ULTRA-HIGH ENERGY COSMIC RAYS P. Tinyakov Université Libre de Bruxelles (ULB), Brussels Odense Winter School, October 2012
Outline 1 Introduction 2 Extensive airshowers 3 Detection methods Surface Detectors Fluorescence Telescopes UHECR experiments 4 Propagation Interaction with radiation backgrounds Deflections in magnetic fields 5 Production of UHECR 6 State of the art Spectrum Composition Anisotropy 7 Summary
Historical remarks 1912 On a ballon at an altitude of 5000 meters, Victor Hess discovered penetrating radiation coming form space. 1929 Using a newly invent cloud chamber, Dimitry Skobelzyn observed the first ghostly tracks left by cosmic rays. 1933 In his cloud chamber, Carl Anderson discovered antimatter in the form of the anti-electron, later called the positron. 1937 Seth Neddermeyer and Carl Anderson discovered the the muon in cosmic rays. This gave birth to the science of elementary particle physics. 1938 Pierre Auger, putting particle detectors far apart high in the Alps, discovered extensive air showers.
Historical remarks
All-particle CR spectrum = detector of 1 m2 has count rate 10 3 per second = detector of 1 m2 has count rate 2 per month Ultra-High Energy CR = detector of 1 km2 has count rate 1 per century
Why it is interesting to study ultra-high energy cosmic rays (UHECR)? These are the highest-energy particles ever observed. The energy range is inaccessible to colliders. Observations of UHECR give a possibility to test laws of Nature at extremely high energies. UHECR carry information about astrophysical processes capable of accelerating particles to extremely high energies. This, potentially, opens a new window into the Universe at highest energies ever. CHARGED PARTICLE ASTRONOMY?
Extensive airshowers Earth atmosphere is used as a calorimeter first interaction primary particle hadronic component air nucleus π + core Cherenkov light π o γ γ γ π e e + ν µ shower maximum γ γ µ ν electromagnetic cascade
Extensive airshowers first interaction hadronic component core Cherenkov light shower maximum electromagnetic cascade Probability of interaction 1 when σ n(x)dl 1 key quantity: depth X = m p n(x)dl First interaction: take σ (100MeV) 2 = m p n(x)dl 100 g/cm 2 Note: vertical atmosphere 1000 g/cm 2 = first interaction at h 10 km Depth of first interaction depends on primary particle: smaller for iron, larger for proton, even larger for photon. However, fluctuations are large. Width of the shower at maximum: a few km. Shower development in not sensitive to primary particle: details average away. Exception: number of muons (smaller for photon-initiated showers).
Extensive airshowers Main features: Most energy is carried by the EM component Shower front is curved; the curvature depends on age: younger showers are more curved = in principle, curvature may be used to measure the depth of the first interaction Very old (very inclined) showers consist of muons which are the most penetrating component (except undetectable neutrinos). These showers have a flat front. = Young very inclined showers = neutrino primary particles
Extensive airshowers Longitudinal shower profile is expressed in terms of atmospheric depth X = ρdl and is approximately given by the Gaisser-Hillas formula: ( ) X X max X 0 ( ) Xmax λ Xmax X N(X) = N max exp X max X 0 λ X max X 0 λ depth of the maximum depth of the first interaction a parameter number of charged particles, arbitrary units 1.2 1 0.8 0.6 0.4 0.2 0 200 400 600 800 1000 1200 1400 depth, g/cm2
Extensive airshowers Lateral distribution (density of charged particles as distance r from the core) is well fitted by the Linsley formula ( ) r α ( S(r) = const 1 + r ) (η α) R 0 R 0 R 0 Moliere radius 80 90 m α, η two shape parameters 1e+10 1e+09 1e+08 1e+07 1e+06 100000 10000 0 200 400 600 800 1000 1200 1400 distance, m
Extensive airshowers Modeling of showers No accurate analytical model of shower exists Monte-Carlo code for simulation of showers: CORSIKA Even numerically, complete calculation is very difficult. A necessary approximation thinning Crucial ingredient of simulations hadronic models. They require extrapolation of hadronic interactions to high energies = source of systematic errors LHC data are useful (actually used) for the extrapolation
Summary I: extensive airshowers Shower development is a statistical process involving a few billion particles. First interaction that carries direct information about the primary particle is never observed. Shower modeling, especially at first stages, involves uncertainties resulting from extrapolation of particle cross sections to CM energies > 100 TeV However, there are indirect observables sensitive to the depth of the first interaction (= cross section), such as shower front curvature, depth of the shower maximum etc.
DETECTION Two basic detection methods: Detecting particles at the ground by an array of the detectors (SD) Detecting fluorescent light emitted (isotropically) by the shower core (FD) Possibility of radio detection is under investigation.
SURFACE DETECTORS
SURFACE DETECTORS Schematic detector design
SURFACE DETECTORS A typical SD event (Telescope Array)
SURFACE DETECTORS Geometry reconstruction: from detector counts & timing information
SURFACE DETECTORS Energy estimation: from full Monte-Carlo
SURFACE DETECTORS Typical parameters Detector spacing: 1 1.5 km Systematic error in energy: 20% Energy resolution: 20% Angular resolution 1 1.5 Energy-independent aperture above certain energy threshold, which is (a few) 10 18 ev for TA
FD event example
FLUORESCENCE TELESCOPES Energy determination: back-of-envelope estimate E = length de dx E = 1 2 N max 1000g/cm 2 2 MeV g/cm 2 E = 1 10 9 evn max Energy determination is robust. Based on center of shower, not tails. Easy to Monte Carlo.
FLUORESCENCE TELESCOPES Energy error 20% Energy determination is close to calorimetric Little dependence on Monte-Carlo Energy-dependent aperture Arrival direction error in monocular mode is very asymmetric, 0.5 5 Possibility of stereo and hybrid observations = substantial improvement in angular resolution
HYBRID LAYOUT
Hybrid event example
UHECR EXPERIMENTS
UHECR EXPERIMENTS Parameters of some SD detectors exposure angular energy # of events (km 2 yr sr) resolution resolution at E > 10 19 ev AGASA 2 500 2.5 25% 775 Auger 23 000 0.9 17% 7800 HiRes 2 500 0.6 14% 378 TA 5 000 1.5 20% 1809 Yakutsk 824 2.5 30% 364 at E = 10 20 ev
PROPAGATION Uniform distribution of arrival directions suggests extragalactic origin Galactic magnetic field cannot confine protons with energy E > 10 18 ev = almost certainly CRs with E > (a few) 10 18 ev are of extragalactic origin = UHECR propagate over large (perhaps cosmological) distances. Issues to consider: scattering off matter deflections in magnetic fields (for charged particles)
Scattering Scattering off protons negligible: L = 1 σn 1031 cm 10 3 R U Scattering off CMB photons important (the Greizen-Zatsepin-Kuzmin, or GZK effect). Characteristic energy: E p m π(m p + m π /2) 2ω γ Greisen 1966; Zatsepin, Kuzmin 1966 (a few) 10 19 ev
GZK processes Cross sections:
GZK processes Total pion photoproduction cross section: Cross section [mubarn] 700 600 500 400 300 200 neutron proton 100 0 0 10 1 10 E [GeV] lab 2 10 3 10
GZK processes Mean free path at 10 20 ev: λ 1 σn 5 Mpc Typical energy loss in single collision: 20% As a function of energy: e + e pair production size of the Universe attenuation length interaction length
GZK processes Energy as a function of distance (protons)
The GZK cutoff For uniformly distributed sources one has flux R 0 1 r 2 r 2 dr R = there should be a cutoff by a factor R U /R GZK 100
What is observed? First measurements were controversial AGASA HiRes Now the controversy is resolved: there is a cutoff
What if UHECR are nuclei? If UHECR are nuclei, the picture remains qualitatively the same Relevant quantity is the γ-factor rather than the energy. For nuclei the γ-factor is smaller (by their atomic number) = pion production on CMB is no longer possible Light nuclei attenuate more than heavy ones (i.g., iron) The process relevant for energy attenuation is spallation, i.e., splitting off of nucleons. CM energies required are of order few MeV = dominated by IR photons The resulting attenuation length is similar to the case of protons (for iron) However, attenuation of nuclei depends on the IR background which is known less accurately than the CMB
Photon backgrounds Actual photon backgrounds: CMB radio IR
What if UHECR are photons? Photons attenuate due to e + e -production Photon Interaction and Attenuation Length [Mpc] 3 10 2 10 1 10 0 10 1 10 2 10 17 10 18 10 19 10 20 21 10 10 E [ev] 22 10 23 10 24 10
What if UHECR are photons? At highest energies they propagate over tens of Mpc at most
Summary of UHECR attenuation Regardless of the UHECR composition (assuming SM particles), their propagation length shrinks from hundreds of Mpc a few tens of Mpc at energies around 10 20 ev = There must be drop (cutoff) in the UHECR spectrum at E 10 20 ev (the GZK cutoff in case of protons) = Sources of highest-energy CRs must be relatively close to us, within 100 Mpc or so
Deflections in magnetic fields Deflections are governed by the Lorentz force Ṗ = q v B In relativistic limit, deflections depend only on the ratio of charge to energy Regular field θ 0.52 q ( E ) 1 ( R ) ( B ) 10 20 ev 1kpc 10 6 G Random field θ 1.8 q ( E ) 1 ( lc R ) 1/2 ( B ) 10 20 ev 50Mpc 2 10 9 G = Need to understand magnetic fields
Extragalactic fields Extragalactic magnetic fields are poorly known. Bounds from Faraday rotation measurements: B < 10 9 G but may be much smaller. From numerical simulations: Dolag, Grasso, Springel, Tkachev 2003; Sigl, Miniati, Enslin 2004 (contradict each other) E = 10 19 ev
Extragalactic fields MFs in clusters are large, but this is irrelevant... Caveat: if we live in a filament, this may no longer be true: Earth Earth...because CRs must come from clusters anyway directions may get isotropized.
Galactic field Galactic magnetic field consists of regular and turbulent components It can be inferred from the Faraday rotation measures of extragalactic and Galactic radio sources (pulsars) Taylor, Stil and Sunstrum, 2009, ApJ, 702, 1230
Galactic field Another RM survey: Kronberg, Newton-McGee 2009
Galactic field The RM data can be well fitted by a model containing both disk and halo components Pshirkov, P.T. 2011 ApJ 738 192 NORTH HALO FIELD SUN DISK FIELD SOUTH HALO FIELD
Galactic field Model vs. observations
Galactic field Basic model parameters: Magnitude of disk field around the Earth: 2µG Pitch: 5 Thickness of the disk: 1 kpc Magnitude of the halo: 4µG Height of the halo above disk: 1.3 kpc Typical uncertainties: 30%
Deflections: magnitude E = 4 10 19 ev, protons
Deflections: directions
Time delays A side effect of deflections in MF is time delays Delays occur for geometrical reasons δl L (δφ)2 Take δφ = 3, (δφ) 2 = 0.003 L = 5 kpc t = 50 yr L = 50 Mpc t = 5 10 5 yr L δl δφ = transient sources are seen as steady
Summary of deflections in MF Deflections in the extragalactic MF are likely to be negligible. Caveat: we may live inside a filament of the large-scale structure where fields can reach 10 8 10 7 G with the correlation length O(Mpc). Then deflections may be large. Deflections of protons in GMF are dominated by the regular field and may be of the order 2 6 at energy E = 10 20 ev depending on the direction = Charge-particle astronomy may be possible at highest energies, but only if UHECR are protons. In case of iron nuclei deflections are 26 times larger and arrival directions are isotropized.
Production of UHECR The production mechanisms can be generally divided into two types bottom-up (e.g., acceleration) top-down (e.g., decay of superheavy particles, topological defects, etc.) The top-down mechanisms are disfavored for two reasons: they are less motivated given that the cutoff in the spectrum is observed they are disfavored by the constraints on the photon fraction in UHECR (will be discussed below).
Acceleration Acceleration can be either stochastic or direct. The general idea behind stochastic acceleration (variations of Fermi mechanism): shock
Acceleration Operates in many astrophysical environments: supernova remnants, active galactic nuclei, GRB, colliding galaxies etc. Naturally gives power spectrum dn/de E γ Known to work at low energies However, unclear whether it can provide energies as high E 10 20 ev
Acceleration There are two general constraints on maximum energy: Particle Larmor radius has to fit within the acceleration cite (Hillas condition) E < qbr Energy losses have to be smaller than energy gains qb > q4 m 4 E 2 B 2
Acceleration Hillas condition + constraints due to losses (proton, E = 10 20 ev): Neutron stars Excluded by energy losses 10 White dwarfs 5 log(b ) G 0 5 AGN GRB Radio lobes Galaxy clusters Excluded by Hillas condition Galactic disk & halo 15 10 5 0 5 log(r ) kpc
Acceleration An alternative direct acceleration by an electric field induced in the vicinity of compact objects (e.g., rapidly-rotating magnetized neutron stars; rotating black holes embedded in a magnetic field)
Acceleration The acceleration to E 10 20 ev is much easier in case of iron than in case of protons Neronov et al, 2009 New J. Phys. 11 065015
Summary of acceleration Stochastic acceleration mechanism: better understood known to operate in various environments (supernova remnants, AGN,...) at lower energies maximum energy is constrained = unclear whether it can accelerate particles to sufficiently high energies, and where this can be achieved Direct acceleration: is less studied (idealized condition discussed before are unlikely in Nature, while realistic conditions are too complicated) but potentially more efficient
STATE OF THE ART New generation of experiments: Pierre Auger Observatory and Telescope Array. Advance: large area + hybrid type of detector. = Increase in statistics by a factor > 10 = Possibility to cross-calibrate the surface and fluorescent detectors
Spectrum The idea to build the hybrid detector originated from the contradiction between the AGASA (SD) and HiRes (FD) results. Hybrid detector allows to cross-calibrate the SD and FD by using the set of events common to both detectors. This program failed: both Auger and TA found that the energies measured by two detectors are systematically different (SD higher). It is not understood why. Note: it is still useful to have hybrid detector since SD has higher statistics, while FD has more straightforward energy determination. Current strategy is to calibrate SD on FD.
Spectrum TA energy calibration by FD
Spectrum The spectrum measured by the TA
Spectrum A mere rescaling brings spectra of different experiments to a good agreement:
Spectrum Remaining issues: The difference between SD and FD energy estimates The differences in energy determination by different experiments What is the absolute energy scale? This is important for the interpretation of the spectrum: What is the origin of the cutoff at around E = 10 20 ev? Is it due to propagation or a mere cutoff in the source? What is the dip at E < 10 19 ev a trace of e + e -pair production or Galactic-to-extragalactic transition?
Composition An observable sensitive to composition is the depth of the shower, X max Both mean value of X max and its variation differ for proton and iron: X max is larger for protons X max is larger for protons The measurement is difficult, because X max is measured by the FD only, and FD event selection is biased with respect to X max
Composition Protons or heavy Nuclei?
Composition Limits on flux of photons
Anisotropy Arrival directions is the most robust observable: independent of hadronic models free from systematic errors exposure angular energy # of events (km 2 yr sr) resolution resolution at E > 10 EeV Auger 21 000 0.9 17% 4727 HiRes 2 500 0.6 14% 378 TA 2 900 1.5 20% 854 Yakutsk 824 2.5 30% 364 at E = 10 20 ev
Global sky distribution Global distribution of UHECR at E > 10 19 ev is compatible with uniform. E>10 EeV 30 25 E>10 EeV 20 15 10 5 P =42% KS 0 0 50 100 150 200 250 300 350 RA, degrees 120 100 P = 37% KS E>10 EeV Equatorial coordinates, TA events with E > 10 EeV (May 2008 September 2010) 80 60 40 20 0 data MC 80 60 40 20 0 20 40 60 80 DEC, degrees
Global sky distribution The same, but at higher energies: 7 6 5 4 E>57 EeV P KS = 30% 3 2 1 0 0 50 100 150 200 250 300 350 RA, degrees Equatorial coordinates, TA events with E > 57 EeV (May 2008 September 2010) 7 6 5 4 3 2 1 0 E>57 EeV P = 6% KS MC DEC, degrees data 80 60 40 20 0 20 40 60 80
Harmonic analysis: dipole amplitude Amplitude of the dipole as a function of energy: Auger, differential Auger, cumulative
Harmonic analysis: dipole phase TA preliminary Auger Yakutsk
Small-scale clustering AGASA 3.5 3 2.5 2 1.5 1 0.5 0 E > 40 EeV 0 5 10 15 20 25 30 35 40 δ TA Auger Auger
Search for point sources: correlations with AGN
Updated Auger analysis Auger collaboration, Astroparticle Phys. 34 (2010) 314 p data 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 p = 0.21 iso Data 68% CL 95% CL 99.7% CL 10 20 30 40 50 Total number of events (excluding exploratory scan) Updated analysis: 21 events correlate out of 55 total. With the whole set, isotropy is excluded at about 3σ.
Search for AGN signal in Telescope Array TA collaboration, 2011 Equatorial coordinates Original estimate of the correlating fraction is not supported Consistent with the updated estimate Consistent with no correlation 3 times the present statistics is needed for a conclusive test
AGNs or Cen A? Tensions within AGN interpretation: Virgo paucity Chemical composition (Fe or p?) Local AGNs are weak Cen A alternative explanation for the correlation signal? Gorbunov et al, 0711.4060, 0804.1088 Fargion, 0801.0227 Cen A is the closest radiogalaxy by chance projected on LSS Is outside of HiRes and TA 70 FOV Data 60 50 68% Isotropic 95% Isotropic 99.7% Isotropic Events 40 30 20 10 0 0 20 40 60 80 100 120 140 160 Cen A
Correlations with BLL 11 correlate at 0.8 while 3 expected Gorbunov et al, JETP Lett. 80 (2004) 145 HiRes Collaboration, ApJ 636 (2006) 680 Not supported by the Auger data Pierre Auger Collaboration, ICRC 07
CORRELATIONS WITH MATTER DISTRIBUTION At highest energies the UHECR propagation distance becomes of order 50 100 Mpc due to GZK suppression The matter distribution at these scales is inhomogeneous, being a network of galaxy clusters, filaments and voids = Anisotropy is expected, unless the deflections are very large = Composition is crucial Example: protons, E > 57 EeV, 6 Gaussian smearing: V Co UM C l=360 Hy N l=180 l=0 PP PI F E
Methods to test the correlations with the LSS: Pick a (complete) catalog of objects which are sufficiently numerous to trace well the matter distribution, and cross-correlate UHECR arrival directions with this catalog. Knowing matter distribution (e.g., from a complete catalogs of galaxies) calculate the predicted UHECR flux and compare to observations.
Correlation function Smoothed map (θ = 5 ) 2MRS catalog of galaxies with K mag < 11.25 data collected after the exploratory scan = distribution following matter is favored over isotropy Results of the likelihood analysis
TA results Expected flux is calculated from the galaxy distribution (extended 2MRS catalog, 109 000 galaxies with K mag < 12.5 within 250 Mpc), with Gaussian smoothing of the angular width θ. Protons are assumed. Compatibility with data is checked with the flux sampling test and presented as the probability as a function of θ Flux map smoothed at θ = 6, protons P-value showing compatibility with isotropy or LSS, as a function of θ 1 10 EeV, no MF 0.1 95 % CL ISO p value 0.01 0.001 STRUCT E > 10 EeV 0.0001 0 5 10 θ 15 20
The same plots, but at higher energies 1 40 EeV, no MF ISO p values 0.1 95% CL STRUCT E > 40 EeV 0.01 0 5 10 15 20 θ 1 57 EeV, no MF. ISO p values 0.1 95% CL E > 57 EeV 0.01 0 STRUCT 5 10 15 20 θ
Summary No hints of new physics in the latest data. The existence of the cutoff in the spectrum is firmly established. However, it is not known whether this is a GZK cutoff, or merely maximum energy in the sources Composition is still unknown (but no large fraction of photons). Results are controversial. Key question for the future of the field: is charged-particle astronomy possible? No definite evidence for anisotropy (including point sources) is found so far. Various hints exist, but not yet sufficiently significant. O(3) O(10) increase in statistics will clear this up.