the Rayleigh s attering, in whi hphotons are s attered by polarizable air mole ules with a smaller size than the pho-tons wavelenght. Another important ontribution omes from the Mie s attering,produ ed when photons intera t with small dust parti les suspended in the air. Fi-nally, Cherenkov photons an be absorbed by the ozone layer (for λ < 290nm) inthe upper part of the atmosphere.The previously dened Cherenkov angle θc an be expressed as a fun tion ofthe altitude. Sin e the refra tion index is not onstant in the atmosphere, therelationship is su h that θ in reases with de reasing height. Therefore, when thelight emitted at dierent heights rea hes the ground, a light pool is reated: thephotons arrive at a similar distan e from the axis of the shower of parti les, and theproje ted ones of light that should be rings be ome smeared out. The Cherenkovlight density is onstant up to 120m away from the shower axis, where the suddendrop o urs.6.1.2 Hadroni and ele tromagneti showersIn order to produ e this Cherenkov light, very energeti osmi and gamma rays areneeded. Depending on the nature of the primary parti le, the development of theshower an be very dierent (i.e., Figure 6.2).If the primary parti le is a gamma-ray, the as ade will be an ele tromagneti one. When a photon intera ts with one atmospheri nu leus, an ele tron-positronpair is generated. These se ondary parti les loose energy produ ing less energeti photons through bremsstrahlung. The shower keeps developing, quite symmetri-

86 Chapter 6. Analysis of MAGIC data ally, alternating between pair produ tion and bremsstrahlung, until a riti al en-ergy is rea hed, 83 MeV, below whi h the main energy loss is due to ionization.This point (usually between 7 and 13 km above sea level) oin ides with the showermaximum, and after that, the as ade dies out. Due to the relativisti energiesinvolved, the parti les from a γ-indu ed shower are strongly olimated along thein ident dire tion.On the other hand, when osmi rays (usually protons) hit the atmosphere, manydierent pro ess and types of se ondary parti les are involved. These hadroni as- ades an generate pions (90%), kaons and antiprotons (10%) when intera tingwith atmospheri nu lei, and in subsequent steps, muons and neutrinos, amongother parti les. The hadroni ore of the shower ontinues to intera t until the en-ergy threshold for pion produ tion is rea hed (1 GeV), then the ionization pro essbe omes dominant and the as ade dies out. The majority of se ondary parti lesare pions, and therefore an initiate an ele tromagneti sub-shower. In their neu-tral form, those pions de ay mainly in two photons. The harged pions de ay inmuons and neutrinos. The latter will not intera t due to their low ross-se tion,whereas muons intera t almost ex lusively through ionization and, thanks to theirlong lifetimes, usually rea h the ground before de aying. Hadrons are able to transfersigni ant momenta to the se ondary generated parti les, thus widening the tran-verse evolution of the shower. Another important hara teristi is that the arrivaldire tions for the hadroni showers are isotropi , ontrary to γ-indu ed showers,whi h originally ome from the pointed astronomi al sour e. These, together withmorphologi al and timing dieren es, are important features when dis riminatinggammas from hadrons dete ted with IAC teles opes.6.1.3 Imaging Air Cherenkov Te hniqueThe goal of the imaging air Cherenkov (IAC) teles opes is to dete t, at groundlevel, very energeti photons oming from astronomi al obje ts, using the Cherenkovee t. These teles opes olle t Cherenkov light on a ree ting surfa e (mirror)and fo us it on a dete tor ( amera). The dete tor's instrumentation onverts thephotons into ele tri pulses and later saves these as a geometri al proje tion of theatmospheri showers.The image of the air shower is the key of the IAC te hnique. It ontains infor-mation about the longitudinal development of the as ade of parti les through thenumber of photons and arrival time of the images. Cherenkov photons emitted atdierent heights rea h the amera at dierent positions, see Figure 6.3. When point-ing to a possible γ-ray sour e, the most energeti se ondary parti les (lo ated at thetop of the shower) will hit losely to the entral part of the amera. Meanwhile, theless energeti lower part of the shower will form images away from the enter of the amera. The amount of dete ted Cherenkov light is an estimator of the total numberof se ondaries, and therefore, of the energy of the primary parti le. Moreover, theorientation and shape of the image of the shower is a good indi ator of the in omingdire tion and the nature of that primary parti le. For instan e and as explained

6.1. Cherenkov te hnique and teles opes 87

Figure 6.3: Image formation s heme in the amera of an IAC teles ope. The valuesare referred to a 1 TeV γ-indu ed shower. The blue part is the image head whereasthe red part is the image tail. The numbers in the pixels orrespond to the numberof in ident photons. [Tes aro 2010

88 Chapter 6. Analysis of MAGIC datain 6.1.2, arrival dire tions of the hadroni showers are uniformly distributed alongthe dete tor and their proje ted image is wide spread, while γ-ray shower imageshave ni e ellipti al shape and their axis point to the enter of the amera. Thismorphologi al dieren es help reje ting hadroni events from the ba kground.In order to extra t γ-ray indu ed showers from the ba kground, an energy thresh-old is imposed to the primary photon:Eth ∼

√

φΩτ

εA(6.4)The energy threshold is inversely proportional to the square root of the quantume ien y of the dete tor (ε) and the mirror surfa e (A). It also depends on the solidangle subtended by the mirror (Ω), the integration time of the signals in the amera(τ) and the ba kground light from the night sky (φ, whi h main sour es are stars,airglow, zodia al light, arti ial man-made light or even moon light). Therefore,IAC teles opes need large mirror surfa es and high sensitivity photodete tors.6.2 The MAGIC teles opesSin e the fall of 2009, the MAGIC experiment, seen in Figure 6.4, observes gamma-ray sour es with two 17-m diameter IAC teles opes. They are lo ated at theObservatorio del Roque de los Mu ha hos, in La Palma, Canary Islands, Spain(28N17W , 2200m above sea level). Their design mainly fullls two aims: a hiev-ing the lowest energy threshold (of 50 GeV for standard observations, or even 25GeV with a spe ial trigger setup), and a fast repositioning (an average of 30 se ondstakes to point the teles opes from one point to another in the sky). The earliestoperations started in 2004 with one single-dish teles ope (MAGIC-I), and lately theinstrument was upgraded to stereos opi observing mode by adding a se ond - inmany ways, lone - teles ope (MAGIC-II) at ∼ 85m from the rst one. In doingso, the system improved the sensitivity, the primary energy re onstru tion and thesour e position determination. A list and des ription of the hara teristi s of theMAGIC experiment is presented in the following se tions.6.2.1 Stru ture and ree torLight arbon ber reinfor ed tubes support the stru ture of both teles opes andallow for a fast repositioning in any dire tion of the sky, for instan e, in ase ofgamma-ray burst (GRB) alerts. The drive system also provides a high degree ofpointing a ura y. Ea h teles ope has two motors of 11 kW of power ea h in theazimuth axis and one for the zenith movement. The mentioned pointing quality isreinfor ed with a starguider system. It onsists on a CCD amera pla ed at the enter of the ree tor, with a 4.6 eld of view (FOV), thus imaging partially the amera and a portion of the sky. Using six LEDs on the amera as a referen esystem, the starguider ompares the position of dete ted stars with a atalogue. In

6.2. The MAGIC teles opes 89

Figure 6.4: Pi ture of the two MAGIC teles opes: MAGIC-I (left) and MAGIC-II(right)this way, orre tions for possible mispointing an be later applied during the analysisof the data.The ree tor has a paraboli urvature in order to minimize the arrival timespread and improve the signal to noise ratio. As it was previously stated, bothteles ope dishes have 17m of diameter, with a total ree tive surfa e of 239m2 and(80− 85)% of ree tivity in the range from 250 to 650 nm. MAGIC-I has 49.5 m ×49.5 m spheri al mirrors, while MAGIC-II has 1m × 1m mirror elements, whi hdoubles the size of the former. Despite the rigidity of the stru ture, the ree torsurfa e an suer distortions depending on the pointing position of the teles opes,therefore a re-adjusting needs to be done by the a tive mirror ontrol (AMC). Ea hmirror panel has a laser at the enter and two me hani al a tuators. A amerare ords the image formed by the lasers when pointing to the losed amera, and thealignment of the mirrors is ompleted via software orders to the motors.6.2.2 CameraThe amera re eives the Cherenkov light, olle ted in the ree tor surfa e, and onverts the signal into photoele trons. The design of the amera is su h that itattempts to have a large FOV (∼ 3.5) and a ne pixelization (to improve thegamma/hadron separation). Photomultiplier tube (PMT) ameras for both tele-s opes are slightly dierent. In the ase of MAGIC-I, it has hexagonal shapeand onsists of 397 small (30mm diameter) inner PMTs, surrounded by 180 big-

90 Chapter 6. Analysis of MAGIC data

Figure 6.5: S heme for the standard trigger onguration in MAGIC I (left) and II(right) ameras ([Meu i et al. 2007, [Cortina et al. 2009).ger (60mm diameter) PMTs. Meanwhile, MAGIC-II amera has a ir ular shapeand 1039 medium (25mm diameter) PMTs. To improve the quantum e ien y(QE) and ompensate the dead spa e between the edges of the rounded PMTs, ea hone is surrounded by hexagonal-shaped, non-imaging light on entrators (WinstonCone); in addition, the hemispheri al entran e window is oated with a milky la -quer doped with a wavelenght shifter. Temperature and humidity inside the ameraare ontrolled by a water ooling system.6.2.3 Readout, trigger and data adquisitionAll the ele troni s of the MAGIC system are kept in a fa ility 150 m away fromthe teles opes, in order to avoid extra weight to the amera and keep it safe frombad weather onditions. The ele troni signal is onverted into an opti al one andtravels from the amera to the readout system via opti al bers. When the signalis re eived, it is transformed into an ele tri urrent by means of a photodiode,and split in two signals. One part is digitized with a multiplex 2GSamples/s FlashAnalog-to-Digital-Converter (FADC) and kept in a ring buer for MAGIC-I. ForMAGIC-II, the Cherenkov pulses are sampled by low-power Domino Ring Sampler hips, working at 2GSamples/s, and stored again in a ring buer. The other partof the ele troni signal is sent to the trigger system.Only a ertain -inner- region of ea h amera (see Figure 6.5) is onsidered forthe trigger, whi h in turn has dierent de ision levels. The rst level (L0T) is adis riminator: if the analog signal of one pixel ex eeds a ertain threshold (whi hdepends on light onditions), the hannel signal goes to the next level. For L1T,if lose pa ked next-neighbour (3, 4 or 5 NN) pixels have a signal in a short (2 −

5ns) time window, the harge is re orded (in mono-observations). There is a L2Tthat would allow further online pattern re ognition later in the analysis, but it hasnot been implemented. The a tual standard mode of operation, i.e., stereos opi ,requires a L3T: if the signal is dete ted at both teles opes in a ertain time window,

6.3. Analysis method 91then the events are re orded.For some dedi ated observations (like pulsar observations), there is another op-tion, alled the analog sum trigger [Aliu et al. 2008a, that lowers the energy thresh-old down to 25 GeV. It onsists of a linear sum of the signals of large pat hes ofpixels. This time, the trigger area is not an inner ir le but a entered doughnutshape.In ase the trigger o urs, the digitization pro ess stops and the information ontained in the ring buers is kept in a dis . The raw data are stored by the dataadquisition (DAQ) system in run les of 2GB maximum. All les are transferred tothe data enter Port d'Informa io Cienti a (PIC, htt://magi .pi .es) in Bar elona.6.2.4 CalibrationAn opti al alibration system is needed to determine how to onvert FADC ountsinto a number of photoele trons (phe). In order to a hieve this, a set of dierentLEDs (emitting at 370, 460 and 520 nm) are red onto the amera to uniformlyilluminate it. When the signal is triggered, the alibration events are re ordedin dedi ated runs, and an be later used in the analysis hain to orre t for gainvariations.6.2.5 Observation modesThe MAGIC teles opes an observe in two modes: on-o and wobble. On theone hand, to observe a sour e in on-o mode, the teles opes rst point to sour eposition in the sky (ON), and then, to a dark region of the sky (with no-knowngamma-ray sour e and similar zenith and general onditions), in order to substra tthe ba kground (OFF). On the other hand, if the wobble mode [Fomin et al. 1994is hosen, two opposite dire tions in the sky, 0.4 degrees away from the sour e, areobserved during 20 minutes ea h. For wobble mode observations, the ON region isdened around the sour e position in the amera, while the OFF region is enteredat a position opposite to the sour e with respe t to the amera enter (so- alled,the anti-sour e position). The standard MAGIC analysis uses also two additionalOFF regions orresponding to the sour e position rotated by ±90 with respe t tothe amera enter, see Figure 6.6. The standard observation mode in MAGIC is thewobble mode be ause, despite the lower sensitivity, it saves observation time andprovides a better ba kground estimation.6.3 Analysis methodThe nal goals of the data analysis software of a ground-based Cherenkov teles opeare: distinguishing between gamma-like and ba kground-like events, determiningthe energy of the primary γ-ray (to derive the energy spe trum of the dete tedemission), al ulating the in oming dire tion of the gamma shower to estimate the

92 Chapter 6. Analysis of MAGIC data

Figure 6.6: Sket h of the denition of the signal (ON) and ba kground (OFF) regionsin wobble observations. The anti-sour e is the OFF position lo ated symmetri allyto the ON position (the red ir le) with respe t to the enter [Mazin 2007.

6.3. Analysis method 93position of the emitting sour e, and determining the arrival time of ea h γ-ray andidate to produ e light urves.In this se tion, the analysis te hnique used in MAGIC will be des ribed in de-tail. The MARS software pa kage (MAGIC Analysis and Re onstru tion Software,[Moralejo et al. 2009) is a olle tion of programs for the analysis of the MAGICdata, written in C++, in the ROOT framework maintained at CERN. Ea h pro-gram orrespond to a ertain step in the analysis pro edure:• allisto: performs the signal extra tion and alibration steps.• star : arries out the image leaning and al ulates the image parameters.• osteria: is in harge of the training of the Random Forest.• melibea: applies the trained matri es to al ulate the hadronnes parameterand the estimated energy of ea h event.• uxl : al ulates the energy spe trum, the ee tive area and the light urveof the sour e.• elestina/ aspar : produ es a skymap of the ex ess events of the region of thesky that ontains the sour e.The re onstru tion of gamma ray initiated air shower hara teristi s requiresdetailed Monte Carlo (MC) simulations of the shower development and of the re-sponse of the teles ope. In the MAGIC analysis, MC gamma showers are also usedto optimize the uts for ba kground reje tion and to estimate the ee tive olle -tion area after all uts, whi h allows onverting ex ess events into a physi al ux of

γ-rays. The MC simulation program of MAGIC uses the software CORSIKA 6.019[He k et al. 1998. Gamma showers are simulated, under the US standard atmo-sphere, and the output Cherenkov photons that arrive to ground level are re orded.As a se ond step, the so- alled ree tor program rst omputes the Cherenkov lightattenuation in the atmosphere, and subsequently, the reexion of the photons onthe dish mirrors is simulated. Next, the amera program simulates the responseof the teles ope PMTs and both trigger and DAQ systems. Camera also performsa smearing of the arrival dire tions a ording to the opti al point spread fun tion(PSF) of the teles ope. After this point, the simulated events are ready to be usedin the MAGIC analysis hain. In the present Thesis, analysis of sour es observedin both mono and stereo mode are presented. The analysis hain is introdu ed inwhat follows.Signal extra tion and alibrationThe Cherenkov photon ashes produ e very short signal pulses when theyrea h the PMTs, whi h are digitized and stored in the raw data. The re- onstru tion of this information is alled signal extra tion. On e the numberof ounts is re overed, the alibration onsists in onverting these into num-ber of photoele trons (phe). Thus, dedi ated alibration les are re orded as

94 Chapter 6. Analysis of MAGIC dataexplained in 6.2.4 (see also [Gaug 2006). Up to this point, analyzers an ob-tain this information dire tly from the data enter at PIC, without personallyrunning the programs or exe utables.Image leaningAt this stage, pixels whi h ontain no signal from the gamma showers arere ognized and eliminated. Only the shower image survives. The standardmethod leans the image in two steps: rst, it sele ts groups of at least twopixels ( ore) with a phe ontent higher than a minimum value q1, then, pixelssorrounding it (boundary) an be added if they ex eed a signal q2 and if theyalso have at least another boundary pixel ex eeding q2. Next, the mean arrivaltime of all ore pixels is al ulated. Only those ore pixels inside the timing oin iden e window (δt1) are kept, and similarly (δt2) for the boundary pixels.The usual image leaning setting used in MAGIC-I analysis has q1 = 6phe,q2 = 3phe, δt1 = 4.5 ns and δt2 = 1.5 ns [Oya 2010. When arrival timeinformation is not used, a harder image leaning is applied (e.g. q1 = 10pheand q2 = 5phe, whi h is the one used in MAGIC-II).Image parametersThe leaned images an be hara terized by the so- alled Hillas parameters[Hillas 1985, that are related to the statisti al moments of the images up tothird order. For a graphi al interpretation of some of these, see Figure 6.7.The following parameters are independent of the position of the sour e in the amera:

• SIZE: Total number of phe in the image. It is roughly proportional tothe primary parti le energy for a xed impa t parameter (distan e of theshower from the teles ope axis, a.k.a. IP) of the shower.• LENGTH: Se ond moment of the light distribution along the major imageaxis.• WIDTH: Se ond moment of the light distribution along the minor imageaxis.• CONC[N: Fra tion of the total amount of phe ontained in the N mostluminous pixels. Usually, N = 2.• leakage: Fra tion of the light of the image ontained in pixels that belongto the outermost ring of pixels of the amera. It is useful for re ognizingimages partially outside the amera.• M3Long: Longitudinal third moment of the distribution of the hargealong the major axis. It is used to resolve the head/tail degenera y.• time RMS: It is the RMS of the arrival times of all pixels that survivedthe image leaning, and measures the time spread of the arrival times.The following image parameters are dependent on the position of the sour ein the amera:

6.3. Analysis method 95

Figure 6.7: Graphi al representation of some of image parameters des ribed in thetext. The nominal position of the observed sour e is (x0, y0) [Mankuzhiyil 2010.

96 Chapter 6. Analysis of MAGIC data• alpha: Angle (α) between the dire tion of the major axis and the line onne ting the image entroid with the sour e position. Images from

γ-ray showers from the sour e will have small alpha values.• Dist: Distan e between the image entroid and the sour e position in the amera. It is orrelated to the impa t parameter of the shower.• time gradient: It measures the magnitude of the time prole of the event.This parameter is obtained from the slope of the linear t applied on thearrival time versus the spa e oordinate along the major axis.Note that the time-related image parameters are not exa tly Hillas parame-ters, but are used to enhan e the analysis performan e [Tes aro et al. 2007,[Aliu et al. 2009. There is also another parameter, Disp, that allows to pro-du e sky-plots even with a stand-alone Cherenkov teles ope and will be intro-du ed in 6.3.1.Quality sele tionSin e observations with a Cherenkov teles ope are performed under partially ontrolled onditions, data quality sele tion uts need to be done before start-ing the analysis of any astronomi al obje t. The most reliable and ee tiveparameter is the trigger rate. Sin e the CR ux is dominated by the onstantisotropi hadron ba kground, the trigger rate is expe ted to be onstant duringthe observation time. Therefore, a ut on the event trigger rate is an ee tivequality ut to avoid, for instan e, bad weather ondi tions like louds or dustin the air ( alima).A typi al MAGIC analysis usually in ludes a few hours of observation of the CrabNebula sour e. These observations are hosen as lose as possible to the observationperiod and zenith angle of the analyzed sour e. Given the strong and steady signalfrom the Crab Nebula ( onsidered a standard andle in γ-ray Astronomy), thesedata are onsidered as a test sample where to he k the onsisten y and sensitivityof the urrent analysis.6.3.1 Mono observationsThe tasks that are going to be explained below try to estimate the main hara ter-isti s of the primary parti le that originated the air shower: the nature, energy anddire tion of the primary γ-ray. The orresponding programs in MARS are osteriaand melibea.Ba kground reje tionIn order to dis riminate images of gamma showers from the mu h more abun-dant images of hadroni origin, as well as those from isolated muons and u -tuations of the night sky ba kground, previously des ribed image parametersare used (also others like the zenith distan e). MAGIC analysis makes use ofthe Random Forest (RF) method for this task. The RF method uses a forest

6.3. Analysis method 97

Figure 6.8: Sket h of a tree stru ture for the lassi ation of an event v via size,length and width. The de ision path through the tree, leading to lassi ation ofthe event as hadron an be followed [Errando 2009.of de ision trees to lassify ea h event. The de ision trees are reated (trained)from MC simulated γ-ray events and real events from hadron events from un-altered data samples (sometimes, also o-data samples). The individual treesare grown by using a list of image parameters with a proven dis riminationpower and nding adequate uts on these parameters (see Figure 6.8). Fromthe training sample, a binary de ision tree an be onstru ted, subdividingthe parameter spa e in two parts, and iteratively repeating the pro ess. Thepro edure stops when a sub sample omposed only by gammas or by hadronsis found (a leaf of the tree is rea hed). The ending leaf is labeled as a 0(gamma) or 1 (hadron).The ba kground reje tion takes pla e after the training. Ea h event is passedthrough all de ision trees and they will s ore 0 or 1, depending on the rea hedleaf label. A mean s ore is omputed averaging over all the dierent separa-tion tree results. This average is alled hadronness (h) and represents a sortof probability for the event to be a hadron (h lose to 1) or a gamma (h loseto 0). There are two kinds of parameters used to train the h parameter: pa-rameters whi h have power to dis riminate between hadrons and γs (WIDTH,LENGTH, CONC, Dist, M3Long, time RMS and time Gradient) and, also,SIZE and zenith distan e be ause the rest of parameters depend strongly onthe size of the shower images. In luding these last two, the uts in RF an bes aled dynami ally with the geometry of the shower image.Energy re onstru ionThe energy of the γ-ray events is also re onstru ted by the Random Forest

98 Chapter 6. Analysis of MAGIC datamethod, using training MC γ-ray samples. A MC simulated γ-ray samplewith a known primary γ-ray energy is filled in bins of logarithmi energy. The lassier will be trained to a ommodate ea h event in a parti ular energybin. After this, ea h tree will assign a spe ifi energy range to ea h event,that will be analogous to the previously des ribed hadronness parameter. Theparameters used normally are WIDTH, LENGTH, SIZE, log(SIZE/(WIDTHLENGTH)), CONC, leakage, ZD, Dist and time Gradient, being the last twosour e position dependent. The obtained energy resolution is about 20% forenergies from 100 GeV to 10 TeV, in reasing at lower energies and de reasingfor higher ones. The energy of the γ-rays is usually overestimated at lowenergies (<100 GeV) and underestimated at higher energies (>10 TeV). Thefinite resolution and bias require that the energy estimation is later orre tedby the spe trum unfolding, using a migration matrix.Analysis approa hes (α and θ2)The analyzer an either de ide to go for an alpha-type analysis or a theta-square-type analysis. Alpha has already been des ribed with the image pa-rameters. Theta, (θ), is the angular distan e in the sky between the expe tedemitting position and the re onstru ted in oming dire tion of the re ordedshower. An ex ess of events at small values of alpha or θ2 is expe ted, as theγ-ray shower images will be oriented towards the enter of the amera. Thealpha-analysis approa h is suitable when the emitting sour e an be onsid-ered as point-like. If the point-like sour e has very well known sky oordinatesa sour e-dependent analysis an be performed: a priori assumptions aboutthe lo ation of the sour e are done. The advantages are that, by using someextra image parameters (like M3long or the time gradient), a better ba k-ground reje tion and an enhan ed sensitivity in the analysis are a hieved.The θ2−analysis is, instead, more suitable when the sour e of gamma raysis extended or when the position of a point-like sour e is not well known. Asour e-independent analysis allows the produ tion of sky maps showing theex ess of gamma rays in the observation eld, but no assumptions about thesour e lo ation are done, resulting in a less sensitive analysis. In the ase ofa single dish teles ope, the re onstru tion of the primary gamma dire tion isdone through the so- alled Disp method.On e this step of the analysis is rea hed, the signal may have been found asarriving from the pointed sour e. If the signal is strong enough, both a spe trumand a light urve an be al ulated with uxl . If it is not statisti ally meaningful,dierential or integral upper limits on the gamma-ray ux an be set. There is alsothe possibility of produ ing a skymap, by proje ting the arrival dire tions of thesele ted gamma-ray andidates into elestial oordinates.Spe trumThe dieren ial energy spe trum of a sour e is dened as the number of γ-raysarriving to Earth, per unit of energy, time and area:

6.3. Analysis method 99dF

dE=

dNγ

dtdAeffdE(6.5)Here Nγ is the number of dete ted γ-rays and t the exposure time. Theee tive olle tion area, Aeff , is the area in whi h air showers an be observedby the teles ope, folded with the effi ien y of all the uts applied in the analysis(εγ). It in reases with the zenithal angle (spe ially above 45 degrees). This ut effi ien y an be estimated using MC gamma rays, and is defined as thenumber of produ ed gammas divided by the number of events that survive theanalysis uts. Loose ba kground reje tion uts are applied in order to omputethe spe trum be ause they redu e the ee t of systemati un ertainties andprovide larger statisti of ex ess events.Due to the nite resolution of the dete tor and the bias introdu ed in theenergy estimation, the measured spe tra be ome distorted. The unfoldingpro edure tries to orre t this distortion. Those biases are due to the fa t thatthe true energy of the in oming gamma-rays is not measured but indire tlyestimated. Several unfolding methods an be hosen in the MAGIC analysis,differing from the algorithm used for the al ulation of the true distribution[Albert et al. 2007 . An agreement within the different methods is requiredto trust the obtained results.The light urves are plots where the integral gamma-ray ux above a ertainenergy is shown as fun tion of the observation epo h. Cuts used for the light urve are the same as for the spe trum al ulation.SkymapAs mentioned previously, sky maps are al ulated by re onstru ting thearrival dire tions of the gamma-like showers in amera oordinates andproje ting them into elestial oordinates. The arrival dire tion of thegamma-ray images is estimated using the Disp method ([Lessard et al. 2001,[Domingo-Santamaría et al. 2005). The method assumes that the arrival di-re tion of the primary parti le lies on the major axis of the shower. The Dispparameter is dened as the angular distan e between the re onstru ted sour eposition in the amera and the enter of gravity of the shower image. Theparameterization used to al ulate Disp is optimized using a MC gamma-raysample, and is a fun tion of the image parameters. One important degenera yin the Disp method are the two solutions found for a sour e lo ation: one inthe head of the image and another in the tail. The M3long parameter is usedto hoose and pla e the sour e position always loser to the head of the image.The program used to perform skymaps is alled elestina. It produ es skyplots with three ba kground estimation methods, on-o, wobble and model,depending on the observation mode (the latter serves for both). Re ently, two-simpler- programs have repla ed elestina: aspar and zin . Zin is to the elestina model skyplots what aspar is to the elestina wobble skyplots. In

100 Chapter 6. Analysis of MAGIC datathe present work, when observations in wobble mode are presented, asparis the hosen program to reprodu e skymaps, and the ba kground estimationwas done from the same data set. However, in one set of observations, the elestina program has been used (on-o observations, see next Chapter for de-tails), due to the fa t that zin was still not available at that moment. Afterthe ba kground is built, a map of ex ess events is generated by subtra ting thedistribution of ba kground events to the γ-ray andidates. The ex ess map issmoothed using the PSF of the teles ope.6.3.2 Stereo observationsSin e fall 2009, MAGIC-II has joined its twin Cherenkov teles ope MAGIC-I in theobserving duty. The two teles opes an be operated independently or in stereo-s opi mode. The stereos opi mode leads to a better re onstru tion of the imageparameters and a stronger ba kground suppression. When an image of the shower isobtained with the two teles opes, image parameters are individually obtained fromthe signal of ea h teles ope, and then ombined to al ulate stereo parameters. Thein oming dire tion of the shower (i.e., the sour e position) is obtained from the in-terse tion of the major axes of the two images of the shower. In addition, the heightof the shower maximum an also be estimated.The stereo analysis, therefore, uses the θ2-approa h as the standard one. The restof the hain onsists basi ally on dupli ating ea h step: two alibration, two randomforests and two star les are generated, both for the data and the MC. After that, anew program alled SuperStar is used to onvert two star les with individual-imageparameters to a Stereo-Parameter le. It performs the stereos opi re onstru tion ofthe shower parameters (dire tion, ground impa t, altitude) and an also re onstru tthe energy from look-up tables. Then, modied versions (to make use of the stereoparameters) of melibea, uxl and aspar/zin omplete the analysis method.

Chapter 7Two Milagro-dete ted, BrightFermi Sour es in the Region ofSNR G65.1+0.6Contents7.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1017.2 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1037.3 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1047.4 Upper limits on the gamma-ray ux . . . . . . . . . . . . . . 1067.5 Interpretation and dis ussion . . . . . . . . . . . . . . . . . . 1097.6 Con lusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1107.1 MotivationSin e the laun h of the Fermi satellite and the publi ation of new sour e list, severalproposals were presented by the dierent IAC teles ope ollaborations to arry outa follow up of these sour es up to the highest energies. In February 2009, the Fermi ollaboration published a list of the most signi ant gamma-ray sour es above 100MeV, dete ted by the large area teles ope (LAT) within 3 months of observation[Abdo et al. 2009 . Among the gala ti sour es dete ted by LAT, 34 were withinthe eld of view of the Milagro gamma-ray observatory ([Abdo et al. 2009i andreferen es within). The sensitivity of this instrument peaks in the energy range10 − 50TeV. Consequently, the Milagro ollaboration re-analysed their previousskymap looking for ounterparts [Abdo et al. 2009i, and laimed 14 new sour eswith onden e levels above 3σ. Sin e many of them were positionally oin identwith Fermi pulsars, they are mostly believed to be Pulsar Wind Nebula (PWN), orsimilar sour es.The MAGIC teles ope (des ribed in the previous Chapter) is a well-suited in-strument to study the energy range between Fermi and Milagro. To sele t possible andidates for MAGIC, an interpolation of the uxes given by both experimentswas al ulated in two dierent ways. The Fermi Bright Sour e List provides anintegral value above 1GeV, whereas the Milagro sour es are listed together withtheir dierential ux points normalized at 35TeV. Firstly, a power law interpolation

102 Chapter 7. MAGIC upper limits in the region of SNR G65.1

Figure 7.1: Radio image of the region around SNR G65.1+0.6. The numberedobje ts are: (1) 3EG J1958+2909, (2) 2 CG 065+00, (3) 0FGL J1954+2838, (4)0FGL J1958.1+2848, (5) region of dierent spe tral index, (6) IRAS 19520+2759,(7) bright ompa t radio obje t. an be done from these two values, resulting in a ux value at 1TeV and a spe tralindex. Sin e the Fermi ux is likely to ontain both pulsar and steady emission,this value is probably an overestimation. Thus, a se ond ux estimation may bedone by extrapolating the Milagro value to 1TeV, assuming a power law index of2.1. The expe ted ux range optimal for MAGIC observations lies in between theseestimated values. Looking for interesting sour es, those that were already observedby MAGIC were ruled out (i.e. those in the H.E.S.S. s an of the gala ti plane),together with the ones that were below the threshold sensitivity. An interesting ob-je t that did not fulll these hara teristi s appears in the region of SNR G56.1+06.It turned out to be the pulsar 0FGL J1954.4+2838, with a al ulated ux range of2%− 4.9% of Crab.In fa t, this region is densely populated, as an be seen in Figure 7.1, andthe obje ts it ontains and their likely asso iations is still a matter of debate.G65.1+0.6 (dash-dotted shape) is a faint supernova remnant (SNR) rst reported by[Lande ker et al. 1990. In [Tian & Leahy 2006, a distan e of 9.2 kp and a Sedovage of 40−140 kyr were suggested. Despite its distan e, it is a very extended obje t(90′ × 50′). Very lose to the already mentioned Fermi sour e (labeled 3 in theFigure 7.1), there is a strong ompa t radio obje t (7). In the southern rim of theshell, a region with dierent spe tral indexes (5), originally thought to be an extra-

7.2. Observations 103gala ti obje t [Seiradakis et al. 1985, an be found. Also, the nearby pulsar PSR1957+2831 (X) may be asso iated to the remnant [Tian & Leahy 2006, due to the ompatibility of its hara teristi age. Looking at longer wavelenghts, the infraredsour e IRAS 19520+2759 (6) has been suggested to be at a similar distan e of theSNR and, given that it has asso iated a CO line, a H2O and OH maser emission,an intera tion between the mole ular loud and the remnant ould be a possibles enario [Arquilla & Kwok 1987.The gamma-ray emission in the region of G65.1+0.6 was rst dete ted by theCOS-B satellite [Swanenburg et al. 1981 as 2CG065+00, and later onrmed bythe EGRET satellite (3EG J1958+2909) in [Hartman et al. 1999, where a possibleextension or multiple sour es were denoted. But it was only re ently that Fermi ould larify their signal as aused by the two sour es 0FGL J1954 + 2838 and0FGL J1958.1 + 2848 (3 and 4 in Figure 7.1).Observing the SNR G65.1+06 region, MAGIC was expe ted to deliver the fol-lowing s ienti results:• Establishing the lo ation of the γ−ray sour e with a resolution ex eeding theone of Fermi. This might larify whether the origin of the radiation omesfrom inside the shell, or even from a ompa t obje t.• Obtaining a omplete spe trum, together with Fermi and Milagro results, to hara terize the sour e at very high energies. On the one hand, the resultmay shed some light to onrm the SNR as the sour e of the emission. Onthe other hand, if no signal is found, it would imply that the Milagro obje tis not identi al to the Fermi Bright sour e. For instan e, Milagro ould beseeing the intera tion between the SNR and the mole ular loud. In this ase,one of the pulsars (1957+2831 or 1954+2838) may not belong to G65.1+0.6.• Proting from the in lusion of other obje ts in the eld of view of the amerato obtain hints to other dete tions. These other obje ts are a pulsar, the rimof the SNR and the infrared region, among others.The results of the study of this region were published in the Astrophysi al Jour-nal in 2010 [Aleksi¢ et al. 2010, and the PhD student is one of the orresponding o-authors in this paper.7.2 ObservationsThe observations were entered on the Fermi Bright sour e 0FGL J1954.4 + 2838(from now on referred as J19540) in July and August 2009. The observations were arried out in wobble mode (see previous Chapter, or [Fomin et al. 1994), whi hyielded two datasets with osets of ±0.4 in RA from this sour e, see Figure 7.2.The wobble position was alterned every 20 min, and the data were taken at zenithangles between 0 and 40 degrees. At the time, the MAGIC-II teles ope was still

104 Chapter 7. MAGIC upper limits in the region of SNR G65.1under omissioning, therefore the analysis performed here uses only the data fromthe stand-alone MAGIC-I teles ope.Summer observation time at the MAGIC site usually implies alima: dust sus-pended in the atmosphere oming from the Sahara desert, that interferes with thequality of the data taking. Therefore, from the total amount of time observed, only24.7 hours survived the quality sele tion uts. Those uts were applied to the dataruns mainly based on the steadiness of the event rate after image leaning, but also onsidering a few parameters that hara terize the transparen y of the atmosphere,su h as the sky temperature and humidity. An additional ut removed the eventswith a total harge (size) of less than 100 photoele trons, thus providing a betterba kground reje tion.As it was previously stated, this region in luded other interesting obje ts. Spe if-i ally, there was another Fermi sour e, 0FGL J1958.1 + 2848 (shortly referred asJ19580), in the eld of view of one of the two wobble positions. 12.6 hours of ee tiveobservation time were analyzed for this obje t.7.3 Data analysisEa h data set was analyzed in the MARS analysis framework (see previous Chapter,[Moralejo et al. 2009), standard software for the analysis of MAGIC data. To lookfor the signal oming from the two Milagro-dete ted Fermi sour es, a standardanalysis was performed, using θ2 plots and skymaps. It shall be noted, though,that when the 1-year atalog of Fermi sour es [Abdo et al. 2010a was released,the exa t oordinates of both now identied pulsars (1FGL J1954.3 + 2836 and1FGL J1958.6 + 2845, in short J1954 and J958, respe tively) hanged respe t tothe ones given in the previous Fermi Bright Sour e list (0FGL J1954 + 2838 and0FGL J1958.1 + 2848, named in this text J19540 and J9580, respe tively). Thewhole analysis had to be repeated, this time using the latest more a urate sour epositions.Both sour es had to be analized dierently. In the ase of J1954, a standardwobble analysis ould be done: in su h ase, the exposure inhomogeneities an elout when omparing the photon ux from the sour e to the one on the oppositeside of the amera ( alled anti-sour e, see Figure 7.2). However, for J1958, a wobbleanalysis using only one of the two wobble positions does not guarantee this an e-lation. Therefore, the analysis was done in ON/OFF mode, using the near wobblesample as ON-sour e data and the far wobble sample as OFF-sour e data. Havingthe OFF-sour e at the same position in relative amera oordinates as the sour e inthe ON sample, the exposure inhomogeneities may an el out.Figure 7.3 shows the θ2 distributions for ea h sour e, where the values in thehorizontal axis represent the squared angular distan es between photon dire tionsand the sour e position. In the ase of having a very high energy (VHE) gamma-rayemission oming from the observed sour e, an ex ess is expe ted to appear for lowθ2 values, see se tion 6.3.1. This ex ess omes from substra ting the ba kground

7.3. Data analysis 105

19.8519.9019.9520.00RA (2000)

28.0

28.5

29.0

29.5

30.0

DEC

(20

00)

W2W1

Fermi J1954

Fermi J1958 (only in W1)

1420 MHz G65.1+0.6 shell

Milagro 3 contour

MAGIC Pointing Positions

MAGIC Off J1954

MAGIC Off J1958 (only in W2)

Figure 7.2: Observation setup for the two Fermi sour es J1954 and J1958 in the ontext of SNR G65.1+0.6 and a Milagro signi an e ontour. J1958 appears onlyin one wobble position (W1), so the OFF data is taken from the other wobblesample, using the same position relative to the pointing dire tion. The outline ofthe remnant is taken from the radio map in [Lande ker et al. 1990. The extensionof the Milagro signi an e ontour [Abdo et al. 2009i is ompatible with their pointspread fun tion.

106 Chapter 7. MAGIC upper limits in the region of SNR G65.1Entries 80393

2θ0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

# E

ven

ts

0

100

200

300

400

500

600

700

Entries 803931FGL J1954.3+2836Entries 39006

2θ0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

# E

ven

ts

0

50

100

150

200

250

300

Entries 390061FGL J1958.6+2845

Figure 7.3: Plots showing the θ2 distributions for J1954 (left) and J1958 (right).The shapes of the ON and OFF distributions agree well with ea h other in bothsour es, whi h means there is no γ−ray signal. The dashed lines indi ate the signalregions.to the number of events around the sour e position (ON events), having previouslyestimated the ba kground by ounting the events around the anti-sour e position(OFF events). The number of OFF regions explored in the analysis of J1954 was oneand three, although only the last option is shown in the Figure 7.3. Therefore, theunlikely ase of an emission o uring by han e at both ON and OFF lo ations ata similar ux level is dis arded. Furthermore, a ut in hadronnes of 0.1 was appliedto improve the ba kground estimation and determine the signal region separatelyin both sour es. No signi ant signal ould be found in any of the sour es.In order to nd a signal using a blind sear h (i.e., regardless of the sour e posi-tion), skymaps in a region of 2.8 × 2.0 were drawn in dierent energy ranges. Thesame uts in hadronnes (0.1) and size (100) as in the θ2 plots were applied to themaps. Figure 7.4 shows the skymap for both sour es above 200 GeV. No signi antsignal ould be found in any of the skyplots in either of the sour es.7.4 Upper limits on the gamma-ray uxSin e no signal ould be retrieved from the SNR G65.1+0.6 region, upper limits were al ulated of the integral and dierential ux. To obtain the dierential upper limit(u.l.) values, the data was divided into three bins of estimated energy (starting from120GeV, going to 375GeV, then 2.8TeV and nally 12TeV). From the θ2-plots, anevent number upper limit was al ulated at 95% onden e level ( .l.) for ea h bin,using the method des ribed in [Rolke et al. 2005. An e ien y systemati errorof 30% was assumed [Albert et al. 2008b. To al ulate the ux upper limit, thenumber of events per energy bin is established to be:N = teff

∫

Φ(E)Aeff (E)dE (7.1)

7.4. Upper limits on the gamma-ray ux 107

Figure 7.4: Skymaps of the ex ess events (top) and signi an e (bottom) for J1954(left) and J1958 (right). The bla k star shows the enter of the pointed sour e, thewhite ir le represents the MAGIC point spread fun tion (PSF) at mid energies.The distribution of signi an es are well-tted with a simple gaussian.

108 Chapter 7. MAGIC upper limits in the region of SNR G65.1Table 7.1: Dierential upper limits for both sour es, for the present ross he kanalysis.Sour e Name Extension Emed Signi an e F95%(deg) (GeV) (σ) (10−12TeV−1 m−2s−1)1FGL J1954.3+2836 ≤ 0.08 221 -1.4 25980 -1.0 0.55686 +2.2 0.0221FGL J1958.6+2845 ≤ 0.08 221 -1.0 76966 -0.9 0.715123 +0.7 0.033where Φ is the ux (photons per unit of time, energy and area), teff is the ee -tive time of observation and Aeff is the ee tive area after uts. To simulate thisAeff (E), a power law energy spe trum with a photon index of −2.1 is assumed forea h energy bin:

Φ = K

(

E

E0

)−2.1 (7.2)where K is a normalization fa tor. The inuen e of the photon index is minor, sin ethe energy ranges are su iently small. On e the upper limit is derived from thenumber of events (Nul), the onversion to ux upper limits uses (in ea h energybin):K ≤

Nul

teff∫

(E/E0)−2.1Aeff (E)dE(7.3)The mean γ−ray energy in ea h bin of re onstru ted energies, after all uts, an beobtained through Monte Carlo simulations, sin e the data an only be sele ted byan estimated energy.Until this point, the analysis was performed assuming the sour es were point-like, i.e. the extension was supposed similar or smaller than the PSF of MAGIC(dened as the sigma of a two dimensional Gaussian fun tion). For this ase, it was onsidered to be 0.08. In this parti ular region, though, γ−rays may ome froman extended sour e (a PWN or the shell of the SNR). Therefore, another upperlimit for the ux was al ulated, assuming an extension of 0.3. This parti ularvalue was hosen be ause the biggest TeV PWNe have sizes of few tens of parse s,whi h at the distan e of G65.1+0.6 would be within this radius. In reasing thesignal integration radius, more ba kground events are in luded and that leads, thus,to higher upper limits. The derived 95% .l. ux upper limits are summarized inTable 7.1. The dieren es between the limits of J1954 and J1958 are all ompatiblewithin the statisti al u tuations.Also, an integral upper limit above 200GeV was derived for ea h sour e. Theresults were 3×10−13 TeV−1 m−2 s−1 for J1954 and 4×10−13 TeV−1 m−2 s−1 forJ1958, whi h an also be translated in ommonly used units as 3% and 2% of CrabNebula ux, respe tively.

7.5. Interpretation and dis ussion 109Table 7.2: Chara teristi parameters of the two Fermi pulsars 1FGL J1954.3+2836and 1FGL J1958.6+2845, refereed as J1954 and J1958, respe tively. See referen esin the text.Parameters 1FGL J1954.3+2836 1FGL J1958.6+2845Age (kyr) 69.5 21Period (ms) 290 92.7Spin-down luminosity (erg/s) 10.48 × 1035 3.39 × 1035Energy ut-o (GeV) 2.9 1.27.5 Interpretation and dis ussionWhen the 1-year Fermi Catalog was released, J1954 and J958 ould be re-ported as γ−ray pulsars thanks to a blind sear h ([Saz Parkinson et al. 2010,[Abdo et al. 2009d and [Abdo et al. 2010b). They appeared to be normal Fermipulsars, as an be seen in their hara teristi parameters listed in the table 7.2.Before this happened, signi an es of 4.3σ for J19540 and 4.0σ for J19580 were laimed as dete tions by Milagro [Abdo et al. 2009i. Given that the angular reso-lution of Milagro is about 0.4 − 1.0, those values are still orre t for the 1-year atalog positions of the sour es, whi h are oset by ≤ 0.1. Flux values were de-termined for a hara teristi median energy of 35TeV. As it is shown in Figure 7.1,the region of SNR G65.1+0.6 is densely populated, and Milagro sour es ould beasso iated with the Fermi pulsars, a possible extragala ti sour e, a mole ular loudintera ting with the SNR or a radio ompa t obje t (see se tion 7.1 for further de-tails). However, the most likely assumption was that they were indeed related to theobje ts observed by Fermi, i.e., that they were PWNe or, in the ase of J1954, theshell of the SNR, whi h surrounds the pulsar. Gamma-rays at TeV energies ouldalso be produ ed in an intera tion of the shell with a oin ident mole ular loud,su h as the infrared sour e IRAS 19520+2759 (gure 7.1).The upper limits al ulated on the ux at 1TeV, at the maximum of the MAGICsensitivity, present values of 3% and 2% of Crab for J1954 and J1958, respe tively.Therefore, the photon index in the energy range of 1 to 35TeV must be harder than2.2 for J1954, and 2.1 for J1958, if an asso iation between the Milagro and the Fermiobje ts is assumed and they are treated as point-like obje ts. The spe tral energydistribution (SED) would be thus likely to peak at energies greater than 1TeV. Onthe other hand, if an extension up to 0.3 is assumed, the orresponding ux limitsin Crab Nebula units are 14% for J1954 and 3% for J1958. In this extended ase,the photon indi es would be limited to values inferior to 2.6 and 2.2, respe tively.A omplete SED, overing from the Fermi to the Milagro energies, an be on-stru ted by adding the MAGIC upper limits (see Figure 7.5). From this generalpi ture, it an be stated that the most likely s enario to explain the gamma rayemission dete ted by Milagro might be the presen e of two PWNe, asso iated withthe Fermi pulsars. Given the old ages of the pulsars (see table 7.2), it is reasonable

110 Chapter 7. MAGIC upper limits in the region of SNR G65.110-1 100 101 102 103 104 105

E (GeV)

10-12

10-11

10-10

Flux ·

E2 (

TeV

cm

2 s

1)

MAGIC (95% U.L.)

MAGIC < 0.3 deg (95% U.L.)

Milagro

Eight blind PSR Fit

Fermi 1-Year Catalog

3% Crab (MAGIC)

10-1 100 101 102 103 104 105

E (GeV)

10-12

10-11

10-10

Flux ·

E2 (

TeV

cm

2 s

1)

MAGIC (95% U.L.)

MAGIC < 0.3 deg (95% U.L.)

Milagro

Fermi PSR Cat. Fit

Fermi 1-Year Catalog

Egret

3% Crab (MAGIC)

Figure 7.5: Compilation of ux measurements and upper limits for: (left) 1FGLJ1954.3+2836 from Fermi [Saz Parkinson et al. 2010, Abdo et al. 2010a , (right)1FGL J1958.6+2845 from EGRET [Hartman et al. 1999, Fermi [Abdo et al. 2010b,Abdo et al. 2010a, together with MAGIC and Milagro [Abdo et al. 2009i data.The 3% fra tion of the MAGIC Crab spe trum [Albert et al. 2008b is shown for omparison.to expe t an inverse ompton omponent that dominates their energy outows, and onsequently the VHE emission to be extended [de Jager & Djannati-Ataï 2009,[Tanaka & Takahara 2009.7.6 Con lusionsAfter the analysis of nearly 25 hours of good quality data in the region of theSNR G65.1+0.6, no signi ant γ-ray emission ould be found in the MAGIC energyrange (from hundreds of GeV to several TeV). In that same region, the Milagro ollaboration had reported the emission of γ-rays, with a median energy of 35 TeV,and the Fermi satellite had dete ted two pulsars, 1FGL J1954.3 + 2836 and 1FGLJ1958.6 + 2845. The la k of dete tion with the MAGIC-I teles ope, both using thea-priori sour e-position analysis and a skymap of the area, yielded three dierentialux upper limits for ea h sour e.In the light of MAGIC results, the ux upper limits support the s enario in whi hthe multi-TeV emission measured by Milagro is aused by a dierent me hanism thanthe emission dete ted by Fermi. Taking into a ount the ages of the pulsars and theSNR, the existen e of two old PWNe powered by the two GeV pulsars is one ofthe likely s enarios.

Chapter 8Simulations of CTA response toparti ular s ien e asesContents8.1 Sorting out dierent layouts and ongurations for CTA . 1118.2 A brief look at starting-up tools . . . . . . . . . . . . . . . . 1138.3 Spe tral studies . . . . . . . . . . . . . . . . . . . . . . . . . . 1158.3.1 Mole ular louds illuminated by CR from nearby SNR . . . . 1168.3.2 Starburst galaxies M82 & NGC 253 . . . . . . . . . . . . . . 1188.4 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120The CTA experiment (Cherenkov Teles ope Array) 1 has been dened as thenext generation of teles opes using the Cherenkov te hnique. In the energy range offew TeVs, the CTA sensitivity will improve by one order of magnitude ompared tothe previous generation experiments (like MAGIC or H.E.S.S.). In addition, the en-ergy range will also be expanded, from 0.01 to 100TeV, thanks to the large numberof dierent sized Cherenkov teles opes. Furthermore, an improvement regarding theangular resolution is expe ted by a hieving a better resolution and higher photonstatisti s that an rene the imaging of the air showers. In order to a hieve full-sky overage, two arrays will be built: one in the northern and one in the southernhemisphere. The northern site is planned to fo us on extragala ti sour es, whereasthe southern one will take are of the gala ti sour es, spe ially sin e it will overthe entral gala ti plane. CTA is planned as an open observatory, with transparenta ess to data, analysis tools and user training. The main design work has beenperfomed [Hofmann et al. 2010, and now it is fa ing the prototyping and onstru -tion phases. Simulations of CTA response to interesting s ien e ases are presentedbelow.8.1 Sorting out dierent layouts and ongurations forCTAAt an early stage of the design of the CTA array, several layouts were generated, ea hof them with dierent distributions and types of teles opes. These ongurations1http://www. ta-observatory.org/

112Chapter 8. Simulations of CTA response to parti ular s ien e ases

Figure 8.1: Konrad Bernlöhrs adapted layout of the original ultra-CTA (from Padova2008 CTA meeting). Four types of teles opes are plotted, see table 8.1 for details,being 1 = red, 2 = green, 3 = blue and 4 = magenta.Table 8.1: Main hara teristi s of dierent types of teles opes for CTA.Area Diameter F.o.V. Pixels1 412m2 23m 5 0.092 100m2 11m 8 0.183 37m2 7m 10 0.254 100m2 11m 10 0.18

8.2. A brief look at starting-up tools 113

Figure 8.2: Layout ongurations of the three representatives: B (left), D ( enter), I(right). The types of teles opes orresponds to the ones detailed in Table 8.1, being1 (red), 2 (green), 3 (blue), 4 (magenta), the number in bra kets orresponds to theeld of view of ea h teles ope.of CTA were originally taken as subsets of the so- alled ultra-CTA (see Figure 8.1),and were sele ted ensuring a similar onstru tion ost. The initial studies for thePhysi s group started with these 11 ongurations (named from A to K). The veryrst step onsisted on grouping and omparing all of them in order to understandtheir intrinsi hara teristi s and make any subsequent analysis easier. The nalgoal was to work only with one representative onguration from ea h group, andanalyze dierent sour es with these representatives.All the ongurations were lassied, in a rst approximation, by looking atthe dierential sensitivity, and maximizing their response at low energies (LE),high energies (HE) and over the whole energy range. In the end, three simulated ongurations were hosen from ea h group with the following hara teristi s:• Compa t distribution with bigger teles opes, optimized at LE.• Extended distribution with mid-size teles opes, optimized at HE.• Mix of the previous two: optimized over the whole energy range.The hoi es were validated omparing ea h representative to the rest of the ongurations of their respe tive groups. Then, deviations from the representativewere al ulated, quantifying the goodness of the representative with respe t to theothers in the same group.Finally, omparing the angular and energy resolution, the hosen ongurationswere still good representatives at LE, HE and over the whole energy range, seeFigures 8.4 and 8.5. On e the ongurations were su essfully lassied, the threerepresentatives (B, D, I) were used to simulate the spe tra and light urves of spe i sour es, whi h ould be interesting ases for CTA.8.2 A brief look at starting-up toolsIn the following se tions, the simulated response of CTA has been al ulated usingthe following ma ros provided by Daniel Mazin:

114Chapter 8. Simulations of CTA response to parti ular s ien e ases

Figure 8.3: Dierential sensitivity of the three representative ongurations: D(red), B (blue), I (green), in Crab units (C.U.) for 50 hours of observation.

Figure 8.4: Angular resolution (80% ontainment radius) of the three representative ongurations, see Figure 8.3, for best resolution (left) and best sensitivity (right).

8.3. Spe tral studies 115

Figure 8.5: Energy resolution of the three representative ongurations, see Figure8.3, for best resolution (left) and best sensitivity (right).• makeCTAspe .C : Simulates CTA spe tral points for a given sour e,• testCTA.C : Calls previous ma ro and draws the spe trum.Moreover, the simulations depend on the layout ongurations. Their performan eles ontain information on the ee tive area, the dierential sensitivity, the ba k-ground rate and the angular resolution (68% and 80% ontainment).The main ingredients for the program makeCTAspe .C are the intrinsi spe tralshape and ux of the hosen sour e, the aorementioned ee tive area and theobservational time. In the ase of sour es outside our Galaxy, the extragala ti ba kground light (EBL) attenuation an be taken into a ount. The sele tion riteriafor data points require more than 3 sigma, a number of ex ess events greater than10 and the ex ess 1% above the ba kground signal. If the number of ex ess events orsigni an e is too low, the spe trum is re-binned. Additionally, the ON events are al ulated randomly a ording to the Poisson statisti s with a given mean value.In the ase of an extended sour e, the ba kground ux is integrated. When thesimulation is ompleted, the ma ro provide the dierential energy spe trum and theintegral ux in a spe i energy range. On the other hand, testCTA.C handles theuser interfa e, by introdu ing the a tual onguration le of the CTA performan e,the fun tion of the sour e spe trum, the observation time and the size of the sour eextension. Also, attenuation an be applied if needed, as well as redening theenergy range.The use of these tools, thus, provide CTA simulated data points for the spe tralenergy distribution of dierent interesting sour es, as shown in subsequent se tions.8.3 Spe tral studiesEvery simulated layout for CTA provides dierent advantages and disadvantageswhen observing dierent obje ts in the sky. A tually testing the behaviour of in-

116Chapter 8. Simulations of CTA response to parti ular s ien e ases

Figure 8.6: Total gamma ray emission from a mole ular loud of mass 105M⊙ lo- ated at a distan e of 1 kp . The distan e between the mole ular loud and theSNR is 50, 100 and 200 p for left, enter and right panel, respe tively. The solid,dotted, and dashed lines refer to the emission at 2000, 8000 and 32000 years afterthe SN explosion. Private ommuni ation from S. Gabi i, based on gure 5 from[Gabi i et al. 2009.teresting gamma-ray sour es with CTA simulations will help to hoose whi h oneof those ongurations is favoured. In the next se tions, initial spe tral studies ondierent obje ts are shown in order to obtain estimations for unexplored energyranges, new spe tral features and model validations. These obje ts are in a ertainway related to osmi ray diusion, either in our galaxy (IC 443, mole ular louds lose to supernova remnants) or in others (starburst galaxies M82 and NGC 253).The ase of IC 443 is detailed in the next Chapter.8.3.1 Mole ular louds illuminated by CR from nearby SNRNon-thermal emission is expe ted to ome from mole ular louds, due to inter-a tions of osmi rays (CR) penetrating the loud. This emission would be en-han ed if the mole ular loud is in the proximity of a supernova remnant (SNR)[Gabi i et al. 2009, [Rodríguez Marrero et al. 2009. The CR spe trum has two ontributions:• Gala ti ba kground: hara terized by a steep spe trum, steady in time, thatpeaks in the GeV energy region;• Runaway CRs from SNR: present a hard spe trum, variable in time. Thisse ond peak appears at TeV energies and moves to lower and lower energies:the higher the energy, the earlier and faster CRs diuse away.When superimposing both ontributions, a urious on ave spe tra appears. ThisV-shape is ree ted in gamma-rays, given that above 100 MeV, the spe trum is dom-inated by emission oming from neutral pion de ay (π0 → γγ). If joint observationsof Fermi and ground based Cherenkov teles opes were arried out, the dete tion ofsu h a shape ould prove the presen e of a CR a elerator lose to mole ular louds.

8.3. Spe tral studies 117

energy E(TeV)0.01 0.02 0.1 0.2 1 2 3 4 5 10 20 30 100

)-1

s-2

dN

/dE

(T

eV c

m2

E

-1410

-1310

-1210

-1110

-1010

-910

intrinsic spectrum at 50 pcintrinsic spectrum at 100 pcintrinsic spectrum at 200 pcexpected spectrum (CTA) at 50 pcexpected spectrum (CTA) at 100 pcexpected spectrum (CTA) at 200 pc

Figure 8.7: Gamma-ray emission from a mole ular loud of 105M⊙ illuminated byan a elerator (whi h explosion o urred 2000 yrs ago) and simulated with layout onguration I, for 20 hours of observation. The mole ular loud is lo ated ata distan e of 1 kp from the observer and pla ed at dierent distan es from thea elerator: 50 (bla k), 100 (blue), 200 (red) p .One of the theoreti al studies on this topi , by [Gabi i et al. 2009, was usedin the present work to simulate a response with CTA. In Figure 8.6, the gammaemission oming from a mole ular loud near a SNR is plotted. The mole ular loud, lo ated at 1 kp from the observer, has a mass of 105M⊙. Considering thatthe main bulk of gamma-ray emission omes from π0 de ay, the urves show the uxat dierent epo hs after the supernova (SN) explosion and at dierent separationsfrom the a elerator. The larger this separation is between the mole ular loud andthe SNR, the lower is the level of CR ux that omes from the a elerator, as seenin the panels from left to right. There is also an evolution in time, proportionalto the square of the separation. In ea h panel, the peak moves to lower and lowerenergies with time (from right to left) due to the fa t that CR with lower and lowerenergies an progressively rea h the loud.Considering a xed time sin e the SN exploded (2000 years, solid urve in Figure8.6), dierent separations between the SNR and the loud are simulated in Figure8.7. Observing 20 hrs, pe uliar features in the spe tra an already be seen at theCTA sensitivity. A broad distribution with big teles opes at the enter (I) appearsas the best hoi e to over low and high energies. An additional omponent shouldarise at lower energies (Fermi range) due to the ontribution of CR from gala ti ba kground.

118Chapter 8. Simulations of CTA response to parti ular s ien e asesCTA will be able to distinguish mole ular louds illuminated by es aping osmi rays from a nearby SNR. The expe ted softening at lower energies, however, is hardlyrea hable, leaving the whole V-shape spe trum to be obtained by a ombination ofFermi and CTA data. In order to get enough statisti s, from 20 to 30 hours ofobservation time are needed. Moreover, the onguration I seems more suitable toallow a better understanding of every possible feature in the spe tra.Promising as these results appear, there are however some aveats inherent tothe model. Despite the large mass of the simulated loud, these type of obje tstend to be even more massive and extended. Therefore treating the sour e as point-like may be a bit optimisti . Also, the largest separation of the loud and theSNR, 200 p , is translated in angular distan e in the sky as ∼ 6 degrees, whi his larger than the average dete tors' eld of view. Thus, the asso iation betweentarget and a elerator be omes more di ult. Finally, the distan e at whi h themole ular loud is lo ated probably will need to be readjusted, given that there arenot so many andidates at 1 kp . W28, an already dete ted gamma-ray sour e, is onsidered a possible asso iation of a mole ular loud and a SNR and it is lo atedat 1.5 kp .8.3.2 Starburst galaxies M82 & NGC 253Starburst galaxies are hara terized by an enhan ed star formation and supernova(SN) explosion rate. Sin e SNe are believed to a elerate osmi rays, wheneverthey are found in dense environments, like the entral regions of starburst galaxies,gamma-ray emission has been predi ted to appear as a result of proton-proton inter-a tion and subsequent neutral pion de ay. Re ently, two of their most representativemembers, M82 and NGC 253, have been dete ted at GeV -TeV energies. In the lowerenergies, the Fermi satellite was responsible of the dete tion of both galaxies above100MeV [Abdo et al. 2010 . At higher energies, both Cherenkov array experimentsVERITAS and H.E.S.S. dete ted M82 above 700GeV [A iari et al. 2009a and NGC253 above 220 GeV [A ero et al. 2009, respe tively. See table 8.2 for details on thosedete tions.The very long observational time required to dete t these two starburst galaxiesby ground based gamma-ray experiments (more than a hundred hours), disesteemedfurther studies on these interesting obje ts. However, sin e these two galaxies are the losest ones to us (therefore, easier to resolve), a deeper knowledge will shade somelight on CR diusion and gamma-ray absorption, among other spe tral pe ularities.Hopefully, CTA will help in these aspe ts and may also open the door for futurepopulation studies on these and other related obje ts, like globular lusters.As a rst approa h to the problem, a simulation of the spe tra of both galaxieswas performed using H.E.S.S. and VERITAS data. The model explained in Chapter5, see also [Domingo-Santamaría & Torres 2005, [de Cea del Pozo et al. 2009b, hasbeen taken as input. In Figure 8.8, the urves that arise from the modeling of bothspe tra are shown, together with the data points obtained by the dierent gamma-ray teles opes. For M82 (left), the spa e parameter studied (green shaded region

8.3. Spe tral studies 119Table 8.2: Summary of the observational dete tions of M82 and NGC 253 by Fermi(at high energies, HE), VERITAS and H.E.S.S. (very high energies, VHE). Theenergy thresholds for the integral ux al ulations are: above 100MeV in the aseof Fermi LAT observations, above 700GeV for VERITAS, and above 220GeV forH.E.S.S. M 82 NGC 253Signi an e (σ) 6.8 (Fermi) 4.8 (Fermi)Time observed 1 year 1 yearHE Integral ux 1.6± 5stat ± 0.3sys 0.6± 0.4stat ± 0.4sys(10−8 m−2s−1)Dierential photon index 2.2± 0.2stat ± 0.05sys 1.95± 0.4stat ± 0.05sysSigni an e (σ) 4.8 (VERITAS) 5.2 (H.E.S.S.)Time observed 137 hrs 119 hrsVHE Integral ux 3.7± 0.8stat ± 0.7sys 5.5± 1.0stat ± 2.8sys(10−13 m−2s−1)Dierential photon index 2.5± 0.6stat ± 0.2sys -

Figure 8.8: Spe tral modeling of the gamma-ray emission of the two losest star-burst galaxies M82 (left) and NGC 253 (right). The overlaid blue box orre-sponds to the range of energies that CTA is predi ted to observe (from a fewtens of GeV to ∼ 100TeV). Data points in both plots orrespond to dete tionsby Fermi (diamond), VERITAS (star) and H.E.S.S. (square). Figures taken from[de Cea del Pozo et al. 2009a. For a omplete explanation, see Chapter 5.

120Chapter 8. Simulations of CTA response to parti ular s ien e asesfrom [de Cea del Pozo et al. 2009b) satisfa torily explains the dete ted high andvery high energy emission oming from the galaxy. One of the possible urves inthat region, stating a rate of 0.2 supernova explosions per year (solid bla k), is usedas input for the simulation. In the ase of NGC 253 (right), given the un ertaintyin the distan e to the galaxy (from 2.5 up to 3.9Mp ), a set of urves an a ountfor the ux at these high energies [de Cea del Pozo et al. 2009a. The urve sele tedfor the simulation orresponds to a distan e of 3.9Mp (solid red), given that it hasthe highest ux normalization at the CTA energy range.Simulating 30 hours of observation with CTA, the spe trum of the starburstgalaxy M82 is learly seen in Figure 8.9. Three dierent array ongurations weretested: the three representatives extra ted in se tion 8.1. The ompa t onguration(B) gives slightly better overage at lower energies (0.1TeV), whereas the moreextended ongurations (D or I) over ni ely the last part of the spe trum (up to10TeV). In ea h ase, a mu h better re onstru ted spe trum is a hieved, omparedto the existing data points from Fermi and VERITAS (in grey).However, when studying the ase of NGC 253, 30 hours were not enough toobtain a ondent dete tion of the galaxy. Indeed, at least 50 hours of observationappear to be the estimated time to guarantee a dete tion and re onstru tion ofthe spe trum of this sour e, as seen in Figure 8.10. Again, the three representativeswere tested (but only one is shown), and the best results are obtained when using anextended array onguration (either I or D). Although surprising as it may seem thatthe H.E.S.S. integral data point is below the CTA sensitivity, it should be remindedthat a spe ial analysis te hnique alled model analysis2 was used to obtain su hdete tion.The results for NGC 253 were not as promising as for M82, regarding a statisi- ally well re onstru ted spe trum in a reasonable time (around 30 hours). Nonethe-less, it seems that CTA will be able to dete t starburst galaxies. For a better under-standing of all possible spe tral features, onguration D or I seem more suitable,i.e., extended array ongurations with or without bigger teles opes at the enterthat provide a good re onstru tion of low energy gamma showers. Future workson this topi ould be determining how far away CTA an dete t other starburstgalaxies, so that a population study an be performed.8.4 Future workThe CTA observatory will open a wider range of energies (from tens of GeV to tensof TeV) to study gamma-ray astronomy, providing better sensitivity and angular res-olution. A thorough study is intended to be arried out, among other topi s, on the2The Model Analysis is based on a omparison and t of observed air shower images with apre omputed library of images [de Naurois & Rolland 2009. It was trained with simulated gammarays and with real osmi -ray data from ba kground elds [Ohm et al. 2009. The algorithm yieldsan improvement by a fa tor 1.5 to 1.7 in the statisti al signi an e of faint sour es ompared withthe standard image analysis [Aharonian et al. 2006a, [A ero et al. 2009, as veried with a numberof other gamma-ray sour es.

8.4. Future work 121

Figure 8.9: M82 spe trum as simulated for CTA for 30 hrs with onguration B(top), D ( enter), I (bottom). The grey points orrespond to observational data: atlower energies, an upper limit (68% .l.) from Fermi, and at higher energies, thedierential data points and an upper limit (99% .l.) from VERITAS.

122Chapter 8. Simulations of CTA response to parti ular s ien e ases

Figure 8.10: NGC 253 spe trum as simulated for CTA in 50 (left) and 70 hours(right) with onguration D. The observational data (grey) orresponds to an up-per limit (68% .l.) from Fermi ; and H.E.S.S. integral data point ( onverted todierential ux)origin of Gala ti CR, their propagation within galaxies, and their intera tion withthe environment. For a long time, supernova remnants (SNR) have been thoughtto be the sour es of CR. To test this hypothesis with CTA, population studies areneeded in order to in rease the number of dete ted SNR and improve statisti s onthese obje ts. In addition to this issue, a dis ussion on sour e onfussion (due to thelarge number of sour es to be dete ted in the gala ti plane) should be addressed.Another test on the SNRs as the origin of CRs ould be done by improving thequality of the spatial orrelation studies between γ-rays and X-rays, and betweenγ-rays and gas distribution at SNRs. This last probe is yet to be arried out withCTA.If SNRs are indeed responsible for the origin of gala ti CRs, then the most ex-tended explanation for their a eleration is via diusive sho k a eleration (DSA).To better understand this pro ess with CTA, some studies have been already per-formed on the uto region in SNR gamma ray spe tra (from IC 443, see nextChapter, and also RX J1713.7-3946 [Aharonian et al. 2004a). The dete tion ofvery high energy photons (i.e. signi antly above 10 TeV) from any SNR wouldhelp to dis riminate between hadroni and leptoni models, sin e the Klein-Nishinaee t would dramati ally suppress the gamma ray ux due to inverse Compton inthis energy domain. Another topi that would shed some light on this aspe t isspatially resolving the spe tra of SNRs possible thanks to the improved angularresolution.To better understand CR propagation it will be useful to study mole ular louds that present gamma-ray emission oming both from the CR ba kgroundand nearby a elerators (i.e. SNRs). In the rst ase, the so- alled passive mole -ular louds ould be used to tra e the spatial distribution of CRs in the Galaxy[Gabi i 2008, Aharonian 2001, Issa & Wolfendale 1981, Casanova et al. 2010. The ase of mole ular louds asso iated with SNR has been presented in se tion 8.3.1.These studies are important be ause the dete tion of the gamma ray radiation om-

8.4. Future work 123ing from these mole ular louds might onstitute an indire t eviden e for the a el-eration of CRs at SNRs and, sin e the hara teristi s of the radiation are expe tedto depend on the CR diusion oe ient, these studies might be used to onstrainits value in the vi inity of CR a elerators. Another interesting work related tothis issue would be studying the ase in whi h the a elerator is pla ed inside themole ular loud.Finally, regarding the studies on starburst galaxies, as previously mentioned, abreakthrough would ome from the dis overy of more sour es of this type, given theimprovement in sensitivity. In any ase, it will also be important to explore thealready dete ted starburst galaxies, not only their spe tra (as in se tion 8.3.2) butalso their extension and morphology. Regarding further spe tra features, any steep-ening at the highest energies ould explain extra omplexity in the radiative parti ledistribution, as presently needed to explain observational data and, ultimately, putlimits on parti le a eleration in these systems. Last, investigating ux variabilitywould give reliability to CR-related emission s enarios in any extragala ti sour e.

Chapter 9IC 443 in MAGIC stereo andprospe ts with CTAContents9.1 Proposal and observations with MAGIC stereo . . . . . . . 1289.2 Analysis and results . . . . . . . . . . . . . . . . . . . . . . . . 1299.3 IC443 as seen in CTA . . . . . . . . . . . . . . . . . . . . . . . 131As already introdu ed in Chapters 3 and 4, IC 443 is a well known supernovaremnant (SNR) in gamma-ray astronomy. One of the most striking fa ts is that the entroids of these dete tions at high and very high energies do not oin ide, as anbe seen in Figure 9.2. There is a lear displa ement between the MAGIC/VERITASsour e and the EGRET/Fermi one. The EGRET entral position is lo ated dire tlytowards the enter of the radio shell of the SNR, where the Fermi sour e 0FGLJ0617.4+2234 is also lo ated. Meanwhile, the MAGIC sour e is (in Gala ti oor-dinates) south of it, lose but beyond the 95% onden e level ( .l.) ontour of theEGRET dete tion (bla k ontours in Figure 9.1). As noted in [Albert et al. 2007b,the MAGIC sour e entroid is positionally oin ident with a giant loud in frontof the SNR (see also 12CO ontours in Figure 9.1). If the Fermi sour e is takenas referen e, the entroid is displa ed 0.05 degrees from EGRET, 0.15 degrees fromMAGIC, and 0.12 degrees from VERITAS entroids.Not only is there a dieren e in the position, but there also seems to be a breakin the spe trum between these two ranges of energies. If an extrapolation of thespe trum of the EGRET/Fermi sour e is performed into the VHE regime, a higherux and harder spe trum than the one observed by MAGIC is obtained. In fa t,su h a hange appears to be ree ted on the Fermi data. For a spe tral energydistribution (SED) from 200MeV up to 50GeV, it seems better explained with,for instan e, a broken power law that rolls over at about 3GeV, steepening at thehighest energies to mat h in slope the one that is found by MAGIC.The region of the sky in whi h IC443 an be found ontains a pulsar wind nebula(PWN) CXOU J061705.3+222127 [Olbert et al. 2001, [Bo hino & Bykov 2001.However, in all energy bands, the entroid of the orrespondingly dete ted sour esis in onsistent with this PWN (and the putative pulsar). For instan e, the entroidof the PWN is 0.26 degrees from the Fermi LAT entroid. Given the above fa tors,a possible way of explaining a relation between the MAGIC sour e and the SNR is

126 Chapter 9. IC 443 in MAGIC stereo and prospe ts with CTA

l [deg]

b [

deg

]

2.6

2.8

3

3.2

3.4

sr

-5ex

cess

eve

nts

/ 10

-20

0

20

40

60

80

100

120

140

160

188.5189189.5

PSF

Figure 9.1: Sky map of gamma-ray andidate events (ba kground subtra ted) in thedire tion of MAGIC J0616+225 for an energy threshold of about 150GeV in gala ti oordinates. Overlayed are 12CO emission ontours ( yan) showing the position oftheir maxima oin iding with the MAGIC dete tion, and ontours of 20 m VLAradio data (green), X-ray ontours from Rosat (pu12rple) and gamma-ray ontoursfrom EGRET (bla k). The white star denotes the position of the pulsar CXOUJ061705.3+222127. The bla k dot shows the position of the 1720MHz OH maser.The white ir le shows the MAGIC I PSF of σ = 0.1. From [Albert et al. 2007b

127

Figure 9.2: Lo ations and extensions of the 4 gamma-ray sour es. Centroid positionsare marked with dierent symbols: EGRET (blue triangle), MAGIC (red downsidetriangle), VERITAS entroid (green star) and Fermi LAT (bla k diamond). Therespe tive lo alization errors are shown as rosses. Best-t spatial extensions of theFermi ( ross-hat hed band) and VERITAS (striped green band) sour es are drawnas rings with radii of θext68 and widths of ±1σ error. The PWN lo ation is shown as amagenta dot. Contours are the lo ations and shapes of the lo al sho ked mole ular louds taken from [Huang et al. 1986. Figure taken from [Abdo et al. 2010g

128 Chapter 9. IC 443 in MAGIC stereo and prospe ts with CTAa hieved through the diusion of osmi rays (CR). MAGIC J0616+225 is onsis-tent with the interpretation of CR intera ting with the giant mole ular loud lyingin front of the remnant, produ ing no ounterpart at lower energies. In the modelby [Torres et al. 2008, the nearby EGRET/Fermi sour e an be produ ed by thesame a elerator, and in this ase, a o-spatial MAGIC sour e is not expe ted. Forfurther details on the model, see Chapters 3 and 4.In order to further bridge the energy gap between Fermi and MAGIC, a proposalfor observing IC443 with MAGIC stereo system was made. Details of the proposaland subsequent analysis are explained in the following se tions.9.1 Proposal and observations with MAGIC stereoAs it has been previously stated, MAGIC I observations yield to the dete -tion of a new sour e of gamma-rays, J0616+225. This sour e is lo ated at(RA,DEC)=(06h16m43s,+2231'48), with a statisti al positional error of 1.5', anda systemati error of 1'. A simple power law was tted to the measured spe tralpoints through:dNγ

dAdtdE= (1.0 ± 0.2) × 10−11

(

E

0.4TeV

)−3.1±0.3

cm−2s−1TeV−1 (9.1)with quoted errors being statisti al. The integral ux of MAGIC J0616+225 above100GeV is about 6.5% of the Crab Nebula.MAGIC observations in stereo mode ould provide, together with Fermi, on-tinuous overage of the sour e from 100MeV up. It should be noted, though, thatby the time the proposal was made, no Fermi results were released. The intendedaims, ba k then, were as follows:• To analyze the position of the sour e as a fun tion of dierent low-to-highenergy uts. Thus, he king whether the sour e moves towards the enter(outskirts) of the SNR with lower (higher) energy threshold.• To nd whether the spe trum of the MAGIC dete tion be omes harder as theenergy threshold de reases.• To extra t morphologi al information (i.e., the extension of the sour e) andinvestigate a possible dependen y on the energy threshold.The whole proposal was driven by the already explained s enario in whi h CRwould es ape from the SNR and intera t with the mole ular loud in front of it.The mole ular loud environment surrounding IC 443 and its possible onne tionwith the EGRET gamma-ray sour e was studied by e.g., [Torres et al. 2003. Thereis a large amount of mole ular mass (∼ 104M⊙) onsistent with the distan e tothe SNR, orresponding to a velo ity range of −20 to 20 km/s. The highest COintensity dete ted is dire tly superimposed on the entral position of the MAGICsour e.

9.2. Analysis and results 129A ording to that s enario, the displa ement between EGRET and MAGICsour es would have a physi al origin. Su h displa ement would ome from thedierent properties of the proton spe trum at dierent lo ations, whi h in turn isprodu ed by the diusion of CRs from the a elerator (IC 443) to the target. Spe i predi tions for future observations an be made as a result of the model presentedby [Torres et al. 2008. At high energies, a morphologi al and spe tral hange fromthe position of the loud (i.e. the enter of MAGIC J0616+225) towards the enterof IC 443 should be seen. At a morphologi al level, when observing at lower andlower energies, the radiation will be dete ted loser to the position of the SNR shell.At a spe tral level: su ient statisti s should show that the lower the gamma-rayenergy, the harder the spe trum is. Measurements by MAGIC stereo and Fermi ould lead to the empiri al determination of the diusion oe ient in the medium.Originally, the observations were planned during winter 2009 - 2010, but due tobad weather onditions, only a small fra tion of the granted time was spent on thedata taking between January and February 2010. A preliminar analysis of the datais presented below.9.2 Analysis and resultsStereos opi observations were arried out in wobble mode, at two positions 0.4away from the enter. The reason for this observational mode was that the skyregion around the sour e has a relatively high and non-uniform level of ba kgroundlight. Within a distan e of ∼1.5 from the sour e, there are two bright stars, one ofthem is brighter than 8 mag, eta Gem with 3.28 mag, and the other (a bit furtheraway) is mu Gem, with 2.88 mag. By hoosing one of the two wobble o-positions oin iding with the brigthest star (eta Gem), a similar sky brightness distribution an be a hieved in both wobble positions.Although the sour e ould be observed at good low zenith angles from November2009 onwards, the stereos opi mode in MAGIC was still under omissioning. Thestart of the data taking was then postponed until De ember 2009. By that time,observations of the Crab Nebula (standard andle in gamma-ray astronomy) wouldhave been perfomed and analysed, and a deeper knowledge about the new system ould have been a hieved. However, due to bad weather, only 7 hours were observedbetween January and February 2010. The proposal was re-submitted in the next y le (2010 - 2011) to obtain more observational time, but unfortunately no moretime was granted.The observations were taken at zenith angles below 30, to a hieve the ne essarylow energy threshold this study requires. At La Palma, IC 443 ulminates at about6 zenith angle (ZA). Sin e the already dete ted spe trum was found to be steep(Γ = −3.1± 0.3), the intended study an better benet from low energy threshold(thus the low zenith angle) and greater senstivity (granted by the new stereos opi system).Given the small amount of time observed (5 hours after quality uts), the

130 Chapter 9. IC 443 in MAGIC stereo and prospe ts with CTA

Figure 9.3: Plot showing the θ2 distribution for the MAGIC sour e in the region ofIC 443. The shapes of the ON and OFF distributions agree well with ea h other.A hint of signal is already seen in barely 3 hours of observations.originally planned study ould not be performed, and a qui k analysis was ar-ried out. The data olle ted was analyzed in the MARS analysis framework([Moralejo et al. 2009), standard software for the analysis 1 of MAGIC data. Due tothe novelty of the stereo system, the analysis tools used were still under developmentby that time.In order to look for gamma-rays oming from the pointed sour e, standard utsfor stereo were applied in hadronness and size (0.2 and 100, respe tively). In the θ2distribution 2 from Figure 9.3, an ex ess at lower angles an barely be seen, implyingthat a hint of signal ( 2σ) an be dete ted. However, it is not enough for a spe tralre onstru tion, nor the detailed study in the spe trum and morphology (in bins ofenergy). The gamma-ray hint of signal is positionally oin ident with the previouslypublished MAGIC J0616+225 sour e. A skymap above 200GeV, with the same utsas the θ2 plot, an be seen in Figure 9.4.1An example of one analysis of MAGIC data an be found at [Aleksi¢ et al. 20102a denition an be found in se tion 7.3

9.3. IC443 as seen in CTA 131

Figure 9.4: Skymap (map of signi an es) of the SNR IC 443 (left) and distributionof signi an es (right) of the same region. The bla k ross marks the position of theMAGIC sour e J0616+225. The white ir le shows the PSF of the MAGIC stereosystem.Table 9.1: Power law (p.l.) index for MAGIC-like or VERITAS-like spe trum forIC 443, when simulating CTA response for 50 hrs in onguration I.Observed p. l. index CTA simulated p. l. indexMAGIC-like spe trum 3.1± 0.3 3.14 ± 0.04VERITAS-like spe trum 2.99 ± 0.38 2.98 ± 0.039.3 IC443 as seen in CTAHaving in mind similar goals as in the MAGIC stereo proposal (see se tion 9.1),a preliminar spe tral study on IC 443 was performed for CTA. The response ofthe instrument was simulated using the tools and the three representative layout ongurations des ribed in the previous Chapter 8. The energy range that CTAwould over goes from a few GeVs to hundreds of TeVs. Initially, the observedspe tra (by satellite and ground based teles opes) were used as the intrinsi one,spe i ally the ones from the MAGIC and VERITAS experiments and the se ondpart of the broken power-law provided by the Fermi satellite. The dened energyrange was from 0.02 to 0.07 TeV for the Fermi spe trum, and from 0.07 to 7TeVfor the MAGIC and VERITAS spe tra. The CTA response was simulated usingFermi/MAGIC and Fermi/VERITAS spe tra to see if CTA an distinguish them.Every intrinsi power law has dierent index: 2.56 ± 0.11 for Fermi, 3.1 ± 0.3 forMAGIC and 2.99±0.38 for VERITAS. As seen in Figure 9.5, depending on the hoi emade for the intrinsi spe trum, CTA will be able to distinguish whether the powerlaw index is similar to the MAGIC's or the VERITAS's slope, when extrapolated tohigher energies. This an also be veried in table 9.1, where the spe i numbersderived from Figure 9.5 are shown.In the simulations, several realisti observational times were tested to establishwhen new results would be a hievable. In order to have better resolution at low

132 Chapter 9. IC 443 in MAGIC stereo and prospe ts with CTA

Figure 9.5: CTA simulated spe trum of IC 443 with onguration I (green urve)and 50 hours. The input spe tra (grey lines) orrespond: at lower energies tothe power law published from Fermi above 3GeV, whereas at higher energies thepower law index is taken from the measurements by MAGIC and VERITAS. Thegrey symbols show the a tual data from previous experiments. The simulated datapoints for CTA above 100 GeV an be tted to a power law, whi h index dependson the assumed initial spe tra (bla k dots for MAGIC, blue ones for VERITAS).

9.3. IC443 as seen in CTA 133Table 9.2: Statisti s of two ttings to the spe trum for IC 443, when simulatingCTA response for 20 hrs in onguration I. The simulated data points use theintrinsi spe tra from Figure 9.6: a power law with a ut o at a few TeV. The twottings, then, are a power law with or without a ut o. The rst number is thet probability, whi h values should be in the range between 5− 60%, the losest to50%, the better. The se ond set of numbers represents the redu ed χ2 (dened asχ2 divided by the number of degrees of freedom), and for a good t, it should be loser to 1. Cut-o at 5TeV Cut-o at 10TeVPower law t 2.21% 9.69%19.38 / 9 14.79 / 9Power law + ut-o t 61.09% 39.7%6.38 / 8 8.38 / 8energies (several GeV), CTA needs at least 20 hours of observation time, indepen-dently of the layout of the used onguration. On the other hand, at high energies(several TeV), CTA needs more than 10 hours to a hieve better results than withalready operating experiments. At these energies, the ompa t onguration (B) isnot as useful as the other more extended ongurations (I or D).Another interesting point fo uses on whether CTA ould distinguish the exis-ten e of a ut-o in the power law spe trum above the energies of MAGIC andVERITAS. The last data points measured by ground based teles opes are between1 and 2 TeV. If a ut-o was to be found lose to this energy, it ould be dete tedeasier by an extended onguration as, for example, onguration D (with the bestsensitivity in this energy range). However, a higher ut-o (above 10 TeV) ttingis hard to dis riminate from a simple power-law tting, independently of the used onguration, see Figure 9.6. To obtain enough statisti s, more than 20 hours ofobservation time are needed. But even in that ase, distinguishing between a power-law and a power-law with a ut-o t is not guaranteed, as an be seen in table9.2.In on lusion, the prospe ts of having CTA as the next generation of Cherenkovteles opes will bring a solution to some interesting physi s ases. In the ase ofthe IC 443 spe trum, CTA will be able to present major overage of the spe trum than the one presentely a hieved by MAGIC or VERITAS and with more a ura y(improving 10% the error bars). With already merely 20− 30 hours of observationtime, new and meaningful results ould be a hieved with enough statisti s. However,if a ut-o were to appear above 5TeV, CTA might not be able to distinguish itfrom a simple power-law. For a better understanding of all its possible spe trafeatures, an array onguration with an extended distribution but also with big sizeteles opes seems more suitable for all this kind of studies (like the onguration I).The next step leading to an understanding of the neighbourhood around the

134 Chapter 9. IC 443 in MAGIC stereo and prospe ts with CTA

energy E(TeV)0.01 0.02 0.1 0.2 1 2 3 4 5 10 20 30 100

)-1

s-2

dN

/dE

(T

eV c

m2

E

-1410

-1310

-1210

-1110

-1010

-910

expected spectrum (CTA) for IC 443

Fermi, MAGIC, VERITAS

power law fit

power law with cutoff (at 5 TeV) fit

energy E(TeV)0.01 0.02 0.1 0.2 1 2 3 4 5 10 20 30 100

)-1

s-2

dN

/dE

(T

eV c

m2

E

-1410

-1310

-1210

-1110

-1010

-910

expected spectrum (CTA) for IC 443

Fermi, MAGIC, VERITAS

power law fit

power law with cutoff (at 10 TeV) fit

Figure 9.6: The spe trum of IC443 is simulated with layout CTA-I for 50 hours ofobservation, assuming an input spe trum in the form of a power law with a ut o of5 (top) and 10 (bottom) TeV. The power law index is the one published by MAGIC(-3.1). The highest energy ut-o that ould be resolved is at 5 TeV, at higherenergies (i.e. 10 TeV) a tting with power law with a ut o (red) is statisti allyundistinguishable from a simple power law (green) in the observed CTA range ofenergies.

9.3. IC443 as seen in CTA 135SNR IC 443 will be a morphologi al study and an a tual he k of a possible energydependen e.

Chapter 10Con lusions and future work10.1 Final remarksOne of the most distinguishing aspe ts on studies about diuse emission is thatsome spe tral signatures may serve as an identi ation of the underlying me hanismprodu ing the gamma rays. In other words, they may help to dis ern whi h kind ofa elerator and under whi h diusion properties osmi rays (CRs) propagate.Following the topi on CR diusion, in Chapters 3 and 4, a theoreti al model hasbeen presented explaining the phenomenology around the SNR IC 443 at energiesabove 100MeV, pla ing an emphasis on the displa ement between the sour es athigh and very high energy (HE and VHE respe tively). The displa ement is gener-ated by the dierent properties of the proton spe trum at dierent lo ations, i.e.,how separated are the SNR and the mole ular loud in front of it. These dieren esare produ ed by the diusion of CRs from the a elerator (IC 443) to the target.The VHE sour e dis overed by the MAGIC teles ope is interpreted as a delayedTeV emission of CRs diusing from the SNR. Whereas the HE sour e an be ex-plained as produ ed by the same a elerator, without produ ing a o-spatial VHEsour e. Some of the predi tions made before the Fermi teles ope was laun hedwere orre tly fulllen. At a morphologi al level, when the energy is lower, thegamma-ray radiation has been dete ted loser to the SNR shell. At a spe tral level,su ient statisti s have shown that the spe trum gets harder when going towardslower gamma-ray energies.However, the Fermi results have extended the energy domain of the SED mu hbeyond what was possible for its prede essor. The spe trum appears harder thanwhat was previously suggested, presenting an almost at SED up to 10GeV, with aroll-over in the spe trum between 10 and 100GeV. The original parameters of themodel needed a sligth modi ation in their values: mainly, the lo ation and masses ofthe overtaken louds. The ombined spe tra measured by VHE and Fermi teles opes an be explained in a s enario where a giant loud produ es a peak at about theFermi spe tral turnover. The distan e between the loud and the SNR shell issmaller than previously predi ted and the CR are ae ted by a faster diusion. Thediusion oe ient is lower and the osmi -ray density is higher than the Earth-values of both magnitudes. Nonetheless, there are left some un ertainties in theamount and the lo alization of the target mole ular mass, and the density of thismole ular material. Those aveats should be addressed in subsequent studies onthis an similar sour es, like the southern W28 dete ted by H.E.S.S., or the northernW51C reported by MAGIC and VERITAS.

138 Chapter 10. Con lusions and future workIn Chapter 5, a model for starburst galaxies is presented and su esfully on-rmed by the more re ent observations. Su h galaxies have an enhan ement bothin the star formation and supernova (SN) explosion rate, and dense environmentsat their enters. Very energeti gamma rays are produ ed due to CRs intera tionswith ambient nu lei and subsequent π0 de ay. The main model des ribed multi-frequen y (from radio to near infrared) and multi-messenger (neutrinos) predi tionsfor the photon emission oming from the entral region of M82. The model pre-sented is onsistent along the ele tromagneti spe trum, and allows to tra k theemission to one and the same original osmi -ray population. This population isa onsequen e of every leptoni and hadroni hannel, in luding both primary andse ondary parti les. The model explains reasonably well both the HE and VHEemission oming from the two losest starburst galaxies M82 and NGC 253, withina range of the explored parameters already mentioned. Now, the opportunity to ontinue working in these obje ts is open, and further studies on CR enhan ement,X-ray predi tions, and features in the spe tra below 300GeV and above 1TeV areexpe ted.10.2 Future workThe Thesis on ludes with a few initial studies on key obje ts with the forth omingCherenkov Teles ope Array (CTA) in Chapter 8:• The ability to distinguish mole ular louds illuminated by es aping osmi rays from a nearby SNR, although the whole V-shape spe trum has to beobtained by a ombination of Fermi and CTA data. The aveats of this studyare a onsequen e of the theoreti al model behind it: su h large masses formole ular louds tend to be extended and morphologi al studies are required.Also, a range of un ertainty is needed for both distan es and mass of mole ular louds, be ause asso iation between SNR and mole ular louds gets harder thewider the separation.• Con erning starburst galaxies, a mu h better re onstru ted spe trum shouldbe a hieved, ompared to the existing one for M82. Further studies are ex-pe ted to report on the possible uto in the proton spe tra aroung TeV en-ergies. However, the other andidate from this sour e- lass, NGC 253, is notso easily dete ted, therefore an eort should be made towards a hieving ananalysis, like the H.E.S.S model analysis that guaranteed its dete tion. On ethis is a omplished, diuse radiation from other starburst galaxies pla edfurther away may be dete ted, so that a population study an be initiated.• The study on SNR IC 443 tried to stablish the energy dependen e of thegamma-ray entroid, the photon index of the spe trum and the extension. Inthe spe tral study it ould be seen that the existing statisti s on photon indexwere signi antly improved. However, the opposite happened respe t to the

10.2. Future work 139possibility to distinguish a ut-o over few TeVs, where su ient statisti swere not a hieved.CTA will have better sensitivity (one order of magnitud at 1TeV) and betterangular resolution. This is right now the future of the ground-based astronomyat the highest energies. The window will be open to unveil the origin of CR andthe SNR as their main a elerator. On the latter obje ts, population studies andspe tral studies looking for a ut-o (to disentangle hadroni /leptoni models)are expe ted to be ome ommon. Related with this topi , the natural targetsfor these a elerators, the mole ular louds nearby, will also be studied. Otherproje ts await, like multifrequen y (X-ray, gas) and multi-messenger (neutrinos)observations and models, together with morphologi al and extension-relatedstudies, where a dependen y on the energy might be tested. To avoid the sour e onfussion, more detailed diusion skymaps will be developed. And with respe t toextragala ti sour es, like the starburst galaxies, extension and morphology will beexplored, apart from detailed spe tra and variability tests (like in blazar and othera tive gala ti nu lei obje ts).All and all, this Thesis is just the tip of the i eberg: the eld is wide and manyprospe ts for future work will ontinue to open as our understanding on gamma-ray astronomy progresses. The studies presented in this Thesis simply representa starting point, and even more ex iting topi s lay ahead. The journey has justbegun. . .

Bibliography[Abdo et al. 2008 A. A. Abdo, M. A kermann, W. B. Atwood, L. Baldini, J. Ballet,G. Barbiellini, M. G. Baring, D. Bastieri, B. M. Baughman, K. Be htoland 175 oauthors. The Fermi Gamma-Ray Spa e Teles ope Dis overs thePulsar in the Young Gala ti Supernova Remnant CTA 1. S ien e, vol. 322,no. 5905, pages 12181221, 2008. (Cited on page 11.)[Abdo et al. 2009a A. A. Abdo, M. A kermann, M. Ajello, B. Anderson, W. B.Atwood, M. Axelsson, L. Baldini, J. Ballet, G. Barbiellini, M. G. Baringand 170 oauthors. Dete tion of 16 Gamma-Ray Pulsars Through BlindFrequen y Sear hes Using the Fermi LAT. S ien e, vol. 325, no. 5942, pages840844, 2009. (Cited on page 11.)[Abdo et al. 2009b A. A. Abdo, M. A kermann, M. Ajello, K. Asano, W. B. At-wood, M. Axelsson, L. Baldini, J. Ballet, G. Barbiellini, M. G. Baring and199 oauthors. A limit on the variation of the speed of light arising fromquantum gravity ee ts. Nature, vol. 462, no. 7271, pages 331334, 2009.(Cited on page 13.)[Abdo et al. 2009 A. A. Abdo, M. A kermann, M. Ajello, W. B. Atwood, M. Ax-elsson, L. Baldini, J. Ballet, D. L. Band, G. Barbiellini, D. Bastieri and 200 oauthors. Fermi/Large Area Teles ope Bright Gamma-Ray Sour e List. As-trophysi al Journal Supplement, vol. 183, no. 1, pages 4666, 2009. (Citedon page 101.)[Abdo et al. 2009d A. A. Abdo, M. A kermann, M. Ajello, W. B. Atwood, M. Ax-elsson, L. Baldini, J. Ballet, G. Barbiellini, M. G. Baring, D. Bastieri and192 oauthors. Dete tion of 16 Gamma-Ray Pulsars Through Blind Fre-quen y Sear hes Using the Fermi LAT. S ien e, vol. 325, no. 5942, pages840844, 2009. (Cited on page 109.)[Abdo et al. 2009e A. A. Abdo, M. A kermann, M. Ajello, W. B. Atwood, M. Ax-elsson, L. Baldini, J. Ballet, G. Barbiellini, M. G. Baring, D. Bastieri and192 oauthors. A Population of Gamma-Ray Millise ond Pulsars Seen withthe Fermi Large Area Teles ope. S ien e, vol. 325, no. 5942, pages 848852,2009. (Cited on page 11.)[Abdo et al. 2009f A. A. Abdo, M. A kermann, M. Ajello, W. B. Atwood, M. Ax-elsson, L. Baldini, J. Ballet, G. Barbiellini, D. Bastieri, B. M. Baughmanand 157 oauthors. Dete tion of High-Energy Gamma-Ray Emission fromthe Globular Cluster 47 Tu anae with Fermi. S ien e, vol. 325, no. 5942,pages 844848, 2009. (Cited on page 11.)

142 Bibliography[Abdo et al. 2009g A. A. Abdo, M. A kermann, M. Ajello, M. Axelsson, L. Baldini,J. Ballet, G. Barbiellini, D. Bastieri, B. M. Baughman and 167 oauthors.Modulated High-Energy Gamma-Ray Emission from the Mi roquasar CygnusX-3. S ien e, vol. 326, no. 5959, pages 15121516, 2009. (Cited on page 10.)[Abdo et al. 2009h A. A. Abdo, M. A kermann, M. Arimoto, K. Asano, W. B.Atwood, M. Axelsson, L. Baldini, J. Ballet, D. L. Band, G. Barbiellini and244 oauthors. Fermi Observations of High-Energy Gamma-Ray Emissionfrom GRB 080916C. S ien e, vol. 323, no. 5922, pages 16881692, 2009.(Cited on page 13.)[Abdo et al. 2009i A. A. Abdo, B. T. Allen, T. Aune, D. Berley, C. Chen, G. E.Christopher, T. DeYoung, B. L. Dingus, R. W. Ellsworth, M. M. Gonzalezand 20 oauthors. Milagro Observations of Multi-TeV Emission from Gala -ti Sour es in the Fermi Bright Sour e List. Astrophysi al Journal Letters,vol. 700, no. 2, pages 127131, 2009. (Cited on pages 101, 105, 109, 110and 175.)[Abdo et al. 2010a A. A. Abdo, M. A kermann, M. Ajello, A. Allafort, E. Antolini,W. B. Atwood, M. Axelsson, L. Baldini, J. Ballet, G. Barbiellini and 225 oauthors. Fermi Large Area Teles ope First Sour e Catalog. Astrophysi- al Journal Supplement, vol. 188, no. 2, pages 405436, 2010. (Cited onpages 104, 110 and 175.)[Abdo et al. 2010b A. A. Abdo, M. A kermann, M. Ajello, W. B. Atwood, M. Ax-elsson, L. Baldini, J. Ballet, G. Barbiellini, M. G. Baring, D. Bastieri and207 oauthors. The First Fermi Large Area Teles ope Catalog of Gamma-rayPulsars. Astrophysi al Journal Supplement, vol. 187, no. 2, pages 460494,2010. (Cited on pages 109, 110 and 175.)[Abdo et al. 2010 A. A. Abdo, M. A kermann, M. Ajello, W. B. Atwood, M. Ax-elsson, L. Baldini, J. Ballet, G. Barbiellini, D. Bastieri, K. Be htol and 161 oauthors. Dete tion of Gamma-Ray Emission from the Starburst GalaxiesM82 and NGC 253 with the Large Area Teles ope on Fermi. The Astro-physi al Journal Letters, vol. 709, no. 2, pages L152L157, 2010. (Cited onpages 13, 58, 74 and 118.)[Abdo et al. 2010d A. A. Abdo, M. A kermann, M. Ajello, W. B. Atwood, L. Bal-dini, J. Ballet, G. Barbiellini, D. Bastieri, B. M. Baughman, K. Be htol and172 oauthors. Fermi Gamma-Ray Imaging of a Radio Galaxy. S ien e,vol. 328, no. 5979, pages 725728, 2010. (Cited on page 12.)[Abdo et al. 2010e A. A. Abdo, M. A kermann, M. Ajello, W. B. Atwood, L. Bal-dini, J. Ballet, G. Barbiellini, D. Bastieri, K. Be htol, R. Bellazzini and 206 oauthors. Gamma-Ray Emission Con urrent with the Nova in the Symbi-oti Binary V407 Cygni. S ien e, vol. 329, no. 5993, pages 817821, 2010.(Cited on page 9.)

Bibliography 143[Abdo et al. 2010f A. A. Abdo, M. A kermann, M. Ajello, L. Baldini, J. Ballet,G. Barbiellini, M. G. Baring, D. Bastieri, B. M. Baughman, K. Be htol and171 oauthors. Gamma-Ray Emission from the Shell of Supernova RemnantW44 Revealed by the Fermi LAT. S ien e, vol. 327, no. 5969, pages 11031106, 2010. (Cited on pages 9 and 169.)[Abdo et al. 2010g A. A. Abdo, M. A kermann, M. Ajello, L. Baldini, J. Ballet,G. Barbiellini, D. Bastieri, B. M. Baughman, K. Be htol, R. Bellazzini and162 oauthors. Observation of Supernova Remnant IC 443 with the FermiLarge Area Teles ope. Astrophysi al Journal, vol. 712, no. 1, pages 459468,2010. (Cited on pages 42, 43, 45, 127 and 177.)[Abdo et al. 2010h A. A. Abdo, M. A kermann, M. Ajello, L. Baldini, J. Ballet,G. Barbiellini, D. Bastieri, B. M. Baughman, K. Be htol, R. Bellazzini and170 oauthors. PSR J1907+0602: A Radio-Faint Gamma-Ray Pulsar Pow-ering a Bright TeV Pulsar Wind Nebula. The Astrophysi al Journal, vol. 711,no. 1, pages 6474, 2010. (Cited on page 43.)[Abdo et al. 2010i A. A. Abdo, D. Parent, J. E. Grove, G. A. Caliandro,M. Roberts, S. Johnston and M. Chernyakova. Fermi LAT Dete tion ofGeV Gamma-Ray Emission from the Binary System PSRB1259-63. TheAstronomer's Telegram, vol. 3085, pages 1+, 2010. (Cited on page 10.)[Abdo et al. 2011 A. A. Abdo, M. A kermann, M. Ajello, A. Allafort, L. Baldini,J. Ballet, G. Barbiellini, D. Bastieri, K. Be htol, R. Bellazzini and 158 oau-thors. Gamma-Ray Flares from the Crab Nebula. S ien e, vol. 331, no. 6018,pages 739742, 2011. (Cited on page 10.)[A iari et al. 2009a V. A. A iari, E. Aliu, T. Arlen, T. Aune, M. Bautista,M. Beili ke, W. Benbow, D. Boltu h, S. M. Bradbury and 83 oauthors.A onne tion between star formation a tivity and osmi rays in the star-burst galaxy M82. Nature, vol. 462, no. 7274, pages 770772, 2009. (Citedon pages 13, 74 and 118.)[A iari et al. 2009b V. A. A iari, E. Aliu, T. Arlen, T. Aune, M. Bautista,M. Beili ke, W. Benbow, S. M. Bradbury, J. H. Bu kley, V. Bugaev and82 oauthors. Observation of Extended Very High Energy Emission from theSupernova Remnant IC 443 with VERITAS. Astrophysi al Journal Letters,vol. 698, no. 2, pages L133L137, 2009. (Cited on page 41.)[A iari et al. 2009 V. A. A iari, E. Aliu, T. Arlen, M. Bautista, M. Beili ke,W. Benbow, S. M. Bradbury, J. H. Bu kley, V. Bugaev, Y. Butt and 383 oauthors. Radio Imaging of the Very-High-Energy γ-Ray Emission Regionin the Central Engine of a Radio Galaxy. S ien e, vol. 325, no. 5939, pages444447, 2009. (Cited on page 12.)

144 Bibliography[A ero et al. 2009 F. A ero, F. Aharonian, A. G. Akhperjanian, G. Anton, U. Bar-res de Almeida, A. R. Bazer-Ba hi, Y. Be herini, B. Behera, K. Bernlöhr,and 161 oauthors. Dete tion of Gamma Rays from a Starburst Galaxy. S i-en e, vol. 326, no. 5956, pages 10801082, 2009. (Cited on pages 13, 118and 120.)[Aharonian & Atoyan 1996 F. A. Aharonian and A.M. Atoyan. On the emissivityof π0-de ay gamma radiation in the vi inity of a elerators of gala ti osmi rays. Astronomy and Astrophysi s, vol. 309, pages 917928, 1996. (Cited onpages 20, 21, 22, 35, 45, 51, 52, 53 and 172.)[Aharonian et al. 2003 F. Aharonian, A. Akhperjanian, M. Beili ke, K. Bernlöhr,H.-G. Börst, H. Bojahr, O. Bolz, T. Coarasa, J. L. Contreras, J. Cortinaand 43 oauthors. Is the giant radio galaxy M 87 a TeV gamma-ray emit-ter? Astronomy and Astrophysi s, vol. 403, pages L1L5, 2003. (Cited onpage 12.)[Aharonian et al. 2004a F. A. Aharonian, A. G. Akhperjanian, K.-M. Aye, A. R.Bazer-Ba hi, M. Beili ke, W. Benbow, D. Berge, P. Berghaus, K. Bernlhr,O. Bolz and 90 oauthors. High-energy parti le a eleration in the shell of asupernova remnant. Nature, vol. 432, no. 7013, pages 7577, 2004. (Citedon page 122.)[Aharonian et al. 2004b F. A. Aharonian, A. G. Akhperjanian, K.-M. Aye, A. R.Bazer-Ba hi, M. Beili ke, W. Benbow, D. Berge, P. Berghaus, K. Bernlöhr,O. Bolz and 90 oauthors. High-energy parti le a eleration in the shell of asupernova remnant. Nature, vol. 432, no. 7013, pages 7577, 2004. (Citedon pages 9 and 169.)[Aharonian et al. 2005a F. Aharonian, A. G. Akhperjanian, K.-M. Aye, A. R.Bazer-Ba hi, M. Beili ke, W. Benbow, D. Berge, P. Berghaus, K. Bernlöhr,C. Boisson and 92 oauthors. Dis overy of the binary pulsar PSR B1259-63in very-high-energy gamma rays around periastron with HESS. Astronomyand Astrophysi s, vol. 442, pages 110, 2005. (Cited on page 10.)[Aharonian et al. 2005b F. A. Aharonian, A. G. Akhperjanian, K.-M. Aye, A. R.Bazer-Ba hi, M. Beili ke, W. Benbow, D. Berge, P. Berghaus, K. Bern-löhr, C. Boisson and 90 oauthors. A New Population of Very High EnergyGamma-Ray Sour es in the Milky Way. S ien e, vol. 307, no. 5717, pages19381942, 2005. (Cited on page 10.)[Aharonian et al. 2005 F. A. Aharonian, A. G. Akhperjanian, K.-M. Aye, A. R.Bazer-Ba hi, M. Beili ke, W. Benbow, D. Berge, P. Berghaus, K. Bernlöhr,C. Boisson and 93 oauthors. Dis overy of Very High Energy Gamma RaysAsso iated with an X-ray Binary. S ien e, vol. 309, no. 5735, pages 746749,2005. (Cited on page 10.)

Bibliography 145[Aharonian et al. 2006a F. Aharonian, A. G. Akhperjanian, A. R. Bazer-Ba hi,M. Beili ke, W. Benbow, D. Berge, K. Bernlhr, C. Boisson, O. Bolz, V. Borreland 97 oauthors. Observations of the Crab nebula with HESS. Astronomyand Astrophysi s, vol. 457, no. 3, pages 899915, 2006. (Cited on page 120.)[Aharonian et al. 2006b F. Aharonian, A. G. Akhperjanian, A. R. Bazer-Ba hi,M. Beili ke, W. Benbow, D. Berge, K. Bernlöhr, C. Boisson, O. Bolz, V. Bor-rel and 105 oauthors. Energy dependent γ−ray morphology in the pulsarwind nebula HESS J1825-137. Astronomy and Astrophysi s, vol. 460, no. 2,pages 365374, 2006. (Cited on page 43.)[Aharonian et al. 2006 F. A. Aharonian, A. G. Akhperjanian, A. R. Bazer-Ba hi,M. Beili ke, W. Benbow, D. Berge, K. Bernlhr, C. Boisson, O. Bolz, V. Borreland 96 oauthors. First dete tion of a VHE gamma-ray spe tral maximumfrom a osmi sour e: HESS dis overy of the Vela X nebula. Astronomy andAstrophysi s, vol. 448, no. 2, pages L43L47, 2006. (Cited on page 25.)[Aharonian et al. 2006d F. A. Aharonian, A. G. Akhperjanian, A. R. Bazer-Ba hi,M. Beili ke, W. Benbow, D. Berge, K. Bernlöhr, C. Boisson, O. Bolz, V. Bor-rel and 113 oauthors. Fast Variability of Tera-Ele tron Volt γ Rays fromthe Radio Galaxy M87. S ien e, vol. 314, no. 5804, pages 14241427, 2006.(Cited on page 12.)[Aharonian et al. 2006e F. A. Aharonian, A. G. Akhperjanian, A. R. Bazer-Ba hi,M. Beili ke, W. Benbow, D. Berge, K. Bernlöhr, C. Boisson, O. Bolz, V. Bor-rel and 92 oauthors. Dis overy of very-high-energy γ-rays from the Gala ti Centre ridge. Nature, vol. 439, no. 7077, pages 695698, 2006. (Cited onpage 10.)[Aharonian et al. 2006f F. A. Aharonian, A. G. Akhperjanian, A. R. Bazer-Ba hi,M. Beili ke, W. Benbow, D. Berge, K. Bernlöhr, C. Boisson, O. Bolz, V. Bor-rel and 92 oauthors. A low level of extragala ti ba kground light as revealedby γ-rays from blazars. Nature, vol. 440, no. 7087, pages 10181021, 2006.(Cited on page 11.)[Aharonian et al. 2009 F. A. Aharonian, A. G. Akhperjanian, G. Anton, U. Bar-res de Almeida, A. R. Bazer-Ba hi, Y. Be herini, B. Behera, W. Benbow,K. Bernlöhr, C. Boisson and 149 oauthors. Dete tion of very high energyradiation from HESS J1908+063 onrms the Milagro unidentied sour eMGRO J1908+06. Astronomy and Astrophysi s, vol. 499, no. 3, pages 723728, 2009. (Cited on page 43.)[Aharonian 2001 F. A. Aharonian. Gamma Rays From Mole ular Clouds. Spa eS ien e Review, vol. 99, no. 1, pages 187196, 2001. (Cited on page 122.)[Akyüz et al. 1991 A. Akyüz, A. Brouillet and M.E. Özel. M82 in gamma-rays.Astronomy and Astrophysi s, vol. 248, no. 2, pages 419420, 1991. (Citedon page 63.)

146 Bibliography[Albert et al. 2006 J. Albert, E. Aliu, H. Anderhub, P. Antoranz, A. Armada,M. Asensio, C. Baixeras, J. A. Barrio, M. Bartelt, H. Bartko and 138 oau-thors. Variable Very-High-Energy Gamma-Ray Emission from the Mi ro-quasar LS I +61 303. S ien e, vol. 312, no. 5781, pages 17711773, 2006.(Cited on page 10.)[Albert et al. 2007a J. Albert, E. Aliu, H. Anderhub, P. Antoranz, A. Armada,C. Baixeras, J. A. Barrio, H. Bartko, D. Bastieri, J. K. Be ker and 130 oauthors. Very High Energy Gamma-Ray Radiation from the Stellar MassBla k Hole Binary Cygnus X-1. The Astrophysi al Journal, vol. 665, pagesL51L54, 2007. (Cited on page 10.)[Albert et al. 2007b J. Albert, E. Aliu, H. Anderhub, P. Antoranz, A. Armada,C. Baixeras, J. A. Barrio, H. Bartko, D. Bastieri, J. K. Be ker and 131 oauthors. Dis overy of Very High Energy Gamma Radiation from IC 443with the MAGIC Teles ope. Astrophysi al Journal, vol. 664, no. 2, pagesL87L90, 2007. (Cited on pages 34, 35, 125, 126 and 177.)[Albert et al. 2007 J. Albert, E. Aliu, H. Anderhub, P. Antoranz, A. Armada,C. Baixeras, J. A. Barrio, H. Bartko, D. Bastieri, J. K. Be ker and 134 oauthors. Unfolding of dierential energy spe tra in the MAGIC experiment.Nu lear Instruments and Methods in Physi s Resear h A, vol. 583, pages494506, 2007. (Cited on page 99.)[Albert et al. 2008a J. Albert, E. Aliu, H. Anderhub, L. A. Antonelli, P. Antoranz,M. Ba kes, C. Baixeras, J. A. Barrio, H. Bartko and 140 oauthors. Very-High-Energy gamma rays from a Distant Quasar: How Transparent Is theUniverse? S ien e, vol. 320, no. 5884, pages 17521754, 2008. (Cited onpage 12.)[Albert et al. 2008b J. Albert, E. Aliu, H. Anderhub, P. Antoranz, A. Armada,C. Baixeras, J. A. Barrio, H. Bartko, D. Bastieri, J. K. Be ker and 132 oauthors. VHE -Ray Observation of the Crab Nebula and its Pulsar withthe MAGIC Teles ope. Astrophysi al Journal, vol. 674, no. 2, pages 10371055, 2008. (Cited on pages 106, 110 and 175.)[Aleksi¢ et al. 2010 J. Aleksi¢, L. A. Antonelli, P. Antoranz, M. Ba kes, J. A. Bar-rio, D. Bastieri, J. Be erra González, W. Bednarek, A. Berdyugin, K. Bergerand 137 oauthors. MAGIC Upper Limits for two Milagro-dete ted, BrightFermi Sour es in the Region of SNR G65.1+0.6. Astrophysi al Journal,vol. 725, no. 2, pages 16291632, 2010. (Cited on pages 16, 103 and 130.)[Aliu et al. 2008a E. Aliu, H. Anderhub, L. A. Antonelli, P. Antoranz, M. Ba kes,C. Baixeras, J. A. Barrio, H. Bartko, D. Bastieri, J. K. Be ker and 139 oauthors. Observation of Pulsed -Rays Above 25 GeV from the Crab Pulsarwith MAGIC. S ien e, vol. 322, no. 5905, page 1221, 2008. (Cited on page 91.)

Bibliography 147[Aliu et al. 2008b E. Aliu, H. Anderhub, L. A. Antonelli, P. Antoranz, M. Ba kes,C. Baixeras, J. A. Barrio, H. Bartko, D. Bastieri, J. K. Be ker and 139 oauthors. Observation of Pulsed γ-Rays Above 25 GeV from the Crab Pulsarwith MAGIC. S ien e, vol. 322, no. 5905, pages 12211223, 2008. (Cited onpage 11.)[Aliu et al. 2009 E. Aliu, H. Anderhub, L. A. Antonelli, P. Antoranz, M. Ba kes,C. Baixeras, J. A. Barrio, H. Bartko, D. Bastieri, J. K. Be ker and 137 oauthors. Improving the performan e of the single-dish Cherenkov teles opeMAGIC through the use of signal timing. Astroparti le Physi s, vol. 30, no. 6,pages 293305, 2009. (Cited on page 96.)[An hordoqui et al. 2003 L. A. An hordoqui, D. F. Torres, T. P. M Cauley, G. E.Romero and F. A. Aharonian. Neutrinos from A reting Neutron Stars.The Astrophysi al Journal, vol. 589, no. 1, pages 481486, 2003. (Cited onpage 55.)[An hordoqui et al. 2004 L.A. An hordoqui, H. Goldberg, F. Halzen and T. J.Weiler. Neutrino bursts from Fanaro Riley I radio galaxies. Physi s LettersB, vol. 600, no. 3-4, pages 202207, 2004. (Cited on page 73.)[An hordoqui et al. 2008 L.A. An hordoqui, D. Hooper, S. Sarkar and A. M. Tay-lor. High energy neutrinos from astrophysi al a elerators of osmi ray nu- lei. Astroparti le Physi s, vol. 29, no. 1, pages 113, 2008. (Cited onpages 57 and 73.)[Arquilla & Kwok 1987 R. Arquilla and S. Kwok. CO observations of IRAS Cir- ular No. 9 sour es 19520 + 2759 and 01133 + 6434 - Regions of star for-mation. Astronomy and Astrophysi s, vol. 173, no. 2, pages 271278, 1987.(Cited on page 103.)[Asaoka & As henba h 1994 I. Asaoka and B. As henba h. An X-ray study ofIC443 and the dis overy of a new supernova remnant by ROSAT. Astronomyand Astrophysi s, vol. 284, no. 2, pages 573582, 1994. (Cited on page 33.)[Atoyan et al. 1995 A. M. Atoyan, F. A. Aharonian and H. J. Völk. Ele trons andpositrons in the gala ti osmi rays. Physi al Review D (Parti les, Fields,Gravitation, and Cosmology), vol. 52, no. 6, pages 32653275, 1995. (Citedon page 21.)[Baring et al. 1999 M. G. Baring, D. C. Ellison, S. P. Reynolds, I. A. Grenier andP. Goret. Radio to Gamma-Ray Emission from Shell-Type Supernova Rem-nants: Predi tions from Nonlinear Sho k A eleration Models. The Astro-physi al Journal, vol. 513, no. 1, pages 311338, 1999. (Cited on page 19.)[Bartel et al. 1987 N. Bartel, M. I. Ratner, A. E. E. Rogers, I. I. Shapiro, R. J.Bonometti, N. L. Cohen, M. V. Gorenstein, J. M. Mar aide and R. A. Pre-

148 Bibliographyston. VLBI observations of 23 hot spots in the starburst galaxy M82. Astro-physi al Journal, vol. 323, no. 1, pages 505515, 1987. (Cited on page 66.)[Bartko & Bednarek 2008 H. Bartko and W. Bednarek. Gamma-ray emission fromPWNe intera ting with mole ular louds. Monthly Noti es of the Royal As-tronomi al So iety, vol. 385, no. 3, pages 11051109, 2008. (Cited on pages 39and 43.)[Bazo Alba et al. 2009 J. L. Bazo Alba, R. Abbasi, Y. Abdou, M. A kermann,J. Adams, J. Aguilar, M. Ahlers, K. Andeen, J. Auenberg, X. Bai, M. Bakerand 240 oauthors. Sear h for neutrino point sour es with I eCube 22-strings.Nu lear Physi s B Pro eedings Supplements, vol. 188, pages 267269, 2009.(Cited on page 73.)[Be k & Krause 2005 R. Be k and M. Krause. Revised equipartition and minimumenergy formula for magneti eld strength estimates from radio syn hrotronobservations. Astronomis he Na hri hten, vol. 326, no. 6, pages 414427,2005. (Cited on pages 71 and 72.)[Be ker 2008 J. K. Be ker. High-energy neutrinos in the ontext of multimessengerastrophysi s. Physi s Reports, pages 173246, Mar h 2008. (Cited on pages 2and 169.)[Bell 1978 A. R. Bell. The a eleration of osmi rays in sho k fronts. II. MonthlyNoti es of Royal Astronomi al So iety, vol. 182, pages 443455, 1978. (Citedon pages 19 and 61.)[Berezhko et al. 2006 E. G. Berezhko, L. T. Ksenofontov and H. J. Völk. Gamma-ray emission expe ted from Kepler's supernova remnant. Astronomy andAstrophysi s, vol. 452, no. 1, pages 217221, 2006. (Cited on page 61.)[Berezinskii et al. 1990 V. S. Berezinskii, S. V. Bulanov, V. A. Dogiel and V. S.Ptuskin. Ginzburg, V. L., Amsterdam, North-Holland, 1990. (Cited onpages 50 and 62.)[Berts h et al. 1993 D. L. Berts h, T. M. Dame, C. E. Fi htel, S. D. Hunter,P. Sreekumar, J. G. Sta y and P. Thaddeus. Diuse Gamma-Ray Emissionin the Gala ti Plane from Cosmi -Ray, Matter, and Photon Intera tions.Astrophysi al Journal, vol. 416, pages 587600, 1993. (Cited on page 20.)[Blattnig et al. 2000a S. R. Blattnig, S. R. Swaminathan, A. T. Kruger, M. Ngomand J. W. Norbury. Parametrizations of in lusive ross se tions for pionprodu tion in proton-proton ollisions. Physi al Review D, id.094030, vol. 62,no. 9, 2000. (Cited on page 51.)[Blattnig et al. 2000b S. R. Blattnig, S. R. Swaminathan, A. T. Kruger, M. Ngom,J. W. Norbury and R. K. Tripathi. Parameterized Cross Se tions for Pion

Bibliography 149Produ tion in Proton-Proton Collisions. Te hni al Report, NASA/TP-2000-210640, vol. 1, 2000. (Cited on page 51.)[Blom et al. 1999 J. J. Blom, T. A. Paglione and A. Carramiñana. Diuse Gamma-Ray Emission from Starburst Galaxies and M31. The Astrophysi al Journal,vol. 516, no. 2, pages 744749, 1999. (Cited on pages 57 and 59.)[Bo hino & Bykov 2000 F. Bo hino and A. M. Bykov. Hard X-ray emission fromIC443: eviden e for a sho ked mole ular lump? Astronomy and Astro-physi s, vol. 362, pages L29L32, 2000. (Cited on page 33.)[Bo hino & Bykov 2001 F. Bo hino and A. M. Bykov. The plerion nebula in IC443: The XMM-Newton view. Astronomy and Astrophysi s, vol. 376, pages248253, 2001. (Cited on pages 33, 39 and 125.)[Bo hino & Bykov 2003 F. Bo hino and A. M. Bykov. XMM-Newton study ofhard X-ray sour es in IC 443. Astronomy and Astrophysi s, vol. 400, pages203211, 2003. (Cited on page 33.)[Braun & Strom 1986 R. Braun and R. G. Strom. The stru ture and dynami s ofevolved supernova remnants - The IC 443 omplex. Astronomy and Astro-physi s, vol. 164, no. 1, pages 193207, 1986. (Cited on page 31.)[Bressan et al. 2002 A. Bressan, L. Silva and G. L. Granato. Far infrared and radioemission in dusty starburst galaxies. Astronomy and Astrophysi s, vol. 392,pages 377391, 2002. (Cited on page 68.)[Brown & Mars her 1977 R. L. Brown and A. P. Mars her. Energeti se ondaryele trons in dense interstellar louds. Astrophysi al Journal, vol. 212, no. 1,pages 659663, 1977. (Cited on page 59.)[Butt et al. 2003 Y. M. Butt, D. F. Torres, J. A. Combi, T. Dame and G. E.Romero. Mole ular Beam Dumps as Tra ers of Hadroni Cosmi RaySour es: the Case of SNR IC 443. In XXIInd Moriond Astrophysi s Meet-ing, pages 323331. arXiv:astro-ph/0206132v2, 2003. (Cited on pages 33, 37and 45.)[Bykov et al. 2000 A. M. Bykov, R. A. Chevalier, D. C. Ellison and Yu. A. Uvarov.Nonthermal Emission from a Supernova Remnant in a Mole ular Cloud.The Astrophysi al Journal, vol. 538, no. 1, pages 203216, 2000. (Citedon page 37.)[Bykov et al. 2005 A. M. Bykov, F. Bo hino and G. G. Pavlov. A Hard ExtendedX-Ray Sour e in the IC 443 Supernova Remnant Resolved by Chandra: AFast Eje ta Fragment or a New Pulsar Wind Nebula? The Astrophysi alJournal, vol. 624, no. 1, pages L41L44, 2005. (Cited on page 33.)

150 Bibliography[Bykov et al. 2008 A. M. Bykov, A. M. Krassil ht hikov, Yu. A. Uvarov, H. Bloe-men, F. Bo hino, G. M. Dubner, E. B. Gia ani and G. G. Pavlov. IsolatedX-Ray-Infrared Sour es in the Region of Intera tion of the Supernova Rem-nant IC 443 with a Mole ular Cloud. The Astrophysi al Journal, vol. 676,no. 2, pages 10501063, 2008. (Cited on page 33.)[Carilli 1996 C. L. Carilli. Free-free absorption towards the nu leus of NGC 253:further eviden e for high pressures in the starburst nu leus. Astronomy andAstrophysi s, vol. 305, pages 402, 1996. (Cited on pages 77 and 174.)[Casanova et al. 2010 S. Casanova, F. A. Aharonian, Y. Fukui, S. Gabi i, D. I.Jones, A. Kawamura, T. Onishi, G. Rowell, H. Sano, K. Torii and H. Ya-mamoto. Mole ular Clouds as Cosmi -Ray Barometers. Publi ations of theAstronomi al So iety of Japan, vol. 62, no. 3, pages 769777, 2010. (Citedon pages 19 and 122.)[Casasola et al. 2004 V. Casasola, D. Bettoni and G. Galletta. The gas ontent ofpe uliar galaxies: Strongly intera ting systems. Astronomy and Astrophysi s,vol. 422, pages 941950, 2004. (Cited on page 63.)[Cavasinni et al. 2006 V. Cavasinni, D. Grasso and L. Ma ione. TeV neutrinosfrom supernova remnants embedded in giant mole ular louds. Astroparti lePhysi s, vol. 26, no. 1, pages 4149, 2006. (Cited on page 55.)[Cherenkov 1934 P. A. Cherenkov. Visible emission of lean liquids by a tion ofgamma radiation. Doklady Akademii Nauk SSSR Ser. A, vol. 2, pages 451455, 1934. (Cited on page 84.)[Chevalier 1999 R. A. Chevalier. Supernova Remnants in Mole ular Clouds. TheAstrophysi al Journal, vol. 511, no. 2, pages 798811, 1999. (Cited onpage 33.)[Cillis et al. 2005 A. N. Cillis, D. F. Torres and O. Reimer. EGRET Upper Limitsand Sta king Sear hes of Gamma-Ray Observations of Luminous and Ultra-luminous Infrared Galaxies. The Astrophysi al Journal, vol. 621, no. 1, pages139145, 2005. (Cited on page 69.)[Claussen et al. 1997 M. J. Claussen, D. A. Frail, W. M. Goss and R. A. Gaume.Polarization Observations of 1720 MHz OH Masers toward the Three Super-nova Remnants W28, W44, and IC 443. Astrophysi al Journal, vol. 489,pages 143159, 1997. (Cited on page 31.)[Corbet et al. 2011 R. H. D. Corbet, C. C. Cheung, M. Kerr, R. Dubois, D. Do-nato, G. A. Caliandro, M. J. Coe, P. G. Edwards, M. D. Filipovi , J. L.Payne and J. Stevens. 1FGL J1018.6-5856: a New Gamma-ray Binary. TheAstronomer's Telegram, vol. 3221, pages 1+, 2011. (Cited on page 10.)

Bibliography 151[Cornett et al. 1977 R. H. Cornett, G. Chin and G. R. Knapp. Observations of COemission from a dense loud asso iated with the supernova remnant IC 443.Astronomy and Astrophysi s, vol. 54, no. 3, pages 889894, 1977. (Cited onpages 33, 34 and 44.)[Cortina et al. 2009 J. Cortina, F. Goebel, T. S hweizer and for the MAGIC Col-laboration. Te hni al Performan e of the MAGIC Teles opes. In Pro .31st International Cosmi Ray Conferen e. (ód¹, Poland). arXiv: astro-ph/0907.1211v1, 2009. (Cited on pages 90 and 174.)[Crut her et al. 1978 R. M. Crut her, D. H. Rogstad and K. Chu. Aperture-synthesis observations of H I in the galaxy M82. Astrophysi al Journal,vol. 225, no. 1, pages 784789, 1978. (Cited on page 63.)[Crut her 1988 R. M. Crut her. Di kman, R. and Snell, R. and Young, J. Eds.,New York, Springer, 1988. (Cited on page 68.)[Crut her 1994 R. M. Crut her. Clouds, ores and low mass stars. In Pro eedingsof the 4th Haysta k Observatory, volume 65 of Astronomi al So iety of thePa i Conferen e Series, page 87. Clemens, D. P. and Barvainis, R., 1994.(Cited on page 68.)[Crut her 1999 R. M. Crut her. Magneti Fields in Mole ular Clouds: ObservationsConfront Theory. The Astrophysi al Journal, vol. 520, no. 2, pages 706713,1999. (Cited on page 68.)[De Angelis 2011 A. De Angelis. Re ent results from VHE gamma astrophysi srelated to fundamental physi s and osmi rays. ArXiv e-prints, 2011. (Citedon pages 8 and 179.)[de Cea del Pozo et al. 2009a E. de Cea del Pozo, D. F. Torres, A. Y. Rodríguezand O. Reimer. Model analysis of the very high energy dete tions of thestarburst galaxies M82 and NGC 253. In Pro . 1st Fermi Symposium (Wash-ington, DC). arXiv: astro-ph/0912.3497v2, 2009. (Cited on pages 16, 58,119, 120 and 176.)[de Cea del Pozo et al. 2009b E. de Cea del Pozo, D. F. Torres and A. Y. Ro-dríguez Marrero. Multi-messenger model for the starburst galaxy M82. TheAstrophysi al Journal, vol. 698, no. 2, pages 10541071, 2009. (Cited onpages 16, 58, 75, 76, 118, 120, 174 and 179.)[de Grijs et al. 2001 R. de Grijs, R. W. O'Connell and J. S. III Gallagher. TheFossil Starburst in M82. The Astronomi al Journal, vol. 121, no. 2, pages768792, 2001. (Cited on page 66.)[de Jager & Djannati-Ataï 2009 O.C. de Jager and A. Djannati-Ataï. Neutronstars and pulsars, volume 357. Astrophysi s and Spa e S ien e Library,(Berlin) Springer, 2009. (Cited on page 110.)

152 Bibliography[de Naurois & Rolland 2009 M. de Naurois and L. Rolland. A high performan elikelihood re onstru tion of -rays for imaging atmospheri Cherenkov tele-s opes. Astroparti le Physi s, vol. 32, no. 5, pages 231252, 2009. (Cited onpage 120.)[De Noyer & Frerking 1981 L. K. De Noyer and M. A. Frerking. Some new resultson sho k hemistry in IC 443. Astrophysi al Journal, vol. 246, pages L37L40, 1981. (Cited on pages 33 and 44.)[Dermer 1986 C. D. Dermer. Se ondary produ tion of neutral pi-mesons and thediuse gala ti gamma radiation. Astronomy and Astrophysi s, vol. 157,no. 2, pages 22322, 1986. (Cited on page 46.)[Di kman et al. 1992 R. L. Di kman, R. L. Snell, L. M. Ziurys and Y.-L. Huang.Stru ture and kinemati s of dense gas asso iated with the supernova remnantIC 443. Astrophysi al Journal, vol. 400, no. 1, pages 203213, 1992. (Citedon pages 33, 34 and 44.)[Dogel & Sharov 1990 V. A. Dogel and G. S. Sharov. Manifestations of osmi raya eleration in giant mole ular louds. Astronomy and Astrophysi s, vol. 229,no. 1, pages 259271, 1990. (Cited on page 19.)[Domingo-Santamaría & Torres 2005 E. Domingo-Santamaría and D. F. Torres.High energy -ray emission from the starburst nu leus of NGC 253. Astronomyand Astrophysi s, vol. 444, no. 2, pages 403415, 2005. (Cited on pages 20,51, 57, 58, 61, 63, 68, 74, 78, 79, 118, 174 and 179.)[Domingo-Santamaría et al. 2005 E. Domingo-Santamaría, J. Flix, J. Ri o, V. S al-zotto and W. Wittek. The DISP analysis method for point-like or extendedgamma sour e sear hes/studies with the MAGIC Teles ope. In Pro . of the29th International Cosmi Ray Conferen e, Pune (India), volume 5, page363. Mumbai: Tata Institute of Fundamental Resear h, 2005. (Cited onpage 99.)[Drury et al. 1994 L. O'C. Drury, F. A. Aharonian and H. J. Völk. The gamma-rayvisibility of supernova remnants. A test of osmi ray origin. Astronomy andAstrophysi s, vol. 287, pages 959971, 1994. (Cited on pages 19 and 59.)[Elias et al. 1978 J. H. Elias, D. J. Ennis, D. Y. Gezari, M. G. Hauser, J. R. Hou k,K. Y. Lo, K. Matthews, D. Naduea, G. Neugebauer, M. W. Werner and W. E.Westbrook. 1 millimeter ontinuum observations of extragala ti obje ts.Astrophysi al Journal, vol. 220, no. 1, pages 2541, 1978. (Cited on pages 77and 174.)[Errando 2009 M. Errando. Dis overy of very high energy gamma-ray emissionfrom 3C 279 and 3C 66A/B with the MAGIC teles ope. PhD thesis, Univ.Autonoma de Bar elona, 2009. (Cited on pages 97 and 175.)

Bibliography 153[Everett et al. 2008 J. E. Everett, E. G. Zweibel, R. A. Benjamin, D. M Cammon,L. Ro ks and J. S. III Gallagher. The Milky Way's Kiloparse -S ale Wind:A Hybrid Cosmi -Ray and Thermally Driven Outow. The Astrophysi alJournal, vol. 674, no. 1, pages 258270, 2008. (Cited on page 59.)[Fatuzzo & Melia 2003 M. Fatuzzo and F. Melia. A Kinship Between the EGRETSupernova Remnants and Sagittarius A East. The Astrophysi al Journal,vol. 596, no. 2, pages 10351043, 2003. (Cited on page 59.)[Fazio 1967 G. G. Fazio. Gamma Ratiation from Celestial Obje ts. Annual Reviewof Astronomy and Astrophysi s, vol. 5, page 481, 1967. (Cited on page 59.)[Fesen & Kirshner 1980 R. A. Fesen and R. P. Kirshner. Spe trophotometry of thesupernova remnant IC 443. Astrophysi al Journal, vol. 242, no. 1, pages10231040, 1980. (Cited on page 31.)[Fi htel et al. 1976 C. E. Fi htel, D. A. Knien, D. J. Thompson, G. F. Bignamiand C. Y. Cheung. Signi an e of medium-energy gamma-ray astronomyin the study of osmi rays. Astrophysi al Journal, vol. 208, no. 1, pages211219, 1976. (Cited on page 20.)[Fomin et al. 1994 V. P. Fomin, A. A. Stepanian, R. C. Lamb, D. A. Lewis,M. Pun h and T. C. Weekes. New methods of atmospheri Cherenkov imagingfor gamma-ray astronomy. I. The false sour e method. Astroparti le Physi s,vol. 2, no. 2, pages 137150, 1994. (Cited on pages 91 and 103.)[Förster et al. 2003 N.M. Förster, M. Sauvage, V. Charmandaris, O. Laurent,P. Gallais, I. F. Mirabel and L. Vigroux. ISOCAM view of the starburst galax-ies M82, NGC253, and NGC1808. Astronomy and Astrophysi s, vol. 399,pages 833855, 2003. (Cited on pages 66 and 173.)[Freedman et al. 1994 W. L. Freedman, S. M. Hughes, B. F. Madore, J. R. Mould,M. G. Lee, P. Stetson, R. C. Kenni utt, A. Turner, L. Ferrarese, H. Fordand 5 oauthors. The Hubble Spa e Teles ope Extragala ti Distan e S aleKey Proje t. 1: The dis overy of Cepheids and a new distan e to M81. As-trophysi al Journal, Part 1, vol. 427, no. 2, pages 628655, 1994. (Cited onpage 62.)[Funk et al. 2008 S. Funk, O. Reimer, D. F. Torres and J. A. Hinton. The GeV-TeVConne tion in Gala ti γ-ray sour es. The Astrophysi al Journal, vol. 679,no. 2, pages 12991314, 2008. (Cited on pages 20, 22, 35, 67, 169 and 173.)[Gabi i & Aharonian 2007 S. Gabi i and F. A. Aharonian. Sear hing for Gala ti Cosmi -Ray Pevatrons with Multi-TeV Gamma Rays and Neutrinos. TheAstrophysi al Journal, vol. 665, no. 2, pages L131L134, 2007. (Cited onpages 19, 22, 28, 35, 37, 40 and 45.)

154 Bibliography[Gabi i et al. 2009 S. Gabi i, F. A. Aharonian and S. Casanova. Broad-band non-thermal emission from mole ular louds illuminated by osmi rays fromnearby supernova remnants. Monthly Noti es of the Royal Astronomi alSo iety, vol. 396, no. 3, pages 16291639, 2009. (Cited on pages 19, 28, 54,116, 117 and 176.)[Gabi i 2008 S. Gabi i. Gamma Ray Astronomy and the Origin of Gala ti Cosmi Rays. In XXI European Cosmi Ray Symposium. arXiv: astro-ph/0811.0836,2008. (Cited on page 122.)[Gaensler et al. 2006 B. M. Gaensler, S. Chatterjee, P. O. Slane, E. van der Swaluw,F. Camilo and J. P. Hughes. The X-Ray Stru ture of the Pulsar Bow Sho kG189.22+2.90 in the Supernova Remnant IC 443. The Astrophysi al Jour-nal, vol. 648, no. 2, pages 10371042, 2006. (Cited on page 33.)[Gaisser et al. 1995 T. K. Gaisser, F. Halzen and T. Stanev. Parti le astrophysi swith high energy neutrinos [Erratum: 1996PhR...271..355G. Physi s Re-ports, vol. 258, pages 173236, 1995. (Cited on page 55.)[Gaisser et al. 1998 T. K. Gaisser, R. J. Protheroe and T. Stanev. Gamma-RayProdu tion in Supernova Remnants. Astrophysi al Journal, vol. 492, no. 1,pages 219227, 1998. (Cited on pages 19, 37 and 39.)[Gallagher & Smith 2005 J. S. Gallagher and L. J. Smith. Observational Ponder-ings on Disk-Halo Interfa es. In Extra-Planar Gas, volume 331 of Astronom-i al So iety of the Pa i Conferen e Series, pages 147154. R. Braun, 2005.(Cited on page 59.)[Gaug 2006 M. Gaug. Calibration of the MAGIC teles ope and Observation ofGamma Ray Bursts. PhD thesis, Univ. Autonoma de Bar elona, 2006. (Citedon page 94.)[Ginzburg & Syrovatskii 1964 V. L. Ginzburg and S. I. Syrovatskii. The origin of osmi rays. New York: Ma millan, 1964. (Cited on pages 2, 21, 35, 59and 61.)[Green 2004 D. A. Green. Gala ti supernova remnants: an updated atalogue andsome statisti s. Bulletin of the Astronomi al So iety of India, vol. 32, pages335370, 2004. (Cited on page 31.)[Halzen & Klein 2010 F. Halzen and S. R. Klein. Invited Review Arti le: I eCube:An instrument for neutrino astronomy. Review of S ienti Instruments,vol. 81, no. 8, pages 081101+, 2010. (Cited on page 14.)[Halzen 2006 F. Halzen. Astroparti le physi s with high energy neutrinos: fromAMANDA to I eCube. The European Physi al Journal C, vol. 46, no. 3,pages 669687, 2006. (Cited on page 54.)

Bibliography 155[Hartman et al. 1999 R.C. Hartman, D. L. Berts h, S. D. Bloom, A. W. Chen,P. Deines-Jones, J. A. Esposito, C. E. Fi htel, D. P. Friedlander, S. D.Hunter, L. M. M Donald and 17 oauthors. The Third EGRET Catalogof High-Energy Gamma-Ray Sour es. Astrophysi al Journal Supplement Se-ries, vol. 123, no. 1, pages 79202, 1999. (Cited on pages 5, 12, 34, 43, 103,110 and 175.)[He k et al. 1998 D. He k, J. Knapp, J. N. Capdevielle, G. S hatz and T. Thouw.Fors hungszentrum Karlsruhe, Hannover (Germany), 1998. (Cited onpage 93.)[Hermsen 1990 W. Hermsen. Highlights from the COS-B mission. Advan es inSpa e Resear h, vol. 10, pages 6979, 1990. (Cited on page 4.)[Hess 1912 V. Hess. Uber Beoba htungen der dur hdringenden Strahlung bei siebenFreiballonfahrten. Physikalis he Zeits hrift, vol. 13, pages 10841091, 1912.(Cited on page 1.)[Hewitt et al. 2006 J. W. Hewitt, F. Yusef-Zadeh, M. Wardle, D. A. Roberts andN. E. Kassim. Green Bank Teles ope Observations of IC 443: The Natureof OH (1720 MHz) Masers and OH Absorption. The Astrophysi al Journal,vol. 652, no. 2, pages 12881296, 2006. (Cited on page 31.)[Hillas 1985 A.M. Hillas. Cerenkov light images of EAS produ ed by primarygamma. In F. C. Jones, editeur, International Cosmi Ray Conferen e, vol-ume 3, pages 445448, 1985. (Cited on page 94.)[Hofmann et al. 2010 W. Hofmann, M. Martinez and the CTA onsortium. DesignCon epts for the Cherenkov Teles ope Array. eprint arXiv:1008.3703v2, 2010.(Cited on page 111.)[Holder et al. 2005 J. Holder, H. M. Badran, G. Blaylo k, I. H. Bond, P. J. Boyle,S. M. Bradbury, J. H. Bu kley, K. Byrum, D. A. Carter-Lewis, O. Celikand 63 oauthors. Re ent Observations of IC443 with the Whipple 10m Tele-s ope. In High Energy Gamma-Ray Astronomy: 2nd International Sympo-sium. Heidelberg (Germany), volume 745 of AIP Conferen e Pro eedings,pages 275280. New York: Ameri an Institute of Physi s, 2005. (Cited onpage 34.)[Huang et al. 1986 Y.-L. Huang, R. L. Di kman and R. L. Snell. IRAS sour esasso iated with sho ked gas regions in IC 443. Astrophysi al Journal, Part 2- Letters to the Editor, vol. 302, pages L63L66, 1986. (Cited on pages 34,127 and 177.)[Hughes et al. 1994 D.H. Hughes, W.K. Gear and E.I. Robson. The Submillime-tre Stru ture of the Starburst Nu leus in M82 - a Dira tion-Limited 450-MICRON Map. Monthly Noti es of the Royal Astronomi al So iety, vol. 270,no. 3, page 641, 1994. (Cited on pages 66 and 173.)

156 Bibliography[Humensky & the VERITAS Collaboration 2008 T. B. Humensky and the VER-ITAS Collaboration. Observation of the Supernova Remnant IC 443 withVERITAS. In Pro . of the 30th International Cosmi Ray Conferen e.Mérida,Yu atán, Mexi o, volume 2, pages 835838. Universidad Na ional Autónomade Méxi o, Mexi o City, Mexi o, 2008. (Cited on page 34.)[Issa & Wolfendale 1981 M. R. Issa and A. W. Wolfendale. Gamma rays from the osmi ray irradiation of lo al mole ular louds. Nature, vol. 292, pages430433, 1981. (Cited on page 122.)[Kamae et al. 2005 T. Kamae, T. Abe and T. Koi. Dira tive Intera tion andS aling Violation in pp>π0 Intera tion and GeV Ex ess in Gala ti DiuseGamma-Ray Spe trum of EGRET. The Astrophysi al Journal, vol. 620,no. 1, pages 244256, 2005. (Cited on page 51.)[Kara hentsev et al. 2003 I. D. Kara hentsev, E. K. Grebel, M. E. Sharina, A. E.Dolphin, D. Geisler, P. Guhathakurta, P. W. Hodge, V. E. Kara hentseva,A. Sarajedini and P. Seitzer. Distan es to nearby galaxies in S ulptor. As-tronomy and Astrophysi s, vol. 404, pages 93111, 2003. (Cited on page 74.)[Kelner et al. 2006 S. R. Kelner, F. A. Aharonian and V. V. Bugayov. Energyspe tra of gamma rays, ele trons, and neutrinos produ ed at proton-protonintera tions in the very high energy regime. Physi al Review D, vol. 74, no. 3,page 034018, 2006. (Cited on pages 20, 52, 53, 54, 59 and 172.)[Keohane et al. 1997 J. W. Keohane, R. Petre, E. V. Gotthelf, M. Ozaki andK. Koyama. A Possible Site of Cosmi Ray A eleration in the SupernovaRemnant IC 443. Astrophysi al Journal, vol. 484, pages 350359, 1997.(Cited on page 33.)[Kheli 2003 B. Kheli. Re her he de sour es gamma par un methode de Maximumde Vraisemblan e: Appli ation aux AGN et aux sour es gala tiques suivis parle teles ope CAT. PhD thesis, University of Caen, Fran e, 2003. (Cited onpage 34.)[Klein et al. 1988 U. Klein, R. Wielebinski and H.W. Morsi. Radio ontinuum ob-servations of M82. Astronomy and Astrophysi s, vol. 190, no. 1, pages 4146,1988. (Cited on pages 66 and 173.)[Kronberg & Sramek 1985 P.P. Kronberg and R.A. Sramek. Dis overy of new vari-able radio sour es in the nu leus of the nearby galaxy Messier 82. S ien e,vol. 227, pages 2831, 1985. (Cited on page 66.)[Kronberg & Wilkinson 1975 P. P. Kronberg and P. N. Wilkinson. High-resolution,multifrequen y radio observations of M82. Astrophysi al Journal, vol. 200,no. 1, pages 430438, 1975. (Cited on page 66.)

Bibliography 157[Kronberg et al. 1985 P.P. Kronberg, P. Biermann and F.R. S hwab. The nu leusof M82 at radio and X-ray bands - Dis overy of a new radio population ofsupernova andidates. Astrophysi al Journal, vol. 291, no. 1, pages 693707,1985. (Cited on page 66.)[Lamb & Ma omb 1997 R. C. Lamb and D. J. Ma omb. Point Sour es of GeVGamma Rays. Astrophysi al Journal, vol. 488, pages 872880, 1997. (Citedon pages 34 and 43.)[Lande ker et al. 1990 T.L. Lande ker, M. Clutton-Bro k and C.R. Purton. G65.1+0.6 - A new supernova remnant of low surfa e brightness. Astronomyand Astrophysi s, vol. 232, no. 1, pages 207214, 1990. (Cited on pages 102,105 and 175.)[Lasker et al. 1990 B. M. Lasker, C. R. Stur h, B. J. M Lean, J. L. Russell,H. Jenkner and M. M. Shara. The Guide Star Catalog. I - Astronomi- al foundations and image pro essing. Astronomi al Journal, vol. 99, pages20192058, 1990. (Cited on pages 31, 32, 34 and 170.)[Leahy 2004 A. D. Leahy. 1420 and 408 MHz Continuum Observations of the IC443/G189.6+3.3 Region. The Astronomi al Journal, vol. 127, no. 4, pages22772283, 2004. (Cited on page 31.)[Lessard et al. 2001 R. W. Lessard, J. H. Bu kley, V. Connaughton and S. Le Bo-he . A new analysis method for re onstru ting the arrival dire tion of TeVgamma rays using a single imaging atmospheri Cherenkov teles ope. As-troparti le Physi s, vol. 15, no. 1, pages 118, 2001. (Cited on page 99.)[Lindner 2003 J. Lindner. Dusty galaxies: modelling spe tral energy distributions.Canadian Undergraduate Physi s Journal, vol. 2, no. 1, pages 712, 2003.(Cited on page 69.)[Lipari 1993 P. Lipari. Tev muons in hadroni showers. Astroparti le Physi s,vol. 1, no. 4, pages 399415, 1993. (Cited on page 55.)[Loeb & Waxman 2006 A. Loeb and E. Waxman. The umulative ba kground ofhigh energy neutrinos from starburst galaxies. Journal of Cosmology andAstroparti le Physi s, vol. 5, no. 003, pages 28, 2006. (Cited on pages 57and 72.)[Longair 1992 M. S. Longair. volume 1. Cambridge University Press, Cambridge,UK, 1992. (Cited on page 83.)[Longair 1994 M. S. Longair. volume 2. Cambridge University Press, 2nd ed.,Cambridge, UK, 1994. (Cited on page 61.)[Lozinskaya 1981 T. A. Lozinskaya. The radio brightness-diameter relation for su-pernova remnants - Some new distan e alibrators. Soviet Astronomy Letters,vol. 7, pages 1719, 1981. (Cited on page 33.)

158 Bibliography[Mankuzhiyil 2010 Nijil Mankuzhiyil. MAGIC gamma-ray observations of distantAGN and a study of sour e variability and the extragala ti ba kground lightusing Fermi and air Cherenkov teles opes. PhD thesis, University of Udine,2010. (Cited on pages 95 and 174.)[Mannheim & S hli keiser 1994 K. Mannheim and R. S hli keiser. Intera tions of osmi ray nu lei. Astronomy & Astrophysi s, vol. 286, pages 983996, 1994.(Cited on page 61.)[Maras hi et al. 1968 L. Maras hi, G. C. Perola and S. S hwarz. X-ray ux andanisotropy produ ed by gala ti osmi -ray ele trons. Il Nuovo Cimento BSeries 10, vol. 53, no. 2, pages 443458, 1968. (Cited on page 59.)[Mariotti 2011 M. Mariotti. MAGIC onrms VHE gamma-ray emission fromHESS J0632+057 between 7-9 Feb 2011. The Astronomer's Telegram,vol. 3161, pages 1+, 2011. (Cited on page 10.)[Marko et al. 1999 S. Marko, F. Melia and I. Sar evi . High-Energy Emissionfrom Relativisti Parti les in Sagittarius A*. The Astrophysi al Journal,vol. 522, no. 2, pages 870878, 1999. (Cited on page 59.)[Mars her & Brown 1978 A. P. Mars her and R. L. Brown. Energeti se ondaryele trons and the nonthermal gala ti radio ba kground - A probe of the mag-neti eld in interstellar louds. Astrophysi al Journal, vol. 221, no. 1, pages588597, 1978. (Cited on page 59.)[Mauersberger et al. 1996 R. Mauersberger, C. Henkel, R. Wielebinski, T. Wiklindand H.-P. Reuter. Mole ular distribution and kinemati s in nearby galaxies.I. NGC 253. Astronomy and Astrophysi s, vol. 305, pages 421434, 1996.(Cited on page 74.)[Mayya et al. 2005 Y. D. Mayya, L. Carras o and A. Luna. The Dis overy of SpiralArms in the Starburst Galaxy M82. The Astrophysi al Journal, vol. 628,no. 1, pages L33L36, 2005. (Cited on page 62.)[Mayya et al. 2006 Y. D. Mayya, A. Bressan, L. Carras o and L. Hernandez-Martinez. The Star Formation History of the Disk of the Starburst GalaxyM82. The Astrophysi al Journal, vol. 649, no. 1, pages 172180, 2006. (Citedon page 63.)[Mazin 2007 D. Mazin. A study of very high energy gamma-ray emission fromAGNs and onstraints on the extragala ti ba kground light. PhD thesis,Te hnis he Universität Mün hen, 2007. (Cited on pages 92 and 174.)[M Leod et al. 1993 K. K. M Leod, G. H. Rieke, M. J. Rieke and D. M. Kelly. M82- The saga ontinues. Astrophysi al Journal, vol. 412, no. 1, pages 111126,1993. (Cited on page 66.)

Bibliography 159[Melo et al. 2002 V. P. Melo, A. M. Pérez Gar ía, J. A. A osta-Pulido, C. MuñozTuñón and J. M. Rodríguez Espinosa. The Spatial Distribution of the Far-Infrared Emission in NGC 253. The Astrophysi al Journal, vol. 574, no. 2,pages 709718, 2002. (Cited on page 74.)[Meu i et al. 2007 M. Meu i, R. Paoletti, R. Pegna, A. N. Stamerra Turini,D. Corti, M. Mariotti and the MAGIC Collaboration. The trigger system ofthe MAGIC teles ope: on-line sele tion strategies for Cherenkov teles opes.IEEE Trans. Nu l. S i., vol. 54, no. 2, pages 404409, 2007. (Cited onpages 90 and 174.)[Moiseev et al. 2007 A. A. Moiseev, R. C. Hartman, J. F. Ormes, D. J. Thomp-son, M. J. Amato, T. E. Johnson, K. N. Segal and D. A. Sheppard. Theanti- oin iden e dete tor for the GLAST large area teles ope. Astroparti lePhysi s, vol. 27, no. 5, pages 339358, 2007. (Cited on page 20.)[Montmerle 1979 T. Montmerle. On gamma-ray sour es, supernova remnants, OBasso iations, and the origin of osmi rays. Astrophysi al Journal, vol. 231,no. 1, pages 95110, 1979. (Cited on page 19.)[Moralejo et al. 2009 A. Moralejo, M. Gaug, E. Carmona, P. Colin, C. Delgado,S. Lombardi, D. Mazin, V. S alzotto, J. Sitarek, D. Tes aro and for theMAGIC ollaboration. MARS, the MAGIC Analysis and Re onstru tionSoftware. In Pro . 31st International Cosmi Ray Conferen e (ód¹, Poland).arXiv: astro-ph/0907.0943, 2009. (Cited on pages 93, 104 and 130.)[Moskalenko & Strong 1998 I. Moskalenko and A. Strong. Produ tion and Propaga-tion of Cosmi -Ray Positrons and Ele trons. Astrophysi al Journal, vol. 493,pages 694715, 1998. (Cited on page 59.)[Mouh ine et al. 2005 M. Mouh ine, H. C. Ferguson, R. M. Ri h, T. M. Brown andT. E. Smith. Halos of Spiral Galaxies. I. The Tip of the Red Giant Bran has a Distan e Indi ator. The Astrophysi al Journal, vol. 633, no. 2, pages810820, 2005. (Cited on page 74.)[Mufson et al. 1986 S. L. Mufson, M. L. M Collough, J. R. Di kel, R. Petre,R. White and R. Chevalier. A multiwavelength investigation of the super-nova remnant IC 443. Astronomi al Journal, vol. 92, pages 13491357, 1986.(Cited on page 33.)[Naito & Takahara 1994 T. Naito and F. Takahara. High energy gamma-ray emis-sion from supernova remnants. Journal of Physi s G: Nu lear and Parti lePhysi s, vol. 20, no. 3, pages 477486, 1994. (Cited on page 19.)[Nolan et al. 2003 P. L. Nolan, W. F. Tompkins, I. A. Grenier and P. F. Mi helson.Variability of EGRET Gamma-Ray Sour es. The Astrophysi al Journalvol. 597, no. 1, pages 615627, 2003. (Cited on page 34.)

160 Bibliography[O'Connell & Mangano 1978 R. W. O'Connell and J. J. Mangano. The entralregions of M82. Astrophysi al Journal, vol. 221, no. 1, pages 6279, 1978.(Cited on page 62.)[Ohm et al. 2009 S. Ohm, C. van Eldik and K. Egberts. γ/hadron separation invery-high-energy γ-ray astronomy using a multivariate analysis method. As-troparti le Physi s, vol. 31, no. 5, pages 383391, 2009. (Cited on page 120.)[Olbert et al. 2001 C. M. Olbert, C. R. Cleareld, N. E. Williams, J. W. Keohaneand D. A. Frail. A Bow Sho k Nebula around a Compa t X-Ray Sour e inthe Supernova Remnant IC 443. Astrophysi al Journal, vol. 554, no. 2, pagesL205L208, 2001. (Cited on pages 33, 39, 43 and 125.)[Ong 2011 R. A. Ong. VERITAS Reports In reased A tivity of HESS J0632+057in Very High Energy Gamma Rays. The Astronomer's Telegram, vol. 3153,pages 1+, 2011. (Cited on page 10.)[Ormes et al. 1988 J. F. Ormes, M. E. Ozel and D. J. Morris. On the high-energygamma-ray signature of osmi -ray sour es. The Astrophysi al Journal,vol. 334, pages 722733, 1988. (Cited on pages 25 and 37.)[Orth & Bungton 1976 C. D. Orth and A. Bungton. Se ondary osmi -ray ele -trons and positrons from 1 to 100GeV in the upper atmosphere and inter-stellar spa e, and interpretation of a re ent positron ux measurement. As-trophysi al Journal, vol. 206, no. 1, pages 312332, 1976. (Cited on page 59.)[Ott et al. 2005 J. Ott, A. Weiÿ, C. Henkel and F. Walter. The Temperature Distri-bution of Dense Mole ular Gas in the Center of NGC 253. The Astrophysi alJournal, vol. 629, no. 2, pages 767780, 2005. (Cited on pages 77 and 174.)[Oya 2010 I. Oya. Observations of A tive Gala ti Nu lei with the MAGIC Tele-s ope. PhD thesis, Univ. Complutense de Madrid, 2010. (Cited on page 94.)[Paglione et al. 1996 T. A. D. Paglione, A. P. Mars her, J. M. Ja kson and D. L.Berts h. Diuse Gamma-Ray Emission from the Starburst Galaxy NGC 253.Astrophysi al Journal, vol. 460, pages 295302, 1996. (Cited on pages 57,59, 63 and 74.)[Pavlidou & Fields 2001 V. Pavlidou and B. Fields. Diuse Gamma Rays fromLo al Group Galaxies. The Astrophysi al Journal, vol. 558, no. 1, pages6371, 2001. (Cited on page 63.)[Persi et al. 2008 M. Persi , Y. Rephaeli and Y. Arieli. Very-high-energy emissionfrom M 82. Astronomy and Astrophysi s, vol. 486, no. 1, pages 143149,2008. (Cited on pages 57, 63, 64 and 69.)[Pun h et al. 1992 M. Pun h, C. W. Akerlof, M. F. Cawley, M. Chantell, D. J.Fegan, S. Fennell, J. A. Gaidos, J. Hagan, A. M. Hillas, Y. Jiang, A. D.

Bibliography 161Kerri k, R. C. Lamb, M. A. Lawren e, D. A. Lewis, D. I. Meyer, G. Mo-hanty, K. S. O'Flaherty, P. T. Reynolds, A. C. Rovero, M. S. S hubnell,G. Sembroski, T. C. Weekes and C. Wilson. Dete tion of TeV photons fromthe a tive galaxy Markarian 421. Nature, vol. 358, no. 6386, pages 477478,1992. (Cited on page 7.)[Ramaty & Lingenfelter 1966 R. Ramaty and R. E. Lingenfelter. Gala ti Cosmi -Ray Ele trons. Journal of Geophysi al Resear h, vol. 71, pages 3687+, 1966.(Cited on page 59.)[Ranalli et al. 2008 P. Ranalli, A. Comastri, L. Origlia and R. Maiolino. A deepX-ray observation of M82 with XMM-Newton. Monthly Noti es of the RoyalAstronomi al So iety, vol. 386, no. 3, pages 14641480, 2008. (Cited onpage 62.)[Reimer 2001 O. Reimer. Gamma-Ray Properties of Unidentied EGRET Sour es.In The Nature of Unidentied Gala ti High-Energy Gamma-Ray Sour es.Tonantzintla, Puebla, Mexi o, volume 267 of Astrophysi s and Spa e S ien eLibrary, pages 1734. Kluwer A ademi Publishers Dordre ht, 2001. (Citedon page 40.)[Rekola et al. 2005 R. Rekola, M. G. Ri her, Marshall L. M Call, M. J. Valtonen,J. K. Kotilainen and Chris Flynn. Distan e to NGC 253 based on the plane-tary nebula luminosity fun tion. Monthly Noti es of the Royal Astronomi alSo iety, vol. 361, no. 1, pages 330336, 2005. (Cited on page 74.)[Rho et al. 2001 J. Rho, T. H. Jarrett, R. M. Cutri and W. T. Rea h. Near-InfraredImaging and [O I Spe tros opy of IC 443 using Two Mi ron All Sky Surveyand Infrared Spa e Observatory. The Astrophysi al Journa, vol. 547, no. 2,pages 885898, 2001. (Cited on page 31.)[Ri e et al. 1988 W. Ri e, C. J. Lonsdale, B. T. Soifer, G. Neugebauer, E. L. Kopan,Lawren e A. Lloyd, T. de Jong and H. J. Habing. A atalog of IRAS obser-vations of large opti al galaxies. Astrophysi al Journal Supplement Series,vol. 68, pages 91127, 1988. (Cited on page 74.)[Rieke et al. 1973 G. H. Rieke, D. A. Harper, F. J. Low and K. R. Armstrong.350-MICRON Observations of Sour es in H II Regions, the Gala ti Center,and NGC 253. Astrophysi al Journal, vol. 183, pages L67L71, 1973. (Citedon pages 77 and 174.)[Rieke et al. 1980 G. H. Rieke, M. J. Lebofsky, R. I. Thompson, F. J. Low andA. T. Tokunaga. The nature of the nu lear sour es in M82 and NGC 253.Astrophysi al Journal, vol. 238, no. 1, pages 2440, 1980. (Cited on page 66.)[Rodríguez Marrero et al. 2008 A. Y. Rodríguez Marrero, D. F. Torres,E. de Cea del Pozo, O. Reimer and A. N. Cillis. Diusion of Cosmi

162 BibliographyRays and the Gamma-Ray Large Area Teles ope: Phenomenology at the1-100 GeV Regime. The Astrophysi al Journal, vol. 689, no. 1, pages213218, 2008. (Cited on pages 16, 19 and 20.)[Rodríguez Marrero et al. 2009 A. Y. Rodríguez Marrero, D. F. Torres andE. de Cea del Pozo. The GeV to TeV view of SNR IC443: predi tions forFermi. In 6th Workshop on S ien e with the New Generation of High EnergyGamma-Ray Experiments (S iNeGHE '08). arXiv: astro-ph/0901.2615v1,2009. (Cited on pages 16, 28, 38, 45, 48, 116 and 171.)[Rolke et al. 2005 W. A. Rolke, A. M. López and J. Conrad. Limits and onden eintervals in the presen e of nuisan e parameters. Nu lear Instruments andMethods in Physi s Resear h Se tion A, vol. 551, no. 2-3, pages 493503,2005. (Cited on page 106.)[Romero et al. 1999 G. E. Romero, P. Benaglia and D. F. Torres. Unidentied 3EGgamma-ray sour es at low gala ti latitudes. Astronomy and Astrophysi s,vol. 348, pages 868876, 1999. (Cited on page 40.)[Rosado et al. 2007 M. Rosado, L. Arias and P. Ambro io-Cruz. Kinemati s of theMole ular Hydrogen Asso iated with the Supernova Remnant IC 443. TheAstronomi al Journal, vol. 133, no. 1, pages 8998, 2007. (Cited on pages 33,34 and 50.)[Sabatini et al. 2010 S. Sabatini, M. Tavani, E. Striani, A. Bulgarelli, V. Vittorini,G. Piano, E. Del Monte, M. Fero i, F. de Pasquale, M. Trifoglio and 51 oauthor. Episodi Transient Gamma-ray Emission from the Mi roquasarCygnus X-1. The Astrophysi al Journal Letters, vol. 712, pages L10L15,2010. (Cited on page 10.)[Sakai & Madore 1999 S. Sakai and B. F. Madore. Dete tion of the Red GiantBran h Stars in M82 Using the Hubble Spa e Teles ope. The Astrophysi alJournal, vol. 526, no. 2, pages 599606, 1999. (Cited on page 62.)[Saz Parkinson et al. 2010 P.M. Saz Parkinson, M. Dormody, M. Ziegler, P. S. Ray,A. A. Abdo, J. Ballet, M. G. Baring, A. Belore, T. H. Burnett, G. A.Caliandro and 27 oauthors. Eight -ray Pulsars Dis overed in Blind Fre-quen y Sear hes of Fermi LAT Data. Astrophysi al Journal, vol. 725, no. 1,pages 571584, 2010. (Cited on pages 109, 110 and 175.)[Seiradakis et al. 1985 J. H. Seiradakis, W. Rei h, W. Sieber, R. S hli keiser andH. Kühr. An extended extragala ti radio sour e behind the Milky Way. As-tronomy and Astrophysi s, vol. 143, no. 2, pages 478480, 1985. (Cited onpage 103.)[Seta et al. 1998 M. Seta, T. Hasegawa, T. M. Dame, S. Sakamoto, T. Oka,T. Handa, M. Hayashi, J.-I. Morino, K. Sorai and K. S. Usuda. Enhan ed

Bibliography 163CO J=(2-1)/J=(1-0) Ratio as a Marker of Supernova Remnant − Mole u-lar Cloud Intera tions: The Cases of W44 and IC 443. The Astrophysi alJournal, vol. 505, no. 1, pages 286298, 1998. (Cited on pages 33 and 44.)[Smith et al. 1991 P. A. Smith, P. W. J. L. Brand, C. M. Mountain, P. J. Puxleyand N. Nakai. The (C-12)O to H2 ratio in the entre of M82. MonthlyNoti es of the Royal Astronomi al So iety, vol. 252, pages 6P11P, 1991.(Cited on page 63.)[Ste ker 1970 F. W. Ste ker. The Cosmi γ-Ray Spe trum from Se ondary Parti leProdu tion in Cosmi -Ray Intera tions. Astrophysi s and Spa e S ien e,vol. 6, no. 3, pages 377389, 1970. (Cited on page 59.)[Ste ker 1977 F. W. Ste ker. Observations of gala ti gamma-rays and their impli- ations for gala ti stru ture studies. Astrophysi al Journal, vol. 212, no. 1,pages 6070, 1977. (Cited on page 59.)[Ste ker 2007 F. W. Ste ker. Upper Limits to Fluxes of Neutrinos and Gamma-Raysfrom Starburst Galaxies. Journal of Physi s: Conferen e Series, vol. 60, no. 1,pages 215218, 2007. (Cited on pages 57 and 73.)[Stephens & Badhwar 1981 S. A. Stephens and G. D. Badhwar. Produ tion spe -trum of gamma rays in interstellar spa e through neutral pion de ay. Astro-physi s and Spa e S ien e, vol. 76, no. 1, pages 213233, 1981. (Cited onpage 51.)[Stri kland et al. 1997 D. K. Stri kland, T. J. Ponman and I. R. Stevens. ROSATobservations of the gala ti wind in M 82. Astronomy and Astrophysi s,vol. 320, pages 378394, 1997. (Cited on page 62.)[Strong & Moskalenko 1998 A. Strong and I. Moskalenko. Propagation of Cosmi -Ray Nu leons in the Galaxy. The Astrophysi al Journal, vol. 509, no. 1,pages 212228, 1998. (Cited on page 59.)[Sturner et al. 1997 S. J. Sturner, J. G. Skibo, C. D. Dermer and J. R. Mattox.Temporal Evolution of Nonthermal Spe tra from Supernova Remnants. As-trophysi al Journal, vol. 490, no. 2, pages 619632, 1997. (Cited on page 19.)[Swanenburg et al. 1981 B.N. Swanenburg, K. Bennett, G. F. Bignami, R. Bu - heri, P. Caraveo, W. Hermsen, G. Kanba h, G. G. Li hti, J. L. Masnou,H. A. Mayer-Hasselwander and 4 oauthors. Se ond COS B atalog of high-energy gamma-ray sour es. Astrophysi al Journal, Part 2 - Letters to theEditor, vol. 243, pages L69L73, 1981. (Cited on page 103.)[Tabatabaei et al. 2007 F. S. Tabatabaei, R. Be k, E. Kr14gel, M. Krause, E. M.Berkhuijsen, K. D. Gordon and K. M. Menten. High-resolution radio on-tinuum survey of M 33. II. Thermal and nonthermal emission. Astronomyand Astrophysi s, vol. 475, no. 1, pages 133143, 2007. (Cited on page 68.)

164 Bibliography[Tamm & Frank 1937 I. E. Tamm and I. M. Frank. Coherent radiation of fastele trons in a medium. Doklady Akademii Nauk SSSR, vol. 14, page 107,1937. (Cited on page 84.)[Tanaka & Takahara 2009 S. J. Tanaka and F. Takahara. A Model of the Spe tralEvolution of Pulsar Wind Nebulae. In Pro . 1st Fermi Symposium (Wash-ington, DC). arXiv: astro-ph/1001.2360, 2009. (Cited on page 110.)[Tatis he & Hernanz 2007 V. Tatis he and M. Hernanz. Eviden e for NonlinearDiusive Sho k A eleration of Cosmi Rays in the 2006 Outburst of theRe urrent Nova RS Ophiu hi. The Astrophysi al Journal, vol. 663, pagesL101L104, 2007. (Cited on page 10.)[Tatis he & Hernanz 2008 V. Tatis he and M. Hernanz. Eviden e for NonlinearDiusive Sho k A eleration of Cosmi Rays in the 2006 Outburst of RSOphiu hi. In RS Ophiu hi (2006) and the Re urrent Nova Phenomenon,volume 401 of Astronomi al So iety of the Pa i Conferen e Series, pages328+. A. Evans, M. F. Bode, T. J. O'Brien, & M. J. Darnley, 2008. (Citedon page 10.)[Tavani et al. 2010 M. Tavani, A. Giuliani, A. W. Chen, A. Argan, G. Barbi-ellini, A. Bulgarelli, P. Caraveo, P. W. Cattaneo, V. Co o, T. Contessiand 52 oauthors. Dire t Eviden e for Hadroni Cosmi -Ray A elerationin the Supernova Remnant IC 443. Astrophysi al Journal Letters, vol. 710,no. 2, pages L151L155, 2010. (Cited on page 42.)[Teles o & Harper 1980 C. M. Teles o and D. A. Harper. Galaxies and far-infraredemission. Astrophysi al Journal, vol. 235, no. 1, pages 392404, 1980. (Citedon pages 69, 74, 77 and 174.)[Teles o et al. 1991 C. M. Teles o, M. Joy, K. Dietz, R. De her and H. Campins.Infrared mapping of M82 - A starburst in an edge-on barred galaxy. Astro-physi al Journal, vol. 369, no. 1, pages 135146, 1991. (Cited on page 62.)[Tes aro et al. 2007 D. Tes aro, H. Bartko, N. Galante, F. Göbel and theMAGIC Collaboration. Study of the performan e and apability of the newultra-fast 2 GSample/s FADC data a quisition system of the MAGIC tele-s ope. ArXiv: astro-ph/0709.1410, 2007. (Cited on page 96.)[Tes aro 2010 D. Tes aro. TeV gamma-ray observations of nearby A tive Gala ti Nu lei with the MAGIC teles ope: exploring the high energy region of themultiwavelength pi ture. PhD thesis, Univ. Autonoma de Bar elona, 2010.(Cited on pages 87 and 174.)[Tian & Leahy 2006 W.W. Tian and D.A. Leahy. Asso iation Of The SNRG65.1+0.6 And The PSR J1957+2831. Bulletin of the Ameri an Astro-nomi al So iety, vol. 455, page 1053, 2006. (Cited on pages 102 and 103.)

Bibliography 165[Torres & Domingo-Santamaría 2005 D. F. Torres and E. Domingo-Santamaría.Some Comments on the High Energy Emission from Regions of Star For-mation Beyond the Galaxy. Modern Physi s Letters A, vol. 20, no. 37, pages28272843, 2005. (Cited on pages 60 and 173.)[Torres et al. 2001a D. F. Torres, G. E. Romero, J. A. Combi, P. Benaglia, H. An-derna h and B. Punsly. A variability analysis of low-latitude unidentiedgamma-ray sour es. Astronomy and Astrophysi s, vol. 370, pages 468478,2001. (Cited on page 34.)[Torres et al. 2001b D. R. Torres, M. E. Pessah and G. E. Romero. On the timevariability of Gamma-ray sour es: a numeri al analysis of variability indi es.Astronomis he Na hri hten, vol. 322, no. 4, pages 223227, 2001. (Cited onpage 40.)[Torres et al. 2003 D. F. Torres, Gustavo E. Romero, T. M. Dame, J. A. Combiand Y. M. Butt. Supernova remnants and γ−ray sour es. Physi s Reports,vol. 382, no. 6, pages 303380, 2003. (Cited on pages 31, 32, 33, 40, 44, 128and 170.)[Torres et al. 2004 D. F. Torres, O. Reimer, E. Domingo-Santamaría and S. W.Digel. Luminous Infrared Galaxies as Plausible Gamma-Ray Sour es forthe Gamma-Ray Large Area Spa e Teles ope and the Imaging Atmospheri Cerenkov Teles opes. The Astrophysi al Journal, vol. 607, no. 2, pages L99L102, 2004. (Cited on pages 57, 63 and 69.)[Torres et al. 2008 D. F. Torres, A. Y. Rodríguez Marrero and E. de Cea del Pozo.MAGIC J0616+225 as delayed TeV emission of osmi rays diusing fromthe supernova remnant IC 443. Monthly Noti es of the Royal Astronomi alSo iety: Letters, vol. 387, no. 1, pages L59L63, 2008. (Cited on pages 16,22, 25, 31, 35, 128 and 129.)[Torres et al. 2010 D. F. Torres, A.Y. Rodríguez Marrero and E. de Cea Del Pozo.The GeV to TeV onne tion in the environment of SNR IC 443. MonthlyNoti es of the Royal Astronomi al So iety, vol. 408, no. 2, pages 12571266,2010. (Cited on pages 16, 41, 49 and 172.)[Torres 2004 D. F. Torres. Theoreti al Modeling of the Diuse Emission of GammaRays from Extreme Regions of Star Formation: The Case of ARP 220.The Astrophysi al Journal, vol. 617, no. 2, pages 966986, 2004. (Citedon pages 45, 57, 58, 63 and 68.)[Troja et al. 2006 E. Troja, F. Bo hino and F. Reale. XMM-Newton Observationsof the Supernova Remnant IC 443. I. Soft X-Ray Emission from Sho kedInterstellar Medium. The Astrophysi al Journal, vol. 649, no. 1, pages 258267, 2006. (Cited on pages 33 and 34.)

166 Bibliography[Turner & Ho 1985 J. L. Turner and P. T. P. Ho. The 1 parse radio ore andpossible nu lear eje tion in NGC 253. Astrophysi al Journal, vol. 299, no. 2,pages L77L81, 1985. (Cited on page 74.)[Völk et al. 1989 H. J. Völk, U. Klein and R. Wielebinski. M82, the Galaxy, andthe dependen e of osmi ray energy produ tion on the supernova rate. As-tronomy and Astrophysi s, vol. 213, no. 1, pages L12L14, 1989. (Cited onpage 68.)[Völk et al. 1996 H. J. Völk, F. A. Aharonian and D. Breits hwerdt. The Nonther-mal Energy Content and Gamma-Ray Emission of Starburst Galaxies andClusters of Galaxies. Spa e S ien e Reviews, vol. 75, no. 1, pages 279297,1996. (Cited on pages 63 and 64.)[Volkova 1980 L. V. Volkova. Energy Spe tra and Angular Distributions of Atmo-spheri Neutrinos. Sov. J. Nu l. Phys., vol. 31, pages 784, 1980. (Cited onpage 55.)[Wang et al. 1992 Z. R. Wang, I. Asaoka, S. Hayakawa and K. Koyama. Hard X-rays from the supernova remnant IC 443. Astronomi al So iety of Japan,Publi ations, vol. 44, no. 3, pages 303308, 1992. (Cited on page 33.)[Waxman & Bah all 1999 E. Waxman and J. Bah all. High energy neutrinos fromastrophysi al sour es: An upper bound. Physi al Review D, vol. 59, no. 2,pages 023002+, 1999. (Cited on page 73.)[Weaver et al. 2002 K. A. Weaver, T. M. He kman, D. K. Stri kland andM. Dahlem. Chandra Observations of the Evolving Core of the StarburstGalaxy NGC 253. The Astrophysi al Journal, vol. 576, no. 1, pages L19L23, 2002. (Cited on page 72.)[Weekes et al. 1989 T. C. Weekes, M. F. Cawley, D. J. Fegan, K. G. Gibbs, A. M.Hillas, P. W. Kowk, R. C. Lamb, D. A. Lewis, D. Ma omb, N. A. Porter, P. T.Reynolds and G. Va anti. Observation of TeV gamma rays from the Crabnebula using the atmospheri Cerenkov imaging te hnique. Astrophysi alJournal, vol. 342, no. 1, pages 379395, 1989. (Cited on page 7.)[Weiÿ et al. 2001 A. Weiÿ, N. Neininger, S. H14ttemeister and U. Klein. The ee tof violent star formation on the state of the mole ular gas in M 82. Astron-omy and Astrophysi s, vol. 365, pages 571587, 2001. (Cited on page 63.)[Wild et al. 1992 W. Wild, A. I. Harris, A. E kart, R. Genzel, U. U. Graf, J. M.Ja kson, A. P. G. Russell and J. Stutzki. A multi-line study of the mole ularinterstellar medium in M 82's starburst nu leus. Astronomy and Astro-physi s, vol. 265, no. 2, pages 447464, 1992. (Cited on page 63.)[Yao et al. 2006 W. M. Yao, C. Amsler, D. Asner, R. M. Barnett, J. Beringer,P. R. Bur hat, C. D. Carone, C. Caso, O. Dahl and 160 o authors. Review

Bibliography 167of Parti le Physi s. Journal of Physi s G Nu lear Physi s, vol. 33, no. 1,page 1232, 2006. (Cited on page 84.)[Young & S oville 1984 J. S. Young and N. Z. S oville. Mole ular louds in M82.Astrophysi al Journal, vol. 287, no. 1, pages 153166, 1984. (Cited onpage 63.)[Zhang & Fang 2008 L. Zhang and J. Fang. Nonthermal Emission from a Radio-bright Shell-Type Supernova Remnant IC 443. The Astrophysi al Journal,vol. 675, no. 1, pages L21L24, 2008. (Cited on page 40.)

List of Figures1.1 All-parti le osmi ray spe trum. From [Be ker 2008. . . . . . . . . 21.2 Main physi al pro esses that an generate gamma-ray photonsthrough dierent intera tionsm, both hadroni (b) and leptoni ( ,d, e). Annihilation an take pla e among leptons or hadrons. . . . . 31.3 Skymap above 100MeV from our Galaxy by the entire EGRET mis-sion (phases 1 to 4), with the main sour es dete ted in gamma rays.Credit: EGRET Team/NASA. . . . . . . . . . . . . . . . . . . . . . 51.4 All-sky map above 300MeV during the rst year of the Fermi LATteles ope. Credit: NASA/DOE/Fermi LAT Collaboration. . . . . . . 61.5 Left: Skymap RX J1713.7-3946 by H.E.S.S., bla k ontours over-plotted show the X-ray brightness that ASCA dete tes from 1 to3 keV. From [Aharonian et al. 2004b. Right: Fermi LAT spe tralenergy distribution (SED) of the SNR W44, ea h urve orrespondsto π0 de ay (solid), ele tron bremsstrahlung (dashed), inverse Comp-ton s attering (dotted) and bremsstrahlung from se ondary ele tronsand positrons (thin dashed). From [Abdo et al. 2010f . . . . . . . . 91.6 Left: Skymap NGC 253 by HESS. Right: Spe trum M 82 by VERITAS. 131.7 A tual design of the I eCube neutrino dete tor with 5160 opti al sen-sors viewing a kilometer ubed of natural i e. The signals dete tedby ea h sensor are transmitted to the surfa e over the 86 strings towhi h the sensors are atta hed. I eCube en loses its smaller prede- essor, AMANDA. From I eCube S ien e Team - Fran is Halzen. . . 152.1 SEDs generated by CR propagation in ISM with dierent properties.Fluxes orrespond to a loud with M5/d

2kpc = 0.5. Curve for D10 =1026, 1027, and 1028 m2 s−1 are shown with solid, dotted, and dashedlines respe tively. Sensitivities of EGRET (red) and Fermi (blue)(both for dierent dire tions in the sky with dierent ba kground ontribution), H.E.S.S. (magenta) (survey mode and pointed obser-vations with typi al integrations), and MAGIC (yellow), are shownfor omparison purposes (see gure 1 of [Funk et al. 2008 for detailson sensitivities). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

170 LIST OF FIGURES2.2 Examples of the model predi tions for a hadroni maxima in the1 − 100GeV regime. The left top (bottom) panel shows the pre-di tions for a loud s aled at M5/d

2kpc = 0.025 (0.04), lo ated at20 (30) p from an a elerator of 104 (3 × 104) yr, diusing with

D10 = 1027 m2 s−1. The right top (bottom) panel urve shows thepredi tions for a loud s aled at M5/d2kpc = 0.08 (0.06) lo ated at10 (20) p from an a elerator of 103 (104) yr, diusing with D10 =1028 m2 s−1. In reasing the ratio M5/d2kpc, the urves move up main-taining all other features. . . . . . . . . . . . . . . . . . . . . . . . . 242.3 For ea h ombination of age and a elerator-target separation, forwhi h more than two thousand spe tra where numeri ally produ ed,the energy of the maximum of su h spe tra are shown in a on-tour plot. The olor of the dierent ontours orresponds to therange of energy where the maximum is found a ording to the olorbar above ea h gure. From top to bottom, plots are reated forthe ase of an impulsive sour e inje ting protons in a medium with

D10=1026 m2 s−1, 1027 m2 s−1 and 1028 m2 s−1. . . . . . . . . . . . 262.4 The parameters for the plots are as follows, (top left) the dashed urveon the left: t = 4×105 yr, R = 5p , M5/d2kpc = 0.01; the dashed urveon the right: t = 104 yr, R = 20p , M5/d

2kpc = 0.1; (top right) thedashed urve on the left: t = 2×106 yr, R=100 p , M5/d

2kpc = 3; thedashed urve on the right: t = 4×103 yr, R = 15p ,M5/d

2kpc = 0.1;(bottom left) the dashed urve on the left: t = 2×106 yr, R=15p ,

M5/d2kpc = 0.004; the dashed urve on the right: t = 103 yr, R =5p , M5/d

2kpc = 1; ( bottom right) the dashed urve on the left: t =2×106 yr, R = 40p , M5/d

2kpc = 0.017; the dashed urve on the right:

t = 6×104 yr, R = 30p ,M5/d2kpc = 2.5. D10 is set to 1026 m2 s−1.

R is the a elerator- loud separation. . . . . . . . . . . . . . . . . . 273.1 CO distribution around the remnant IC 443 (G189.1+3.0). The 3EGgamma-ray sour e J0617+2238 is plotted with white ontours. Theopti al boundary of the SNR is superimposed as a bla k ontour[Lasker et al. 1990. The opti al emission seems to fade in regionswhere CO emission in reases. This indi ates that the mole ular ma-terial is likely lo ated on the foreground side of the remnant, absorb-ing the opti al radiation. Plot taken from [Torres et al. 2003, gure10. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323.2 CR spe trum generated by IC 443 at two dierent distan es, 10 (solid)and 30 (dashed) p , at the age of the SNR. Two types of a eleratorare onsidered, one providing a ontinuous inje tion (bla k) and otherproviding a more impulsive inje tion of CRs (red). The horizontal linemarks the CR spe trum near the Earth. The Y-axis units have been hosen to emphasize the ex ess of CRs in the SNR environment. . . . 36

LIST OF FIGURES 1713.3 MAGIC and EGRET measurement of the neighborhood of IC 443(stars and squares, respe tively) as ompared with model predi -tions. The top (bottom) panel shows the results for an impulsive( ontinuous) ase. At the MAGIC energy range, the top panel urvesshow the predi tions for a loud of 8000M⊙ lo ated at 20 (1), 25(2), and 30 (3) p , whereas they orrespond to 15 (1), 20 (2), 25(3), and 30 (4) p in the bottom panel. At lower energies, the urveshows the predi tion for a few hundred M⊙ lo ated at 3− 4p . TheEGRET sensitivity urves (in red) are shown for the whole lifetimeof the mission for the Gala ti anti- entre (solid), whi h re eived thelargest exposure time and has a lower level of diuse gamma-rayemission, and for a typi al position in the Inner Galaxy (dashed),more dominated by diuse gamma-ray ba kground. The Fermisensitivity urves (in blue) show the simulated 1-year sky-survey sen-sitivity for the Gala ti North pole, whi h orresponds to a positionwith low diuse emission (solid), and for a typi al position in theInner Galaxy (dashed). These urves were taken from http://www-glast.sla .stanford.edu/software/IS/glast_latperforman e.html.From [Rodríguez Marrero et al. 2009. . . . . . . . . . . . . . . . . . 383.4 Integrated photon ux as a fun tion of time above 100MeV and100GeV, solid (dashed) lines orrespond to the ase of the loudlo ated at 10 (30) p . The horizontal lines represent the values ofintegrated uxes in the ase that the CR spe trum intera ting withthe loud is the one found near Earth. The verti al line stands forthe SNR age. EGRET and Fermi initially predi ted integral sen-sitivity are shown, onsistent in value and olor oding with those inFigure 3.3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394.1 Earlier MAGIC and EGRET (stars and diamonds, respe tively), andre ent Fermi LAT and VERITAS (squares and upper trianges, re-spe tively) measurements of the neighborhood of IC 443 as omparedwith model predi tions for an impulsive and a ontinuous a elera-tor, as onsidered in Chapter 3. The nominal values of parametersfor these models are the same as in Figure 3.3, although here thedierent ontributions are summed up. See the text for details. . . . 444.2 As in Figure 4.1, summed results (right) are produ ed by two main omponents (left) oming from a giant loud in front of the SNR,whi h is at least partially overtaken by the diusing osmi rays(∼5300M⊙ at 10 p ) and a loser-to-the-shell loud (at 4 p , with350M⊙), similar to the previous examples. The dotted and dashedlines at the VHE range orresponds to dierent normalizations,whi h an also be understood as intera ting masses of ∼4000 and∼3200M⊙ at the same distan e. The diusion oe ient is as be-fore, D10 = 1026 m2 s−1. . . . . . . . . . . . . . . . . . . . . . . . . . 46

172 LIST OF FIGURES4.3 Cosmi -ray spe trum generated by the impulsive a elerator (IC 443)at the two dierent loud distan es onsidered in Figure 4.2: 10 (solid)and 4 p (dashed), at the age of the SNR, as a fun tion of energy.Dierent olors show results for dierent diusion oe ients (bla k,D10 = 1026 m2 s−1; and red, D10 = 1027 m2 s−1). The right panelshows the ratio between the osmi -ray spe tra of the left panel, andthe osmi -ray spe trum near Earth, as a fun tion of energy. . . . . 474.4 Example of a model output with D10 = 1027 m2 s−1. The dierent urves represent results for the lo ation of the giant mole ular loudat 10, 15, 20, 25, 30 p from the SNR shell, whereas the lose-to-the-SNR loud is at 4 p . Neither in this nor in any other of the studiedmodels, the VHE sour e spe trum an be reprodu ed by varying theparameters with su h a diusion oe ient s ale. Furthermore, theresulting SED in the Fermi LAT range is not hard enough to mat hthe data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474.5 Contour plot depi ting the position of the peak of the SED gener-ated by a 30 kyrs old inje tion intera ting with louds at dierentdistan es, for a range of diusion oe ient s ale, D10. . . . . . . . . 484.6 Examples of solutions around the dis ussed main values, exploring thedegenera ies (or un ertainties) in determining the numeri al values ofmodel parameters mat hing the observational data. The order of thepanels in this plot, top to bottom and left to right, orresponds withthe parameters des ribed in Table 4.5, being ea h olumn one of thethree groups therein. Adapted from [Torres et al. 2010. . . . . . . . 494.7 Comparing gamma-ray yields with dierent δ parameters, from leftto right δ = 0.4, 0.5, 0.6, and 0.7. The other parameters are as inFigure 4.2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 524.8 Comparison (left) and ratio (right) of the ross se tion parame-terizations used in the previous Figures, the δ-fun tional form by[Aharonian & Atoyan 1996, with that of [Kelner et al. 2006. . . . . 534.9 Left: Gamma-ray ux results using the [Kelner et al. 2006 approx-imation for dierent distan es from the shell to the TeV-produ ing loud, 10 (solid), 15 (dotted) and 20 (dashed) p . The lose-to-the-remnant loud is xed at 4 p and ontains 350M⊙ hanges in thislatter value do not improve the overall t. Right: Gamma-ray uxresults using the parameterization from [Kelner et al. 2006 for dier-ent values of the giant loud mass pla ed at a xed distan e of 10 p (see text for details). . . . . . . . . . . . . . . . . . . . . . . . . . . 534.10 Ele trons (ele trons and positrons are shown together), photons andtwo avors of neutrinos produ ed within the louds onsidered nearbyIC 443, using a set of parameters shown in Figure 4.9 (right panel)with mass of the giant loud equal to 7272M⊙. The νµ and νe neu-trino urves show both the parti le and the anti-parti le ux. Datashould only be ompared with the photon urve. . . . . . . . . . . . 54

LIST OF FIGURES 1735.1 Code ow of Q-diffuse, adapted from[Torres & Domingo-Santamaría 2005 to take into a ount newupdates developed for this work. . . . . . . . . . . . . . . . . . . . . 605.2 Left: Comparison of the steady population of ele trons that wouldresult after solving the diuse loss equation inje ting only primary(solid blue) and only se ondary (dashed blue) parti les from kno k-on and pion de ay in the inner region of M82. The total steadyele tron population (solid bla k), resulting from the inje tion of bothprimary and se ondary ele trons, is also shown. Parameters usedin this Figure oin ide with those presented later within the samemodel (Figures 5.3 5.4). Right: Steady proton (solid) and ele tron(dashed) distributions in the innermost region of both M82 (bla k)and NGC 253 (red). . . . . . . . . . . . . . . . . . . . . . . . . . . . 655.3 Multi-frequen y spe trum of M82 from radio to infrared. The ob-servational data points orrespond to: [Klein et al. 1988 (triangles),[Hughes et al. 1994 ( ir les) and [Förster et al. 2003 (squares), andreferen es therein in ea h ase. The results from modelling orre-spond to: syn hrotron plus free-free emision (dashed), dust emis-sion (dotted) splitted in a ool (blue, Tc = 45K) and a warm(purple,Tw ≃ 200K) omponent, and the total emission from radioand IR emission (solid). . . . . . . . . . . . . . . . . . . . . . . . . . 665.4 Left: Energy distribution of the dierential gamma-ray uxes, explor-ing a range of un ertainties in supernova explosion rate and utosin the primary energy, as it is explained in the text. The sensi-tivity urves for EGRET (red), Fermi (blue), MAGIC (purple), allfrom [Funk et al. 2008, and the intended one for the forth omingCerenkov Teles ope Array (CTA, violet) are shown. Right: Dieren-tial neutrino ux predi tions from the inner region of M82, total andseparated in dierent hannels. The neutrino predi tions make useof the same explored parameters already presented in the left paneland explained in the text. . . . . . . . . . . . . . . . . . . . . . . . . 675.5 Comparison of the multi-wavelength predi tions for dierent initialspe tral slope in the inje tion, see Figure 5.3 for further details. Thebla k urve orrespond to the model with proton inje tion spe trump = 2.1 and magneti eld B = 130µG, whereas the red urve orre-spond to the results of modelling with p = 2.3 and B = 170µG. Thegamma-ray emission from the −2.3 model in the ase of the highestSN explosion rate is shown against those obtained with the harderinje tion (the green shadow, oming from the un ertainties des ribedbefore). Main dieren es appear at high energies. . . . . . . . . . . . 715.6 Dierent ontributions to the radio emission (syn hrotron + free-free) by the steady primary-only (blue) and se ondary-only (yellow)ele tron population, also ompared to the total radio emission of thewhole ele tron population (bla k). . . . . . . . . . . . . . . . . . . . 72

174 LIST OF FIGURES5.7 Energy distribution of the dierential gamma-ray uxes of M82, ex-ploring a range of un ertainties in supernova explosion rate and ef- ien y to inje t energy from SN to CR. The shaded green area orresponds to the original model presented in this Chapter and[de Cea del Pozo et al. 2009b. Data points and upper limit orre-spond to both VERITAS (stars) and Fermi (diamonds) dete tions. . 765.8 Multifrequen y spe trum of NGC 253 from radio to infrared.The observational data points orrespond to: [Carilli 1996 (tri-angles), [Elias et al. 1978 ( ir les), [Rieke et al. 1973 (asterisks),[Ott et al. 2005 (diamonds) and [Teles o & Harper 1980 (squares),and referen es therein. The results from modelling orrespond to:syn hrotron plus free-free emission (dashed), dust emission (dotted)splitted in a ool (blue, Tcold = 45K) and a warm (purple, Twarm

∼ 200K) omponent, and the total emission from radio and IR emis-sion (solid). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 775.9 Energy distribution of the dierential gamma-ray uxes of NGC 253,exploring the un ertainty in distan e, a range of times ale diusion(τ0) and possible utos in the proton inje tion spe trum. The orig-inal model from [Domingo-Santamaría & Torres 2005 is also shownfor omparison, as well as data points from Fermi dete tion (dia-monds) and the integral ux from the H.E.S.S. dete tion (square),transformed in dierential ux (assuming a range of inje tion spe tra). 786.1 Polarization of the medium by a harged parti le with (a) low velo ity,v < c/n, and (b) high velo ity, v > c/n. Huygens onstru tion ofCherenkov wavefront in ( ) . . . . . . . . . . . . . . . . . . . . . . . 846.2 S heme of a γ-ray indu ed ele tromagneti shower (right) and ahadron-indu ed shower (left) developing in the atmosphere . . . . . . 856.3 Image formation s heme in the amera of an IAC teles ope. Thevalues are referred to a 1 TeV γ-indu ed shower. The blue part is theimage head whereas the red part is the image tail. The numbers inthe pixels orrespond to the number of in ident photons. [Tes aro 2010 876.4 Pi ture of the two MAGIC teles opes: MAGIC-I (left) and MAGIC-II(right) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 896.5 S heme for the standard trigger onguration in MAGIC I (left) andII (right) ameras ([Meu i et al. 2007, [Cortina et al. 2009). . . . . 906.6 Sket h of the denition of the signal (ON) and ba kground (OFF)regions in wobble observations. The anti-sour e is the OFF positionlo ated symmetri ally to the ON position (the red ir le) with respe tto the enter [Mazin 2007. . . . . . . . . . . . . . . . . . . . . . . . . 926.7 Graphi al representation of some of image parameters des ribed inthe text. The nominal position of the observed sour e is (x0, y0)[Mankuzhiyil 2010. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

LIST OF FIGURES 1756.8 Sket h of a tree stru ture for the lassi ation of an event v via size,length and width. The de ision path through the tree, leading to lassi ation of the event as hadron an be followed [Errando 2009. . 977.1 Radio image of the region around SNR G65.1+0.6. The numberedobje ts are: (1) 3EG J1958+2909, (2) 2 CG 065+00, (3) 0FGLJ1954+2838, (4) 0FGL J1958.1+2848, (5) region of dierent spe -tral index, (6) IRAS 19520+2759, (7) bright ompa t radio obje t. . 1027.2 Observation setup for the two Fermi sour es J1954 and J1958 in the ontext of SNR G65.1+0.6 and a Milagro signi an e ontour. J1958appears only in one wobble position (W1), so the OFF data is takenfrom the other wobble sample, using the same position relative tothe pointing dire tion. The outline of the remnant is taken from theradio map in [Lande ker et al. 1990. The extension of the Milagrosigni an e ontour [Abdo et al. 2009i is ompatible with their pointspread fun tion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1057.3 Plots showing the θ2 distributions for J1954 (left) and J1958 (right).The shapes of the ON and OFF distributions agree well with ea hother in both sour es, whi h means there is no γ−ray signal. Thedashed lines indi ate the signal regions. . . . . . . . . . . . . . . . . 1067.4 Skymaps of the ex ess events (top) and signi an e (bottom) forJ1954 (left) and J1958 (right). The bla k star shows the enter of thepointed sour e, the white ir le represents the MAGIC point spreadfun tion (PSF) at mid energies. The distribution of signi an es arewell-tted with a simple gaussian. . . . . . . . . . . . . . . . . . . . . 1077.5 Compilation of ux measurements and upper limits for: (left)1FGL J1954.3+2836 from Fermi [Saz Parkinson et al. 2010,Abdo et al. 2010a , (right) 1FGL J1958.6+2845 from EGRET[Hartman et al. 1999, Fermi [Abdo et al. 2010b, Abdo et al. 2010a,together with MAGIC and Milagro [Abdo et al. 2009i data. The 3%fra tion of the MAGIC Crab spe trum [Albert et al. 2008b is shownfor omparison. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1108.1 Konrad Bernlöhrs adapted layout of the original ultra-CTA (fromPadova 2008 CTA meeting). Four types of teles opes are plotted, seetable 8.1 for details, being 1 = red, 2 = green, 3 = blue and 4 =magenta. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1128.2 Layout ongurations of the three representatives: B (left), D ( en-ter), I (right). The types of teles opes orresponds to the ones de-tailed in Table 8.1, being 1 (red), 2 (green), 3 (blue), 4 (magenta), thenumber in bra kets orresponds to the eld of view of ea h teles ope. 1138.3 Dierential sensitivity of the three representative ongurations: D(red), B (blue), I (green), in Crab units (C.U.) for 50 hours of obser-vation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

176 LIST OF FIGURES8.4 Angular resolution (80% ontainment radius) of the three representa-tive ongurations, see Figure 8.3, for best resolution (left) and bestsensitivity (right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1148.5 Energy resolution of the three representative ongurations, see Fig-ure 8.3, for best resolution (left) and best sensitivity (right). . . . . . 1158.6 Total gamma ray emission from a mole ular loud of mass 105M⊙lo ated at a distan e of 1 kp . The distan e between the mole u-lar loud and the SNR is 50, 100 and 200 p for left, enter andright panel, respe tively. The solid, dotted, and dashed lines referto the emission at 2000, 8000 and 32000 years after the SN explo-sion. Private ommuni ation from S. Gabi i, based on gure 5 from[Gabi i et al. 2009. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1168.7 Gamma-ray emission from a mole ular loud of 105M⊙ illuminatedby an a elerator (whi h explosion o urred 2000 yrs ago) and sim-ulated with layout onguration I, for 20 hours of observation. Themole ular loud is lo ated at a distan e of 1 kp from the observerand pla ed at dierent distan es from the a elerator: 50 (bla k), 100(blue), 200 (red) p . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1178.8 Spe tral modeling of the gamma-ray emission of the two losest star-burst galaxies M82 (left) and NGC 253 (right). The overlaid blue box orresponds to the range of energies that CTA is predi ted to observe(from a few tens of GeV to ∼ 100TeV). Data points in both plots orrespond to dete tions by Fermi (diamond), VERITAS (star) andH.E.S.S. (square). Figures taken from [de Cea del Pozo et al. 2009a.For a omplete explanation, see Chapter 5. . . . . . . . . . . . . . . 1198.9 M82 spe trum as simulated for CTA for 30 hrs with onguration B(top), D ( enter), I (bottom). The grey points orrespond to observa-tional data: at lower energies, an upper limit (68% .l.) from Fermi,and at higher energies, the dierential data points and an upper limit(99% .l.) from VERITAS. . . . . . . . . . . . . . . . . . . . . . . . . 1218.10 NGC 253 spe trum as simulated for CTA in 50 (left) and 70 hours(right) with onguration D. The observational data (grey) orre-sponds to an upper limit (68% .l.) from Fermi ; and H.E.S.S. integraldata point ( onverted to dierential ux) . . . . . . . . . . . . . . . . 122

LIST OF FIGURES 1779.1 Sky map of gamma-ray andidate events (ba kground subtra ted)in the dire tion of MAGIC J0616+225 for an energy threshold ofabout 150GeV in gala ti oordinates. Overlayed are 12CO emission ontours ( yan) showing the position of their maxima oin iding withthe MAGIC dete tion, and ontours of 20 m VLA radio data (green),X-ray ontours from Rosat (pu12rple) and gamma-ray ontours fromEGRET (bla k). The white star denotes the position of the pulsarCXOU J061705.3+222127. The bla k dot shows the position of the1720MHz OH maser. The white ir le shows the MAGIC I PSF ofσ = 0.1. From [Albert et al. 2007b . . . . . . . . . . . . . . . . . . 1269.2 Lo ations and extensions of the 4 gamma-ray sour es. Centroid po-sitions are marked with dierent symbols: EGRET (blue triangle),MAGIC (red downside triangle), VERITAS entroid (green star) andFermi LAT (bla k diamond). The respe tive lo alization errors areshown as rosses. Best-t spatial extensions of the Fermi ( ross-hat hed band) and VERITAS (striped green band) sour es are drawnas rings with radii of θext68 and widths of ±1σ error. The PWN lo a-tion is shown as a magenta dot. Contours are the lo ations and shapesof the lo al sho ked mole ular louds taken from [Huang et al. 1986.Figure taken from [Abdo et al. 2010g . . . . . . . . . . . . . . . . . 1279.3 Plot showing the θ2 distribution for the MAGIC sour e in the regionof IC 443. The shapes of the ON and OFF distributions agree wellwith ea h other. A hint of signal is already seen in barely 3 hours ofobservations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1309.4 Skymap (map of signi an es) of the SNR IC 443 (left) and distribu-tion of signi an es (right) of the same region. The bla k ross marksthe position of the MAGIC sour e J0616+225. The white ir le showsthe PSF of the MAGIC stereo system. . . . . . . . . . . . . . . . . . 1319.5 CTA simulated spe trum of IC 443 with onguration I (green urve)and 50 hours. The input spe tra (grey lines) orrespond: at lower en-ergies to the power law published from Fermi above 3GeV, whereasat higher energies the power law index is taken from the measure-ments by MAGIC and VERITAS. The grey symbols show the a tualdata from previous experiments. The simulated data points for CTAabove 100 GeV an be tted to a power law, whi h index dependson the assumed initial spe tra (bla k dots for MAGIC, blue ones forVERITAS). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1329.6 The spe trum of IC443 is simulated with layout CTA-I for 50 hours ofobservation, assuming an input spe trum in the form of a power lawwith a ut o of 5 (top) and 10 (bottom) TeV. The power law index isthe one published by MAGIC (-3.1). The highest energy ut-o that ould be resolved is at 5 TeV, at higher energies (i.e. 10 TeV) a ttingwith power law with a ut o (red) is statisti ally undistinguishablefrom a simple power law (green) in the observed CTA range of energies.134

List of Tables1.1 Mission parameters of the three latest spa e teles opes in the MeV GeV range: EGRET, AGILE and Fermi LAT. The sensitivity above100MeV is onsidered for a 2-year survey at high latitudes. . . . . . 61.2 Performan e of the three main Cherenkov experiments in the GeV TeV range: H.E.S.S., MAGIC and VERITAS. In the title: ♯ Tels.stands for number of teles opes, Tels. Area is the area of ea h tele-s ope, f.o.v. is the eld of view, Tot. Area is the total area of the arrayof telse opes, Eth is the energy threshold, Ang. res. means angularresolution, and Sensitivity 50 h onveys the sour e ux in 50 hours ofobservation respe t to the Crab Nebula ux with a signi an e of 5sigma. From [De Angelis 2011. . . . . . . . . . . . . . . . . . . . . . 82.1 Dependen e of the SED (E2F vs. E) on various parameters. Imp.( ont.) stands for the impulsive ( ontinuous) a elerator ase. De-penden es upon loud parameters su h as density (nCl), mass (MCl),and radius (RCl) are obvious and related. . . . . . . . . . . . . . . . 234.1 Main model parameters for solutions shown in Figure 4.6, ex eptmodel 3, whi h instead is shown in Figure 4.2. DGMC and dsnr arethe distan e to the GMC and the loser-to-the-SNR mole ular louds.The three quoted f values dene MGMC = (1/f) 8000M⊙. The threegroups explore dierent degenera ies: in the position of the GMC, inthe position of the smaller loud, and on the diusion oe ient. . . 515.1 Physi al parameters used in the multi-wavelength model of M82, pre-sented in the previous se tion and in [de Cea del Pozo et al. 2009b,together with spe i values that try to mat h the emission from theVERITAS dete tion. In any ase, the (small) variations exploredbelow are within the former predi tions of the original model. Thelist of parameters is divided in se tions: observational values, derivedfrom observational values, obtained from modelling, and assumed.SB stands for starburst. . . . . . . . . . . . . . . . . . . . . . . . . . 755.2 Physi al parameters used in the multiwavelength modelof NGC 253, as presented both in the previous[Domingo-Santamaría & Torres 2005 study and in this se tion,but exploring some variations allowed within the model in orderto mat h the emission from the H.E.S.S. dete tion. The list ofparameters is divided in se tions: observational values, derived fromobservational values, obtained from modelling, and assumed. SBstands for starburst. . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

180 LIST OF TABLES7.1 Dierential upper limits for both sour es, for the present ross he kanalysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1087.2 Chara teristi parameters of the two Fermi pulsars 1FGL J1954.3 +

2836 and 1FGL J1958.6+2845, refereed as J1954 and J1958, respe -tively. See referen es in the text. . . . . . . . . . . . . . . . . . . . . 1098.1 Main hara teristi s of dierent types of teles opes for CTA. . . . . . 1128.2 Summary of the observational dete tions of M82 and NGC 253 byFermi (at high energies, HE), VERITAS and H.E.S.S. (very high en-ergies, VHE). The energy thresholds for the integral ux al ulationsare: above 100MeV in the ase of Fermi LAT observations, above700GeV for VERITAS, and above 220GeV for H.E.S.S. . . . . . . . 1199.1 Power law (p.l.) index for MAGIC-like or VERITAS-like spe trumfor IC 443, when simulating CTA response for 50 hrs in onguration I.1319.2 Statisti s of two ttings to the spe trum for IC 443, when simulatingCTA response for 20 hrs in onguration I. The simulated data pointsuse the intrinsi spe tra from Figure 9.6: a power law with a ut oat a few TeV. The two ttings, then, are a power law with or withouta ut o. The rst number is the t probability, whi h values shouldbe in the range between 5 − 60%, the losest to 50%, the better.The se ond set of numbers represents the redu ed χ2 (dened as χ2divided by the number of degrees of freedom), and for a good t, itshould be loser to 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

LaTeX template based on: http://olivier. ommowi k.org/thesis_template.phpCover design: SaiSoto.