PDF hosted at the Radboud Repository of the Radboud University

Nijmegen

The following full text is a publisher's version.

For additional information about this publication click this link.

http://hdl.handle.net/2066/151781

Please be advised that this information was generated on 2023-10-29 and may be subject to
change.
 Cosmic Radiation

Reconstruction of Cosmic-Ray Properties from Radio Emission

of Extensive Air Showers

Johannes Schulz
 c 2016, Johannes Schulz
Cosmic Radiation
Thesis, Radboud University Nijmegen, The Netherlands
Illustrated; with bibliographic information and English, Dutch, and German summaries

ISBN: 978-94-028-0017-3
Printed by Ipskamp Drukkers, Enschede

This work was supported by the Netherlands Research School for Astronomy (NOVA).
 Cosmic Radiation

Reconstruction of Cosmic-Ray Properties from Radio Emission

of Extensive Air Showers

Proefschrift

ter verkrijging van de graad van doctor

aan de Radboud Universiteit Nijmegen
op gezag van de rector magnificus,
volgens besluit van het college van decanen
in het openbaar te verdedigen op donderdag 11 februari 2016
om 14:30 uur precies

door

Johannes Schulz

geboren op 23 oktober 1986

te Lemgo, Duitsland
P ROMOTOR: Prof. dr. H. D. E. Falcke

C OPROMOTOR: Dr. habil. J. R. Hörandel

M ANUSCRIPTCOMMISSIE: Prof. dr. S. J. de Jong (voorzitter)

Prof. dr. S. C. M. Bentvelsen

Universiteit van Amsterdam

Prof. dr. M. Erdmann

RWTH Aachen University, Duitsland

Prof. dr. P. J. Groot

Dr. T. Huege
Karlsruher Institut für Technologie, Duitsland
 Cosmic Radiation

Reconstruction of Cosmic-Ray Properties from Radio Emission

of Extensive Air Showers

Doctoral Thesis

to obtain the degree of doctor

from Radboud University Nijmegen
on the authority of the Rector Magnificus
according to the decision of the Council of Deans
to be defended in public
on Thursday, February 11, 2016 at 14:30 hours

by

Johannes Schulz

born on October 23, 1986

in Lemgo, Germany
S UPERVISOR: Prof. dr. H. D. E. Falcke

S UPERVISOR: Dr. habil. J. R. Hörandel

D OCTORAL THESIS COMMITTEE: Prof. dr. S. J. de Jong (chair)

Prof. dr. S. C. M. Bentvelsen

University of Amsterdam

Prof. dr. M. Erdmann

RWTH Aachen University, Germany

Prof. dr. P. J. Groot

Dr. T. Huege
Karlsruhe Institute of Technology, Germany
 Contents

Introduction 1

1 Cosmic Rays and Extensive Air Showers 3

1.1 Cosmic Rays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.1 Energy Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.2 Composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.1.3 Arrival Direction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.1.4 Propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.1.5 Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2 Extensive Air Showers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2.1 Longitudinal Development . . . . . . . . . . . . . . . . . . . . . . 11
1.2.2 Heitler Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.2.3 Monte Carlo Simulations . . . . . . . . . . . . . . . . . . . . . . . 15
1.2.4 Radio Emission . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.2.5 Detector Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2 The Pierre Auger Observatory 27

2.1 The Surface Detector Array . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2 The Fluorescence Detector . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.3 The Low-Energy Detector Extensions - Infill and HEAT . . . . . . . . . . . . 33
2.4 The Auger Engineering Radio Array - AERA . . . . . . . . . . . . . . . . . 35
2.4.1 AERA24 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.4.2 AERA124 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.4.3 AERA153 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.4.4 Research and Development Stations . . . . . . . . . . . . . . . . . . 40
2.4.5 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.4.6 Wireless Communication System . . . . . . . . . . . . . . . . . . . 41
2.5 System Integration for AERA124 . . . . . . . . . . . . . . . . . . . . . . . 44
2.5.1 Solar Power System . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.5.2 Radio Frequency Interference Tests . . . . . . . . . . . . . . . . . . 46
2.5.3 Tests of the Wireless Communication System . . . . . . . . . . . . . 50

3 Data Acquisition and Hybrid Shower Reconstruction 55

3.1 Background for Radio Air Shower Detection . . . . . . . . . . . . . . . . . . 55
3.2 Triggering and Data Acquisition . . . . . . . . . . . . . . . . . . . . . . . . 55
3.2.1 Self-Trigger . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
viii

3.2.2 External-Trigger . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.2.3 Internal Particle-Trigger . . . . . . . . . . . . . . . . . . . . . . . . 57
3.2.4 Periodic-, Random-, and Airplane-Triggers . . . . . . . . . . . . . . 58
3.2.5 Data Handling and Merging . . . . . . . . . . . . . . . . . . . . . . 58
3.3 Shower Reconstruction Software Framework . . . . . . . . . . . . . . . . . 59
3.4 Baseline Detector Shower Reconstruction . . . . . . . . . . . . . . . . . . . 59
3.4.1 Surface Detector Array . . . . . . . . . . . . . . . . . . . . . . . . 59
3.4.2 Fluorescence Detector . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.4.3 FD-SD Energy Calibration with Hybrid Showers . . . . . . . . . . . 64
3.5 Radio Air Shower Reconstruction . . . . . . . . . . . . . . . . . . . . . . . 65
3.5.1 Radio Detector Description . . . . . . . . . . . . . . . . . . . . . . . 66
3.5.2 Initialisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.5.3 Antenna Channel Level . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.5.4 Station Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.5.5 Event Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.5.6 Station and Event Pre-, Intermediate-, and Post-Selection . . . . . . . 70
3.5.7 RdObserver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.6 AERA Dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.7 Radio Simulations for the AERA Detector Geometry . . . . . . . . . . . . . 75
3.8 Generic Radio Simulations on a Star-Shape Pattern . . . . . . . . . . . . . . 76

4 AERA Scintillators 79
4.1 Scintillator Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.1.1 Scintillator Module . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.1.2 Scintillator Electronics . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.1.3 Scintillator-Trigger . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.1.4 Trigger Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.2 Scintillator Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.2.1 Scintillator Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.2.2 Detector Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.2.3 Lateral Distribution Function . . . . . . . . . . . . . . . . . . . . . . 90
4.3 Reconstruction of Measured Air Showers . . . . . . . . . . . . . . . . . . . 92
4.3.1 Dataset and Reconstruction . . . . . . . . . . . . . . . . . . . . . . 92
4.3.2 Measured Particle LDF . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.3.3 Direction Reconstruction . . . . . . . . . . . . . . . . . . . . . . . . 97

5 The Radio Lateral Distribution Function 101

5.1 LDF Parametrization for AERA . . . . . . . . . . . . . . . . . . . . . . . . 101
5.2 Relation between LDF Parameters and Air-Shower Properties . . . . . . . . 103
5.2.1 Air-Shower Simulation Set . . . . . . . . . . . . . . . . . . . . . . . 103
 ix

5.2.2 Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

5.2.3 Core Position . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
5.2.4 Depth of the Shower Maximum . . . . . . . . . . . . . . . . . . . . 113

6 Radio Energy Reconstruction 121

6.1 Data Selection and Event Reconstruction . . . . . . . . . . . . . . . . . . . . 121
6.1.1 Preselection of Cosmic Ray Candidates . . . . . . . . . . . . . . . . 121
6.1.2 Reconstruction of Radio Data . . . . . . . . . . . . . . . . . . . . . 122
6.1.3 Selection of Radio Signals Induced by Cosmic Rays . . . . . . . . . 123
6.1.4 Uncertainties on the Energy Density in a Single Radio Station . . . . 125
6.2 Energy Estimator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.2.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.2.2 Shower-by-Shower Uncertainties . . . . . . . . . . . . . . . . . . . 127
6.2.3 Absolute Scale Uncertainties . . . . . . . . . . . . . . . . . . . . . . 129
6.3 Energy Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
6.3.1 Uncertainties of the Reconstructed Cosmic-Ray Energy . . . . . . . . 130
6.3.2 Precision and Possible Improvements of the Reconstruction . . . . . 130
6.3.3 The Energy Content of Extensive Air Showers in MHz Radiation . . 131
6.4 Relevance for the Absolute Energy Scale Determination . . . . . . . . . . . 132

7 Radio Xmax Reconstruction 135

7.1 Xmax from the 2D Radio LDF . . . . . . . . . . . . . . . . . . . . . . . . . 135
7.1.1 Calibration with Full Detector Simulations . . . . . . . . . . . . . . 135
7.1.2 Comparison of Simulations and Multi-Hybrid Data . . . . . . . . . . 137
7.2 Xmax from Simulated Energy-Density Profiles . . . . . . . . . . . . . . . . . 140
7.2.1 Basic Principle of the Method . . . . . . . . . . . . . . . . . . . . . 141
7.2.2 Individual Shower Simulations . . . . . . . . . . . . . . . . . . . . . 142
7.2.3 Data Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.2.4 Fit Procedure for the Reconstruction of Xmax . . . . . . . . . . . . . 145
7.2.5 Application to Multi-Hybrid Data . . . . . . . . . . . . . . . . . . . 145
7.2.6 Uncertainty of the Reconstructed Value . . . . . . . . . . . . . . . . 150
7.2.7 Direct Comparison between both Methods . . . . . . . . . . . . . . . 154
7.3 Xmax Reconstruction using RD-SD Hybrid Data . . . . . . . . . . . . . . . . 154

8 Conclusions and Outlook 159

A Appendix 165
A.1 Wireless Communication Hardware and Settings . . . . . . . . . . . . . . . 165
A.2 Module Sequence RdObserver . . . . . . . . . . . . . . . . . . . . . . . . . 167
A.3 Quality Cuts for the RD-SD-FD Multi-Hybrid Dataset . . . . . . . . . . . . 169
x

A.4 Scintillator Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170

A.5 LDF Width Parameter Correlation . . . . . . . . . . . . . . . . . . . . . . . 172
A.6 Likelihood Function of the Energy Calibration . . . . . . . . . . . . . . . . . 173

Bibliography 174

Index 189

Summary 193

Samenvatting 197

Zusammenfassung 201

Acknowledgments 205

About the Author 207

 Introduction

The Large Hadron Collider can accelerate charged particles to enormous energies of more
than 1014 eV [1]. However, the existence of particles with more than one-hundred-thousand
times higher energies has already been known for more than 50 years [2]. The discovery of
these ultra highly-energetic particles, which are called cosmic rays, marks about half time
between the first observations that radiation of high penetration power enters the atmosphere
from above ("Strahlung von sehr hoher Durchdringungskraft von oben her in unsere Atmo-
sphäre eindringt") [3] by Victor Hess in 1912 and today where the understanding of its nature
is still incomplete. During these 100 years, the scientific impact of cosmic rays rapidly devel-
oped to a large field of fundamental research. Major steps were taken in measuring the flux,
the composition, and the arrival direction of the cosmic rays over a broad range of energies,
but nevertheless the sources and the acceleration processes remain unknown.
Most of the contemporary ultra high-energy cosmic-ray observatories employ a ground based
detector array which is sensitive to the particle cascade of millions of secondary particles de-
veloping when a cosmic ray collides with a nucleus in the Earth’s atmosphere. These particle
cascades are called extensive air showers and were first described by Pierre Auger and Werner
Kolhörster [4, 5]. The particles can be detected with 100% duty cycle and a good resolution in
reconstructing the energy and the arrival direction of the original cosmic ray can be achieved.
A key to improve the understanding of the sources of the cosmic rays is however to identify the
nature of these particles. This information will improve the picture of their acceleration and
propagation and will help to exclude possible sources. Nowadays, optical detectors are used
to measure the shower development and extract information about the composition of cosmic
rays. These detectors are sensitive to the fluorescence afterglow of the atmosphere when ni-
trogen is excited by the particle cascade. The amount of light produced is very low and only
extremely sensitive cameras are able to detect a signal. Already moonlight is outshining the
fluorescence light by orders of magnitude which limits the duty cycle of these optical detec-
tors. Therefore, they can only provide information about the particle type for a fraction of the
cosmic rays arriving at Earth.
Shifting from observations in the optical regime to MHz radio frequencies allows to detect
the particle showers as well. This method is independent of almost all environmental condi-
tions and thereby allows for a close to 100% duty cycle. After being introduced in the 1960s
with promising results but many technological limitations, the radio technique hibernated for
decades until being revived to help answering the still remaining questions about cosmic rays.
One of these modern radio air-shower detectors is the Auger Engineering Radio Array (AERA)
built in the heart of the Pierre Auger Observatory, which is the world’s largest air shower ob-
servatory.
2

Coexisting with a classical particle detector array and a fluorescence light detector, the radio
technique has been developed and calibrated towards the necessary precision to provide high
resolution air shower data with 100% duty cycle. The development of methods for the recon-
struction of the interesting air shower properties from data recorded with AERA is the main
subject of this thesis. During the last 3 years, AERA was also extended twice and parts of the
necessary hardware work are presented as well.
In this thesis, an overview of cosmic-ray and air-shower physics is given in the beginning,
followed by the description of hardware and software tools for simultaneous multi-detector
air-shower measurements as well as the discussion of the measurements of the energy and the
trend of the cosmic-ray composition as determined from the radio data, and future prospects
on coming developments in the end. In chapter 1, an introduction to cosmic rays and ex-
tensive air showers is given together with information about the current detection techniques.
Chapter 2 is a description of the Pierre Auger Observatory in general and AERA in particular.
Here, additional information about the topic of system integration for the upgrades of AERA
is presented. The design of the wireless communication system and the tests for the system
integration are discussed in detail (original work). The third chapter contains a presentation of
the data acquisition of AERA and the reconstruction of radio, Surface, and Fluorescence De-
tector data. Additionally, new simulation strategies are introduced and discussed together with
new functionality added to the reconstruction framework (collaborative work). In chapter 4,
the reconstruction of air shower parameters based on small particle counters installed in the
radio stations for triggering purposes is discussed. Data from these detectors are reconstructed
and analysed for the first time (original work). Chapter 5 contains a detailed discussion about
simulation studies of the lateral distribution of the radio signal produced for the geographical
position of AERA and its sensitivity to air shower parameters (original work). The results are
used in chapter 6 and 7 to extract information about air shower properties. This information
with respect to the cosmic-ray energy is employed in chapter 6 to perform a calibration be-
tween the Surface Detector array of the Pierre Auger Observatory and AERA (collaborative
work). Furthermore, this procedure opens the door towards an individual energy calibration
of cosmic-ray detector arrays based on first principle calculations. In chapter 7, two meth-
ods for the determination of the depth of maximum air-shower development based on radio
measurements are presented. This depth is an indicator for the particle type of the cosmic ray.
Both approaches are discussed and for the first time, the two methods are compared to an inde-
pendent measurement taken from the Fluorescence Detector of the Pierre Auger Observatory.
Additionally, the results of the two methods are compared among each other and one of them
is used to determine the average depths of the shower maximum for air showers measured with
AERA as a function of the energy (original work). These values are compared to experimental
data from other experiments for the given energy range as well as to theoretical predictions.
The presented work is concluded in chapter 8 and an outlook on future short, mid, and long
term developments is given.
 Cosmic Rays and Extensive Air Showers
1
A cosmic ray is an energetic particle travelling through outer space. Even though this defini-
tion applies to photons the same way as to neutrinos and charged particles, mostly the latter
are referred to as cosmic rays. The Earth is constantly hit by cosmic rays and various research
projects are measuring these particles to identify their origin. The most energetic cosmic rays
are subject of the studies in this thesis. This chapter gives an overview of the main observable
properties of cosmic rays: energy, arrival direction, and particle type. Furthermore, the prop-
agation and possible sources are discussed, followed by a description of the phenomenon of
extensive air showers and its experimental observation.

1.1 Cosmic Rays

The energy spectrum of cosmic rays fascinates by the wide range it spans over tens of decades
in decreasing flux with increasing energy. Covering these ranges experimentally is only pos-
sible by using multiple detection strategies. A complication of measuring cosmic rays is their
interaction with nuclei in the Earth’s atmosphere, which makes direct detection of the cosmic
rays at ground impossible. For the lower energetic particles (E < 1012 eV), every square meter
is exposed to several hundred cosmic rays per second. This high rate allows for space or air
borne experiments on satellites or balloons to directly detect the particles and measure their
energy, charge/mass, and arrival direction e.g., [6, 7, 8, 9, 10]. As the maximum payload on
space- and airborne carriers is limited, large sensitive areas cannot be instrumented above the
Earth’s atmosphere and ground based detector systems are employed to detect cosmic rays
with energies above 1015 eV corresponding to a flux of less than one particle per square meter
per year. These detectors detect the cascade of secondary particles produced in the interaction
between the cosmic ray and the atmosphere, called extensive air shower (EAS) [4, 5, 11] (see
section 1.2).

1.1.1 Energy Spectrum

A great achievement is the determination of the energy spectrum up to energies of 1020 eV by
several ground based experiments despite the indirect nature of the measurements at energies
higher than 1015 eV. The spectrum is shown in figure 1.1. For energies below 1010 eV, the flux
is suppressed by solar magnetic fields and modulated by solar activity [12]. Above this limit,
4 Chapter 1. Cosmic Rays and Extensive Air Showers

the flux decreases monotonously and the spectrum is close to a single power law

dN
∝ E −γ (1.1)
dE
with a spectral index γ = 2.7. For energies above 1015 eV, the spectral index exhibits changes
at certain energies. Around 4.5 × 1015 eV, the spectrum steepens to γ = 3.1. This break
is called the "knee" and it is followed by the "second knee" around 4 × 1017 eV where the
spectral index further increases to γ = 3.3. Close to the end of the energy range, the spectrum
flattens again at the "ankle" at around 4 × 1018 eV. These features are displayed in the inset
of figure 1.1. At the end of the spectrum around 1020 eV, a steep cut-off in flux is observed.
The exact origin of this cut-off and the spectral breaks remain unexplained. However, many
observations, such as different spectral breaks for different primary particle types [13] suggest
a connection between the shape of the spectrum and the sources and propagation mechanisms
of the cosmic rays, e.g., [14, 15]. Especially, the transition from galactic to extra galactic
cosmic rays and the end of the maximal reachable acceleration at the sources are favoured
explanations. For the cut-off at the highest energies, propagation effects are also considered as
explanation which will be discussed later.

1.1.2 Composition
Cosmic radiation is mainly composed of nuclei of the elements of the periodic table. The most
abundant species are protons and helium nuclei. The relative abundances of these elements
however change with energy. At low energies (1 − 100 × 109 eV), the cosmic radiation shows
relative abundances of elements similar to what is found in the solar system. Some elements
are however more abundant in cosmic rays. Most of these elements have slightly smaller
atomic numbers than the CNO elements, iron, or lead. They are believed to be produced in
spallation reactions of cosmic rays of the named groups with interstellar matter [24].
For higher energy measurements above 1015 eV, the determination of the atomic number of an
individual cosmic ray is not possible any more as these cosmic rays are only detected through
the extensive air showers they initiate in the atmosphere. Therefore, our knowledge about the
composition of the cosmic radiation at higher energies is based on composition-sensitive ob-
servables of the air showers such as the ratio between different particle types or the atmospheric
depth of the air shower maximum. These observables are interpreted using Monte Carlo sim-
ulation techniques and different hadronic interaction models which are extrapolations from
interaction cross sections measured at particle accelerators. Changes in the composition ap-
pear around the same energies as where the spectral index changes and are likely to be linked
to each other.
The depth of the air shower maximum Xmax is used as the standard quantity in the experimental
composition determination. In a simplified picture, the cross section σ for cosmic-ray inter-
actions in the atmosphere is proportional to the mass number of the cosmic ray (σ ∝ A2/3 ).
1.1. Cosmic Rays 5

10-6 PAMELA
ATIC2
10-9 CREAM
Tibet air shower array
KASCADE
10-12 KASCADE Grande
HiRes I
10-15 HiRes II
Pierre Auger Observatory
flux (m−2 sr−1 s−1 eV−1 )

10-18
10-21
flux ·E3 (m−2 sr−1 s−1 eV2 )

10-24 1025 knee 2nd knee ankle

10-27
10-30 1024

10-33
1023
10-36 1015 1016 1017 1018 1019 1020 1021
10-39 8 energy E (eV)
10 10 10 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021
9 10
energy E (eV)

Figure 1.1: The energy spectrum of cosmic rays measured by various space- and airborne as well
as ground based experiments. The flux multiplied by the energy to the power of 3 is shown in the
inset revealing the breaks in the spectrum above an energy of E = 1014 eV. The data is taken from
[16, 17, 18, 19, 20, 21, 22] and the figure is modified from [23].

Therefore, iron nuclei interact higher up in the atmosphere and the initiated shower develops
at a smaller atmospheric depth leading to a smaller Xmax . Results from different experiments
for which the average depth of the shower maximum hXmax i is plotted as a function of the
energy together with different model predictions for protons and iron nuclei from the knee to
the highest energies are displayed in figure 1.2. The comparison with the different models sug-
gests a light composition at the knee which then changes from light to heavy below the second
knee. From the second knee to the ankle an opposite change towards a lighter composition is
indicated. For the highest energies, the hXmax i accuracy is affected by the low flux, but a trend
towards a mixed composition can be observed.
The transition from light to heavy above the knee is often suggested to be caused by a mass-
or rather charge-dependent cut-off starting with protons at around 5 × 1015 eV and going up
to iron at around 1017 eV. Two main causes which might coincide are usually discussed. One
is the rigidity dependent leakage from the galaxy and the other is the rigidity dependent maxi-
mum acceleration power of the sources (e.g., [15, 25, 26]). Both interpretations are suggesting
a change from galactic to predominantly extra galactic cosmic rays around 1017 − 1018 eV.
6 Chapter 1. Cosmic Rays and Extensive Air Showers
K.-H. Kampert, M. Unger / Astroparticle Physics 35 (2012) 660–678

〈Xmax〉 [g/cm2]
TA, preliminary, 〈∆〉=17 g/cm2

〈Xmax〉 [g/cm2]
850
HiRes, 〈∆〉=26 g/cm2 840
800 HiRes/MIA
CASA-BLANCA 820
750 Yakutsk
Tunka 800
700 Auger
780
650
ton 760
600 pro
740
550 QGSJetII
Sibyll2.1 720
500
EPOSv1.99 700
450
iron
1015 1016 1017 1018 1019 1020 10
18 19
10
E [eV]

Fig. 8. Measurements of hXmaxi with non-imaging Cherenkov detectors (Tunka [118,126], Yakutsk [127,128], CASA-BLANCA [123]) an
Figure
[129],1.2:
HiResMeasurements of the
[130], Auger [131] andaverage
TA [132])depth of the
compared toshower maximum
air shower hX[133]
simulations from hadronic
max i using differentinteraction
experi- models [36,38,37]. Hi
ments for energies
for detector effectsabove the knee.
as indicated ThehDlines
by the indicate
i values the predictions
(see text). fromshows
The right panel different hadronic
a zoom interaction
to the ultra-high energy region.

models for protons and iron nuclei. The data from the HiRes and the TA experiment are corrected for
detector effects as discussed in [25] where the figure is taken from.
energy deposit [PeV/(g/cm2)]

50 Auger event
shower maxima deeper than 800, 924
photon with zenith angles of 0, 30 and 60 de
Particle interactions at energies up to 3 × 10 eV (lab
40 16
proton this acceptance
frame) are still accessible bias an additional rec
at the Large
Hadron Collider (LHC) and predictions for the hadronic iron interactions from ent that canmodels
the current further distort the meas
There are two ways to deal with su
have been30found to be in good agreement with the data [27]. Therefore, changes ested inin the under- the data to air sho
comparing
lying particle interaction processes are unlikely to cause the composition changes. primary For higherthen the biased dat
particles,
20 air shower predictions that include t
energies up to the end of the known spectrum, the interpretation of the observed composition
For this purpose the full measuremen
transitions is challenging due to the increasing uncertainties. However, an explanation for a
10 including the attenuation in the atm
changing composition at the highest energies analogue to the changes at and the knee might be to obtain a predic
reconstruction
an energy dependent maximum in acceleration power of the extragalactic sources shower[28]. Other hXmaxiobs. Anothe
maximum,
0
explanations200 are employing data sample to shower geometries fo
400 600propagation
800 effects
1000 to explain
1200 1400 the observed behaviour.
1600
is small (e.g. by discarding vertical
slant depth [g/cm2] remaining reconstruction effects to o
Fig. 9. Example of a longitudinal air shower development as measured with
ment of hXmaxi in the atmosphere.
1.1.3 Arrival Direction
fluorescence telescopes. Data points are taken from [145] (E = (30 ± 2) EeV) and Whereas the former approach max
compared to ten simulated [133] air showers for three different primary particle latter allows the direct comparison of
The arrival
types using direction
the hadronicofinteraction
ultra high-energy
model EPOS1.99cosmic
[36]. rays (UHECR) at Earthsimulations
can be measured
even for models that were
within about 1 [29]. This however does not mean that the position of the
◦ publication.
sources can Moreover,
be only measurem
the detector-specific distortions due t
determined within 1◦ . Magnetic fields change the direction of charged particles during their
ima for showers with similar energy contains the maximum infor- tion can be compared directly.
propagation.
mation about Thecomposition
measured arrival directions
that can of cosmic
be obtained from rays are nevertheless The
fluorescence fundamental
HiRes and toTA collaborations f
detectors.the
understand Given enough
origin of these statistics
particles.and an exactbyknowledge
If emitted of thepowerful
a few extremely hXmax iobs [130,132]
sources within and to compare
expected small
a relativity distributions
distanceforofdifferent
the orderprimaries,
of tens ofit Mpc,
shouldthebe skymap
possibleof theshower simulations.
measured arrival In the HiRes ana
to extract composition groups (see e.g. [150]) similar to what is to assure an Xmax-bias that is consta
directions
done forshould
surfaceshow a clustering
detectors. of cosmic however,
In the following, rays emitted by the
we will same source.
con- The recent
for different primaries and hadronic i
skymap
centrateof on
UHECR measured
the first by the Pierre
two moments of the Auger Collaboration
Xmax-distribution, is shown
hXmax i in figureTA1.3.
minary The uses only minim
analysis
and r (X
plot also contains
max ). the positions of active galactic nuclei (AGN) taken from the Swift-BAT biases. The Aug
dependent detection
For the determination of the average shower maximum, exper- age shower maxima that are withou
iments bin the recorded events in energy and calculate the mean of the quoted systematic uncertainties [
the measured shower maxima. For this averaging not all events are volume cuts. Yakutsk derives Xmax in
used, but only those that fulfill certain quality requirements that tween the slope of the Cherenkov-LD
vary from experiment to experiment, but all analyses accept only maximum (cf. Section 3.2). This relati
profiles for which the shower maximum had been observed within simulations and is universal with re
1.1. Cosmic Rays 7

Figure 6. Cross-correlation of events with the AGNs in the Swift-BAT catalog. The top-left panel shows the values of fmin and P as a function of the maximum
distance, D, to the AGNs considered. The top-right panel shows the results of the scan in ψ and E th for the value D = 80 Mpc corresponding to the minimum values in
Figure 1.3:plot.Arrival
the top-left The bottom direction ofdistribution
plot shows the sky cosmic(inrays Galacticwith energies
coordinates) above
of the events with E ⩾58 EeV
58 EeV asdots).
(black measured
Red circles ofwith the
1° radius Pierre
are drawn
around the BAT AGNs closer than 80 Mpc.
Auger Observatory (dots). In addition, the positions of AGNs with distances less than 80 Mpc are
taken
map from the Swift-BAT
of events catalogue toand
and objects corresponding theindicated
minimum with circles
reported X-rayofband1◦ luminosity,
radius. The  X , and
solid lineradio-galaxy
for the represents
found, i.e., E ⩾ 72 EeV and D = 90 Mpc. Circles of radius sample the reported radio luminosity, R , computed per
the edges
4 . 75 areof
◦
thearound
drawn field every
of viewradio of theand
galaxy Pierre Auger
the events are Observatory
logarithmic energy and the dashed
bin at 1.1 GHz.line
For corresponds
Swift, we scan from to the
indicated with black dots.  X 10 42 up to 1044 erg s−1, while for the radio galaxies we
Super-Galactic plane. Figure
While the cross-correlation taken
analysis does from [31].
not provide us with scan from  1039 up to 1041 erg s−1, considering three
R
a significant indication of excesses of pairs with any of the
logarithmic steps per decade, for a total of seven luminosity
catalogs considered, at any energy, distance, and angle, we note
that they all yield minima for similar maximum distances to the values in each case. These luminosity values cover most of the
catalogue
objects ([30]. Various
80–90 Mpc), detailed
although studiesthreshold
for different have beenrangeperformed
spanned by theas forluminosities
actual example in AGNs
of the [31]that
present in the catalogs (just 10 AGNs from the Swift sample
andare an
energies and angular scales. The fact that the distances are
excess
similarofforcosmic rays with
the three catalogs arrival
is actually directions
expected have  X back
given the pointing tos−1the
10 42 erg Centaurus
, while A region
only three AGNs from theand
existing correlations between catalogs, since AGNs are radio-galaxy sample have R 1039 erg s−1). Given the
bright close-by
preferentially AGNs
located has
in regions withbeen found.
a high density The statistical
of galaxies. studies
additional scan show
performed in ,however no coarser
we do a slightly significant
scan
On the other hand, the preference toward D 80 Mpc is in D, using 20 Mpc steps to cover from 10 up to 190 Mpc.
evidence
mostly dueforto the
anisotropy and
fact that, for this more
value, dataCentaurus
the whole need to beConsidering
acquired first for further
the Swift catalog,investigation of the
we show in the top-left
Supercluster gets included and in this region there is an excess panel of Figure 8 the resulting values of fmin as a function of
excesses observed
of high-energy events. so far. the maximum AGN distance and the minimum adopted
luminosity, min , in the respective bands (the white region in
6.2. Cross-correlation with Bright AGNs the top-left corner of the plot is due to the absence of nearby
We present here the results of a scan over the minimum objects above those luminosity thresholds). The values of fmin
1.1.4 Propagation
source luminosities, considering for the Swift AGNs the are obtained after scanning on Ψ and Eth , as in the previous

10
To interpret the composition and arrival direction measurements in the context of the sources,
the propagation of the particles needs to be taken into account as the composition of cosmic
rays detected at Earth might not be the same as emitted by the sources. Especially for ultra
high-energetic cosmic rays (E > 1017 eV), the universe is not transparent [32]. Interactions
with photons (γ) from the cosmic microwave background (CMB) or the UV/optical/IR back-
ground photons are predicted to significantly attenuate cosmic rays. Several different types
of interactions are expected with varying energy thresholds. For protons with energies below
5 × 1018 eV for instance, the dominant interaction process is the electron (e− ) positron (e+ )
pair production which can occur for every nucleus N as

N + γ → N + e+ + e− . (1.2)

For larger nuclei also the photo-disintegration contributes significantly to the attenuation as

N + γ → N ∗ + X, (1.3)

where N ∗ is the remainder of the initial nucleus and X the split up fragment. The cross section
for photo-disintegration depends on the type of nucleus and can be the dominant interaction
8 Chapter 1. Cosmic Rays and Extensive Air Showers

process for cosmic rays with energies above 1017 eV. For the most energetic cosmic rays
(e.g., Eproton > 5 × 1019 eV) interactions with CMB photons have a significant cross section
and can produce pions via the delta resonance [33, 34] as
p + γCM B → ∆+ → p + π 0
(1.4)
→ n + π+.
The pions decay further according to
π 0 → γγ
(1.5)
π + → µ + + νµ
and the neutron decays according to
n → p + e− + ν¯e . (1.6)
This process is called GZK-effect after Greisen, Zatsepin, and Kuzmin, who independently
proposed a cut-off in the cosmic-ray energy spectrum for protons around 1020 eV. For heav-
ier nuclei, the energy threshold for photo-pion production is significantly higher, however the
photo-disintegration also leads to a cut-off in the spectrum of these elements. Therefore no
conclusions about the composition can be drawn from observations of a cut-off only. The in-
fluence of the propagation effects are summarized in figure 1.4 for the case of nitrogen and
iron nuclei.
During the propagation, charged particles are not only experiencing interaction, but also de-
flection magnetic fields. Except for the highest energetic protons in the cosmic radiation, the
detected arrival direction can therefore be far from pointing back to the source. Extra- and
intergalactic magnetic fields of different strengths are deflecting the cosmic rays during their
propagation. The mechanism is the same as for confining the particles in their sources as
discussed in section 1.1.5, but the magnetic fields which the particles encounter during their
propagation are much smaller. Therefore, the particles are not trapped, but deflected propor-
tional to their rigidity
pc
R= , (1.7)
Ze
where p is the momentum and Z is the number of elementary charges e. The rigidity of a
particle depends on its charge which means iron nuclei of the same energy as protons will
be effected much stronger by deflections during propagation. Unfortunately, the charge of
individual high-energy cosmic rays cannot be measured. The degree of deflection also depends
on the strengths and directions of the magnetic fields the particles encounter which are not
know in detail either.

1.1.5 Sources
Cosmic rays must originate from objects which are able to accelerate individual particles to
enormous energies. The acceleration however does not need to take place in one go. It can be
1.1. Cosmic Rays 9

104 104
energy loss length [Mpc]

energy loss length [Mpc]

103 103

102 102
all processes all processes
redshift losses redshift losses
photopion production (CMB) photopion production (CMB)
10 photopion production (EBL) 10 photopion production (EBL)
pair production (CMB) pair production (CMB)
pair production (EBL) pair production (EBL)
photodisintegration (CMB) photodisintegration (CMB)
1 1
photodisintegration (EBL) photodisintegration (EBL)

18 19 20
10 10 10 21
10 22
10 1018 1019 1020 1021 1022
energy [eV] energy [eV]

Figure 1.4: Propagation lengths of 14 N (left) and 56 Fe (right) as a function of initial primary cosmic
ray energy. The energy loss length is defined as the travelled distance over which the particle energy
decreases by 1/e due to the shown energy loss processes. Figure taken from [35].

distributed in multiple steps as long as the particles encounter enough of these accelerations,
as it is also the case for man-made ring accelerators. A meanwhile established mechanism for
the acceleration process based on multiple acceleration steps is called Fermi acceleration [36].
The idea is that charged particles can gain energy by statistical acceleration due to multiple
scattering processes in strong turbulent magnetic fields. These fields occur in shock waves of
the interstellar plasma such as the expanding shells of supernova explosions or around active
galactic nuclei. The particles are accelerated when crossing the shock wave heads on. For each
encounter, the energy gain is equal to

∆E 4
≈ βshock , (1.8)
E 3
where βshock = vshock /c is the shock velocity. To encounter enough shock front crossings to
gain the observed energies of UHECR, the particles must be trapped long enough in the source
region. Charged particles can be confined by magnetic fields forcing the particles on a circular
trajectory (in a simplified picture). The size of the trajectory is characterized by the Larmor
radius
1 E B
rL = 1.08 15
pc (1.9)
Z 10 eV µG
with the number of elementary charges of the cosmic ray Z, the magnetic field of the source
B, and the cosmic ray energy E [37]. The requirement of containing particles that travel on
such radii puts constrains on the properties of possible sources and also on the composition
of the cosmic rays they emit. As the Larmor radius is proportional to 1/Z, light particles will
leak from the sources first, followed by heavier and heavier particles. This picture can be used
to explain the spectral break and the changing composition around the knee for galactic and
above the ankle for extra galactic cosmic rays [38]. Based on these arguments, a condition for
the maximum energy to which a source can accelerate particles is given by the Hillas criterion
    
B L 2 Emax
> , (1.10)
µG pc Z βshock 1015 eV
 10 Chapter 1. Cosmic Rays and Extensive Air Showers
1.1. Cosmic Rays 7

1012 G Neutron stars

White dwarfs

Magnetic field strength

106 G
LHC
TEVATRON AGNs
SppS GRBs

Sun spots Magnetic

10
stars

20
1G

eV
10
20

pr
ot
eV
10

on
12

ir
on
eV
SNRs Radio galaxy
Interplanetary
6 space
10 G Galactic disk Galactic
cluster
Galactic halo
IGM

1 km 106 km 1 pc 1 kpc 1 Mpc 1 Gpc

1 AU
Size

Figure 1.3:1.5:
Figure The Hillas
The HillasPlot.
Plot.The figure
In this shows
figure, which potential
the maximum astronomical
energy which particlessources are able totoaccel-
can be accelerated
erate(according
particles totoequation
the highest
1.10)detected
in variousenergies of 10 objects
astronomical
20 eV, assuming theyvelocity
is shown. The containβshocks with velocity
shock is assumed
=to1.beThe figure
equal to 1.illustrates
The regionsthatofsources have lines
the diagonal to fulfill a combined
indicate criterion
the parameter spaceofallowing
source extension,
for particle mag-
neticacceleration
field strength
to a and charge
certain energy.of the
For particle that is
protons with anaccelerated. It eV,
energy of 1020 alsoonly
shows
verythe
fewenergy
sourcesthat current
fulfil
accelerators canMan
the criteria. deliver.
made The figure isare
accelerators based
shownonasequation 1.2.
well for reference. Figure taken from [39], originally
published in [37].

For energies
where L is thearound
size of10the eV
shocktypical
regioncandidates
and Emaxare
15
supernova
is the maximum remnants
energy [34]. They rays
for cosmic are large
enough
from tothecontain particles
particular source.and
To also provide
accelerate the necessary
particles shock waves.theNo
to ultra high-energies, directneed
sources correlation
to
of the arrival direction has been measured, mostly due to the deflection
be either very large, have a very strong magnetic field, or the combination of both needs to be of the propagating
particles in thelarge.
sufficiently Galactic
Figuremagnetic
1.5 gives an field. Indirect
overview of theobservations of -rays
maximum reachable as by-products
energy in particle of
the acceleration
accelerations forhave,
knownhowever,
astronomicalstrengthened theirasstatus
objects as well someas candidates
man-made [35]. Other
accelerators possible
as refer-
sources
ence. at this energy are pulsar wind nebulae or X-ray binaries [36]. At higher energies
the As
sources are essentially
supernova explosions unknown.
are known to There are shock
produce only few frontssources that in
expanding theoretically
the interstellarfulfill a
sufficient
medium, combination of thecandidate
they are a typical size andfor magnetic field criterium,
particle acceleration [40].as shown inthefigure
Following Hillas1.3.
crite-In this
figure,
rion,the potential
they sources protons
could accelerate are classified according
up to around to their
1015 eV. Othermagnetic
objects suchfield
as strength
pulsars orandthe size,
andgalactic
therebydisktheir
arepower to accelerate
also reasonable particles.
candidates Only few
at energies sources
between the reach
knee andthe the
lines that For
ankle. indicate
the the
required
highestacceleration
energies abovepower.
the ankle, only a few sources are powerful enough. The termination
Searches
shocks for sources
associated areofhindered
with jets by the
active galactic fact(AGN)
nuclei that charged cosmic
for example rays are
are suited sitesdeflected
for the in
the acceleration
magnetic fields whilerays
of cosmic propagating from their
up to the highest sources
energies. to Earth.
Gamma-ray In the
bursts (GRB) same areway
also that
pro- their
acceleration
posed to be ingood
sources is a function
candidates for shockof charge, magnetic
acceleration field
initiated strength,
by the burst. and size of the
Furthermore, source,
radio
so isgalaxy
theirlobes,
deflection angle.originating
and shocks The larger fomthe distances
collisions the particles
of galaxy clusters propagate and criterion
fulfil the Hillas the stronger
the for
magnetic
UHECRfields, the further
acceleration. theysources
All these will beare
deflected fromdistributed
isotropically their original
on thetrajectory. The angle
largest distance
# atscales.
whichBy they willpropagation
taking be deflected depends
effects on the charge-momentum
as discussed in subsection 1.1.4 into ratio with the particle
account,
horizon however shrinks significantly and theq source · B · d candidates are not distributed isotropi-
#⇠ , (1.3)
p
where q is the charge, B the magnetic field, d the propagation distance and p the relativistic
impulse, which does not depend much on the mass of the particle. The deflection angle is
therefore significantly different between the two extreme cases of iron nuclei (Z = 26) and
1.2. Extensive Air Showers 11

cally within this volume. The deflections of the charged particles are also on these scales
not negligible. In order to further constrain the type of possible sources, a multi messenger
approach can be employed by combining results from gamma-ray observation as well as high-
energy neutrino detections with cosmic-ray measurements. High energetic gamma rays and
neutrinos are produced in interactions of charged cosmic rays with matter (e.g., around the
sources) or background light photons (see equations 1.4 and 1.5). Both particle types are not
deflected in magnetic fields and point back to their sources. Gamma rays with energies of
more than 1014 eV have been observed from supernova remnants (SNR) supporting their role
as particle accelerators at energies around the knee [41, 42]. For higher cosmic-ray energies,
no corresponding gamma rays have yet been detected in the PeV range. A couple of neutrinos
with energies exceeding 1015 eV were recently detected by the IceCube collaboration hinting
at astrophysical high-energy neutrino sources [43, 44].

1.2 Extensive Air Showers

Primary cosmic rays cannot reach the surface of the Earth because of their interactions with
atmospheric molecules. In the collisions, secondary particles are produced which undergo fur-
ther interactions thereby creating a particle cascade. Depending on the type of the primary
particle, different kinds of shower developments occur. In the following, the air shower devel-
opment will be discussed for the case of a primary proton. Based on this, the description of
an electromagnetic shower is given and the results are transferred to also extract information
about heavier primaries.

1.2.1 Longitudinal Development

Nuclei in the cosmic radiation are interacting hadronically with atmospheric nuclei. For cos-
mic ray energies up to about 1017 eV interactions have been studied at accelerators and ex-
trapolations are used for higher energies where no experimental data are available. In general,
the development of the initiated particle cascade can be characterized by three main compo-
nents referred to as the hadronic, the muonic, and the electromagnetic component. In the first
interaction of the primary particle, mainly pions (π) and kaons (K) are produced and also
the remainders or fragments (p, X) of the collision partners (p, N ) are available for further
interactions as in
p + N → p + X + π ± + π 0 + K ± + K 0 + ... . (1.11)

After the first interaction, the primary particle (here p) is called leading particle and still car-
ries about 50% of its initial energy on average. In the next step, the produced particles can
either interact again or decay. The high energetic charged pions and kaons live long enough
to undergo further interactions continuing the hadronic cascade. After several interactions, the
energy of the kaons and charged pions has decreased so much that it becomes more likely for
12 Chapter 1. Cosmic Rays and Extensive Air Showers

primary

π0
γ
K+ π0
p
p π+ γ
−
π
e+
π0 γ e−
γ
K0
n p n µ+ γ
π+ −
ν̄µ µ
n e−
n p
π− e+
γ
γ γ

e−
µ+ νµ νµ γ
ν̄µ νe

e+ e+
e−
ν̄µ µ− e − γ

hadronic muonic component electromagnetic

component component

Figure 1.6: Three components of an extensive air shower initiated by a high-energy hadronic primary
particle. Figure taken from [23].

them to decay than to interact again. The kaons decay into all three kinds of pions and the
(−)
charged pions decay according to π ± → µ± + νµ into muons and (anti)-neutrinos building up
the muonic component. The neutral pions however are decaying into two photons (π 0 → γγ)
after a very short lifetime (τ0 ≈ 10−16 s). These photons are the starting point of the elec-
tromagnetic cascade as they undergo electron-positron pair production and the generated e±
subsequently emit photons through bremsstrahlung which again generate high energy e± . A
(−) (−)
small number of the muons also decays (µ± → e± + νe + νµ ) and the electrons enter the
electromagnetic cascade. The first steps in the air shower development and the building up
of the three shower components are depicted in figure 1.6. During the development of the
cascade, the particles are confined in a relatively thin but laterally extended disk, referred to as
the shower front.

1.2.2 Heitler Model

An analytic description of the physics involved in the development of electromagnetic cascades
in the atmosphere was developed by Heitler [45] and later extended to the hadronic cascade
1.2. Extensive Air Showers 13

by Matthews [46]. Following the latter, the basic principle of the model is a repeated doubling
of the particle number in each interaction by either emission of bremsstrahlung or electron-
positron pair production. The atmosphere is thereby described in layers spaced by the splitting
distance d which is related to the radiation length λem in the ambient medium by d = λem ln 2.
After each step, every outgoing particle is carrying half the initial energy and after n steps the
cascade has developed over a distance

x = nλem ln 2 (1.12)

containing N = 2n particles, each with a fraction of the primary energy E0 /N . Once this
fraction is smaller than the critical energy, splitting stops and energy losses due to ionization
start to dominate. The critical energy Ec = 85 MeV in air sets a limit on the maximum number
of particles
E0
Nmax = = 2nc (1.13)
Ec
with nc as the number of interaction steps until the particles reach the critical energy. Thereby,
the depth of the maximum shower development Xmax considering equation 1.12 can be for-
mulated as
E0
Xmax = nc λem ln 2 = λem ln . (1.14)
Ec
With this simple model it can be shown that Xmax depends on the primary energy. Its increase
is called the elongation rate and is given by

dXmax
Λ≡ . (1.15)
d log10 E0

In order to be compatible with results from accelerator experiments, the maximum number of
particles needs to be multiplied by a correction factor g ≈ 1/13 [47]. This is caused by a num-
ber of simplifications, e.g., ranging out of the charged particles is not taken into account and
neither is radiation of multiple photons during a bremsstrahlung process. Nevertheless, even
in its simple form, the model allows for the general description of electromagnetic cascades.
Also, its predictions for the relation between number of particles and energy as well as the
elongation rate are qualitatively in agreement with observations and full Monte Carlo models
of the shower development [46].
In figure 1.7 a sketch of the described model for electromagnetic showers is displayed on the
left hand side. On the right hand side, the application of the modified model to hadronic cas-
cades in extensive air showers is sketched. The basic idea is that hadrons also interact after
travelling one hadronic interaction length d = λH ln2. After this distance, the hadron produces
a number of pions of which 1/3 are neutral and 2/3 are charged. While neutral pions decay
immediately, charged pions interact again after travelling a distance d producing new pions.
Similar to the critical energy for the electromagnetic cascade, charged pions below a certain
"critical" energy decay to muons before they can interact. After n steps the energy in the
14 Chapter 1. Cosmic Rays and Extensive Air Showers

γ p
n=1 n=1
e− e+ π± π0

n=2 n=2

n=3 n=3

Figure 1.7: Sketches of the Heitler model for the development of extensive air-showers. Left: Elec-
tromagnetic cascade initiated by a photon. Right: Hadronic cascade initiated by a proton. The layers
stand for one interaction distance in both sketches. After the second layer not all resulting particles are
plotted. Figure adapted from [46].

hadronic cascade is equal to  n

2
Ehad = E0 (1.16)
3
and the energy per pion is given by

Ehad E0
Eπ = = 3 . (1.17)
(Nπ± )n ( 2 Nπ± )n

When Eπ drops below the "critical" energy for pions (Ec = 20 GeV), the charged pions decay
to muons and their number can be estimated from
 
E0 lnNπ±
lnNµ = ln . (1.18)
Ec ln( 32 Nπ± )

In contrast to the electromagnetic shower, the number of muons does not scale linearly with the
energy. This fact provides information about the type of the primary particle under the assump-
tion of the superposition principle introduced in [48]. The superposition principle describes
an air shower initiated by a nucleus with mass A as a superposition of A showers initiated by
protons with an energy of EA,0 /A each. This assumption is reasonable as long as the energy
regime is significantly above the binding energy of the primary nucleus. Another application
is the calculation of Xmax
A
for a nucleus based on Xmax
p
for a proton initiated shower with the
same energy which is
A
Xmax p
= Xmax − λem lnA . (1.19)

The interaction cross section however scales with A2/3 as mentioned in section 1.1.2. Both
statements in equations 1.18 and 1.19 provide keys to determine the mass A of the primary
particle. One observable is Nµ which is a factor A0.15 larger for nuclei with mass A than for
protons. The second observable is Xmax which is a factor of λem ln A smaller (i.e., higher up in
1.2. Extensive Air Showers 15

the atmosphere) for nuclei compared to protons. Both statements are in qualitative agreement
with extensive Monte Carlo simulations for air showers but they are only applicable on a
statistical sample and not on a shower-by-shower basis.

1.2.3 Monte Carlo Simulations

The Heitler model is very successful in describing the phenomenology of particle cascades
and extensive air showers. However, it does not allow for precise calculations and predictions
of air shower observables such as the shower development or the particle distribution on the
ground as they are needed for e.g., cross checks with measured air showers or to define ini-
tial parameters for air-shower reconstruction. These challenges are addressed by Monte Carlo
simulations which are able to track every particle that is created during the air shower de-
velopment. The interactions and decays of all particles are stochastically generated based on
measured cross sections and life times. For higher energies, the cross sections are extrapolated
from the measured values and are therefore a source of increasing uncertainties. Frequently
used interaction models are EPOS-LHC [49], FLUKA [50], QGSJET-II [51] and SIBYLL
[52]. Current air shower simulation codes like CORSIKA [53] and AIRES [54] have been
proven to be valuable tools resampling the measured properties of extensive air showers. To
generate simulated showers, the environmental conditions of the detector site such as height
above sea level, atmospheric profile, and geomagnetic field can be loaded into or defined in
the simulation codes.
An approach for air shower simulations using a combination of analytic descriptions and full
Monte Carlo methods following every particle is realized in CONEX [55]. Here, only the
most energetic particles in the shower are tracked as it is done in CORSIKA. For lower en-
ergetic particles, the further development of sub-cascades is calculated analytically following
the cascades equations (e.g., [56, 57, 58]). This procedure is very advantageous in terms of
computation time and CONEX is used for a wide range of applications to calculate the ba-
sic shower parameters or in cases where for example a detailed distribution of the particle
distribution on the ground is not needed.

1.2.4 Radio Emission

A couple of years after the development of the Heitler model, Askaryan proposed the emission
of radiation by electromagnetic cascades developing in a medium [59]. The origin of the radia-
tion with frequencies in the MHz-regime is explained with an excess of negative charges in the
air shower front and called charge excess. As the shower develops in the medium, electrons
are dragged from molecules in the medium along with the shower particles leaving positively
charged ions behind. This results in a charge separation along the shower axis. During the
shower development, the number of particles first in- and then decreases and both effects to-
gether can be seen as a time varying net-current or a changing dipole moment. Additionally,
16 Chapter 1. Cosmic Rays and Extensive Air Showers

charge excess geomagnetic charge excess

x
~
B

geomagnetic
~
B
x

Figure 1.8: Illustration of the two emission mechanisms of coherent MHz pulses from extensive air
showers. Left: The charge excess emission is produced as positive and negative charges are separated
along the shower axis when electrons are dragged with the particles in the shower front. Center: The
geomagnetic emission originates from the separation of mainly positrons and electrons based on their
charge in the geomagnetic field. For both mechanisms, the changes in the particle number during the
shower development are of main importance. Right: The polarisation of the radiation from the two
emission mechanisms as seen in the shower plane around the shower axis (marked by the star) when
the direction of the shower development goes into the paper plane.

positrons in the shower front can annihilate with electrons in the medium amplifying the charge
separation. The radiation emitted by the dipole is coherent as long as the wavelength is larger
than the thickness of the shower front. The polarization of the radiation is determined by the
dipole moment and parallel to the shower axis. An observer at ground however sees it as being
radially polarized with respect to the shower axis. The radiation is beamed in the direction of
the shower development due to the boosted frame of the particles.
Another mechanism of generating radiation in extensive air showers was proposed by Kahn
and Lerche [60]. The emission process in this case is based on the separation of electrons
and positrons due to the Lorentz force in the geomagnetic field and therefore often referred to
as the geomagnetic emission. As for the charge excess mechanism, the changing number of
particles during the shower development leads to a varying net-current. The changing dipole
moment is oriented in the direction of ~v × B, ~ i.e. perpendicular to the shower axis ~v and the
direction of the geomagnetic field B.~ The radiation measured on the ground is polarized in
~
~v × B-direction and coherently emitted as it also originates in the shower front. A sketch of
both emission mechanisms is shown in figure 1.8.
In general, the amplitude of the electric field scales with the number of electrons and positrons
Ne± as the emission is coherent. The radiated energy then scales with Ne2± which is directly
1.2. Extensive Air Showers 17

related to the energy of the primary particle by equation 1.13. The strength of the geomag-
netic emission in addition scales with the sine of the angle between the shower axis and the
geomagnetic field (geomagnetic angle) and the relation of the amplitude and radiated energy
needs to be corrected accordingly. The shape of the emission pattern is further influenced by
relativistic time compression effects due to the super-luminous particle speeds in the shower
front. These compression effects extend the frequency content of the radio pulses from the
MHz to the GHz range. The attenuation of electromagnetic radiation in this frequency regime
in the atmosphere is negligible and the radio pulses from extensive air showers can be mea-
sured as broadband pulses of nanosecond scale duration. The exact shape of the pulse thereby
depends on the shower development and geometry as described in [61].
The electric field measured at ground level is a superposition of the two electric fields emitted
by the different emission mechanisms and the direction of the electric field vector depends
on the strength of the individual components. Starting with the simple picture on the right
side of figure 1.8, the vectorial addition of the two components yields an asymmetric pattern.
~
Going along the ~v × B-direction against the electric-field vector of the geomagnetic emission
reveals an opposed polarization in the two components adding up destructively on this side
~
of the shower axis. Going further in the ~v × B-direction the charge excess emission experi-
ences a sign-flip at the position of the shower axis and the two components start adding up
constructively with an overall maximum on this side. For other positions, in the plane spanned
by ~v × B~ and ~v × (~v × B),
~ the interplay of the components works accordingly and a bean like
shape of the measured amplitude pattern is expected. Note, that in this coordinate system, the
overall bean shape of the pattern is always oriented in the same way.
More than 40 years ago, the first observations of radio pulses originating from extensive air
showers (e.g., [62, 63, 64]) were reported, accompanied by the proposals of various theories to
explain the phenomenons. But only recently, measurements of both, radially and linearly po-
larized emission in the radiation discussed above, support the theory of the two different emis-
sion mechanism as origins. The CODALEMA experiment [65] first reported results which
were not in agreement with a sole geomagnetic emission mechanism [66]. The ratio of the
amplitudes of the radially polarized electric field |EC | originating from the charge excess
mechanism and the linearly polarized electric field |EG | originating from the geomagnetic
mechanism, corrected for the influence of the geomagnetic angle α, is called charge excess
fraction
|EC |
a ≡ sin(α) . (1.20)
|EG |
A quantification of the average charge excess in the measured radiation was presented by
AERA [67] and amounts to 14 ± 2% [68] for the specific geographic site and the underlying
data sample. The individual values for the measured showers are shown in figure 1.9 (left).
With the high antenna density of LOFAR [69], the average contribution of radially polarized
emission has been measured for different zenith angle bins and distances of up to 250 m in
more detail, figure 1.9 (right). The results show a dependency of both parameters raising from
 r0 ✓ = [0 , 20 ) ✓ = [20 , 40 ) ✓ = [40 , 60 )
0 50 m (8.1 ± 2.1)% (6.6 ± 0.8)% (3.3 ± 1.0)%
50 100 m (13.6 ± 0.6)% (10.9 ± 0.3)% (5.4 ± 0.5)%
100 150 m (16.4 ± 0.9)% (12.7 ± 0.3)% (9.0 ± 0.6)%
150 200 m (18.2 ± 0.8)% (14.9 ± 0.3)% (9.9 ± 0.6)%
18 Chapter
200 250 m
1. Cosmic Rays and Extensive Air Showers
(20.3 ± 1.3)% (15.9 ± 0.5)% (9.6 ± 0.8)%

60
measurement nr.

40

20

0
-1 -0.5 0 0.5 1
a
FIG. 9. Distribution of most probable values of a (see Eq. (10)) and their
Figure 14.uncertainties for the
Charge-excess fraction as a function of distance from the shower axis for three di↵erent
Figure 1.9: Charge excess fraction a as measured with
zenith
AERA24 data set (see Appendix B for details). The 68% confidence beltangle
aroundbins. Uncertainties
the mean value AERA (left)
are calculated for individual
as discussed air showers and
in section 6.2.
of a is shown as with
the solidLOFAR
blue line; the (right)
value a = 0 for several
is indicated bins
with the inredzenith
dotted angle θ and distance to the shower axis. Figures taken
line; see text
for further details.
from [68, 70]. 8 Discussion and conclusions

Polarized radio emission from a sample of 163 air showers measured with the LOFAR radio
telescope has been analyzed.
below 5% for inclined showers In close
total 18to
airthe shower
showers whereaxis to about
excluded 20%
from this for vertical
analysis showers
due to coinciding at
thunderstorm
activity. The strong atmospheric electric fields present during thunderstorms are expected
large distances from the shower axis [70].
to significantly a↵ectAtherapid decrease
charged particle of the charge
distributions andexcess fraction
therefore close of the
the polarization
to the shower axis is observed
emissionby[35,
LOFAR
36]. Thisase↵ect
expected
will be due to the insign
investigated flip publication.
a future at the shower axis
The measured emission is strongly polarized, with a median degree of polarization of
position. nearly 99%. Because the geomagnetic and charge-excess emission are linearly polarized in
di↵erent directions, their relative contributions can be determined by polarization analysis.
35
In all measured air showers the geomagnetic emission mechanism clearly dominates the
In parallel to the experimental achievements,
polarization the theoretical
pattern. However, description
a sub-dominant of the
charge-excess radio emission
component can also be seen,

of extensive air showers has developed rapidly and simulation codes for the radio signals are in
qualitative agreement with observations [67, 71]. These simulation codes can be categorized
– 18 –
into macroscopic and microsopic descriptions.
Macroscopic models describe the phenomena in terms of dipole moments initiated by time
varying charge separation on large scales as explained above. These dipole moments can be
described in terms of classical electro-dynamics and allow for an analytic calculation of the
resulting electric fields. The basic shower parameters needed as input for the calculations can
be obtained from CONEX simulations of the corresponding air shower. Examples for such
macroscopic models are MGMR [72] and EVA [73].
Microscopic models calculate the emission of every single particle in the cascade and the in-
terference between their contributions. The complex interplay between different mechanisms
of the radio emission and considerations of second order effects are thereby automatically in-
cluded in the calculations. The radio simulations are mainly used as add-ons to full Monte
Carlo codes determining the radio emission while the particles are tracked in the simulations.
Example codes using the microscopic approach are ZHAires [74] and CoREAS [75].
For simulations produced in the context of this thesis, the CoREAS simulation code was em-
1.2. Extensive Air Showers 19

ployed which will be described in more detail here. The underlying principle to calculate the
radio emission in CoREAS is the end point formalism [76]. In this approach, the trajectories of
individual particles are described as a sequence of straight track segments with instantaneous
acceleration processes (positive and negative respectively) occurring at the beginning and the
end of each segment (endpoints). The radiation emitted during an instantaneous deceleration
at the end of one segment mainly cancels out with the one emitted during the instantaneous
acceleration at the beginning of the next track segment starting at the same endpoint. On a
track, the momentum of the particles is constant and only at the endpoints of the track changes
of momentum occur. Due to these changes of the particle momentum at the endpoints, some
fraction of the radiation does not cancel out and contributes to the overall emission of the par-
ticle trajectory. The exact calculation for the emissions is derived directly from the Maxwell
equations via the Liénard-Weichert potentials and therefore general and independent of any
model of macroscopic particle motion. For the calculations of the electric field contributions
of individual particles, the near-field term in the electric field equation is neglected as it gives
only small corrections to the observed electric fields. According to [76], the radiation emitted
from a charged particle under instantaneous acceleration to the velocity β~ = β~ ∗ at time t00 can
then be calculated in terms of the electric field at a position ~x for a frequency ν, as
0
ikR(t0 ) 0

~ ± (~x, ν) = ± q e e2πiνt0
E r̂ × [r̂ × β~ ∗ ]. (1.21)
c R(t00 ) 1 − nβ~ ∗ · r̂
Here, q is the charge of the particle in c.g.s. units, r̂ is the unit vector to the observer posi-
tion, R is the distance from the endpoint to the observer, n is the medium refracting index,
and k = 2π/λ = 2πνn/c. The "±" accounts for the two directions of acceleration in the
endpoint picture and is positive for acceleration from rest and negative for acceleration to rest.
Analogues, the electric field averaged over a time bin ∆t can be derived as
!
q r̂ × [r̂ × ~ ∗]
β
E~ ± (~x, t) = ± . (1.22)
∆tc (1 − nβ~ ∗ · r̂)R

The width of ∆t determines the time resolution of the resulting electric field trace at the ob-
server position.
The decision which simulation code to use should be made according to the physics question
to be answered. The CoREAS simulation code for example employs an approach of simulat-
ing the electric fields which is independent of models for the emission mechanisms and relies
only on full Monte Carlo air-shower simulations. Every particle in the simulation is consid-
ered and the emission is calculated requiring considerably more computation time than e.g.,
macroscopic approaches. Furthermore the emission mechanisms cannot easily be simulated
separately. One approach to nevertheless investigate the different origins of the radiation is to
produce two simulations with the same start parameters except for the fact that the magnetic
field is switched off for one of the cases. The radiation seen in the second simulation is then
exclusively originated from the charge excess mechanism. The approach however requires an
20 Chapter 1. Cosmic Rays and Extensive Air Showers

additional air-shower simulation and the geomagnetic emission can only be singled out as the
difference between the simulation with and without geomagnetic field [77]. The particle show-
ers in these two simulations will also be slightly different which complicates the interpretation
of the resulting radiation patterns. Aside from these considerations, an advantage of using the
combination of CoREAS and CORSIKA is the detailed particle and radio hybrid output and
the large number of additional options in both packages. Among others, CoREAS allows for
the definition of atmospheric electric fields and has an option to just consider particles within
a certain geometrical region for the calculation of the resulting electric fields.

1.2.5 Detector Systems

Since the discovery of extensive air showers, many different detection strategies have been
developed. Pierre Auger for example used counters with various spacings operated in a coin-
cidence mode. This technique of detectors placed in arrays on the ground is still the backbone
of the majority of modern air shower experiments. However, by using this technique only
a snapshot of the air shower at the time of arrival at ground is available. Therefore, various
additional techniques addressing the measurement of the air shower development are also em-
ployed. These methods are based on the detection of radiation that is directly or indirectly
produced by the charged secondary particles mainly in the electromagnetic component. The
amount of radiation produced at a certain stage in the shower development is thereby related
to the number of particles in the electromagnetic cascade at that stage. The three types of ra-
diation observed from extensive air showers are fluorescence light, Cherenkov radiation, and
(MHz-GHz) radio pulses.

Particle Detector Arrays

Ground based particle detector arrays sample the density of particles in the shower front at
different distances from the shower axis. The lateral extent of the shower front depends on
the energy of the primary particle as well as the stage of shower development and the types
of particles considered. The discretely sampled lateral particle distribution can in general
be continuously described by a lateral distribution function (LDF) developed by Nishimura,
Kamata and Greisen (NKG-function) [78, 79] as
Γ(4.5 − s)
 
Nch  r (s−2)  r (s−4.5)
ρ(Nch , r) = 1+ . (1.23)
2πR2 R R Γ(s)Γ(4.5 − 2s)
Here, Nch is the total number of charged particles, r is the radial distance to the axis, R is
the parameter of the characteristic radius, s is the shower age which accounts for the stage of
shower development and Γ denotes the Gamma-function. For pure electromagnetic showers,
the characteristic radius is the Molière radius. The exact form of this parametrization depends
on the individual detector setup and different versions are used for different experiments [80].
An example of the lateral particle distribution as measured at the Pierre Auger Observatory
1.2. Extensive Air Showers 21

together with a fit of a modified NKG function is shown in figure 1.10 (left). The energy of the
primary particle is in principle connected to the total number of charged particles Nch . The ex-
act relation between the two numbers however also depends on the type of primary particle and
the shower development. This introduces further uncertainties in the reconstruction of shower
parameters. Many experiments operate a certain type of detector and therefore the typical de-
tector responses to different particle types have a large influence on the LDF, see section 4.3.2.
As the number of muons in the air shower is an indicator for the primary particle type (sec-
tion 1.2.2), the separation of the electromagnetic and the muonic component is a powerful tool
for cosmic-ray composition studies. Modern particle detector arrays are employing differen-
tial techniques to realize a separation of the components. One approach is a layered detector
structure, where an absorber is placed between two detector layers absorbing a large fraction
of the electromagnetic particles in the shower front as realized in the KASCADE experiment
[81]. Another approach is used for the upgrade of the Pierre Auger Observatory, AugerPrime.
Here, two detector systems, water Cherenkov detectors and plastic scintillators will be stacked
on top of each other to disentangle the shower components through the different detector re-
sponses. A somewhat different approach is the usage of underground muon counters which
are shielded against all electromagnetic particles by meters of soil in combination with radio
detection stations.
Following from the lateral particle distribution, the energy threshold for an air shower array is
given by the spacing of the detectors, the sensitive detector area, and the type of detectors. The
steep spectrum of cosmic rays additionally puts a lower limit on the size of the instrumented
area to gain sufficient statistics in the desired energy range. Typical combinations of detector
spacings and instrumented areas are e.g., 137 m and 700 × 700 m2 in the KASCADE-Grande
experiment [82] targeting cosmic rays of energies between 1016 eV and 1018 eV or 1500 m and
> 3000 km2 at the Pierre Auger Observatory [29] targeting UHECR above 1018 eV.
The arrival direction of cosmic rays is inferred from the axis of the air shower which in terms
can be reconstructed from the arrival times of the particles in the shower front at different
detector positions on the ground. Efforts are ongoing in studying the correlations between the
shape of the shower front and the shower development, mainly targeting Xmax (e.g., [83]).
Studies focussing on the reconstruction of the energy or the shower development based on the
signal measured with a particle detector array depend on a calibration with either a different
detector technique or Monte Carlo simulations. Particle detector arrays can reach a duty cycle
of 100% and they are an established and cost effective tool in cosmic-ray research. Further
information about particle detector arrays is given in the description of the Surface Detector
array of the Pierre Auger Observatory in chapter 2.1 and in chapter 4 where an analysis based
on the scintillation detectors of AERA is presented.
22 Chapter 1. Cosmic Rays and Extensive Air Showers

Fluorescence Detectors

The detection of fluorescence from extensive air showers allows for a detailed measurement of
the longitudinal shower development. For a cosmic ray, the atmosphere acts like a calorimeter
and a large part of the primary energy is deposited by creating cascades of secondary particles
which are subsequently absorbed in the atmosphere. A part of the energy deposit is caused
by the excitation of atmospheric nitrogen molecules which are de-exciting via the emission
of UV photons. This fluorescence light is emitted isotropically and can be detected with spe-
cial telescopes. The number of emitted photons is proportional to the energy deposited in the
atmosphere. The ratio of photons emitted per unit of energy deposited in the atmosphere by
an electron is called fluorescence light yield. For the measurement of this yield, a dedicated
experiment named AIRFLY [84] was performed as the conversion factor is highly important
for the energy calibration of the detector systems. The fluorescence yield depends on the pres-
sure and temperature of the atmosphere, and the measured photon numbers need to be rescaled
accordingly. To perform the correction, the atmospheric conditions and the emission region
of the photons must be known. Atmospheric state variables are monitored continuously and
fluorescence detectors can be equipped with multi-pixel cameras which allow actual spacial
tracking of the shower development and geometric shower reconstruction. Furthermore, they
are often combined with particle detector arrays providing additional information about the
geometry. Therefore, the fluorescence light measurements can be corrected for propagation
effects as scattering and absorption (this requires further calibration procedures as discussed
in section 2.2). Including all these corrections, the longitudinal profile in terms of deposited
energy dE(X)/dX at an atmospheric depth X can be reconstructed from the detected pho-
tons. The shape of the profile can be parametrized with the Gaisser-Hillas function [85, 86]
as (Xωmax −X0 )/λ
X − X0

fGH (X) = ωmax e(Xωmax −X)/λ , (1.24)
Xωmax − X0
where Xωmax is the atmospheric depth where the maximum energy deposit ωmax is reached and
X0 and λ are shape parameters. X0 hereby represents the depth of the first interaction and λ
represents the mean free path. The integral
Z ∞
Ecal = fGH (X)dX (1.25)
0

over the profile gives the calorimetric energy deposit in the atmosphere. Besides a fraction
of about 10% that goes into muons and neutrinos, the calorimetric energy Ecal corresponds
to the energy of the primary particle if the shower is fully developed in the atmosphere. As
the deposited energy on its part is proportional to the number of ionizing particles, Xωmax can
also be seen as the atmospheric depth of maximum shower development (Xωmax = Xmax ). A
measured longitudinal shower-profile is shown in figure 1.10 (right).
Examples of cosmic-ray experiments employing fluorescence detector systems are HiRES
[87], the Telescope Array [88], and the Pierre Auger Observatory [89].
1.2. Extensive Air Showers 23

dE/dX [PeV/(g/cm2)]
Signal [VEM]
χ2/Ndf= 60.5/47
χ2/Ndf: 9.6/ 18
4
candidates
102
not-triggered
3

2
10

1 0

500 1000 1500 2000 2500 400 600 800 1000 1200 1400
r [m] slant depth [g/cm2]

Figure 1.10: Hybrid detection of an extensive air shower. Left: Lateral particle density distribution as
measured with the Surface Detector array of the Pierre Auger Observatory. The timing of the detected
signals is color coded from yellow to red. Non-triggered stations are plotted as triangles.The distribution
is fitted with a modified version of equation 1.23. Right: Longitudinal shower development in terms of
energy deposit as function of the traversed atmospheric depth measured with the Fluorescence Detector
of the Auger Observatory. The line indicates a fit with the Gaisser-Hillas function (equation 1.24). The
reconstructed parameters of the displayed shower are E = 2.7 × 1018 eV, Xmax = 676 g/cm2 , and
zenith angle θ = 58◦ .

Air Cherenkov Detectors

Next to the fluorescence light, Cherenkov radiation is produced by the ultra-relativistic air-
shower particles in the atmosphere with its non-unity index of refraction. The lateral pattern
of the Cherenkov light contains information about the shower development and especially
Xmax which can be used in dedicated experimental setups. The TUNKA experiment [90] em-
ploys such a non-imaging Cherenkov detector array for extensive air showers.

As fluorescence and Cherenkov telescopes detect photons in the ultraviolet (UV) wavelength
regime, they can only be operated during clear moonless nights which has a large impact on
the duty cycle being limited to less than 15%.

Radio Detector Arrays

Besides fluorescence and Cherenkov light, the third type of radiation generated by an extensive
air shower are MHz (and GHz) radio pulses as discussed in section 1.2.4. The observation of
those pulses is not limited to any special atmospheric conditions and allows for almost 100%
duty cycle only limited in case of strong atmospheric electric fields (e.g., during thunderstorm
conditions). The signal distribution at ground level can be measured by arrays of radio de-
tectors similar to particle detector arrays. The individual radio detection stations are thereby
equipped with antennas (most often in two polarization directions) and fast digitizing elec-
tronics to sample the ultra-short pulses. Starting the renaissance of radio detection, LOPES
[91] and CODALEMA proved that the developments in electronics were advanced enough
24 Chapter 1. Cosmic Rays and Extensive Air Showers

to allow observations of extensive air showers on large scales. Today’s radio detectors like
LOFAR, TUNKA-REX [92] and AERA are already approaching a competitive performance
in the reconstruction of shower parameters such as arrival direction, energy, and Xmax , see
e.g., [71, 93].
The reconstruction of the shower geometry is based on the arrival times of the radio pulses.
By analysing this timing information, it has been found that the wave front of the radio signal
has a hyperbolic shape [94] which is proposed to be related to the shower development [95].
The lateral distribution of the signals, which might be taken as the maximum amplitude of the
electric field or the energy deposited per area (which we call energy density from here on) in
the pulse, is more complex than in the case of the particle detectors due to the two emission
mechanisms and cannot be properly described by a radially symmetric function. This became
at latest obvious from the measurement of the densely populated antenna array of LOFAR. A
new approach is using a two-dimensional function empirically developed on simulations ac-
counting for the seen asymmetries [96]. This function describes the signal P at every position
~ vs. ~v × (~v × B)
(x0 , y 0 ) in the ~v × B ~ plane as

−[(x0 − X+ )2 + (y 0 − Y+ )2 ]
 
0 0
P (x , y ) = A+ · exp 2
σ+
−[(x0 − X− )2 + (y 0 − Y− )2 ]
 
−A− · exp 2
, (1.26)
σ−

where A± are scaling parameters (A+ > A− ), X± and Y± are location parameters, and σ± are
shape parameters. Two examples of the fit of modified functions to LOFAR and AERA data
are shown in figure 1.11. A detailed discussion about the function in its original and modified
version and the correlation between function parameters and air shower properties is given in
chapter 5 and the following.

Multi Hybrid Detector Setups

Even though most detection methods show some sensitivity to all relevant air shower prop-
erties, many cosmic-ray observatories are built in a way that allows measurements of the air
showers with different techniques simultaneously. Thereby, the different advantages can be
combined to an accurate air shower reconstruction. It is furthermore possible to cross calibrate
various detector systems which can be used to reduce the systematic uncertainties of individ-
ual experimental setups. The major present radio detectors are used in combination with a
ground based (particle) detector array. In the case of TUNKA-REX this array consists of air
Cherenkov detectors. In addition to the acquisition of complementary information about the
same air shower, trigger information can be provided to the radio detectors during run time.
This guarantees efficient triggering and allows for a significant suppression of the detection of
pulses from background sources.
The introduction of radio detectors to cosmic-ray experiments offers the measurement of the
7.2. Discussion of fit behavior 111
1.2. Extensive Air Showers 25

position in vx(vxB)-direction [m]

energy density [eV/m2]

600
400

400 350

300
200
250
0
200

-200 150

100
-400
50
-600 0
-600 -400 -200 0 200 400 600
position in vxB-direction [m]

Figure 1.11: Measured intensity patterns of radio emission in the shower plane spanned by ~v × B ~ and
~ Left: Air shower footprint measured with LOFAR. Every coloured circle stands for the
~v × (~v × B).
integrated pulse power in one radio antenna station. The background color shows the fit of a modified
version of equation 1.26 to the data. Right: Air shower footprint measured with AERA. In addition to
the signal stations with signals above the signal-to-noise threshold, sub-threshold antenna stations are
indicated by a "+". The color scale in this case represents the measured energy density. The air shower
displayed on the right side is the same as shown in figure 1.10.

depth of the shower maximum [71] with almost 100% duty cycle. One example is the Pierre
Auger Observatory, where AERA is not only taking data together with the Surface Detector
array, but it is operated in a multi-hybrid setup which also includes the Fluorescence Detector
Figure 7.2: Air showers
and underground as measured
muon counters. with LOFAR
An example with the
of a multi bestdetection
hybrid fits to the data (equation
is already given in(6.10)).
Left: Pattern projected
this chapter where theintodatathe
in shower plane.
the figures The
1.10 andcircles
1.11 indicate
are takenthefrom
measurements,
the same airthe background
shower
indicates
measured thebyfit.three
Thedifferent
integrated total pulse
detectors. power is
It should encoded
also in color.
be stressed, thatRight: Pulse power
the information as a function
gathered
of the distance to the shower axis. The open black squares indicate the measurements, the full dark red
by the radio detectors might very well be used to improve the established techniques of particle
circles show the fit to the data.
and fluorescence air shower detection.

to have measured some structure in the event. It should be noted that there are also some air
showers that were only measured with one station and can still be reconstructed.
Another difficult set of events is air showers with a very small footprint. These are mostly
vertical and their width indicates a large Xmax , i.e. a very short travel time of the radio signal
through the atmosphere. Two examples are shown on top in figure 7.4. While the regions from
50 m and further out are nicely represented, the fit function is unable to account for the highest
signals near the shower axis. One is tempted to disregard these data points as outliers, however
their systematic occurrence in vertical showers, points at a real physical phenomenon. The
same was indicated in the slightly worse 2 for vertical showers. The corresponding trend was
already observed when studying the full set of simulations. There, however, the uncertainty
was always dominated by the large number of antennas that did not measure a signal. It seems
that the parameterization in its current form is less well-suited for signals measured close to the
axis in vertical events. However, these events also have the smallest likelihood to be detected
 The Pierre Auger Observatory
2
Parts of this chapter have been published in:

J. Schulz for the Pierre Auger Collaboration

Status and Prospects of the Auger Engineering Radio Array
Proceedings of the 34th International Cosmic Ray Conference, The Hague, The Nether-
lands, (2015) [67]

J. Schulz, P. Dolron, R. Habraken, J. R. Hörandel, T. Wijnen

EMC-test of the AERA124 reference station
Auger Internal Publication, GAP2013_063 [97]

With an instrumented area of more than 3000 km2 , the Pierre Auger Observatory is leading in
the field of ultra high-energy cosmic-ray research. It is located on the Argentinian high plain
"Pampa Amarilla" (∼ 1400 m a.s.l.) north of the city Malargüe in the province of Mendoza.
The observatory consists of two main detector systems employing different methods to detect
extensive air showers and extract complementary information (from the same shower). The
two complementary detectors, which are described in the following sections (2.1 and 2.2), are
the Surface Detector array and the Fluorescence Detector. Profiting from these two established
and well calibrated baseline detectors, add-ons were introduced to the Observatory over time
to extent the detection capabilities or to test new detection methods. The extensions relevant
for this work will be introduced together with the Observatory’s radio detector, the Auger
Engineering Radio Array which this work is focussing on. Figure 2.1 gives an overview of the
layout of the site with the two baseline detectors, AERA, and the low-energy extension HEAT
(High Elevation Auger Telescope) with the associated dense Surface Detector region called
Infill.

2.1 The Surface Detector Array

The Surface Detector array (SD) covers the 3000 km2 with more than 1600 individual water
Cherenkov detector stations. The stations are deployed on a triangular grid with a spacing of
1500 m. This grid in combination with the station’s technical specifications and the trigger
conditions explained in the following allows for fully efficient air shower detection above a
28 Chapter 2. The Pierre Auger Observatory

Figure 2.1: Map of the Pierre Auger Observatory. The individual stations of the Surface Detector array
are indicated as dots. The Fluorescence Detector buildings are shown at the edges of the grid with the
different field of views of the single telescopes. Close to the Coihueco Fluorescence Detector building
where also HEAT is based, the densified region of the Infill as well as AERA are located. In the middle
of the array, two laser facilities used for atmospheric monitoring (CLF, XLF) are shown. The central
campus in the city of Malargüe is only a few kilometers south-west of the instrumented area. Figure
taken from the Pierre Auger Collaboration.

primary particle energy of 3 × 1018 eV.

The basic design of these detector stations was adapted from the Haverah Park experiment [98]
and comprises a polyethylene tank of a diameter of 3.6 m and a height of 1.6 m. Each tank is
filled to a height of 1.2 m with 12 m3 of purified water which is confined in a liner. All other
station hardware, which is mainly the photomultiplier tubes (PMTs), analogue-to-digital elec-
tronics, wireless communication, global positioning system (GPS) antenna, and a solar power
system, is attached to the main water compartment from the outside for easy access. With
these components, every station forms a self-sufficient unit with all necessary hard- and soft-
ware for power harvesting, data collection, and transmission. A photograph with indications
of the components is shown in figure 2.2. Each station has a power consumption below 10 W
and runs self-sufficient powered by a solar power system including two 55 Wp1 solar panels.

1
Wp = Output in (W) under 1000 W/m2 irradiation and standardized test conditions.
2.1. The Surface Detector Array 29

Figure 2.2: Photo of a Surface Detector station of the Pierre Auger Observatory. Indicated are the indi-
vidual hardware components as water compartment, batteries, global positioning system (GPS), com-
munication antenna, solar panel, photomultiplier tubes, and electronics. Photo courtesy of R. Smida.

The provided power is managed by a commercial charge controller connected to two 105 Ah
12 V-batteries allowing for 99% uptime [99].
All stations are connected via wireless Local Area Network (LAN) data links to the obser-
vatory’s central data acquisition system (CDAS). Every Surface Detector station is equipped
with a custom-designed wireless communication unit . The communication is handled ac-
cording to the Time Division Multiple Access (TDMA) protocol, where each subscriber has a
fixed time slot to send and receive data. The unit is connected to the station’s main electronics
on one side and a 12 dBi Yagi antenna on top of the solar panel support frame (3 m above
ground) on the other side. The unit operates in the 902 − 928 MHz frequency band and has
a power consumption of less than 1 W. The wireless connection allows for remote control of
the station as well as software changes. A group of detector stations forms a unit connected
to an access point (AP) located at one of the four Fluorescence Detector buildings. The typi-
cal number of stations connected to one AP is 57 (upper limit: 65) allowing for a single link
bandwidth of 1.2 kbit/s. To provide predominantly line of site for the connections (even for
the stations in the center of the array), the AP antennas are mounted at a height of 20 − 50 m
on communication towers. The AP antennas are sector antennas with an acceptance of 90◦
and a gain of 17 dBi. In this configuration, the setup allows for long distance connections of
more than 30 km between stations and their APs. The AP wireless unit is the same as in the
stations with slight modifications to allow for the larger data throughput. In a second layer,
all APs of one tower location are grouped together and connected to the other locations and to
the central campus (where CDAS is based) with a backbone link. In this case, a commercial
30 Chapter 2. The Pierre Auger Observatory

Loma Amarilla -40

-50

Radio signal strength (dBm)

o
ec
ihu

-60
Co

-70

s
do
ora
-80

sM
Lo
Central ! -90
Campus
Los Leones
-100

Figure 2.3: Map indicating the wireless communication layout of the Pierre Auger Observatory. The
dots indicate the Surface Detector stations. Color coded is the radio signal strength of the connections
between a station an its access point at one of the Fluorescence Detector sites. Exemplary marked by
an additional square in the lower part of the array are those stations connected to the access points at
Los Leones. The big dots denote the backbone wireless network stations at the central campus and the
four Fluorescence Detector locations. These main wireless links are indicated as dotted line.

7 GHz wireless network with a bandwidth of 34 Mbps per link is used. The links are arranged
in a U-shape with the central campus in the middle and two connections to both sides around
the array (see figure 2.3). This network infrastructure as a whole allows the transmission of
all necessary data packages between the individual SD stations and CDAS. Furthermore, data
from the Fluorescence Detector and the Observatory’s extensions can be handled. Detailed
information about the wireless network can be obtained from [100].
The Surface Detector data emanates from three nine inch photomultiplier tubes (PMTs) which
are installed at the top of the tank facing downwards through transparent windows in the liner.
The PMT signals indicate Cherenkov light induced by super luminous particles traversing the
water. The output is digitized by 40 MHz flash analogue-to-digital converters (FADCs) and
stored in a ring buffer for 10 s. The data are automatically calibrated for the detector response
on station level to the average signal induced by a vertical equivalent muon (VEM). This cali-
bration is updated every minute and also added to the station data for later analysis.
To select interesting events, two different trigger conditions are defined on the single-station
level (T1/T2). The time over threshold (ToT) condition requires 13 out of 120 FADC bins in
a sliding window to be above a threshold of 0.2 VEM in two of the three PMTs. The thresh-
2.2. The Fluorescence Detector 31

old crossing (TH) condition requires all three PMTs to have a signal larger than 3.2 times the
VEM signal height. The local T2 triggers are sent to CDAS where the set of T2 triggers from
different SD stations is checked for spatial and temporal coincidence. In case the requirements
to form a trigger for the Surface Detector array (T3) are met by the combination of T2 triggers,
the data are requested by CDAS and stored centrally for off-line analysis. The exact condi-
tions for T3 triggers and further information about the T1 and T2 trigger settings are presented
in [101].
For the data reconstruction the software framework Off line [102] is used. A deeper insight to
the framework is given in chapter 3. From the Surface Detector data, mainly the direction of
the shower axis and the energy of the primary particle are determined. This energy is inferred
from a cross calibration with the Fluorescence Detector discussed in the following section. The
calibration reveals a resolution of 16% for low-energy showers and 12% for showers with the
highest energies [29]. Additionally, 14% systematic uncertainty are inherited from the Fluo-
rescence Detector energy scale, [103]. The resolution on the reconstructed incoming direction
depends on the number of stations above the threshold (signal stations) and is 1.6◦ for showers
detected with 3 stations and 0.9◦ for showers detected with more than 5 stations [29].

2.2 The Fluorescence Detector

The Fluorescence Detector (FD) covers the volume above the Surface Detector with 24 indi-
vidual telescopes. Four observation sites at the edges of the SD array are hosting six of these
telescopes each. The telescopes are arranged next to each other in a way that their individual
coverage of 30◦ × 30◦ gives each FD site a 180◦ azimuthal field of view (FOV) facing inwards
to the center of the observatory. The elevation of the FOV is minimally 1.5◦ above the hori-
zon. The FD sites and the fields of view in the North-South East-West plane are shown in
figure 2.1. Inclined shower tracks can cross more than one individual FOV and in case of very
high-energetic showers and depending on the geometry, "stereo" ("triple" and even "quadru-
ple") events detected by telescopes at multiple FD sites can be observed.
The telescopes are installed in a closed, clean, and climate controlled room. The individual
setups consist of an optical and an electronics system. The optical part comprises a shutter,
an aperture system, an ultraviolet (UV) bandpass filter window, and a 10 m2 mirror. For the
readout, a 440 pixel (22 × 20) camera with one PMT per pixel is used and connected to fast
digitizing electronics. A photograph of a FD building and a schematic drawing of a telescope
are shown in figure 2.4. The shutter in front of the aperture system is closed during day time,
in case of bad weather conditions (high winds, rain, etc.) or moonlight in the FOV. The band-
pass of the UV filter window ranges from 310 nm to 390 nm (transmission above 50%) which
covers the nitrogen fluorescence emission spectrum and limits the background from visible
light photons. The other optical components are arranged in form of a Schmidt telescope. The
segmented spherical mirror focusses the light onto a curved focal plane where the PMT cam-
32 Chapter 2. The Pierre Auger Observatory

Figure 2.4: The Fluorescence Detector of the Pierre Auger Observatory. Left: Photograph of the
Los Leones FD building during data acquisition. Three of the telescope bays are captured where one
shutter is closed because the moon is in the FOV of the particular telescope. In the background, the
communication tower is visible. Photo courtesy of R. Smida. Right: Sketch of a fluorescence telescope
including the main components of the optical system. Figure taken from [89].

era is installed. Different types of mirrors are used, all providing a reflectivity (at 370 nm) of
more than 90% in a clean state. The grid frame of the PMT camera is specially designed to
guide photons to the PMTs using Winston cones. With the overall design of the telescopes,
the angular spread of light from a distant point source falling onto the camera is well below
the size of a single hexagonal pixel (40 mm side to side). This is important, as an imaged air
shower activates a line of pixels on the camera.
The readout of the PMT signals is performed independently per telescope and the signals are
filtered, digitized, and passed through a multiple stage trigger system (for details see [29, 89]).
Additionally, a ground impact time is estimated for the shower candidate and a trigger is sent
to CDAS. These triggers initiate the request for SD data recorded around this time to form
a hybrid shower candidate. The threshold for full efficiency in hybrid detection of showers
throughout the whole array is 1019 eV. However, depending on the geometry, hybrid detection
of showers with much lower energies is possible. As a part of the SD array is readout after an
FD trigger, valid hybrid events can also contain just a single SD station, lowering the energy
threshold to 1018 eV. The average FD event rate is roughly 0.01 s−1 per FD site. Information
about the reconstruction techniques of the data will be given in chapter 3.
The fluorescence technique has its main strength in determining the atmospheric (slant) depth
of the shower maximum Xmax and the energy of the shower (see section 1.2.5). The resolution
on Xmax is better than 20 g/cm2 and the energy resolution is 7.6% with an additional 14% on
the absolute energy scale for energies above 1018 eV [29, 103]. For hybrid showers indepen-
dently detected with the SD and the FD, the combined angular resolution improves to 0.6◦ and
the resolution on the position of the shower axis is about 50 m.
In order to reconstruct the energy of a cosmic ray based on the measurement of the number
2.3. The Low-Energy Detector Extensions - Infill and HEAT 33

of fluorescence photons detected by an FD telescope, a precise calibration of the detectors is

necessary in addition to the knowledge of the fluorescence yield. Furthermore, environmental
influences on the measurement need to be eliminated or accounted for. The hardware of the
telescopes is calibrated with a known light source that can be mounted to the aperture system
of the telescope about once per year. More frequently, calibrations with a portable laser posi-
tioned at a few kilometer distance to the FD sites are performed shooting straight upwards into
the field of view of the telescopes. In addition every telescope is relatively calibrated before
and after each run with a fixed installed calibration light system. Details on the procedure of
the fluorescence telescope calibration can be found in [89, 104, 105].
As for extensive air shower detection the whole atmosphere acts as a calorimeter, several prop-
erties need to be monitored in detail to be able to correct for the continuously changing detector
medium. Exemplary, only a few key aspects will be mentioned here to demonstrate the com-
plexity of the method.
The atmosphere is often not a clear detector medium and the produced fluorescence photons
are scattered and absorbed by aerosols and clouds. To correct the measured light intensities at
the FD sites, the amount and distribution of aerosols and their influence on the light propaga-
tion need to be measured in-situ and continuously during data acquisition. For this purpose,
laser facilities (XLF and CLF, see figure 2.1) are centrally located in the Surface Detector array
and used to provide a set of (vertical to inclined) UV-light pulses every 15 minutes. The pho-
tons of these laser shots are scattered by the aerosols and can be detected with the fluorescence
telescopes at each site. As the laser beam power is calibrated, the amount of scattered light
at the FD sites is used to correct for aerosols. The complete description of the method and
further information can be found in [106]. Furthermore, the atmospheric state variables are
influencing the shower development. Due to quenching effects, the temperature and pressure
profiles of the atmosphere need to be taken into account to reconstruct the longitudinal shower
development [84, 89]. After several years of direct airborne atmospheric measurements with
weather balloons and radio soundings, the usage of atmospheric data based on the Global Data
Assimilation System (GDAS) became the standard at the Pierre Auger Observatory in 2010
[107]. The influence on the reconstructed value of Xmax is found to be smaller than 1 g/cm2
[108] while it makes the great experimental efforts with balloon flights superfluous.

2.3 The Low-Energy Detector Extensions - Infill and HEAT

To extend the energy threshold of the SD to the second knee of the cosmic ray spectrum at
4 × 1017 eV, two low-energy extensions to the baseline detector system were installed in 2011.
One is a denser populated region in the SD additionally equipped with underground muon
detectors, called the Auger Muon and Infill Ground Array (AMIGA) [109]. The other one is
an extension of the FD consisting of three additional telescopes with an elevated field of view,
the High Elevation Auger Telescopes (HEAT) [110]. Both extensions are depicted in figure 2.5
34 Chapter 2. The Pierre Auger Observatory

SD array 1500m

Coihueco FD

HEAT

Infill array 750m

42 additional detectors
Area ~ 23.5 km2

Figure 2.5: Sketch of the SD Infill array together with the position of the Coihueco FD and HEAT.
The dots indicate regular SD station positions whereas the circles denote the positions of the additional
tanks together forming the Infill. Also indicated are the fields of view for the Coihueco and HEAT
telescopes. Figure adapted from [110].

and a photograph of the High Elevation Auger Telescopes is shown in figure 2.6.
The SD Infill array covers an area of 23.5 km2 and is embedded in the regular SD 6 km to the
East of the Coihueco FD building. Additional Surface Detector stations placed with a spacing
of 750 m enable the Infill to be fully efficient at primary energies above 3 × 1017 eV for zenith
angles smaller than 55◦ . To also profit from the hybrid approach of detecting air showers at
these lower energies, the HEAT telescopes were designed to allow hybrid detections with the
Infill array down to 1017 eV. These three telescopes are therefore based at the Coihueco FD side
but work as an independent detector system. Their basic design does not differ significantly
from the general FD telescopes, but the HEAT telescopes are installed separately from each
other in individual shelters. These shelters can be tilted upwards by 29◦ so that their FOVs
lay above the ones of the Coihueco FD telescopes and just slightly overlap. Taking both
systems together, an elevation range from about 2◦ to 58◦ is covered. The same way an inclined
shower track can range over the FOVs of neighbouring FD telescopes, a vertical shower can
range over the FOVs of the HEAT and the baseline telescopes. In this case the Coihueco and
the HEAT telescopes can be paired up in the reconstruction to one unit. For calibration and
cross-checks with the baseline FD, the HEAT telescopes can be operated in the horizontal
position as well. A special feature of the HEAT telescopes is the faster readout electronics
which are optimized to measure air showers close to the telescope [29]. Both enhancements
use the baseline detector hardware but work in a different parameter space in terms of e.g.,
geometry and energy, therefore the established reconstruction techniques can be used but need
to be adapted to suit the Infill/HEAT data. As AERA is located within the SD Infill, the
enhancements play a major role for the reconstruction as well as the analysis of radio data.
From here on the term "Infill" will be dropped and the term SD denotes the SD Infill. In some
cases the term Infill will however be recalled for clarification. Analogously, the term FD will
also include HEAT if not explicitly stated otherwise.
2.4. The Auger Engineering Radio Array - AERA 35

Figure 2.6: The High Elevation Auger Telescopes. Left: Photograph of the 3 HEAT telescopes in the
upwards orientation. Right: Sketches of a HEAT telescope in the horizontal orientation for calibration
and maintanance (top) and the upwards orientation for data acquisition (bottom). Both figures are taken
from [29].

2.4 The Auger Engineering Radio Array - AERA

AERA is an extension of the Pierre Auger Observatory designed to measure the radio emission
of extensive air showers between 30 and 80 MHz. It is located inside the SD Infill array and
underneath the FOV of the Coihueco and HEAT FD telescopes. The project is carried out
in three phases. At first, an array of 24 radio detection stations (RDS) called AERA24 was
deployed in September 2010 to proof the feasibility of hybrid air shower measurements and to
develop the techniques to extract air shower parameters from the radio measurement. Based
on the experience gained with AERA24, the second stage, AERA124, was deployed in May
2013 with a modified station and antenna design [111]. With different grid sizes, multiple
improved hardware and trigger concepts and an instrumented area of 6 km2 , AERA124 is
accumulating cosmic-ray data at a rate of a couple of thousand showers per year with energies
ranging roughly from below 1017 eV up to exceeding 1019 eV. In the third stage deployed in
April 2015, AERA153 now covers about 17 km2 . The targets of this extension are mainly
horizontal air showers (> 60◦ zenith angle). Their large footprint allows for the detection
with a sparsely instrumented array and the hybrid measurements of these air showers contain
complementary information in radio and particle signals. The layout of AERA including all
three stages is shown in figure 2.7. The work for this thesis is mainly based on data from
AERA24 and AERA124, therefore these two stages will be described in more detail in the
following subsections.
36 Chapter 2. The Pierre Auger Observatory

AERA24 LPDA, ext. trig. CRS

AERA24 LPDA, int. trig. FD site
AERA124 butterfly, int. trig. FD FOV
AERA124 butterfly, ext. trig. HEAT FOV
AERA153 butterfly, ext. trig. beacon
SD station

1 km

Figure 2.7: Schematic overview of the Auger Engineering Radio Array and part of the Infill region of
the Surface Detector array. The Coihueco Fluorescence Detector telescopes and the HEAT extension
are also indicated with their respective field of view (FOV).

2.4.1 AERA24
The AERA24 stations are arranged with a distance of 144 m from each other on a triangular
grid covering 0.4 km2 . Each station operates self-sufficient and consists of five main com-
ponents: the radio antenna, the fast digitizing electronics, the solar power system, the GPS
timing system, and a communication unit. A photograph of an AERA24 station is shown in
figure 2.8. The solar power system consists of four Isofóton solar panels with 60 W power
each, two Victron Energy gel batteries with 150 Ah capacity each and a Morningstar SS-20L
charge controller. The two panels are mounted on a triangular frame and all station electronics
is housed in a water tight metal box underneath which also shields radio frequency interference
(RFI) . The total power consumption of an AERA24 station is provided by the solar power sys-
tem the whole year around.
The antenna is a logarithmic-periodic dipole antenna (LPDA), with two sensitive horizontal
planes [111]. One is oriented in the direction of the magnetic field and the other is oriented
perpendicular to it. For high-precision polarisation measurements, the alignment of the anten-
nas has been carried out with an accuracy of better than 0.6◦ [112]. The antenna is optimized to
operate in the frequency range from 27 − 84 MHz which was found to be relatively radio quiet
in terms of terrestrial background radiation on site [113]. Connected to the foot of the antenna,
a low-noise amplifier (LNA) is used to amplify the signals before they are introduced into a
2.4. The Auger Engineering Radio Array - AERA 37

Figure 2.8: Photograph of an AERA24 station. The logarithmic-periodic dipole antenna is mounted
well above the ground with the solar panels on top of the electronics compartment underneath. The
compartment houses the electronics as well as the batteries and the power controller.

filter-amplifier and digitization chain. There, they are bandpass filtered between 30 − 80 MHz
and fed into a low-gain (+29 dB) and a high-gain (+49 dB) channel each extending the dynamic
range. In the next step, the signals are digitized by 12-bit flash analog-to-digital converters
(FADCs) and further processed by a field programmable gate array (FPGA) and a central pro-
cessing unit (CPU). This chain is referred to as the "digitizer" and it exists in various forms
to test and optimize the technical realization of different detection strategies. For AERA24,
mainly two digitizer versions were deployed focussing on different triggering strategies. One
has a sampling rate of 180 MHz and features a 4 GB ring buffer which can hold data from two
channels for ~7 s or from four channels for ~3.5 s. The ring buffer allows for external trigger-
ing by other detectors of the Pierre Auger Observatory, mainly the SD and the FD. The time
needed to form an external trigger from the SD or the FD via CDAS can be up to 7 s. There-
fore, mostly only the high-gain channels are used for externally triggered data acquisition. In
case of an external trigger, the ring buffer is read out, and a time series of 56.89 µs around
the requested trigger time is sent to the central radio DAQ to be stored for offline analyses. In
addition, running in a self-trigger mode is also possible with this hardware. The second type
of digitizer has a sampling rate of 200 MHz and is designed for self-triggering and therefore
features no large buffer medium. In case of self-triggering, a first trigger decision can be made
in the FPGA based on the ADC values. This trigger information is then forwarded to the CPU
where further processing and signal verification can be done. Remaining triggers are GPS
timestamped and sent to the central radio DAQ where all timestamps of triggered stations are
processed. In case of coincidences, the data are requested from the triggered stations. For
further information on triggering, data aquisition and reconstruction see chapter 3.
The AERA24 stations use a fibre-based scheme to communicate with the central Radio DAQ
[114]. Therefore, a media converter is used for the connection between the Ethernet and the
38 Chapter 2. The Pierre Auger Observatory

Wi-Fi
GPS antenna antenna

physics
electronics antenna
compartment

solar panel

Figure 2.9: The AERA124 station. Technical drawing with indications of the main components.

fibre in the station electronics. The stations are daisy chained in two loops starting at the
Central Radio Station (CRS). Compared to the 100 MBits/s Ethernet connections used on both
ends, the fibre does not limit the bandwidth and therefore provides the opportunity for trigger
development with high rates and data throughput. From the CRS, a wireless link is operated
to the Coihueco FD building connecting to the Auger network, see also chapter 2.4.6.

2.4.2 AERA124
The AERA124 stations are arranged on a triangular grid with two different spacings of 250 m
and 375 m around the dense core of AERA24. The spacings are chosen to enlarge the instru-
mented area and to increase statistics, but also to explore geometries for future cosmic-ray
experiments. With a larger instrumented area and more stations, the challenges of semi-mass
production, deployment, and operation become important. To account for that and also to in-
clude technical expertise from operating AERA24, the layout of the AERA124 stations was
redesigned substantially. The new station design is shown in a rendered mechanical drawing
in figure 2.9. The new setup includes various fundamental changes in the mechanical and
electronic parts. The design was modified in such a way that all components are mounted to
a central pole. The antenna for cosmic-ray measurements is mounted on top of this pole and
all station electronics and the solar power system are housed in one compartment underneath.
This electronics compartment also features an RFI tight Faraday cage as interior to prevent
the station electronics from radiating to the outside. A test of the shielding capability of the
Faraday cage is presented in section 2.5.2. On top of the physics antenna, a 5 GHz grid an-
tenna for wireless communication and the GPS antenna are mounted on a secondary pole. For
2.4. The Auger Engineering Radio Array - AERA 39

cost effectiveness, the solar power system for the new setup consists of only one 135 W Ky-
ocera KD140 solar panel in combination with one of the 150 Ah gel batteries already used for
AERA24 and a new charge controller (IVT SCD-20). This limits the available power and puts
constraints on the power consumption of all components. Part of the redesign goals of the new
station electronics has therefore been a power consumption of less than 12 W.
The antennas used for the AERA124 stations are active bowtie antennas also called "Butter-
fly" [111, 115]. This antenna type has a simple design and has been tested successfully in the
CODALEMA experiment [65] and in an earlier R&D setup at the Pierre Auger Observatory
[116]. In contrast to the LPDA antenna, the Butterfly antenna is sensitive towards the ground
which enhances the signal. This, on the one hand increases the antenna gain, but on the other
hand leads to an influence on the antenna characteristic by the (changing) ground conditions.
The general layout of the digitizer chain is similar to the one used in AERA24 even though
every component has been redesigned. In total, there are 60 AERA124 stations equipped
with externally triggered hardware and 40 stations employing internal triggering. Two major
changes were made to the digitizers without a ring buffer. A new FADC with a depth of 14-
instead of 12-bits increases the dynamic range. Therefore, the duplication of data in high- and
low-gain channels becomes unnecessary and the two anciently low-gain channels are used to
introduce a new internal trigger next to the self-trigger on the radio signals. This internal trig-
ger is based on scintillation counters in the electronics compartment of the radio station itself.
Two (or one) small plastic scintillators are integrated in the 40 AERA124 stations that are not
equipped with ring buffers (see figure 2.7). A detailed technical description of the digitizing
electronics of the self-/scintillator-triggered AERA124 stations is given in [117]. More infor-
mation about the scintillators and also their potential as a stand-alone cosmic-ray detector is
given in chapter 4.
The stations communicate with the DAQ system via a wireless network operating in the 5 GHz
range. The network was set up as part of this work and all components as well as the network
scheme are presented in section 2.4.6. Tests for the system integration of the wireless system
are presented in section 2.5.3. The DAQ system was changed from AERA24 to AERA124
to accommodate the additional stations and mainly moved to the Coihueco FD building. The
DAQ of the internally triggered stations is still based at the CRS. A description of the data
acquisition is given chapter 3.

2.4.3 AERA153

The AREA153 stations are technically very similar to the AERA124 stations equipped with a
Butterfly antenna. Only the digitization electronics has been modified to account for technical
improvements. The main difference to compared AERA124 is the increased station spacing.
One row of seven stations has a spacing of 375 m and the other 18 stations are placed on a
750 m grid and the radio detection stations are located close to the Infill SD stations.
40 Chapter 2. The Pierre Auger Observatory

Measurement
0 15 0 15 0 15 Simulation
13.5 10.5 6.4
30 30 30
12.0 5.6

Ze 45

Ze 45

Ze 45
9.0

ni

ni

ni
10.5 4.8

th

th

th
7.5
9.0

✓
|Ha, | [m]

|Ha, | [m]

|Ha, | [m]
4.0
7.5 6.0

60

60

60
3.2
6.0 4.5
4.5 2.4
3.0
75

75

75
3.0 1.6
1.5 1.5 0.8
90

90

90
0.0 0.0 0.0
f = 35.0 MHz f = 55.0 MHz f = 75.0 MHz

Figure 2.10: Comparison of the measured and the modelled vector effective length of the LPDA an-
Figure 10: The VEL of the Small Black Spider LPDA as a function of zenith angle for three different
tennas of AERA24 for three different frequencies. The data were acquired in situ with a balloon-based
frequencies in measurement and simulation. Data points are given for all zenith directions that we
reference source. Figure taken from [111].
were able to access with the setup shown in Fig. 8 . The uncertainty of the measurement is indicated
by the bars analog to Fig. 9.
2.4.4 Research and Development Stations
suchanasengineering
As LPDAs (cf. Eq. A.8).
array, The additional
AERA variations
is also used of the
to test newVEL within the frequency
technologies and hardwareband occur
com-
due to the interplay of the LPDA’s dipole elements which resonate at different
ponents. Two subjects were addressed with dedicated test stations. One subject is the em- frequencies. We
observe that the simulation reproduces the bandwidth and the overall size of the measured VEL.
ployment of an additional antenna sensitive to the vertical component of the electric field to
For a set of three frequencies, the dependence of the VEL on the zenith angle is shown in Fig.
measure the full three-dimensional electric-field vector. Extra stations equipped with an ad-
10. For the low frequencies, the antenna is most sensitive to zenith angles around 45 . At higher
ditional antenna
frequencies, are placed
a side-lobe offevolves
pattern grid inwith
the up
middle
to twooflobes
AERA124 and another
at the highest 5 of the regular
frequencies.
internally triggered
The primary stations
cause for thein thelobes
side South-East of AERA124
is the constructive were modified
and destructive accordingly
interference with
of the direct
different
wave andvertical
the waveantenna
reflectedtypes. Theground
from the other subject is the detection
at the position of cosmic
of the antenna with itsrays at low
lowest MHz
dipole at
frequencies. ForNote
a height of 3 m. this that
purpose, one additional
a conclusion radio detection
on the reception station
of transient sensitive
signals can onlyto be
frequencies of
drawn if the
1wide
− 10band
MHz combination
was installedof these
in thepatterns
field. including their respective phasing is regarded, as we will
do in Sec. 5.1.
With respect to the shape of the side lobe pattern we note a good agreement between measure-
2.4.5
ment andCalibration
simulation. For the combination of all zenith directions and all frequencies within the
bandwidth we observe an agreement of the simulated and measured VEL of better than ±20%.
Great efforts were made to calibrate the different AERA stations concerning the absolute sig-
The phasing of Ha,f reveals the group delay induced by the Small Black Spider to the transmit-
nal response and relative timing. All components of the analogue signal processing chain were
ted signal. The group delay t is given by:
calibrated individually before being deployed and the production tolerances of the individual
d
parts are small. The contribution oft(w) the =uncertaintyarg(Hofa,f
the) measured
. values induced by the(4.7) sig-
dw
nal chain components is estimated to be 2.5%. For the antennas, detailed electromagnetic sim-
In Fig. 11
ulations the group
were delay and
performed for the zenithair-borne
various direction calibration
is displayed.campaigns
Within the were
measurement bandwidth
carried out on-site
from 30 to 80 MHz the group delay decreases by ⇠ 50 ns, where measurement and simulation agree
to calibrate the direction and frequency-dependent responses on an absolute scale [111]. The
on the functional dependence.
simulated vector effective length (VEL) is shown in figure 2.10 for three different frequencies
As will be discussed in detail in Sec. 5 a non-constant group delay induces the dispersion of
over the rangetransient
the observed of zenith angles
signal andcompared
thus reducesto measurements
its peak amplitude.takenThe
in the fielddelay
group for the AERA24
displayed in
LPDA antennas. The overall agreement between the measurements and the simulations
Fig. 11 results from a combination of the delay introduced by the logarithmic periodic structure of is bet-
ter
the±20%. Taking
Small Black the average
Spider over all
and the delay frequencies,
introduced by thethe differences
filter elements amounts to ±5%.
of the amplifier. For the
Although
the amplifier
Butterfly has been
antennas of designed
AERA124, to suppress
calibrationthe measurements
signal receptionwere outside the bandwidth,
performed it causes a
and preliminary
non constant group delay also within the measurement bandwidth especially at
analyses of the data show a similar result. The calibration is however more challenging due to lower frequencies.
Thesensitivity
the most recentofversion of thetowards
the antenna AERA readout electronics
the ground with itsis non
ableconstant
to compensate digitally
properties. Theforexact
the
modelling of the antenna pattern is work in progress and also aims at including changing en-

– 18 –
2.4. The Auger Engineering Radio Array - AERA 41

vironmental conditions.
For the reconstruction of air-shower properties, the arrival time of the signals is an important
parameter. The reconstruction of the shower axis, interferometric approaches to enhance the
detector sensitivity, and analyses based on the radio wavefront shape are depending on ac-
curate timing. The commercial GPS systems however only provide a coarse resolution of a
couple of nanoseconds and also experience drifting over time. Therefore, the relative timing
of the AERA stations is continuously monitored via their phase offset measuring sine waves
from a reference beacon [118]. The beacon is installed at the Coihueco FD site and emits four
precisely defined sinusoidal signals. Additionally, signals from commercial aircraft detected
by the AERA stations are used to determine the timing. In combination, a precise timing
calibration of about 2 ns is possible, whereas the built-in GPS-clocks exhibit drifts of tens of
nanoseconds. Further information on the timing calibration is available in [31].

2.4.6 Wireless Communication System

For AERA the step from a custom made wireless system as used for the SD to a commer-
cial system manufactured by Ubiquity was made. The system operates in the 5 GHz regime
and supports the standard Ethernet protocol as well as a TDMA protocol which is used in the
AERA setup. As for the network of the Surface Detector array, the AERA network is set up
in a point-to-multipoint configuration, where multiple RDSs are connected to a central base
station access point (AP) .
A sketch of the AERA153 site including all RDSs and all AP locations is shown in figure 2.11.
The network consists of two point-to-point and four point-to-multipoint wireless links and also
includes the fibre-network of AERA24. The part of the AERA124 array to the East of the CRS
(including all internally triggered stations) is connected to two APs at the CRS. There, 23 RDSs
in the north-eastern part are connected to one AP (CRS East) and 17 RDSs in the south-eastern
part are connected to the other AP at the CRS (CRS West). The part of the AERA124 array to
the West of the CRS is connected to an AP at the Coihueco FD site. The same applies for the
25 stations deployed for AERA153 in the South of the AERA153 array for which an additional
AP was installed at the Coihueco site. For the connection between the CRS and Coihueco, the
two dedicated point-to-point links are used for data transmission and maintenance login.
Independently of the assigned AP, all RDSs using wireless communication are equipped with
a combination of a Ubiquiti Bullet M5 and a 27 dBi L-COM grid antenna optimized for the
5 GHz range. For the APs at the CRS, the Ubiquiti Rocket M5 GPS is used in combination
with an Ubiquiti Hi-Gain Airmax Sector 5G-90-20 antenna. For the two APs at Coihueco
also Ubiquity Rocket M5 are used together with the Ubiquiti Hi-Gain Airmax sector anten-
nas. Here, one of the used Ubiquity Rocket is a slightly modified Titanium version. The two
point-to-point links are set up with Ubiquity Bullets M5 and 30 dBi Poynting high-gain grid-
antennas. The 24 RDSs of AERA24 are connected to the CRS via a Draka Comteq Telecom
42 Chapter 2. The Pierre Auger Observatory

AP CO I to AP CO I
AP CO II to AP CO II
AP CRS east to AP CRS east
AP CRS west to AP CRS west
CRS fibre to CRS
CO FD site 2 links CRS-CO

1 km

Figure 2.11: Overview of the AERA communication network. The AERA24 RDSs connected via fibre
to the CRS are indicated by the red dots, AERA124 and AERA153 RDSs connected via wireless links
by triangles. In addition the CRS and the Coihueco Fluorescence Detector building are represented by a
square and a circle, respectively. At both of these two spots, two wireless access points are installed and
their FOV is roughly indicated by the wedges. The color of the station marker corresponds to the AP the
station is connected to. The two access points at the CRS are called East and West for historical reasons,
even though this does not describe their orientation. Between the CRS and the Coihueco building, two
point-to-point links are installed.

SM-LEJEN optical fibre [114]. Furthermore, an experimental system designed by the Insti-
tute for Data Processing and Electronics at the Karlsruhe Institute of Technology (KIT-COMs)
also employing a TDMA protocol was temporarily installed in a part of the array [119]. After
a successful test period, the experimental system was replaced by the commercial Ubiquity
system. The experimental KIT-COMs system was realized as point-to-multipoint connection
with one access point at the Coihueco communication tower and 29 subscriber units [119].
The system operated in the 2.4 GHz range and used a commercially available L-Com 27 dBi
grid dish antenna at each RDS.
For completeness, it is mentioned here that for AERA24 only one point-to-point link between
the CRS and Coihueco was operated using a 2.4 GHz system called Viper manufactured by Ho-
pling Technologies. The Viper was taken down and replaced during deployment of AERA124.
2.4. The Auger Engineering Radio Array - AERA 43

Figure 2.12: User interface of the Ubiquity web-based network management server application Air-
Control. The tool can be used to manage and monitor Ubiquity wireless devices.

Settings

The settings for the individual wireless hardware components are given in the tables in ap-
pendix A.1. In general, all links operate on not-overlapping frequency channels with 40 MHz
width in the range of 5.600 − 5.805 GHz. Each of the six wireless subnetworks has its own
service set identification (SSID) and all components are locked to their sub-network SSID.

Monitoring

For monitoring and maintenance of the more than 150 wireless communication devices, the
web-based Ubiquity network-management server-application called AirControl is used. With
this tool, all components can be managed regarding local and global system settings. It is fur-
thermore possible to send commands to all units simultaneously to e.g., upgrade the firmware
or backup the system settings. The AirControl user interface is shown in figure 2.12.

System Performance

Due to the shifting focus in trigger strategies towards particle based triggers (internally or ex-
ternally detected), the effective data-rates of the AERA153 setup are much below 200 kbps per
station as predicted by Kelley [120]. For the externally triggered stations, the data transmis-
sion rate is about 40 kbps per station and the actual data receiving rates at the two APs at the
Coihueco FD building are roughly 2.5 Mbps (CO I) and 1.0 Mbps (CO II). For the internally
triggered stations, the data transmission rate is about 25 kbps and the receiving rates of the
APs are roughly 570 kbps (CRS East) and 420 kbps (CRS West). For the point-to-point link
between the CRS and Coihueco serving the AERA24 stations the rate is around 700 kbps.
44 Chapter 2. The Pierre Auger Observatory

30
36

radio signal strength (dBm)

42
48
54
60
66
72
78
84
90

Figure 2.13: Overview of the signal strength for the connections of all AERA RDSs with wireless
communication. The gray lines in the background indicate the directions of the links between the
stations and the corresponding access points. The two dimensions of the plot are scaled arbitrarily.

In general, all stations have a stable and high-quality connection to their APs with signal
strengths ranging from −78 dbm to −32 dbm. The noise floor in the used frequency range is
on average −90 dBm which is well below the minimal signal strength reached by all stations.
The causes for a lower signal strengths are various, the most obvious ones are not properly
aligned grid antennas at the RDSs, disturbed line of sight, and link directions outside of the
main acceptance region of the sector antenna as it is the case for one RDS connected to the
CRS East access point. The signal strengths of the wireless subscriber units of all RDSs in the
AERA153 array are shown in figure 2.13.

Due to the positive experience gained with the Ubiquity system, it has been suggested to also
use this kind of hardware for the upgrade of the Pierre Auger Observatory: AugerPrime

2.5 System Integration for AERA124

The AERA124 station design was revised after the first stage of AERA to incorporated the
gained experience concerning deployment and operation of the AERA24 stations. Due to
these changes, the way of integrating every individual component such as e.g., the digitizer
or the power control unit (PCU) into the station had to be reconsidered. Furthermore, tests of
all individual hardware components and an efficient assembly and deployment scheme were
necessary to build up the large number of new RDS. Most of these tests were performed
with a complete prototype of the AERA124 station installed at the Radboud University in
Nijmegen. In this section, three aspects of the system integration are discussed, namely the
new solar power system, the electromagnetic compatibility (EMC) of the new electronics, and
the wireless communication system.
2.5. System Integration for AERA124 45

2.5.1 Solar Power System

The power budget, i.e., the power consumption and the size of the solar power system is a cru-
cial aspect for the design of self-sufficient detector stations. For AERA24, the requirements on
the available power per station were formulated based on the combined power consumption of
all individual units to be 20.4 W [121]. The size of the solar power system was planned based
on simulations using different hardware combinations and considering the solar irradiation for
the location of Malargüe, Argentina, calculated with the PVSYST software package [122]. In
the final design, four 60 W solar panels and two 150 Ah batteries were installed.
Both, solar panels and batteries contribute significantly to the costs of an RDS. For larger scale
arrays, the whole system should be cost optimized. Therefore, one of the design requirements
for the new RDSs was to halve the size of the solar power system which in turn also means
halving the power consumption. The simulations for AERA24 showed that the state-of-charge
of the batteries should never be dropping below 40% during the whole year. To be able to
resize the solar power system for AERA124, the design goal for the power consumption was
set to 12 W encouraged by simulations. This load in combination with one 135 W solar panel
and one 150 Ah battery were simulated and found to be sufficient to keep the battery at least
40% charged [123].
During tests with the prototype station, it turned out, that under normal conditions complete
draining of the battery accrued leading to considerable downtime of the station. Therefore, a
solar power test was performed with the goal to obtain the actual necessary solar irradiance to
continuously operate the station given a constant load of 12 W and compare this to the expected
minimum irradiance value at the AERA site. An overview of the values of the solar irradiance
and the corresponding charge input during 42 days between the 21st of August and the 3rd of
October 2012 is shown in figure 2.14 together with the average solar irradiance for the location
of AERA during June of 3.7 kW/m2 [124] indicated by the vertical line (calculated for a panel
with 50◦ inclination and including local cloud coverage). The solid line shown is a fit to the
data in the interval 0 − 5 kWh/m2 /d. The large fluctuations above that range were found to be
caused by reaching a plateau in the charging state of the battery. From this, the minimum solar
irradiance necessary to keep a steady-state charge level is 3.3 ± 0.5 kWh/m2 /d. Therefore the
solar power system should be able to provide enough power to run a 12 W load continuously
throughout the year, including a safety margin for periods of bad weather. Environmental con-
ditions can however influence the situation. Dust and other dirt can decrease the performance
of the solar panel and also the fences of the RDSs caste a slight shadow on the panel during
some periods of the day. Low temperatures can decrease the capacity of the battery by 35%
and also high temperatures influence the performance of the battery negatively. Even though
the battery is explicitly made for deep discharging cycles, these have an impact on the lifetime.
The used gel batteries are specified for a life time of 600 cycles of more than 50% discharging
[125]. During the first 2 years of operation, the internally triggered stations have been found
to stay in the provided power budget as expected from the simulations and the described tests.
46 Chapter 2. The Pierre Auger Observatory

charge Q (Ah)

40
30
20
10
0
-10
-20
1 2 3 4 5 6 7 8
solar irradiance (kWh/m2/d)

Figure 2.14: Charge input into the battery as a function of the solar irradiance. The horizontal line
indicates a steady state and the vertical line indicates the average solar irradiance onto a surface with
an inclination of 50◦ including modelled cloud coverage at the AERA site in June [124]. The fit to the
data (solid line) is performed in the interval 0 − 5 kWh/m2 /d.

For the externally triggered stations, the power consumption is slightly higher and a small
amount of power related downtime has been experienced for these stations. The status of the
batteries is continuously monitored for all RDS.

2.5.2 Radio Frequency Interference Tests

For AERA124, improvements on several parts were made changing the appearance of the
RDS. Regarding these changes, the EMC of the AERA124 station had to be checked before
commissioning. Similar tests were performed for the AERA24 station [126] and their results
are taken as reference. During these tests, the monitoring hardware was found to produce a
significant amount of RFI and was therefore disabled for later station operation. The resulting
spectrum taken for the AERA24 station is shown in figure 2.15. In contrast to those tests,
where just the electronics in the compartment were tested, it was now possible to assemble
the complete AERA124 reference station inside the EMC chamber. This leads to a complete
measurement of the emission produced by the station and allows to check to which amount the
station picks up its own emissions.

Test Setup

The tests were performed in the ASTRON EMC test facility2 , which is a Faraday-chamber
certified for measurements of full anechoic conducted immunity in the frequency range of
2
www.astron.nl/r-d-laboratory/competence-and-support-groups/technical-support/rf-and-emc-test-
facility/emc-test-fac
2.5. System Integration for AERA124 47

25
20
amplitude (dBµV/m)

15
10
5
0
5
30 40 50 60 70 80 90 100
frequency (MHz)
Figure 2.15: Measurement of the RFI emission of the AERA24 station tested at ASTRON in 2010.
The noise floor (solid black), the station emission with monitoring (dashed red) and the situation with
disabled monitoring (doted green) are shown in the 30 − 100 MHz range. Figure adapted from [126].

150 KHz − 230 MHZ and full anechoic radiated immunity in the frequency range of 30 MHz −
3 GHZ. The test object can be screened with a vertically aligned combination Yagi-bowtie an-
tenna (Schwarzbeck biconilog VULB-9163), which is connected to a spectrum analyzer (Ag-
ilent Type E7405A). The setup acquires data during a sweep over the frequencies of 8.34 s.
For the tests, the complete station was mounted inside the test chamber, which included the
Butterfly antenna, the main pole, the electronics compartment, the solar panel, the battery, the
Wi-Fi grid antenna, two scintillators, the charge controller, and the digitizer together with all
cables and the GPS antenna, see figure 2.16. The station was powered by the internal battery
and connected to the DAQ system running outside the chamber via the build-in wireless com-
munication module (Ubiquiti Bullet M5, see section 2.4.6). First, a Bullet inside the digitizer
was connected to another one outside via an attenuated coaxial-cable. Later on, a point-to-
point wireless link was established to check the emission of the communication hardware, see
figure 2.16 (right).
Furthermore, it was checked whether activity in the solar panel as well as in the charge con-
troller leads to emission. Therefore, a 500 W halogen light was placed in the chamber, illumi-
nating the solar panel. This way, about 2 W electrical power were introduced into the system
by the halogen lamp leading to the requested activity.

Tests and Results

To fully identify emitting parts and to characterize the circumstances under which RFI radi-
ation is visible, the elements of the station were individually powered on. For some of the
measurements, amplitude traces were acquired with the digitizer in addition to the taken spec-
tra. This allows to check whether the emitted radiation is picked up by the station again. Fur-
thermore, data traffic was generated on the connection between the components (e.g., RS232,
48 Chapter 2. The Pierre Auger Observatory

Figure 2.16: Test setup of the AERA124 EMC test in the Faraday-chamber at ASTRON. The whole
station was set up inside the chamber (left). In addition, a halogen lamp illuminated the solar panel, and
a wireless link was established for data transfer (right).

wireless link), as cables and connectors were found to be a substantial noise source for the
AERA24 setup. The individual components were powered and tested one after another and
the detailed results of these test can be found in [97].
In summary, none of the components was found to significantly emit radiation in the AERA
frequency range from 30−80 MHz. The internally triggered digitizers were found to show mi-
nor emission at 40 MHz of about 5 dBµV/m above the baseline. The corresponding spectrum
is given in figure 2.17 and represents the final emission of a fully running station equipped with
an internally-triggered digitizer and two scintillator modules. As the 40 MHz emission lies in
the region of interest, it was checked whether the radio station picks up the radiation coming
from inside the digitizer itself. Therefore, traces were acquired with the digitizer and analysed
in addition to the externally measured spectra. The resulting averaged frequency spectra of
1100 traces for all channels are shown in figure 2.18. Channel 1 and 2 are connected to the
antennas, channel 3 and 4 are connected to one of the scintillator modules each. As expected,
the two radio channels show the shape of the bandpass filter between 30 − 80 MHz in the noise
floor whereas the two scintillator channels do not. Noticeable in all channels are the ampli-
tudes at 40 MHz and 80 MHz. As the scintillator channels were connected to the PMTs via
shielded cables, the emission is most likely picked up in the digitizer itself. In further tests
with terminated signal inputs, it was discovered that the noise is picked up partly on the board.
Even though the emission is pronounced in this test, it is only a problem if it has an amplitude
2.5. System Integration for AERA124 49

15
amplitude (dBµV/m)

10
5
0
5

30 55 80 100 125 150 175 200 225 250 275 300

frequency (MHz)
Figure 2.17: Result of the emission measurement of the final setup configuration with an internally
triggering digitizer. In this measurement, all functionalities of the station were running and a wireless
link for communication was established inside the Faraday chamber. The dashed red line represents the
measurement of the emission of the station, the solid black line represents the noise floor baseline of
the Faraday Chamber.

above the background noise level in the field. In figure 2.19, averaged spectra acquired at the
AERA site with the stations equipped with an internally-triggered digitizer are shown. The
corresponding traces were taken shortly after the AERA124 deployment. The lines at 40 MHz
and 80 MHz are buried in the background noise. After zooming in, the emission at 80 MHz
becomes barely visible. Compared to the amplitudes of background noise in other frequencies,
it is negligibly small. The positions right at the edge of our range of interest makes them even
less problematic.
In addition to the single-part test and the full-setup test with the internally triggered digitizer,
a full-setup test was performed with the externally triggered digitizer as well. For this, the two
scintillators were taken out of the system and the externally triggered digitizer was put into the
Faraday cage inside of the electronics compartment. There are no major emission lines visible
in the taken spectra shown in figure 2.20. At 40 MHz and 80 MHz, minimal peaks are present.
Similar to the previews measurement, traces were taken with the complete with the digitizer
which do not show self-induced noise.

Electromagnetic Compatibility Statement

The measurements show that a comparable or even lower level of emission is achieved with
the new layout compared to AERA24. As the former results can be taken as a reference for the
AERA124 station, the measured level of RFI emission is acceptable. The biggest improvement
was made in shielding the monitoring hardware to read out currents, voltages, and tempera-
tures from the charge controller. Summarizing the performed tests and the obtained spectra, no
broad emission arises from any version of the digitizer in the range of interest of 30 − 80 MHz.
The station equipped with the internally triggered digitizer shows emissions at 40 MHz in the
50 Chapter 2. The Pierre Auger Observatory
Averaged Spectrum Ch 1 Averaged Spectrum Ch 2
22 22
Amplitude (a.u.)

Amplitude (a.u.)
20 20
18 18
16 16
14 14
12 12
10 10
8 8
6 6
4 4
2 2
0 0
10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90 100
Averaged Spectrum Ch 3 Frequency (MHz) Averaged Spectrum Ch 4 Frequency (MHz)
22 22
Amplitude (a.u.)

Amplitude (a.u.)
20 20
18 18
16 16
14 14
12 12
10 10
8 8
6 6
4 4
2 2
0 0
10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90 100
Frequency (MHz) Frequency (MHz)

Figure 2.18: Averaged frequency spectra (1100 time traces) of the radio channels (top) and the scin-
tillator channels (bottom) taken during the final full-setup test inside the Faraday chamber with the
internally triggered digitizer. For the radio channels, the bandpass filter is imprinted in the spectrum.

recorded spectrum. Internally, 40 MHz and 80 MHz are visible in the recorded traces and are
most likely generated by the internal clock of the FPGA. They are almost completely disap-
pearing in the noise floor for measurements in the field. No influence on the emission was
found by neither the wireless communication module nor the two scintillators.
The same conclusion holds true for the setup running with the externally triggered digitizer.
There, emissions at frequencies of 40 MHz and 180 MHz are minimally visible in the recorded
spectra. The traces recorded internally with the digitizer itself do not show an increased am-
plitude at 40 MHz. Here, the proper functioning of the Faraday-cage is crucial.

2.5.3 Tests of the Wireless Communication System

The Ubiquity wireless hardware has been tested in Europe and at the AERA site for about a
year during AERA24 operation. The goal was the verification of the performance in terms
of data throughput, power consumption, general reliability and project feasibility. The band-
width requirements were based on the upscaled data rates of AERA24 and estimated to be
200 kbps/RDS. The maximum power budget was demanded to be 2 − 3 W for the wireless
subscriber unit. Regarding reliability, an uptime fraction off 100% without maintenance was
desired. A detailed description of the requirements and tests can be found in [120]. There, the
2.5. System Integration for AERA124 51
Averaged Spectrum Ch 1 Averaged Spectrum Ch 2
Amplitude (a.u.)

Amplitude (a.u.)
60 60

50 50

40 40

30 30

20 20

10 10

0
10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90 100
Averaged Spectrum Ch 1
Frequency (MHz) Averaged Spectrum Ch 2
Frequency (MHz)
30
Amplitude (a.u.)

Amplitude (a.u.)
34
29
33 28
32 27

31 26
25
30
24
29
23
28 22
27 21
36 38 40 42 44 36 38 40 42 44
Averaged Spectrum Ch 1
Frequency (MHz) Averaged Spectrum Ch 2
Frequency (MHz)
Amplitude (a.u.)

Amplitude (a.u.)

18 14
16
12
14
10
12

10 8

8
6
6
4
76 78 80 82 84 76 78 80 82 84
Frequency (MHz) Frequency (MHz)

Figure 2.19: Averaged frequency spectra (1294 time traces) of the radio channels (Ch 1 and Ch 2) taken
with the AERA124 stations after the deployment in the AERA field (top). Underneath are close-up
views of the frequency ranges around 40 MHz and 80 MHz.

hardware was evaluated positively under the given requirements performing the tests with a
30 dBi Poynting K-GRID-003-06 antenna. As this antenna is rather heavy and expensive, Kel-
ley already proposed to use a 27 dBi L-COM HG5827EG antenna [120]. From his link budget
calculations, 27 dBi is the minimum gain of the antenna needed for AERA124. Therefore, two
of the tests were repeated with the smaller antenna in Europe, as described in the following.
As a test base station, a Ubiquity Rocket M5 connected to an Airmax Sector 5G-90-20 antenna
was set up at the Radboud University in Nijmegen. The height of the installation was approxi-
mately 90 m above ground and due to the flat landscape it was visible from relatively large dis-
tances. For the first test link, a Ubiquity Bullet M5 connected to a 27 dBi L-COM HG5827EG
52 Chapter 2. The Pierre Auger Observatory

15
amplitude (dBµV/m)

10
5
0
5
1030 55 80 100 125 150 175 200 225 250 275 300
frequency (MHz)
Figure 2.20: Result of the emission measurement of the final setup configuration with the externally
triggered digitizer. In this measurement, all functionalities of the station were running and a wireless
link for communication was installed inside the Faraday chamber. The dashed red line represents the
measurement of the emission of the station, the solid black line represents the noise floor baseline of
the Faraday chamber.

antenna via a 3 m TWS-400UF cable and a lightning protector (L-COM AL6-NFNFBW-9)

was used. The device was connected to a laptop and powered via a power-over-Ethernet in-
jector. The distance between the first test location and the base station was 2.4 km. The line
of sight is outside of the main lobe of the gain pattern of the sector antenna as shown in fig-
ure 2.21 (short path, green squares). For this small distance, the smaller antenna gain is not
an issue. The situation of smaller distances and directions outside of the nominal main lobe of
the sector antenna is interesting to consider as the positive test increases the possible layouts
for the connections at the AERA site. The transmission rate was tested via a build-in network
speed test program. The reached transmission rate was 64 Mbps, while a signal-to-noise ratio
of 30 dB and a link quality of 99% was measured.
For the second test link, the same test setup was used. A location close to the one used for
earlier tests with the 30 dBi antenna was chosen which is 10.3 km away from the base station,
see figure 2.21 (long path, marked violet). Here, the line of sight falls into the maximum of
the gain pattern of the sector antenna. In this case, the transmission rate was 5.8 Mbps, while
a signal-to-noise ratio of 13 dB and a link quality of 99% was measured. The power consump-
tion was monitored during this measurement and had a value of 4.2 W at the maximum data
transmission rate. The rates are significantly lower than for the larger grid antenna (TX speed:
74 Mbps at 10.7 km), however the performance is sufficient for the use at the AERA site with
a modified link layout with four APs. By limiting the maximum distance between RDSs and
access points to 6 km, as well as the total number of RDSs subscribed to one AP to 65, a
sufficient bandwidth can be provided (only 50 kbps/RDS are necessary for AERA153 at the
moment). Thereby, the advantages of the smaller and cheaper antenna concerning the wind
load and deployment out-range the limited performance.
2.5. System Integration for AERA124 53

Figure 2.21: Indication of the two wireless test links. The short link (marked green) had a range
of 2.4 km from Nijmegen Heyendaal to Nijmegen Hatert. The long link (marked violet) had a range
of 10.3 km from Nijmegen Heyendaal to Beers. The two line of sights are plotted as solid lines. In
addition, the gain pattern of the sector antenna at the base station is indicated by the dashed lines (for
better readability the pattern is sketched in two sizes).
 Data Acquisition and Hybrid Shower
3
Reconstruction

Depending on the type of hardware and thereby the possible types of triggers, the radio data
are acquired with two different systems. After acquisition, the radio data are merged with the
data recorded by the other detectors of the Pierre Auger Observatory and reconstructed in a
hybrid approach. Thereby, the reconstructed information from one detector can be used as
input for the reconstruction of the data from the other detectors. The radio DAQ as well as the
hybrid data reconstruction pipeline will be described in the following.

3.1 Background for Radio Air Shower Detection

The impact of broad- and narrowband noise differs for the specific steps in the acquisition
and reconstruction pipeline. The overall noise floor for these measurements is set by the radio
emission of the Galactic plane. This changes over time when the Galactic Center moves over
the sky. Additionally, the anthropogenic emission which increases the background further
varies over time, depending on its origin. In case of a threshold crossing trigger, the variations
in the emission require a flexible threshold which can be adjusted e.g., based on the RMS of
the background. More challenging is the appearance of pulsed broadband noise which can
have a similar characteristic compared to the radio pulses from extensive air showers. This
pulsed RFI originate from various sorts of electrical equipment, e.g., generators, broken power
line isolators or transformers. Therefore, most of the pulsed emission is produced close to the
horizon and in case events are recosntructed with zenith angles close to 90◦ they can easily
be rejected in offline analyses. More troublesome are RFI pulses detected in single stations
in coincidence with pulses from air showers in other stations. Various techniques have been
developed to reject RFI pulses already during data acquisition or if that is not possible during
offline analysis. These techniques are discussed in the following section 3.2 and in chapter 6.

3.2 Triggering and Data Acquisition

In the spirit of AERA as an engineering array, data are acquired with various triggering strate-
gies. Self-triggers, external-triggers, and internal particle-triggers, are used for air shower de-
56 Chapter 3. Data Acquisition and Hybrid Shower Reconstruction

tection. Furthermore periodic-, random-, and airplane-triggers are used to collect background
and calibration data.

3.2.1 Self-Trigger
For stand-alone operation of a radio cosmic-ray observatory, a reliable trigger mechanism
needs to be implemented based on the radio measurements themselves. In principle, a basic
signal-over-threshold trigger can be employed. Due to the background of broad- and narrow-
band noise, a more advanced trigger strategy is necessary to be able to pre-select triggered
pulses. Otherwise, the high frequency with which pulsed background noise appears chal-
lenges the data handling capacities of the stations, the network, and the DAQ systems. For
this selection, the voltage time series are online processed in the FPGA and additional trigger
conditions are applied. In principle, both types of digitizers are able to perform self-triggered
data recording. The exact conditions of an additional pre-selection however differ for the two
systems. The digitizers without ring buffers mainly select pulses based on their shape which
is realized by e.g., the number of crossings of a secondary threshold. All self-trigger condi-
tions for this type of hardware are discussed in [127]. For the digitizers with a ring buffer, a
similar procedure is implemented as described in [128]. Once a pulse is selected in the RDS,
the GPS timestamp of the pulse is sent to the central DAQ. There, it is checked for coincident
timestamps from other RDSs and if found, a multi-station event is formed. Hereby, at least
three stations need to participate. For the RDSs operating without a ring buffer, an additional
requirement came in action after the deployment of AERA124. Now, at least two internally
triggered (on the scintillation counters) RDSs need to participate in every event. An exception
are event candidates were more than 10 RDSs are triggered simultaneously. In the central
DAQ of the RDSs with ring buffer a basic direction reconstruction of the wave front is per-
formed. Based on the direction, a cut on the maximum zenith angle and a cone algorithm
are used to veto RFI background. The cone algorithm vetos events when there are multiple
pulses coming from the same direction in a certain time window [128]. This can for exam-
ple be used to reject RFI originating from airplanes. Finally, the voltage time trace data are
requested from the RDSs by the DAQ and written to disk for offline analysis. AERA24 has
proven that air-shower detection in a self-trigger mode is possible. Also, there are methods to
further discriminate miss-triggered RFI pulses in an offline analysis [129, 130, 131].

3.2.2 External-Trigger
External-triggers from the other detector systems of the Auger Observatory are still the main
source of successfully reconstructed multi-station radio air showers. An advantage is that data
are only written to disk when an air shower has been confirmed externally as the coincidence
requirement lowers the probability of selecting random noise pulses. Thereby, the data rate is
kept at a manageable level. This does however not mean that all externally triggered events
3.2. Triggering and Data Acquisition 57

contain a radio signal. Also noise pulses can occur in coincidence with an EAS which need
to be identified and rejected in offline analyses. One of the limitations of employing the SD
or the FD as trigger source is the time needed for the trigger to be formed in CDAS and sent
to AERA. This can take up to about 7 s in case of an FD trigger and is usually of the order of
3 s for an SD trigger. To be able to use a trigger which is received a couple of seconds after
measuring the radio signal, large ring buffers are necessary to store the data in the meantime.
After a trigger, the whole buffer is read out and saved to disk leading to large data volumes. A
second limitation is the poor knowledge of the signal position within the radio voltage trace.
This leads to rather long radio traces, as the signal region can only be roughly determined
by the external timestamps. In an offline analysis, the radio signal needs to be identified in a
relatively large window of sometimes µs length which is placed with respect to the externally
reconstructed core position and arrival direction and their uncertainties. This increases the
chance to pick a random coincident pulse in the signal window.
After an external trigger, it is not known in which of the RDSs a radio pulse was measured,
and therefore all stations are read out. In case of AERA153, this means 107 RDSs with ring
buffers are read out simultaneously for every trigger, where most of them only contain noise.
This is challenging due to the large data-volumes created for every event. At the same time this
situation offers an increase in information as also stations that measured no or a sub-threshold
signal contain useful information for high-level analysis, e.g., concerning the radio LDF.

3.2.3 Internal Particle-Trigger

A slightly different approach which is also interesting for possible future stand-alone radio
detector arrays is the employment of an internal particle trigger. It combines advantages of
external- and self-triggering, and allows self-sufficient and clear identification of the radio
pulses associated with an EAS. Therefore, scintillation counters are designed as an integrated
part of the radio stations and the particle detection on station level can directly trigger the radio
DAQ of the station. Furthermore, the trigger time is very close to the time of the expected
radio pulse in this station giving strong constrains on the radio signal-window. By triggering
the stations individually, it is also not necessary to read out the whole radio array for every
trigger. Details about the scintillation counters, the trigger settings, and the relative trigger
efficiency are given in chapter 4.
As the internal particle-trigger also works on station level, a similar multi-stage trigger scheme
is used as for self-triggering RDSs. After a particle detection in the scintillation counters,
a time stamp is sent to the DAQ and there a search for coincidences is performed and if a
multi-station event is found, the data are requested. The similarity between internal particle-
trigger and self-trigger allows a combination of both to form a multi-station event. This is
advantageous for air-shower detection without being dependent on the efficiency of non-radio
detectors (e.g., the SD or the scintillation counters).
58 Chapter 3. Data Acquisition and Hybrid Shower Reconstruction

3.2.4 Periodic-, Random-, and Airplane-Triggers

Calibration and monitoring of the performance of AERA are important topics for the operation
of the array. Furthermore, a realistic background library is extremely valuable for simulation
studies. These aspects are addressed by using data acquired with special trigger settings. On
periodic triggers, the full voltage time-trace data are requested from all RDSs every 100 s
(every 10 s till July 2014). The data can be used as noise for simulations, allowing for studies
with realistic background conditions. In case measured showers are simulated, noise traces
are available from within 50 s around the time of the measured shower. The periodic data are
also used for background studies. As the background is dominated by the Galactic emission,
the amplitude levels of this background are used to relatively calibrate the stations among each
other [61]. For random triggers, a station sending a GPS timestamp to the DAQ is chosen to
send the time-trace data based on a trigger counter. For the internally triggered RDSs every
40000th timestamp is selected by the random-trigger, for the RDSs equipped with a ring buffer
it is every 999th timestamp. In principle, commercial aircraft flying over the array often emit
RFI and have to be seen as background sources. However, their emission can also be used for
a timing calibration. To record this data in self-trigger mode, the cone algorithm is deactivated
for a region of 5◦ around the position on the sky where the aircraft is located. This position
is inferred from information broadcast by the aircraft as part of the ADS-B information [132].
In addition to these automatic triggers a manual-trigger can be invoked to request data at any
time.

3.2.5 Data Handling and Merging

The data acquired with AERA are written to disk as binary files. To allow more flexibility,
these binary data are converted in a following step to a ROOT-based file format [133, 134].
During the conversion it is possible to separate the different data streams based on the trigger
flags. This allows for example to split up the files used for monitoring purposes or the airplane
data and handle these files separately.
The data of every detector component of the Pierre Auger Observatory are written to individual
ROOT-based raw files. For the SD and the FD, these raw files are merged directly in CDAS
to a combined hybrid file on a daily basis. In the next step the hybrid files can be combined
with the data from the AMIGA muon counters. The AERA data are merged with this SD-
FD-AMIGA files, but they can also be merged with all individual detector raw files separately,
e.g., to SD-radio hybrid files. For events where one or more of the detectors did not trigger, no
information of the corresponding detector system is merged. The exact conditions of merging
radio data depend on the DAQ these data were acquired with. The AERA data acquired with
the DAQ system in Coihueco, including all stations with ring buffers, is directly merged at the
FD site. The AERA data acquired with the DAQ system at the CRS is so far transferred to
Europe and merged there. In the near future, the two radio DAQ systems will be combined
3.3. Shower Reconstruction Software Framework 59

together in Coihueco.
For cosmic-ray data analysis daily multi-hybrid files (RD-SD-FD-AMIGA) are produced for
the time from noon to noon of the next day. Thereby, the data of the nightly FD observations
are contained in one file.

3.3 Shower Reconstruction Software Framework

The data of all detector systems at the Pierre Auger Observatory are processed with the soft-
ware framework Off line [102]. It was originally designed for (hybrid) reconstruction of the
SD and the FD data and/or corresponding air shower simulations. The framework is struc-
tured in three separate parts. One part are the event data which contain the recorded raw
data or the simulation output, respectively. During the reconstruction the data are processed
and reconstructed information as well as results are added to the event data. The second part
is the detector description which contains information about the detector configuration and
performance at any given time of operation. This also includes calibration data as well as
atmospheric monitoring data. The third part of the framework is the sequence of analysis
modules which contain all the algorithms needed during the whole reconstruction procedure.
In general, it links the event data and the detector description. Within the modules, information
from the event data can be requested. For individual events, the detector description is used to
infer the corresponding detector configuration, and reconstruct intermediate or final quantities
of interest. These values are then also added to the event data. Modules cannot communicate
with each other, which limits inter-module dependencies. The event data and the detector de-
scription differ conceptionally from each other. One allows read and write access from the
modules and accommodation of reconstructed values on an event-by-event basis. The other
is a read-only description of the detector. A schematic overview of the Off line framework is
shown in figure 3.1.

3.4 Baseline Detector Shower Reconstruction

During the hybrid reconstruction of the SD and the FD data, the SD information is processed
independently of the FD information. The FD event reconstruction however relies on timing
information of at least one SD station to reconstruct the shower geometry. A hybrid shower
where the data from both detectors were reconstructable is displayed in figure 3.2. The recon-
struction methods for both measurements are described in the following sections.

3.4.1 Surface Detector Array

The reconstruction of air-shower parameters from the measurements of the SD are based on
the recorded signal traces from the PMTs in the SD stations. The information from the PMTs
60 Chapter 3. Data Acquisition and Hybrid Shower Reconstruction

Antennas PMTs Telescopes

Stations Stations FD Sites

RD Event SD Event FD Event Reco Event Sim Event

Event

... Module Module Module ...

Detector

RD (AERA) SD FD Atmosphere

Stations Stations FD Sites

Antennas PMTs Telescopes

Figure 3.1: Structure of the Off line framework. See text for details. Figure adapted from [135].

is combined to obtain the signal time as well as the signal size in vertical equivalent muon
(VEM) per station. At the beginning of the reconstruction, quality criteria are checked in
addition to the trigger requirements already presented in section 2.1. Here a stricter selection
on the station timing is made to reject random stations which were triggered by muons not
related to the air shower. Furthermore, only showers for which the ground impact point of the
shower axis (shower core) is contained in the detector array are considered for highest level
analysis. This is enforced by requiring that the detector station with the highest signal was
surrounded by a hexagon of active detector stations at the time of the air shower (6T5 trigger).
The arrival direction of the air shower can be reconstructed by fitting a wavefront model to
the signal start times in the detector stations ti . The most basic model is a plane wave front.
This is e.g., used in case of low station multiplicity showers. The direction is then found by
minimizing
X [ti − t(~xi )]2
χ2 = 2
, (3.1)
i
σ ti

where ~xi are the detector positions, ti and σti are the measured time and its uncertainty, and
t(~xi ) is the calculated time at position ~xi . For higher station multiplicity, the wavefront is
approximated with an inflating sphere model as

c(ti − t0 ) = |~xsh − ~xi |, (3.2)

where ~xsh and t0 are a virtual origin and start time for the air shower development [29]. A
sketch of the underlying model is shown in figure 3.3 (left) together with a measured footprint
3.4. Baseline Detector Shower Reconstruction 61

Figure 3.2: Hybrid shower geometry display, combing the measurements of the particle footprint on
the ground with the observed longitudinal shower-profile. The signal times in the particle detectors are
indicated by color (yellow = early, red = late) and the time distribution of the pulses in the FD pixels are
indicated as lines to their emission region (purple = early, red = late). The triangles indicate the FOV
of the individual FD telescopes and the gray dots indicate non-triggered SD stations. The circle on the
reconstructed shower axis is the position of Xmax .

where the timing distribution of the stations and the reconstructed direction is shown (right).
The core position as well as the energy of the shower are inferred from fitting a modified NKG-
function to the measured signals projected onto the shower plane. The LDF used for the SD
reconstruction is a simplified version of equation 1.23,
 β  β+γ
r r + r1
S(r) = S(ropt ) . (3.3)
ropt ropt + r1

Here, ropt is the optimum distance, r1 is a constant, S(ropt ) is a fitted scaling parameter, and
β and γ are shape parameters. The optimum distance is the distance from the shower axis at
which S(ropt ) on average shows the smallest sensitivity to the shower geometry and the shower
development. This is important as the value of S(ropt ) is also directly related to the energy and
is used as an energy estimator. The values for ropt and r1 depend on the spacing of the detector
stations. For the regular 1500 m grid the values are ropt = 1000 m and r1 = 700 m. For the
Infill array the optimum distance is smaller, ropt = 450 m. The exponents can only be fitted if
the station multiplicity is large enough to support the number of free parameters. Otherwise,
parametrized values are used as β = 0.9sec(θ) − 3.3 and γ = 0, where θ is the zenith angle
 62 Chapter 3. Data Acquisition and Hybrid Shower Reconstruction

18

17

16

South-North (km)
15

14

13

12

−31 −30 −29 −28 −27 −26 −25 −24

West-East (km)

Figure
0: Reconstruction 3.3: Direction
of shower reconstruction
geometry: schematic based on
representation of signal timing.of Left:
the evolution Sketch of the shower front model
the shower
in the reconstruction of the direction of the air shower axis. See text for details. Figure taken from
[29]. Right: Particle footprint of an air shower. The signal timing is indicated by the color coding
time difference to a plane shower front [ns]

1400

(yellow = early, red = late). The size of the pentagons represents the measured signal size. The black
1200

1000
cross and the black line represent the core position and the reconstructed azimuthal direction.
800

600

of the 400
shower [136]. A display of the fitted lateral distribution of an air shower detected with
the Infill
200 array is shown in figure 1.10. Depending on the station multiplicity, the geometry

of the shower
0
500
can be reconstructed
1000 1500 2000
with
2500
an average uncertainty of 50 m for the core position
[137]. The resolution on the reconstructed incoming direction depends on the number of signal
distance to axis [m]

stations
1: Reconstruction and
of shower is 1.6◦dependence
geometry: for showers detected
of signal with(relative
start times 3 stations and 0.9
to the timing of a for showers detected with
◦

wer front) on perpendicular

more than 5distance to the
stations shower
[29]. axis. The shaded
A complete line is theofresulting
description fit of the
the reconstruction of showers measured
model and its uncertainty.
with the SD is given in [136].

to reject signals produced by random muons by searching for time-compatible

3.4.2 Fluorescence Detector
guarantee the selection of well-contained events, a fiducial cut (called the 6T5
is applied so that only events in which the station with the highest signal is
For the reconstruction of the Fluorescence Detector data, first the ADC traces of all pixels of
ded by all 6 operating neighbors (i.e., a working hexagon) are accepted. This
on assures ana accurate
camera are corrected forofthe
reconstruction thebackground
impact pointand on then converted
the ground, andtoatphotons at the aperture. In
case for
me time allowing multiple
a simple telescopes werecalculation
geometrical triggered hosted
of the at one FD site, the reconstructed data at the
aperture/exposure
mportant for, telescope
e.g., the spectrum analysis [38]. For arrival-direction
level are merged to form a combined pixel array. studiesThea less
arrival times of the photons as
ut can be used (5T5 or even 4T5).
a function of the viewing angle are combined with the timing of the highest-signal SD station
to reconstruct the plane containing the shower axis and the FD telescope (visualized by the
hower geometry
colored lines
ough approximation for the in arrival
figure 3.2). There
direction of isthe
also workisinobtained
shower progressby to fitting
replace the SD station input with
e.g.,
t times of the information
signals, from a RDS.
ti , in individual An air-shower
SD stations image
to a plane as seen
front. with two combined FD cameras
For events
ough triggered stations,
is shown inthese
figuretimes
3.4. are
Withdescribed by a more
the geometry detailed concentric-
information, every time bin in the recorded traces
al model, seecanfigure 30, which approximates the evolution
be projected onto a segment of the shower axis. A segmentof the shower front
in turn corresponds to a cer-
speed-of-light inflating sphere,
tain height and a certain atmospheric depth. The depth is inferred from the integration over
the atmospheric c(ti density
t0 ) = profile
|~xsh ~xfor
i |, the given direction. This allows (8) to reconstruct the amount
of of
x are positions light
the emitted
stations on along the shower
the ground axis from
and where~ x andthe tnumber of photons
are a virtual ori- at the aperture. The re-
i sh 0
a start-time of the shower development (see figure 31). From this 4-parameter

54
3.4. Baseline Detector Shower Reconstruction 63

elevation [deg]
30

25

20

15

10

0
120 110 100 90 80 70 60
azimuth [deg]

Figure 3.4: Combined camera image of an air shower detected by two of the Coihueco fluorescence
telescopes. The color indicates the time distribution of the pulses in the shown pixels (purple = early,
red = late). The line indicates the reconstructed plane spanned by the shower axis and the fluorescence
telescopes.
detected light [photons/m2/100ns]

data
total light
2500 Mie Cherenkov
direct Cherenkov
Rayleigh Cherenkov
2000 multiple scattered

1500

1000

500

0
270 280 290 300 310 320
time slots [100 ns]

Figure 3.5: Reconstructed contributions of (scattered) fluorescence and Cherenkov light to the total
amount of light detected. The displayed air shower is the same as shown in figure 1.10.

construction includes the response of the optical system as well as propagation losses due to
Mie- and Rayleigh-scattering in the atmosphere. These corrections are based on the detailed
calibration measurements of the atmosphere as well as of the detector system discussed in
section 2.2. Depending on the geometry, a part of the detected photons is not originating from
the fluorescence process, but from the Cherenkov radiation produced by the high-energetic
air-shower particles. The two contributions are disentangled and handled independently [86].
An example of the detected light for a measured air shower including the different contribu-
tions can be seen in figure 3.5. The number of electrons and positrons in the air shower is
proportional to the energy deposit in the atmosphere which is fitted by the Gaisser-Hillas func-
tion (equation 1.24). Similar to the procedure for the fit of the NGK-function to the SD data,
the two shape parameters X0 and λ are fixed for the reconstruction of air showers where a
64 Chapter 3. Data Acquisition and Hybrid Shower Reconstruction

four parameter fit to the longitudinal profile is not applicable (depending on the observed track
length and the number of photons). For the Fluorescence Detector, these parameters are taken
as the average values of detected air showers for which a four parameter fit was successfully
performed. The values are hX0 i = −121 g/cm2 and hλi = 61 g/cm2 .
In case telescopes at the Coihueco FD site are triggered in coincidence with telescopes of the
HEAT extension, the data from both sites are reconstructed separately as well as in a combined
approach.
The calorimetric energy (and from that the cosmic-ray energy) and the depth of shower maxi-
mum is obtained for the Gaisser-Hillas fit as discussed in section 1.2.5. The resolution on Xmax
is energy dependent and ranges from 30 g/cm2 at 1017 eV to 15 g/cm2 at 1020 eV [138, 139].
The energy resolution is discussed in the following section. The description of the FD recon-
struction presented here is following [139].

3.4.3 FD-SD Energy Calibration with Hybrid Showers

The SD air shower reconstruction works independently from the FD reconstruction which
needs a signal in one SD station and runs thereby in hybrid mode by default. A special category
of high-quality hybrid showers where the measurements of both detectors fulfil certain quality
criteria are called golden events. These quality cuts are e.g., applied on the fiducial FOV to
ensure an equal trigger probability of both detectors independently of the primary particle.
The golden events are used for a cross calibration between the two detector systems. Here,
an additional criterion is a minimum energy reconstructed by the FD of 3 × 1018 eV for the
calibration of the regular SD. In the following, the energy calibration of the Surface Detector
array will be discussed briefly.
The values of S(ropt ) are used as an energy estimator from the SD. The values of S(ropt )
however show a strong zenith-angle dependency. A larger zenith angle corresponds to a longer
development of the air shower in the atmosphere and therefore a larger attenuation of the
shower particles. Additionally, there is an impact of the track length of the particles in the SD
stations which also depends on the zenith angle. Both effects need to be corrected out during
reconstruction. This can be done under the assumption that the total flux of cosmic rays at the
top of the atmosphere is constant. The corrected values S38 (S35 for the Infill) for a specific
shower can be seen as the value S(ropt ) for the case that the shower axis would have had a
zenith angle of 38◦ (35◦ ). This corrected energy estimator is directly calibrated against the
energy determined with the Fluorescence Detector. The calibration plots for both, the regular
Surface Detector array and the Infill array are shown in figure 3.6. The data are fitted with a
power law
E = A · (S3x )B (3.4)

and the fit values are given in table 3.1. The correlations between the energy measured by the
FD and the energy estimators for the SD and the Infill array are almost linear. The resolution
3.5. Radio Air Shower Reconstruction 65

103
S 35 / 15 VEM
S 38 / 5 VEM

102

S 35 , S38 101

100

10 −1
1018 1019 10 20
E FD (eV )
Figure 3.6: Calibration of the energy indicators S38 and S35 against the FD energy measurement. For
the S38 calibration a minimum energy of 3 × 1018 eV is required leading to the sharp lower energy
cut-off. Figure adapted from [140].

Table 3.1: Parameters of the energy calibration of the SD and the Infill array against the FD. Table
adapted from [140].

detector SD (S38 ) Infill (S35 )

A (EeV) 0.190 ± 0.005 (1.21 ± 0.007) · 10−2
B 1.025 ± 0.007 1.03 ± 0.02
Ethreshold (eV) 3 × 1018 3 × 1017
dataset 01/2004-12/2012 08/2008-12/2012

of the FD energy determination is 7.6%. The systematic uncertainty is 16% at 1017.5 eV and
decreases to 14% at 1018 eV and above [29, 103, 140]. The resolution of the SD energy mea-
surements depend on the energy of the cosmic ray ranging from 22% at 1017 eV over 16% at
3 × 1018 eV to 12% at the highest energies and shows an additional direction dependence. The
systematic uncertainty of the SD energy reconstruction is inherited from the FD reconstruction.

3.5 Radio Air Shower Reconstruction

For the reconstruction of the radio data recorded with AERA, the Off line framework was
extended to integrate the radio part in a multi-hybrid reconstruction procedure together with
the SD and the FD (and their extensions) reconstruction and to make use of functionalities
specifically interesting for the radio reconstruction [135]. These functionalities for example
include operations on the recorded data in the frequency domain. The general structure is
adapted from the SD, where the class for a Surface Detector station corresponds to the one for
66 Chapter 3. Data Acquisition and Hybrid Shower Reconstruction

a radio station and the class for the PMT signals corresponds to the one for the radio channels.
The same holds true for the structure of the detector description which is similar to the one of
the SD. The actually performed reconstruction can strongly depend on the planned analyses.
A general description of the reconstruction steps on channel, station, and event level is given
in the following. The standard multi-purpose reconstruction sequence used for AERA, the
RdObserver, is described subsequently.

3.5.1 Radio Detector Description

To unfold the detector response from the measured data, every component in the signal chain
needs to be corrected for. The values for the channel corrections as well as for the antenna
pattern are available in the detector description. The responses of all individual components in
the signal chain apart from the antenna and the LNA were measured in the laboratory before
deployment. A time dependent detector description was introduced to the Off line framework
for AERA. The information is stored in a database and retrieved during run-time of the re-
construction. In addition, information about the status of the hardware and environmental
conditions is stored in a separate database.

3.5.2 Initialisation
The first step in the radio reconstruction procedure is the initialization of the event. There, the
shower-geometry source (SD, radio, simulation, etc.) and the signal and noise windows are
set. Furthermore, the trigger source is read from the raw file and the events can be grouped
according to this information.

3.5.3 Antenna Channel Level

The radio channels contain the recorded data from each antenna in ADC counts. The ADC
values are converted to a voltage. Depending on the hardware version, the RDSs acquire data
with two or three channels (four in case of high and low gain channels). In the further descrip-
tion the situation for two channels is discussed as most of the RDSs of AERA have two active
channels. If data were recorded for the low and high gain channels, the low gain channels are
selected in case of saturation in one or both of the other two.
The voltage time traces still contain the imprint of the detector response which needs to be
corrected for during the reconstruction. This is partly done on channel level where all charac-
teristics of the hardware parts like cables, filters, and amplifiers are taken into account.
For AERA a Beacon is used for an accurate timing calibration of the individual stations. The
necessary steps for this calibration during reconstruction are performed on the channel data, a
detailed description of the procedure can be found in [31]. The emission lines of the Beacon
are useful for calibration, nevertheless they are RFI in terms of the detection of cosmic rays.
3.5. Radio Air Shower Reconstruction 67

To clean the spectrum concerning narrowband RFI sources the corresponding frequency am-
plitude bins can be set to zero. This also affects information about the signal. Another more
accurate but also computationally more expensive cleaning method is to fit the contributions
of the Beacon lines in the time traces [141]. This fit is performed on parts of the trace which
do not contain the air-shower signal with a superposition of four sine waves. The resulting
amplitudes can then be subtracted from the data. In case of strong narrowband RFI from other
sources e.g., radio or TV stations, the sine-wave suppression is less applicable due to the mod-
ulation in frequency. In those cases, the frequency bins are set to zero. To reduce leakage
effects in the frequency spectrum, a window function is applied to the traces. If desired, the
channel traces can be up-sampled by padding zeros to the frequency spectrum outside of the
physical filter range. This allows to determine the positions of features in the traces as e.g.,
the pulse maximum or the rise time of the pulse with a higher precision. The corrected and
to voltage converted channel-traces, are used to reconstruct the electric field measured on the
station level.

3.5.4 Station Level

The measured channel data and the incoming electric field are linked by the antenna of the
RDS. For AERA, the two antennas are aligned in east-west and north-south direction and the
corrected data of the two channels represent the product of the electric field vector in these
directions and the vector effective length (VEL) of the antenna. The effective antenna length
depends on the incoming direction of the radiation as well as the frequency. To reconstruct
the electric-field vector from the channel data, the antenna response is unfolded. The unfold-
ing procedure requires knowledge of the incoming direction of the radiation, as the antenna
pattern is directional sensitive. For a known direction of propagation ên from a source at posi-
tion (θ, φ) on a sphere, the incoming electromagnetic wave E(t) ~ can be described by the two
components êθ , êφ perpendicular to ên . This so called on-sky coordinate system is depicted in
figure 3.7. The radiation has no component in the direction of propagation when the far field
approximation applies. The time dependent electric-field vector is then given as

~
E(t) = êθ Eθ (t) + êφ Eφ (t) . (3.5)

The conversion from the electric field to the generated voltage at the antenna base is given by
a multiplication with the Jones matrix [142]
!
JNS,θ JNS,φ
J= . (3.6)
JEW,θ JEW,φ

Here, the components give the complex responses of the antennas orientated in the East-West
(EW) and North-South (NS) directions for linearly polarized radiation in the êθ , êφ directions
~
at a given frequency f taken from the VEL (H(θ, φ, f )). To use this relation to unfold the
68 Chapter 3. Data Acquisition and Hybrid Shower Reconstruction

Zenith, z

ên
êφ
θ
êθ

φ
North, y
East, x

Figure 3.7: Geometry used for the reconstruction of the electric field from the measured voltage at
the antenna bases. The on-sky coordinate system (êθ , êφ , ên ) allows to drop the ên component for the
description of the incoming electromagnetic wave in the far field approximation. Additionally indicated
is one of the two AERA antennas (East-West polarization).

antenna response, the channel data (VNS , VEW ) are transformed to the frequency domain where
the inverse Jones matrix is applied as
   
Êθ −1 V̂NS
=J . (3.7)
Êφ V̂EW
An inverse Fourier transformation results in the time dependent electric field vector of the in-
cident wave E~ θ,φ,n (t) which is then converted to the common coordinate system E ~ x,y,z (t). The
antenna response is modelled with the software NEC2 [143]. The used model is generated
according to the geometry of the RDS and ground effects are taken into account. The VEL is
directly calculated by the software for excitation by a plane wave with a predefined polariza-
tion. An equivalent method is to simulate the antenna in transmission mode as in both cases
the VEL is equivalent. This approach is used for the Butterfly antennas, where a sinusoidal
current source is connected to the terminal of one of the two antennas and a load matching the
input impedance of the LNA is connected to the terminal of the other antenna. The output of
the simulation gives the electric field vector (E~ NEC (θ, φ, f )) for a given frequency which can
be converted to the VEL by

~ 2c ~
H(θ, φ, f ) = i ENEC (θ, φ, f ) . (3.8)
It νf
Thereby, c is the vacuum speed of light, It is the current at the antenna terminal and ν is the
vacuum impedance [111, 144]. The direction of the incoming wave is usually taken as the
direction of the air shower axis which is a good approximation for radiation emitted far away
from the antennas.
The direction of the shower axis however might not be known a priori depending on the re-
construction scheme. If not taken from the reconstruction of another Auger component, it is
3.5. Radio Air Shower Reconstruction 69

calculated from the electric field data. Therefore, an iterative approach can be chosen were the
electric field is first calculated based on an initial guess of the shower axis. Based on this, the
antenna response is unfolded and stations with a signal above a certain signal-to-noise ration
(SNR) are selected. From the timing of the selected stations, the direction of the shower axis
is determined analogue to the procedure described for the SD reconstruction in section 3.4.1.
With this direction information, the antenna response is again unfolded and the shower axis
is recalculated. The procedure is terminated when the direction reconstruction converges to a
final value.
In general, most parameters for higher-level analysis are derived from the electric field traces.
The signal time as well as the signal amplitude are determined from the maximum of the
Hilbert-envelop of the time traces which represents the instantaneous amplitude (see example
in section 6.1.2). Depending on the analysis, the total electric-field trace, one of its compo-
nents, or a projection is used. Another important quantity which is determined based on the
electric-field trace is the energy deposited per area hereafter referred to as the energy den-
sity. This is calculated via the integral over the absolute values of the Poynting vector. In
the reconstruction, the absolute electric-field traces are squared and integrated (summed for
discrete sampling) in a window [t1 , t2 ] around the signal time. Outside of this signal window,
the integral over the squared amplitude of the electric field in a window [t3 , t4 ] represents the
background contribution which is subtracted from the integral over the signal region. The
energy density u in units of eV/m2 is then given as
t2 t4
!
X t
~ i )|2 − tbin 2 − t 1
X
~ i )|2 .
u = ε0 c tbin |E(t |E(t (3.9)
t
t4 − t 3 t
1 3

Where ε0 and c are the vacuum permittivity and the vacuum speed of light and tbin is the size
of a time bin. For a pure white noise background, the uncertainty of the energy density can be
calculated. If the true background is not white noise a more practical approach is to approxi-
mate the uncertainty as the energy density of the noise scaled to the signal size window.
The polarization of the cosmic-ray pulse can be determined from the electric-field data as the
direction of the electric-field vector. The polarization is a key to study the emission mecha-
nisms but it can also be used to deselect false signal stations that have only measured a noise
pulse. This method is applied and discussed in chapter 6.
From the quantities derived on station level for all RDSs, air-shower parameters are recon-
structed on the event level.

3.5.5 Event Level

On the event level, the direction of the shower axis, the two-dimensional LDF, and the wave-
front shape can be reconstructed. The direction reconstruction is performed as discussed above
using either a basic plane wave or a more advanced model, depending on the station mul-
tiplicity and the required precision. The two-dimensional LDF is determined by fitting the
70 Chapter 3. Data Acquisition and Hybrid Shower Reconstruction

amplitudes or energy densities with a modified version of equation 1.26. The details on this
procedure and the obtained air shower parameters which are energy, Xmax , and the shower
core position are presented in detail in the following chapters. In addition, the shape of the
wave front can be used to obtain the core position as well as an estimate of Xmax [95]. Here,
a hyperbolic wave front is used to fit the signal times of the RDSs. The opening angle of the
hyperbola thereby shows a correlation with the position of the emission region.

3.5.6 Station and Event Pre-, Intermediate-, and Post-Selection

At the start of the reconstruction procedure R&D stations as well as stations with known hard-
ware issues are excluded. During the reconstruction further RDSs might be excluded based on
there timing residuals, the shower geometry, or the polarization of the measured electric field
vector. In case of a non-sufficient number of signal stations or a failed reconstruction step the
processing of the event is terminated. A selection of the events based on the quality of the
reconstruction or certain parameter values is performed after the processing.

3.5.7 RdObserver
Despite the variety of possible reconstruction schemes tailored for certain analyses, the RdOb-
server is used as a standard reconstruction scheme for hybrid and multi-hybrid showers. Due to
the hybrid approach, reconstructed shower parameters from the SD reconstruction can be used
as input for the radio reconstruction. Thereby, the antenna model is directly unfolded with the
actual arrival direction of the cosmic ray which saves a significant amount of computing time
for the otherwise necessary iterative direction determination. For a successfully reconstructed
shower, the SD reconstruction as well as the radio reconstruction must be successful. The
FD reconstruction is optional if the event contains the corresponding data. A successful FD
reconstruction is however not required in this case.
At present, the RdObserver is only run on data recorded with the externally triggered stations
of AERA. Merged RD+SD (+FD) files are used as input and three types of output are created.
First, steering files for simulations using the reconstructed parameters such as core position,
energy, and arrival direction are generated. Second, the raw data of the reconstructed showers
is written to a separate file for easier data transfer of the selected showers and the possibil-
ity to reconstruct these showers with a modified module sequence. At last, Advanced Data
Summary Tree (ADST) files which are the standard output files of the Off line framework con-
taining all parameters of the reconstructed showers are created. If the RdObserver is run on
all recorded raw data, the RFI is suppressed by setting the frequency amplitude bins to zero, if
the RdOberserver is re-run on the selected events only, the sine wave fitting suppression of the
Beacon emission lines is used to improve the reconstruction of shower parameters for some
of the later presented analyses. A general overview of the steps in the RdObserver module se-
quence and the used settings is given in the tables 3.2 and 3.3. The complete sequence is given
3.6. AERA Dataset 71

Table 3.2: General overview of the reconstruction steps of the RdObserver.

Detector Reconstruction step

General Event read-in, preselection and quality checks
SD PMT signal reconstruction
Station signal reconstruction
Direction and LDF fit
Horizontal reconstruction if θ > 55◦
Radio Event initialisation, station selection, and signal window setting
Electronic channel response correction
Beacon timing calibration
RFI cleaning
Antenna response correction
Station signal reconstruction
Electric field vector calculation
Top-down station selection and plane fit (loop)
Event selection
Radio LDF fit
FD (try) Calibration and telescope data merging
Coihueco and HEAT data merging
Aperture light calculation
Energy deposit reconstruction
Longitudinal profile fit (Gaisser-Hillas)
Radio Time trace cutting
Radio simulation steering file preparation
General Event file exporting (ROOT file format)
Output file generation

in the appendix A.2. The dataset obtained from the RdObserver is presented in section 3.6. It
is used in the analyses discussed in chapter 7.

3.6 AERA Dataset

AERA has recorded a large radio dataset of extensive air showers. Until beginning of Au-
gust 2015, 6569 air showers were detected by three or more antenna stations equipped with
a ring buffer in coincidence with the SD (RD+SD hybrid showers) and reconstructed with
the RdObserver. The data recorded with the internally trigger stations is not yet considered
in the RdObserver reconstruction. The arrival directions and the angular difference between
the shower axis determined from the SD data and the axis determined from the radio data are
72 Chapter 3. Data Acquisition and Hybrid Shower Reconstruction

Table 3.3: Settings of the reconstruction parameters of the RdObserver.

Parameter Setting
Signal window determined from the SD geometry reconstruction
Noise window 2000 − 5000 ns
Shower geometry taken from the SD reconstruction
Deselected R&D stations 87, 91, 98, 107
Up-sampling factor 4
Signal reconstruction on Hilbert envelop, projection on XY direction
Minimum signal-to-noise ratio 10 - definition: (Signal/RMS)2
Station is lonely no signal station within 400 m or
only one signal station within 800 m
Minimum number of signal stations 3
Maximum deviation RD-SD axis 20◦
Energy density integration window 100 ns

shown in figure 3.8. The distribution has a mean of 1.69◦ and can be fitted with a Rayleigh
function with σ = 0.54◦ . Most of the air showers are detected with 3 RDSs, but there are also
air showers which are detected with more than 40 RDSs. The distribution of signal stations
per shower is shown in figure 3.9 (left). The event rate for the presented dataset is shown in
figure 3.9 (right). During the stable operation of AERA24, about 50 air showers were recorded
with three or more RDSs per month which has increased to about 250 air showers per month
for AERA124 after a short commissioning phase. The energy distribution of the measured
hybrid events is shown in figure 3.10 (left). In addition to the hybrid showers, 370 air show-
ers were simultaneously detected with the Fluorescence Detector (RD+SD+FD multi-hybrid
showers). For these showers, Xmax is within the FD field of view and the measured profile
allows for an FD-based energy reconstruction. The resulting Xmax distribution is shown in
figure 3.10 (right). Air showers in coincidence with the AMIGA muon counters were also
detected. The dataset contains more than 500 RD+SD+AMIGA multi-hybrid showers and
more than 50 RD+SD+FD+AMIGA multi-hybrid showers. The numbers are summarized in
table 3.4. The reconstructed shower parameters obtained by using the RdObserver are used as
basis for the analyses targeting the depth of the shower maximum presented in chapter 7. Also
the results presented in chapter 6 are obtained with a very similar reconstruction sequence
which however slightly differs in details for historical reasons. On top of the quality crite-
ria applied to the data at the different stages of the reconstruction, additional quality cuts are
necessary depending on the individual analysis carried out on the data. These selections are
discussed in the corresponding chapters.
For the analysis targeting the reconstruction of Xmax , the RD-SD-FD multi-hybrid showers are
3.6. AERA Dataset 73

500

400 σ = 0.54 ∘
mean = 1.69 ∘

number of showers
300

200

100

0
0 2 4 6 8 10
∘
angular difference ( )

Figure 3.8: Air showers measured with AERA and reconstructed with the RdObserver. Left: Skyplot
of all events. The shown arrival direction is taken from the SD reconstruction. The direction of the
magnetic field axis is indicated with a star. Right: Distribution of the angular difference between the
air-shower axis reconstructed from the SD and from radio data with a mean of 1.69◦ . The line indicates
a fit of a Rayleigh function with σ = 0.54◦ .

Table 3.4: Air shower statistics for coincident detection with various detector combinations.

Detector combination Number of air showers

RD-SD 6569
RD-SD-FD 370
RD-SD-AMIGA > 500
RD-SD-FD-AMIGA > 50

used for the calibration of radio observables against the Fluorescence Detector measurements.
For these calibration the quality of the FD data is essential and therefore a high reconstruction
quality and a reliable reconstruction of Xmax are required. This is achieved by applying addi-
tional quality cuts corresponding to the ones used for the SD-FD energy calibration [140]. The
most important ones needed to ensure a reliable reconstruction are given in table 3.5. The full
list of used cuts can be found in appendix A.3. In total 105 reconstructed showers fulfil these
criteria.
For the FD based Xmax determination, the proper knowledge of the light scattering in the at-
mosphere is crucial. Therefore extended calibration efforts are made and the properties of the
atmosphere concerning the scattering are incorporated during reconstruction. The databases
for the calibration are filled once per year. Unfortunately this database is at the moment only
filled until the end of 2013. Therefore, individual atmospheric conditions can not be consid-
74 Chapter 3. Data Acquisition and Hybrid Shower Reconstruction

104 400
May 2013 April 2015
350

103 300

shower rate (month-1 )

number of showers

250

102 200

150

101 100

50

100 0
0 5 10 15 20 25 30 35 40 45 50 01/12 01/13 01/14 01/15 01/16
number of signal stations per event month

Figure 3.9: AERA showers reconstructed with the RdObserver. Left: Station multiplicity for all suc-
cessfully reconstructed air showers. The y-coordinate is plotted in a log scale. Right: Shower rate in
showers per month as of January 2012. Until the deployment of AERA124 in May 2013, data were
acquired with AERA24. AERA153 was installed in April 2015.

Table 3.5: Most important quality cuts applied to the FD-SD data of the multi-hybrid event set.

detector cut
FD Xmax in expected field of view
FD σXmax ≤ 40 g/cm2
SD zenith angle ≥ 55◦
SD rec-level ≥ 3
SD minimum Trigger = T5

ered in the reconstruction for most of the RD-SD-FD multi-hybrid showers. These showers
are then reconstructed with a model for the scattering based on the atmospheric measurements
in earlier years. This generic and less accurate description of the atmospheric condition intro-
duces an uncertainty of 6.2 g/cm2 on the reconstructed value of Xmax [145]. This uncertainty
is added to the uncertainty obtained from the FD reconstruction when no information about
atmospheric conditions concerning the light scattering properties is available. This uncertainty
obtained from the FD reconstruction is determined on shower-by-shower basis from the pho-
ton statistics, the shower geometry, and atmospheric conditions. The systematic uncertainty in
the energy range of AERA is 8 − 11 g/cm2 [138].
3.7. Radio Simulations for the AERA Detector Geometry 75

103 80
70
60

number of showers
number of showers

2
10 50
40
30
101
20
10
100 0
16.016.517.017.518.018.519.019.520.0 400 600 800 1000 1200 1400
log10(energy (eV)) Xmax (g/cm2 )

Figure 3.10: Air showers reconstructed with the RdObserver. Left: Energy distribution of the measured
air showers as reconstructed from the SD data. Right: Xmax distribution of the RD+SD+FD multi-
hybrid showers.

3.7 Radio Simulations for the AERA Detector Geometry

For most of the air showers in the AERA124 dataset two simulations were produced using the
measured arrival direction and energy as input parameters. One of the two simulations has a
proton as primary particle, the other one an iron nucleus. The electric field is calculated for
the actual AERA station positions and can in this case directly be read into Off line. There,
the detector description is used to generate ADC time traces from the simulations including
the imprint of the actual channel and antenna responses for the individual stations. This is
achieved by applying first the antenna and then the channel response backwards. At this stage,
noise recorded in the vicinity of the shower (maximum 50 s offset) is added to the simulated
electric field traces. The reconstruction of the simulated data follows the steps outlined in
section 3.5.7 except for the timing calibration which is not needed for simulations. As in the
reconstruction of real data, at least three stations with signal above the SNR threshold are
required. If the number is smaller, the shower is skipped and the reconstruction continues
with the next shower. Stations for which no noise data of the time of the event are available
are skipped as well. This allows accounting for the changing grid configuration when stations
experience downtime. It should be noted that the simulated signals and the noise are added one
to one and no scaling is applied, assuming the simulations give the correct absolute amplitude
in the radio signals. This assumption is encouraged by the results presented in [146]. It is
however in some tension with other observations [147].
For the simulations, CORSIKA 7.400 is used together with the hadronic interaction models
FLUKA 2011.2b [50] and QGSJET-II-04 [148].
76 Chapter 3. Data Acquisition and Hybrid Shower Reconstruction

Off line also provides a similar functionality to generate SD and FD data from air shower
simulations which can be included in the reconstruction. For the presented dataset, simulated
data are not generated for the SD and FD and the dataset is reconstructed only with the radio
part of the RdObserver reconstruction. Simulations set up according to this scheme are used
for calibration purposes in the Xmax analysis in the chapter 7.

3.8 Generic Radio Simulations on a Star-Shape Pattern

In the classical approach of simulating radio emission of extensive air-showers, a shower-core
location is set and the electric-field traces are simulated for fixed observer positions relative
to this location. The observer positions are often chosen with respect to a regular grid on the
ground. In the method introduced in [71], the positions of the observers are arranged in the
plane spanned by ~v × B ~ and ~v × (~v × B)
~ and then projected onto the ground to determine the
input for the simulations. Therefore, the grid on the ground is distorted but once transformed
back into the shower plane, it always shows the same shape. A second innovation is the layout
of the antenna grid in the ~v × B ~ vs. ~v × (~v × B)
~ plane which has a regular star-shape with
eight arms and 20 antenna positions per arm with 25 m spacing. The arms are oriented in
~
such a way that one of them is parallel and one antiparallel to the ~v × B-direction and one
~
is parallel and one antiparallel to the ~v × (~v × B)-direction. In this way, the "bean"-shape
of the lateral distribution pattern is always sampled in the regions where the polarizations of
the geomagnetic and the charge excess emission are parallel/anitparallel and where they are
perpendicular to each other. Thereby, it is guaranteed that the minimum and the maximum
of the density pattern are sampled. Thus, parameters related to the signal strength can be
interpolated. The simulated energy densities on the star shaped grid on the ground and in the
shower plane are displayed in figure 3.11 (top row). The interpolation is performed with a
radial basis function method employing a quintic function. The combination of the star-shape
pattern and the interpolation was found to allow the prediction of the energy density at any
point within the simulated star-shape with a difference of less than 2.5% of the maximum
energy density compared to dedicated simulations at those positions [71]. This method is
used throughout the work presented in this manuscript. The interpolated density pattern is
illustrated in figure 3.11 (bottom). For the simulations, CORSIKA 7.4387 is used together
with the hadronic interaction models FLUKA 2011.2b [50] and QGSJET-II-04 [148]. The
CORSIKA showers are thinned at a level of 10−6 and optimized weight limitation [149] is
used. The magnetic field is chosen according to the site of AERA. The energy density at each
simulated position is determined according to equation 3.9 after filtering the simulated electric
field traces to the frequency range of AERA of 30 − 80 MHz. Simulations on the star-shape
station-grid are used in two different analyses presented in chapter 5 and 7. The properties of
the simulation sets used are discussed in the two chapters individually.
3.8. Generic Radio Simulations on a Star-Shape Pattern 77

10.5 10.5
1000

 )-direction (m)
position in North-direction (m)

9.0 400 9.0

energy density (eV/m2 )

energy density (eV/m2 )
500 7.5 7.5
200
6.0 6.0

position in v ×(v ×B
0 0
4.5 4.5
−200
−500 3.0 3.0
−400
1.5 1.5
−1000 −400 −200 0 200 400
−1000 −500 0 500 1000
position in East-direction (m)  -direction (m)
position in v ×B

10.5

400 9.0
 )-direction (m)

energy density (eV/m2 )

7.5
200

6.0
position in v ×(v ×B

4.5

−200
3.0

−400
1.5

−400 −200 0 200 400 0.0

 -direction (m)
position in v ×B

Figure 3.11: Simulation of the energy density of the radio emission of an extensive air-shower. Top left:
Observer positions at which the electric-field traces are simulated on the ground. Top right: Observer
~ vs. ~v × (~v × B)
positions transformed into the ~v × B ~ plane forming a regular eight-arm star. Bottom:
Spatial interpolation of the simulated energy densities with a radial basis function (background color
map). The simulated air shower has an energy of 4.4 × 1017 eV, the zenith angle is 56◦ , and the azimuth
angle is 3◦ .
 AERA Scintillators
4
Parts of this chapter have been published in:

J. Schulz, J. R. Hörandel, S. Jansen, A. Nelles, C. Timmermans

The Scintillators of the Auger Engineering Radio Array
Auger Internal Publication, GAP2014_122

With the deployment of the second stage of AERA not only 100 new radio detection stations
were deployed, but also 75 small plastic scintillators were included in 40 RDSs. The scintil-
lator modules are placed inside the station electronics compartment on top and underneath the
lead battery (Top-, Bottom-scintillator). The distance between the detector stations is 250 m
around and south of the CRS, whereas the stations further to the East and North are at 375 m
distance from each other. The positions of the AERA scintillators are shown in figure 4.1.
Besides triggering, the particle density as function of the distance to shower axis (in the shower
plane) can be utilized to determine the shower-core position. This is of interest due to the high
dependency of nearly all radio analyses on the shower core position. Especially low-energy
showers, triggering only 3 stations of the Surface Detector array suffer from relatively large
uncertainties on the reconstructed core position. These uncertainties can be of the order of
hundreds of meters in the worst cases.
A fit of the measured particle densities at ground with a NKG lateral distribution function (see
equation 1.23) is used here to determine the core position purely based on the scintillator mea-
surements using a maximum likelihood fit for showers sampled with high station multiplicity.
For future analyses, the scintillator measurements can be included in a general reconstruction
that combines core-sensitive data from the scintillators, the SD, and the radio measurements
to constrain the shower geometry.

4.1 Scintillator Hardware

For a better understanding of the applied reconstruction, a short description of the hardware of
the particle detectors is given in the following.
80 Chapter 4. AERA Scintillators

SD station CRS Scintillation counter

Figure 4.1: Overview of the AERA scintillation counters together with the SD stations and the Central
Radio Station (CRS). The distance between the particle counters is 250 m around and south of the CRS,
whereas further to the East and North the distance is 375 m.

4.1.1 Scintillator Module

The heart of the particle detectors is a 457 × 172 × 24 mm3 plastic scintillator [117]. From the
scintillator, the photons are collected and guided to a SensTech P30CWS12-05 photomultiplier
tube (PMT) by three BCF 91-A optical fibres that are glued into grooves in the scintillator
surface. A technical drawing of the detector unit and the installed setup is shown in figure 4.2.
The PMT base includes a high voltage (HV) generator which uses a 12 V input and a control
voltage around 1 V to supply 300 − 1200 V to the photo cathode. The 12 V are provided from
the charge controller and daisy-chained through the scintillator module(s) to the digitizer. The
high voltage setting is controlled via the same cable which the signals are transmitted through.
This, on the one hand simplifies connections and cabling, on the other hand requires an AC
coupling of the signal introducing positive overshoots after a negative signal.

4.1.2 Scintillator Electronics

For the readout of the PMTs, the same electronic components as for the radio channels are
used except for the analog filters. This provides a very flexible data-acquisition system, in
which a scintillator channel can easily be exchanged for a third radio channel. It also provides
powerful hardware with high sampling rate and good amplitude resolution for the particle de-
tectors. Employing a 14 bit ADC with a resolution of about 13 µV/ADC-unit and 5 ns time
binning, the scintillator pulses are recorded in great detail.
4.1. Scintillator Hardware 81

Figure 4.2: Technical drawing of the particle detector unit. Left: Visible are the plastic scintillator,
the fibre connection as well as the PMT. In addition, parts of the housing are shown together with the
PMT holder and electronics. Right: The combination of Top- and Bottom-scintillator with the lead-gel
battery in between (gray).

4.1.3 Scintillator-Trigger

The trigger condition for the scintillator modules is a simple threshold crossing. After a short
commissioning phase of the AERA124 stations, the trigger threshold for the scintillator mod-
ules was fixed to 150 ADC-units. In section 4.2.2, the distribution of signal pulse amplitudes
and their corresponding integrated signals in units of vertical equivalent muons is presented
verifying the chosen threshold.
Due to the slightly different propagation velocities of ultra-relativistic particles in the front of
the air shower and the emitted electromagnetic radiation, one expects the radio signal to arrive
with a delay at the observer with respect to the particle shower front. The hardware for particle
and radio detection introduces additional time delays. On the one hand, the scintillator, the
fibres, and the PMT have a specific rise time, on the other hand, the low-noise amplifier and
the bandpass filters in the radio chain delay the signals. Different cables and cable lengths
also have an influence. During the reconstruction of the radio signals, the responses of all
components in the electronics chain are taken into account and the corresponding time delays
are corrected for. These corrections are not applied to the scintillator chain, as its components
were not measured individually. Therefore, the reconstructed radio signal usually appears
ahead of the scintillator pulse in the station data. The trigger crossing of the scintillator signal
is set to 2315 ns in the voltage time trace. In figure 4.3, the trigger time is shown as blue line
together with the corresponding time distributions of radio pulses before (red) and after (black)
correcting for the radio channel response. Only radio pulses with an amplitude more than 3.5
times above the full trace RMS are taken into account. To avoid possible biases from a too
narrow signal time window, the period of acceptance is set between 2100 and 2500 ns.
82 Chapter 4. AERA Scintillators

1400

number of peaks
1200 before correcting for the channel response
after correcting for the channel response

1000

800

600

400

200

0
2000 2050 2100 2150 2200 2250 2300 2350 2400 2450 2500
peak time (ns)

Figure 4.3: Distribution of radio pulse times for scintillator triggered signals with amplitudes 3.5 times
above the full trace RMS. The vertical line indicates the time of trigger crossing in the particle detector.
Corresponding to this, peak times of the radio pulses are shown in red (light) before correcting for the
channel response and in black (dark) after the correction. After the correction for the channel response,
the radio signals appear ahead of the particle signal as those are not corrected for the response of the
electronics chain. Figure taken from [145].

4.1.4 Trigger Efficiency

Triggering individual radio stations only based on the detection of particles with the scintilla-
tors allows a precise determination of the signal windows. The efficiency in triggering however
is lower compared to using an external trigger from the SD because of the small size of the
particle detectors (100 times smaller effective area than a SD station). This effect is propor-
tional to the cosine of the zenith angle due to the projection of the surface area. Therefore, the
radio self-trigger was installed in addition to the scintillator trigger.
An absolute value for the efficiency cannot be given because the real number of showers falling
into the instrumented aera of AERA is not known. Nevertheless, a comparison with the number
of air showers detected by the SD Infill gives the relative efficiency compared to an external
trigger. The number of showers detected with the SD together with the number of showers
detected with both, the SD and the scintillators, as a function of the primary energies (left)
and zenith angles (right) are presented in figure 4.4. In addition, the number of radio showers
as reconstructed with the basic RdHybridReconstrution pipeline of the Off line framework is
shown in both plots. The datasets were acquired in 2013 and the considered time periods are
set according to the uptime of AERA (downtimes of less than 6 hours are not accounted for).
A large relative difference for the number of showers with energies below roughly 1017 eV
4.1. Scintillator Hardware 83

104 104
SD Events SD Events
Scint Events Scint Events
RD Events RD Events
103 103

102 102
#

#
101 101

0
1016.0 100 0
16.5 17.0 17.5 18.0 18.5 19.0 10 20 30 40 50 60 70 80
log(SD energy/eV) zenith angle ( ◦ )

Figure 4.4: Number of air showers seen by the SD together with the number of those showers also
seen by the scintillators and in the radio stations as a function of the energy (left) and the zenith angle
(right). For the SD data, two different reconstruction schemes are used, depending on the zenith an-
gle. The standard reconstruction is used up to 55◦ . For larger angles, a special horizontal air shower
reconstruction is used.

detected with the scintillators compared to the SD is observed. The impact on the radio detec-
tion of air showers is however small as 1017 eV is the threshold for the radio detection in 3 or
more stations for the given array configuration (see figure 4.4 left). For increasing energies,
the efficiency of the scintillator array is also increasing almost reaches one at the highest en-
ergies. With increasing zenith angle, both distributions of the particle detectors increase due
to the increasing solid angle up to a maximum around 30◦ . Larger angles lead to a dominant
decrease of the projected detector area and thereby to a decrease in the distribution. The de-
crease of the distribution for the scintillators is steeper as their effective area decreases more
rapidly than the one of the Surface Detector stations which are not flat detectors. Additionally,
an asymmetry between the rise and the fall of the distributions is observed due to the increase
in shower attenuation with growing zenith angles. For inclined showers (zenith angle larger
than 60◦ ), the small effective area of the scintillators combined with the attenuation of the
electromagnetic cascade results in the detection limit for showers with these detectors. The
SD data reconstruction employs two different methods depending on the zenith angle of the
detected air shower. Up to 55◦ , the standard reconstruction as presented in section 3.4.1 is
used. For more inclined air showers, a special reconstruction scheme is applied [150].
The distribution of radio events is roughly centred around 45◦ , but it is expected to continue to
more inclined showers than observed here, as the footprint increases with the distance to the
emission region enabling the detection with a relatively sparse grid.
As the efficiency is depending on energy as well as on the zenith angle, the fraction of showers
detected by the scintillators and the SD compared to the fraction of showers which triggered
84 Chapter 4. AERA Scintillators

90 1.0 1.0
80 0.9 17500 0.9
70
0.8 17000
0.8
0.7 0.7
60
zenith angle ( ◦ )

0.6 16500 0.6

NScint/NSD

NScint/NSD
50
0.5

Y (m)
16000 0.5
40
0.4 0.4
30 15500
0.3 0.3
20
0.2 15000 0.2
10
0.1 0.1
0 16.5 14500
17.0 17.5 18.0 18.5 0.0 26000 25500 25000 24500 0.0
log(SD energy/eV) X (m)

Figure 4.5: Fraction of the showers detected by the SD that also triggered the scintillators. Left:
Fraction in bins of energy and zenith angle. Right: Fraction in bins of X (East) and Y (North) position
of the shower axis position for energies above 1017 eV and zenith angles below 60◦ . The positions of
the radio stations are indicated as dots.

the SD (NScint /NSD ) is plotted as a function of both parameters in figure 4.5 (left). For energies
below 1017 eV, the fraction is smaller than 0.5 increasing to 0.9 − 1.0 for most of the bins
above 1017.6 eV and below 60◦ . Due to the irregular spacing of the radio stations, the effi-
ciency changes over the scintillator array. In figure 4.5 (right) the fraction NScint /NSD is shown
depending on the core position of the showers as determined with the SD for energies above
1017 eV and zenith angles below 60◦ . The fractions for the different array regions show a rela-
tion between the ratio NScint /NSD and the grid size. For the region with a spacing of 250 m, the
efficiency is between 0.5 − 1.0 and decreases for the spacing of 375 m to < 0.1 at the outer
edge.

4.2 Scintillator Calibration

An array of 40 scintillator stations spread over an area of about 3 km2 can be considered as
a cosmic-ray detector on its own. Even though, this was never intended and also the size of
the scintillators (< 0.1 m2 ) makes the array unattractive for standalone detections, valuable
information on the shower geometry as input for analyses of the radio emission of extensive
air showers can be extracted.

4.2.1 Scintillator Signals

The two main observables in case of the scintillators are the signal time and the deposited
energy in the detector. The time can directly be inferred from the signal trace itself according
to the chosen definition. Here, the peak time of the signal is used. For the determination of the
4.2. Scintillator Calibration 85

0.002
0.000
−0.002
signal (V)

−0.004
−0.006
−0.008
−0.010
2000 2200 2400 2600 2800 3000
time (ns)

Figure 4.6: Typical time trace recorded with the scintillation detector. The negative signal is followed
by a positive overshoot due to the AC-coupling of the signal line.

deposited energy, the linear proportionality between energy deposit, number of scintillation
photons, and charge output of the PMT are used. This can be calculated from the measured
signal trace by integrating the output voltage U (t) over the sample time dt and accounting for
the 50 Ω input impedance. Z tf
U (t)
Q= dt (4.1)
ti R
For discrete sampling, this is equal to
nsamples
X Ui ∆t
Q= . (4.2)
i=1
R

Before the integration, the trace is corrected for a possible DC offset and fluctuations on
smaller time scales that could influence the signal reconstruction. As mentioned in sec-
tion 4.1.1, the control voltage for the HV setting on the photocathode is sent via the signal
line. Therefore, the signal line is AC-coupled to the electronics. This leads to a positive over-
shoot following the negative signal due to the recharging of the coupling capacity. A signal
time trace as it is typical for air shower particles traversing the scintillator is shown in fig-
ure 4.6. The signal integration window is defined as the range with negative signal around
the first threshold crossing. For two consecutive signals, the second signal may fall within the
overshoot of the first. These delayed pulses are not considered at present because the signals
were not found to have a common pulse shape which can be parametrized and subtracted from
the trace in order to recover delayed signals.

4.2.2 Detector Calibration

The detectors are calibrated with atmospheric muons. For this purpose, the double layer con-
sisting of Top- and Bottom-scintillator is used in combination with the data acquired in the
random trigger mode. In this mode, every 40000th timestamp received by the central DAQ is
selected and the station data are written to disk. This minimizes the chances to select signals
86 Chapter 4. AERA Scintillators

# events
140

120

100

80

60

40

20

0
0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.6 4.0
charge output (fC)

Figure 4.7: Histogram of charge output of one of the Top-scintillators exposed to single muons. The
distribution is corrected for the impact of electronic noise indicated by the dashed line. The distribution
is fitted by a convolution of a Landau and a Gaussian function and the charge output for 1 VEM is taken
as the peak position of the Landau distribution.

affiliated with air showers which would be strongly influenced by the electromagnetic compo-
nent.
In the next step, the calibration data for one scintillator are selected by taking events with a
trigger crossing in the other scintillator of the station. From the resulting distribution, the back-
ground entries are subtracted. For that purpose, data taken every ten seconds are processed in
the same way as the calibration data and the resulting distribution of the electronic noise is
scaled to the same number of entries in the first bin. The cleaned distribution of the integrated
signals is fitted with a convolution of a Landau and a Gaussian function for every individual
scintillator. The integrated signal corresponding to one VEM is then defined as the peak po-
sition of the Landau distribution. A histogram showing the distribution of integrated single
muon signals in one of the Top-scintillators is shown in figure 4.7. Stations which contain
just one scintillator are not taken into account because it is not possible to select only almost
vertical muons. For theses five stations, the combined distribution of the other 35 stations is
used.
The distributions of the Landau-peak positions for all particle detectors placed on top and un-
derneath the battery are shown in figure 4.8. The values for all Top- and Bottom-scitillators
are similar within 10 − 20%. Between the Top- and the Bottom-scintillator, there is an overall
shift of about 15%. The origin of this discrepancy is most likely the individual HV supplied to
the PMTs. This voltage is software controlled to keep the trigger rate constant under changing
temperature and pressure conditions and a lower HV-setting corresponds to a smaller inte-
4.2. Scintillator Calibration 87

#
25
16

14
20

12

10 15

8
10
6

4
5
2

0 0
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
charge output (fC) charge output (fC)

Figure 4.8: Distribution of the charge output corresponding to 1 VEM for all particle detectors. The
variation between the individual detector units is 16 ± 3% for the scintillators at the bottom (left)
and 12 ± 2% for the detectors placed on top of the battery (right). Note that the scintillators are not
expected to have exactly the same signal output for single muons because of the different individual
characteristics of the PMTs.

grated signals. A possible reason for an increased trigger rate in the Bottom-scintillator is
pre-showering of particles in the lead battery.

Energy Calibration

To cross check the calibration of the vertical equivalent muon and to additionally obtain a cal-
ibration for the deposited energy, a GEANT4 [151] simulation of the scintillator modules is
used. As a simplified model, two scintillator modules are implemented with a spacing given
by the height of the battery. These modules contain the plastic scintillator block surrounded
by styrofoam inside the metal housing. For the battery, a block of lead corresponding to the
estimated amount of lead inside the battery is placed on top of the lower scintillator. In the
real battery, the lead is arranged in vertical sheets but information about the exact construction
details is not available. Therefore, the simplified implementation is chosen.
For the calibration, muons with a cos2 θ zenith-angle distribution and an energy of 4 GeV are
used. Analogue to the data driven calibration, only muons going through the bottom detector
are selected for the top detector calibration and vice versa. The distribution of deposited en-
ergy can be seen in figure 4.9. Note that the plotted range does not include the noise peak.
With 4.6 MeV energy deposit for single almost vertical muons, the simulations agree with the
expected value of 2.02 MeV/cm of traversed scintillator material [152]. The observed differ-
ence between Top- and Bottom-scintillator is not reproduced in the simulations using single
muons. However, detector simulations including also particles of the electromagnetic cascade
show hints for pre-showering in the battery.
After fitting the distribution of simulated single-muon signals, a comparison between the
88 Chapter 4. AERA Scintillators

2000
# events

# events
2200
1800
2000

1800 1600

1600 1400

1400 1200
1200
1000
1000
800
800
600
600
400
400

200 200

0 0
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
deposited energy (MeV) deposited energy (MeV)

Figure 4.9: Distribution of deposited energy from single vertical muons as calculated with a simplified
detector model in GEANT4 for Bottom-scintillators on the left and Top-scintillators on the right. The
distribution is fitted with a Landau function leading to an energy deposit of 4.49 ± 0.01 MeV for the
Bottom- and 4.61 ± 0.01 MeV for the Top-detector at the peak position.

model and the measurements can be made. The measured and the simulated single muon
peak positions give the relation between the integrated signal and the deposited energy. To
account for noise, a Gaussian-distribution is added to the simulation according to the width of
the measured noise distribution in every individual detector. Normalizing both distributions
to an area of one allows to match the measured and the simulated distribution. Both distri-
butions as function of the deposited energy are in good agreement as shown in figure 4.10.
For the Bottom-scintillators, varying deviations between the measurement and the simulation
are seen. Given these issues, the Bottom-scintillators are not used in the reconstruction of air
showers at the moment.
The presented calibration is used to convert integrated measured signals in VEM for all in-
dividual detector units of stations with two scintillators. For the stations containing only a
Top-scintillator the combined calibration of all other Top-scintillators is used in this case. The
results are shown in appendix A.4.

Cross Check of Trigger Settings and Silent Stations

To avoid unnecessary processing of data inside the CPU of the station, the scintillator trigger
is implemented on ADC level as a threshold-crossing trigger. The deposited energy, however,
is reconstructed from the integral of the pulse since it correlates with the charge output of
the PMT. The distribution of signals from all Top-scintillators triggered on Top- or Bottom-
scintillator in the random- and air-shower-trigger mode is shown in figure 4.11. The trigger-
ing on the Bottom-scintillators, allows the consideration of sub-threshold events. The trigger
threshold is set to 150 ADC units, which corresponds to about 2 mV depending on the charac-
teristics of the individual electronic components. With the chosen threshold setting, there are
4.2. Scintillator Calibration 89

normalized # events
0.2

0.18

0.16

0.14

0.12

0.1

0.08

0.06

0.04

0.02

0
0 2 4 6 8 10 12 14 16 18 20
deposited energy (MeV)

Figure 4.10: Histogram of single muon signals converted into the deposited energy in one of the Top-
scintillators using the calibration obtained by GEANT4 simulations. The distribution is corrected for
noise and normalized to an area of 1. The red solid line denotes the fit of a convoluted Landau-Gaussian
function to the data. The blue dotted line shows the fitted distributions of the simulated single-muon
signal combined with a Gaussian distribution according to the measured width and normalized to an
area of 1.

only a few signals below threshold which were reconstructed to have an integrated signal of
1 VEM or more. The exact fraction of these events slightly varies due to the mentioned dif-
ferent electronic components. In figure 4.12, the same plot is shown for data taken every ten
seconds. Here, the distribution of signal height as function of the corresponding reconstructed
number of VEM is plotted, representing the background. For these data, almost all the signals
are reconstructed noise fluctuations which are concentrated below 300 µV and 0.1 VEM with a
tail to higher VEM fractions. A negligible number of signals is found in the expected signal re-
gion, representing particles measured in the periodically-triggered data by chance. Due to this,
the electronic background does not need to be considered in terms of triggering. In addition to
the previous results, the upper limit on the integrated signals in a station which did not trigger
can also be inferred. Therefore, all sub-threshold signals are projected onto the x-axis and the
bin content is summed up beginning with the low VEM values until the sum reaches 95.4%
of the total number of entries. This procedure is repeated for all Top- and Bottom-scintillators
individually and the results are given in table A.7 in the appendix. The mean values for Top-
and Bottom-scintillators are 0.58 VEM and 0.70 VEM. Applying a similar procedure to the
noise distributions reveals that more than 99% of all noise events have an integrated signal of
less than 0.07 VEM giving no sign for a large influence of offsets and fluctuations. As the
background fluctuations are well below 0.5 mV, the trigger threshold could be lowered in or-
90 Chapter 4. AERA Scintillators

0.02
peak height (V)

#
0.018
104
0.016

0.014
103
0.012

0.01

0.008 102

0.006

0.004 10

0.002

0 1
0 1 2 3 4 5 6 7 8 9 10
VEM

Figure 4.11: Distribution of the peak height and the corresponding integrated signal in vertical equiva-
lent muon (VEM) for all Top-scintilltors. The data were taken in random- and air-shower-trigger mode.
Per station a trigger in the Top- or Bottom-scintillator was required to also count sub-threshold events.
The red solid line indicates the trigger-threshold of about 2 mV. Below the threshold, the number of
events with signals corresponding to more than 1 VEM is negligible.

der to increase the sensitivity. Nevertheless, the higher threshold is justified as only very few
signals above 1 VEM stay undetected and the trigger rate can be reduced.

4.2.3 Lateral Distribution Function

In order to reconstruct the core position of the air shower, the lateral distribution of the parti-
cles with respect to the shower axis is used. For the fit, the NKG function introduced in equa-
tion 1.23 is employed as lateral distribution function (LDF). Together with the two coordinates
of the core position encoded in the radial distance r to the shower axis, the parametrization
has five free parameters. The number of scintillators with a signal above threshold is normally
not large enough to fit 5 free parameters and even if, the fitting procedure is rather unstable
for these cases. Therefore, parameters are fixed as it is also done in other experiments, e.g.,
KASCADE [153]. To fit the core position, X, Y and Nch are free parameters and R as well as
s are fixed in the reconstruction presented here.

LDF Fit Parameters

To find appropriate values to fix the parameters, simulations are used. More than 400 showers
simulated with the CORSIKA simulation package are used in combination with the GEANT4
model of the scintillators sketched in section 4.2.2. To generate a lateral distribution from
4.2. Scintillator Calibration 91

peak height (V)

#
0.002 104

0.0018

0.0016
103
0.0014

0.0012

0.001 102

0.0008

0.0006
10
0.0004

0.0002

1
0 0.1 0.2 0.3 0.4 0.5 0.6
VEM

Figure 4.12: Distribution of the pulse height and the corresponding integrated signal in vertical equiva-
lent muon (VEM) for all Top-scintillators. The data were taken in the 10 second trigger mode recording
mostly electronic noise. The red line indicates the threshold of about 2 mV. The majority of the entries
is concentrated below 300 µV and 0.1 VEM with a tail to higher VEM fractions.

the CORSIKA output, the particles are binned in rings of 5 m radial thickness in the shower
plane. All particles are shot onto an area underneath the detector which is slightly bigger than
the plastic scintillator. this procedure takes effects of edge cutting particles into account. The
original directions of the shower particles are preserved. The deposited energy is weighted
according to the particle weight assigned by CORSIKA and normalized to the area of the ring
and the impact area. The resulting histograms are fitted with equation 1.23 and an example
is shown in figure 4.13. Note that the distribution is fitted only in the range from 2 − 600 m
as the asymptotic behaviour very close to the axis is unphysical and the function is known to
deviate from the real particle distribution for larger distances to the axis. The distribution of
the parameters from all fits can be used to fix the free parameters making also fits of measured
data with only a handful of data points on a shower-by-shower basis possible. For R and s
these distributions are shown in figure 4.14. Inclined showers with zenith angles above 45◦
show a large spread for both parameters. This might be due to the fact that the electromagnetic
part of those showers dies out before reaching the detector leading to partly unphysical fit
results. Therefore, a cut on the zenith angle of 45◦ is applied to the simulations. From both
distributions, the mean values are taken as fixed values for the corresponding parameters:
R = 34.1±2.9 m and s = 1.70±0.02 . In contrast to the theoretically calculated characteristic
radius for purely electromagnetic showers which is of the order of 80 m, for hadronic showers
a value of R around 30 − 40 m and a fixed shower age have been proven to be suitable fit
conditions for various experiments [153, 154] and simulations.
92 Chapter 4. AERA Scintillators

105

particle density (VEM/m2)

104

103

102

10

0 50 100 150 200 250 300 350 400 450 500

distance to shower axis (m)

Figure 4.13: Simulated particle density as a function of the distance to the shower axis using a simpli-
fied model of the detector. The particles were generated with CORSIKA and the number of particles
in VEM/m2 per distance bin is inferred via the deposited energy given by a GEANT4 simulation and
the detector calibration. The line indicates a fit with the NKG function (equation 1.23) in the range of
2 − 600 m.

4.3 Reconstruction of Measured Air Showers

By applying the methods discussed in the previous section to data acquired with the particle
detectors of AERA, the core position of air showers is reconstructed. The reconstruction of the
arrival direction is analogous to the one for the SD. In the following sections, the used dataset
is described and the reconstructed shower parameters are presented.

4.3.1 Dataset and Reconstruction

The data used were recorded with AERA124 from May 2013 until November 2014. During
this time, data were acquired with different trigger conditions. Among those were the require-
ments to have at least 3 triggered scintillator stations, or 2 triggered scintillator stations and
one radio self-trigger. In addition, variations of the acceptance for Bottom-scintillator triggers
occur in the data.
In a basic coincidence check with the SD data, only scintillator events are selected which have
an SD counterpart close to AERA within one second. The data are merged and reconstructed
with the Off line framework in a modified version of the example application RdHybridRecon-
struction. The used modules for the scintillator reconstruction can be seen in the following
whereas the SD part of the reconstruction stays unchanged:
4.3. Reconstruction of Measured Air Showers 93

250 2.2
54.00 54.00

200 48.00 2.0 48.00

characteristic radius (m)

42.00 42.00

zenith angle ( ◦ )
zenith angle ( ◦ )
150 36.00 1.8 36.00

shower age
30.00 30.00
100 1.6
24.00 24.00

18.00 18.00
50 1.4
12.00 12.00

6.00 6.00
0 1.2 8
108 109 1010 10 109 1010
energy (GeV) energy (GeV)

Figure 4.14: Fit parameters for the characteristic radius R (left) and the shower age s (right) for all
simulated showers according to equation 1.23. The resulting mean values after applying a cut on the
zenith angle of 45◦ are R = 34.1 ± 2.9 m and s = 1.70 ± 0.02 .

<module> EventFileReaderOG </module>

<module> RdEventPreSelector </module>
<module> RdEventInitializer </module>
<module> RdChannelADCToVoltageConverter </module>
<module> RdChannelSelector </module>
<module> RdChannelPedestalRemover </module>
<module> RdScintSignalReconstructor </module>
<module> RdScintPlaneFit </module>

In case of triggering on Top- or Bottom-scintillators, the signal in the Top-scintillator could

be below threshold and therefore a cut on the minimum signal height in the Top-scintillator
of 2 mV is applied. After the determination of the integrated signal of the PMTs, the values
are directly converted into units of VEM employing the calibration values obtained in sec-
tion 4.2.2. In total, 23208 showers have been successfully reconstructed with this method.
For the reconstruction of the shower core position, the results from analysing simulated show-
ers in section 4.2.3 are transferred to the data analysis. Before fitting individual showers, a
few selection criteria on shower and station basis are applied. As for the simulations, a cut on
the zenith angles reconstructed with SD of 45◦ is applied. Due to ambiguities concerning the
scintillator direction reconstruction outlined in section 4.3.3, a cut on the difference in arrival
direction between scintillators and SD is not applied in this analysis. To ensure reliable results
from SD, a successful LDF reconstruction (RecLevel ≥ 3) and a 6T5-trigger are required.
For the Top-scintillators, a minimum number of 4 stations with a measured signal of 1 VEM
or more is necessary to consider the shower. In total 3908 showers are selected with these
criteria. The number of cut showers for individual reasons is shown in table 4.1.
94 Chapter 4. AERA Scintillators

Table 4.1: Gerneral cuts on the used AERA scintillator dataset. The cuts select 3908 of the 23208
successfully reconstructed coincident showers.

cut number of showers after cut

SD Quality (6T5 + RecLevel ≥3) 17455
SD zenith angle < 45◦ 15836
Number Top-scintillators with signal > 4 3908

4.3.2 Measured Particle LDF

The LDF of the SD differs from the one of the scintillators due to the different detector re-
sponses. The profiles of the measured signals in both detectors as function of distance to the
SD shower axis and the SD energy are shown in figure 4.15. Water Cherenkov detectors are
sensitive to charged particles with energies above the Cherenkov threshold and photons pro-
ducing those charged particles in the detector. In comparison, scintillators are sensitive to
charged particles that are stopped in the detector and also to minimum ionizing through-going
particles. Therefore, close to the shower axis, the induced signal is higher in the scintillators,
as there are many low-energy charged particles that deposit a large fraction of their energy
in the detector. Further away from the axis, the muons dominate the particle content in the
shower equalizing the measured signals [155]. Unfortunately, the small size of the scintillators
does not allow measurements at low particle densities i.e., large distances. Also visible are the
different behaviours of the detectors for different energies. Here, the water Cherenkov signals
spread out with higher energy and larger distance, whereas the scintillators show a reversed
pattern even though it is less pronounced due to low statistics.
For the case of reconstructing individual showers, it turns out that the number of showers with
high enough station multiplicity to properly fit rM is rather small and therefore, the character-
istic radius is fixed together with the shower age to the values of R = 34.1 m and s = 1.70
calculated based on simulations (see section 4.2.3). With these settings, the fit can be per-
formed with only four stations maximizing the number of reconstructable showers. It should
also be noted that the LDF shape changes with energy as shown in figure 4.15, which is not
accounted for in the fixed parameter fit.

Maximum Likelihood Function for the LDF Fit

To determine the LDF, equation 1.23 is fitted to the data with fixed R and s in a maximum
likelihood approach analogue to the SD reconstruction [136]. The likelihood function includes
the contributions of small signal stations (fsignal ) and sub-threshold stations (fsilent ) in
Y Y
L= fsignal (ni , µi ) fsilent (ni , µi ). (4.3)
i i
4.3. Reconstruction of Measured Air Showers 95

103 103

1.00 1.00
particle density (VEM/m2 )

particle density (VEM/m2 )

102
0.80 0.80

energy (Eev)

energy (Eev)
102

0.60 0.60
101

0.40 0.40

100 0 50 100 150 200 250 300 350 400 101 0 50 100 150 200 250 300 350 400
distance to SD shower axis (m) distance to SD shower axis (m)

Figure 4.15: Particle density profiles as a function of the distance to the SD shower axis for SD (left)
and Top-scintillators (right) for different shower energies (color code). The two plots show the different
detector responses for the water Cherenkov detectors and the plastic scintillators. The measured signal
in the scintillation detectors is higher, as they also measure a signal from low energetic particles below
the Cherenkov threshold. These low-energetic particles are mostly concentrated around the shower
axis.

Here, the product is calculated over station i at position (x, y) where ni is the number of de-
tected particles and µi is the expectation value from equation 1.23. The log-likelihood function
is then
X X
`= ln fsignal (ni , µi ) + ln fsilent (ni , µi ), (4.4)
i i

respectively.
Poissonian statistics is employed to calculate the likelihood contributions for signal stations
with
ni
X
ln fsignal (ni , µi ) = ni ln µi − µi − ln j. (4.5)
j=1

For non-triggered stations below a threshold of 1 VEM, the contribution is:

ln fsilent (1, µi ) = −µi + ln(1 + µi ). (4.6)

In order to include non-triggered stations into the reconstruction, the precise status of all sta-
tions at the time of the shower is necessary to exclude non operational detectors. The technical
framework in form of a bad-period database is work in progress. During the period of data
96 Chapter 4. AERA Scintillators

20

particle signal (VEM)

SD core
Scint core
10 Scint COG

5
4
3
2

4 10-1
3 10-1
2 10-1

10-1
200 300 400 500 600 700 800
distance to shower axis (m)

Figure 4.16: Lateral particle distribution in vertical equivalent muon (VEM) per detector of an individ-
ual air shower measured with the AERA scintillators. The line and points in black, are plotted using the
SD core position. The red dotted line and triangles are plotted using the scintillator core position. The
signals are fitted with equation 1.23 and with fixed shower age and characteristic radius in both cases.
The purple circles indicate the distance to the scintillator center of gravity (COG) used as start values of
the core position for the LDF fit. The distance between the SD core position and the found scintillator
core position is 41 m.

acquisition, a significant number of stations have experienced downtime. Therefore, the fol-
lowing LDF fits will currently not include sub-threshold station contributions.

LDF Fits

In the approach presented here, the shower core position in (x, y) coordinates and the number
of charged particles Nch are fitted by maximizing the log likelihood function. Additionally, it
was tried to include the shower-age parameter as a free parameter in an iterative procedure for
high station-multiplicity events. However, no stable fit process could be established for this
procedure and therefore the shower age is kept fixed as well. As start value for the core posi-
tion, the reconstructed center of gravity of the Top-scintillator stations is used. The direction
of the shower axis is taken from SD. An example of a fitted shower is shown in figure 4.16.
The distribution of the distances between SD and scintillator core for all 2012 successfully
fitted showers is shown in figure 4.17. Here, a distinction is made between showers where the
reconstructed core is contained within the signal stations (dashed red, 410 events) and events
where the core is not contained (blue, 1602 events). For not contained events, the core recon-
struction can suffer from a weakly constrained geometry [101]. The fit does not converge for
1896 events. As the quality of the fit does not provide information about the absolute quality
4.3. Reconstruction of Measured Air Showers 97

entries
core contained

70 core uncontained

60

50

40

30

20

10

0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
distance SD-core Scint-core (m)

Figure 4.17: Distribution of the distances between the reconstructed SD and scintillator core-position
for contained (dashed red) and uncontained (blue) showers.

of the core reconstruction, this has to be evaluated in future work by using air shower simu-
lations and a detector simulation of the whole scintillator array. Far away from the axis, the
particle density in the shower gets smaller than one particle per effective scintillator area and
the chance increases that detected particles are not associated with the shower. Therefore, it
is helpful to introduce a cut at the distance where the expected number of particles is one per
scintillator. This distance is estimated as function of the energy based on the simulations used
in section 4.2.3. The corresponding plot is shown on the left side of figure 4.18. The red line
indicates a quadratic fit to the data shifted upwards by +150 m. This shift accounts for the fact
that the SD core is used to calculate the distance of the stations during the reconstruction. On
the right side of figure 4.18, the resulting distances between the reconstructed SD and scintil-
lator core are shown. The number of events decreases as more events fail the four station cut
(552 uncontained, 185 contained). The width of the distributions also decreases significantly.
As for the reconstruction without a distance cut, the quality of the core reconstruction can
however only be determined from detailed simulations.

4.3.3 Direction Reconstruction

The arrival direction of the air showers is reconstructed using the timing information of the
signals. In a simple approach, a plane wave is fitted to the peak times of the minimum signal
in the first negative peak below threshold. The reconstruction itself is analogue to the direction
reconstruction for Surface Detector array discussed in section 3.4.1. In this analysis a plane-fit
is chosen to maximize the number of reconstructed events as it only requires three stations
detecting particles. The reconstructed directions are compared to those measured by the SD
98 Chapter 4. AERA Scintillators

entries
40.00 core contained
1400
50 core uncontained
35.00
1200
30.00 40
1000

zenith angle ( ◦ )
25.00
r1VEM (m)

30
800
20.00

600 15.00
20

400 10.00 10

200 5.00
0
8.0 8.5 9.0 9.5 10.0 0 200 400 600 800 1000 1200 1400 1600 1800 2000
log10(Energy) (log10(GeV)) distance SD-core Scint-core (m)

Figure 4.18: On the left site, the distance from the shower axis at which the particle density is equal
to one particle per scintillator is shown as a function of the shower energy. The distance is determined
from simulations. The red line indicates a quadratic fit to the data offset by +150 m taken as a quality
cut on the distance. The right plot shows the corresponding distribution of the distances between the
reconstructed SD and scintillator core for contained (dashed red) and uncontained (blue) showers after
applying the distance cut.

and the resulting differences are displayed in the histogram in figure 4.19 (black dotted line).
The distribution of the angular differences has a rather long tail which differs from a Rayleigh-
like distribution. To check whether this is due to stations which were coincidentally triggered
by single particles unrelated to the shower, the direction is fitted for various selections of
stations. All distributions have a very similar shape and none of the selected cases improves
the agreement to the SD direction significantly. Using the station trigger time instead of the
peak time does also not alter the result. The seen differences are not dominantly caused by the
GPS system as analysing the radio direction reconstruction employing the same GPS system,
results in less deviation from the SD measurements. The small size of the scintillation detector
is a likely reason as it only allows for a coarse sampling of the shower particles and thereby
leads to fluctuations in the timing of the signals.
4.3. Reconstruction of Measured Air Showers 99

entries

all scint with signal

1400 3 scint with highest signal
3 closest scint to SD core
3 closest scint to RD barycenter
1200
scint with initial residual < 220ns
Rayleigh distribution with σ = 4.5
1000

800

600

400

200

0
0 5 10 15 20 25 30 35 40 45
angular difference Scint-SD (°)

Figure 4.19: Distribution of the angular difference between the fitted direction of the shower axis
derived from the SD and the scintillator data. The green dotted line shows the distribution based only
on the plane-fit with the three highest signal stations, the blue dash-dotted line represents the fit with the
three stations closest to the SD core position and the red dash-dotted line represents for the fit utilizing
the three stations closest to the RD Barycenter. For the distribution shown with the dashed cyan line,
a cut on the time residuals of 220 ns with respect to the initial plane-fit is made and the directions are
refitted. The solid line indicates the fit of a Rayleigh function with σ = 0.45.
 The Radio Lateral Distribution Function
5
Parts of this chapter are based on the publication:

A. Nelles, J. Schulz, J.R. Hörandel

A parameterization for the radio signal measured with AERA
Auger Internal Publication, GAP2014_073 [156]

The shape of the lateral distribution of the radio signals differs from the classical particle
lateral distribution function. This is due to the superposition of radiation originating from the
two different emission mechanisms as discussed in chapter 1. As a result of the interplay of
the two mechanisms, the general shape of the radio footprint depends on the geometry which
challenges modelling the data as well as the production of simulation libraries. This issue can
be overcome by transforming the coordinate system into the ~v × B ~ and ~v × (~v × B)
~ plane,
see section 1.2.4 and 3.8. In this coordinate system, the lateral distributions have a common
overall shape for all geometries. Thereby, a general two-dimensional parametrization becomes
feasible. Besides that, the approach is used to generate air-shower simulations in a generic way.
In this chapter, the two-dimensional lateral distribution function is discussed, focussing on the
application to AERA data and the correlation of function parameters with air-shower/cosmic-
ray properties.

5.1 LDF Parametrization for AERA

From the shape of the density profiles generated with the air shower simulations described in
section 3.8, the general two-dimensional function has been deduced as

−(~r − r~+ )2 −(~r − r~− )2

   
u(~r) = A+ · exp 2
− A− · exp 2
, (5.1)
σ+ σ−

where A± are amplitude scaling parameters (A+ > A− ), ~r± are location parameters, and σ±
are width parameters (rewritten from equation 1.26). To check the usability of the parametriza-
tion as fit function in an air-shower reconstruction, the function was first tested on LOFAR data
as described in [96]. For the application, the number of free parameters was reduced to five
by determining correlations between parameters based on the underlying simulation dataset.
For the application of the function to AERA data, these correlations had to be redetermined on
102 Chapter 5. The Radio Lateral Distribution Function

Table 5.1: LDF parameters C0 - C4 of equation 5.2 for the corresponding zenith-angle range.

zenith angle C0 C1 (m) C2 (m) C3 (m) C4 (m−1 )

0◦ − 60◦ - - - 16.25 0.0079
0◦ − 10◦ 0.41 −8.0 ± 0.3 21.2 ± 0.4 - -
10◦ − 20◦ 0.41 −10.0 ± 0.4 23.1 ± 0.4 - -
20◦ − 30◦ 0.41 −12.0 ± 0.3 25.5 ± 0.3 - -
30◦ − 40◦ 0.41 −20.0 ± 0.4 32.0 ± 0.6 - -
40◦ − 50◦ 0.46 −25.1 ± 0.9 34.5 ± 0.7 - -
50◦ − 60◦ 0.71 −27.3 ± 1.0 9.8 ± 1.5 - -

simulations which include the height above see level and the direction and strength of the mag-
netic field at the AERA site. It is further desirable to decrease the number of free parameters
in order to obtain a stable fit even for a low station multiplicity. Both aspects are addressed in
work presented in [156]. After additional modifications, the AERA radio LDF is parametrized
in the shower-plane by
" #
−(~r + C1 ~e~v×B~ − ~rcore )2 −(~r + C2 ~e~v×B~ − ~rcore )2
  
u(~r) = A exp − C0 exp . (5.2)
σ2 (C3 eC4 σ )2

Here, A is the amplitude, σ is the width of the distribution, ~rcore is the position of the shower
core and C1−4 are the constants arising from the simulation-based parameter reduction. The
five parameters are given in table 5.1. Three of them are zenith-angle dependent and their
values are given for bins of 10◦ . The ratio of the amplitudes of the two Gaussians C0 can also
be described as a function of the squared sine of the geomagnetic angle for zenith angles up to
50◦ as

C0 = 0.337 × sin2 (α)−0.218 . (5.3)

As the energy in the geomagnetic emission scales with sin2 (α) this is a physically motivated
parametrization. The fit quality however does in general not improve in this case, most likely
because the interplay between the two Gaussian functions is not only determined by the emis-
sion mechanisms but also accounts for relativistic beaming effects.
An example of the parametrization fitted to the simulated energy-density pattern is shown in
~
figure 5.1 in the shower plane and in a one-dimensional projection of all points onto the ~v × B
axis. The fit re-samples the asymmetry of the energy-density pattern and also the overall am-
plitude and fall off. The dip in the pattern due to the destructive interference of geomagnetic
and charge excess emission is overestimated in this case due to the binned C-parameters. All
together the distribution is sufficiently described and the method marks a significant improve-
ment when compared to one-dimensional approaches.
5.2. Relation between LDF Parameters and Air-Shower Properties 103

12
10.5
simulation
AERA LDF fit
400 9.0 10
 )-direction (m)

energy density (eV/m2 )

energy density (eV/m2 )
200 7.5
8

6.0
position in v ×(v ×B

0 6

4.5

−200 4
3.0

2
−400
1.5

−400 −200 0 200 400 0

0.0 −400 −200 0 200 400
 -direction (m)
position in v ×B  -direction (m)
position in v ×B

Figure 5.1: Application of the parametrized LDF used in AERA (equation 5.2) to a star-shape sim-
ulation. Left: Two-dimensional representation of the fitted LDF (background color) to the simulated
~ axis for
energy-density pattern (circles). Right: Simulated and fitted values projected onto the ~v × B
~ axis.
different slices along the ~v × (~v × B)

5.2 Relation between LDF Parameters and Air-Shower Prop-

erties
From the free fit parameters in the formulation of equation 5.2, air-shower parameters can be
directly inferred. An alternative is the cross calibration with values obtained with a different
detection technique or if this is not possible with values obtained from simulations.
The shower-core position is directly accessible from the fit result. The width of the energy-
density distribution is determined by the beaming of the radiation from the emission region to
the ground. Therefore, the width of the LDF σ is correlated with the distance to the emission
region. This distance is approximated by the distance to Xmax , defined as

Dmax = Xtotal / cos(θ) − Xmax , (5.4)

where Xtotal is the total thickness of the atmosphere above the Pierre Auger Observatory and θ
is the zenith angle of the arrival direction. A simple sketch of the effect is shown in figure 5.2.
In earlier work [96, 157], the amplitude parameter C0 (or A+ in the original equation 5.1) was
found to be correlated with the primary energy when corrected with sin2 (α) (the corrected
parameter will be denoted by an asterisk in the following).

5.2.1 Air-Shower Simulation Set

To determine the resolution of reconstructed air shower parameters based on the two-dimensional
LDF in a general way and also for the case of AERA, an extended set of simulations is used.
In total 3150 simulated air showers with different incoming directions, energies, and Xmax
104 Chapter 5. The Radio Lateral Distribution Function

distance to Xmax
footprint width footprint width

Figure 5.2: Sketch of the relation between the width of the footprint of the radio signal and the distance
to the shower maximum.

values are used. The corresponding distributions are shown in figure 5.3. 896 simulations
were performed for iron nuclei and 2254 initiated by protons. All simulated air showers have
a geomagnetic angle of more than 10◦ . To study the resolution in general and for the case of
the detector geometry of AERA, various analyses are performed for energy, core position, and
Xmax and presented in the following subsections.

5.2.2 Energy

The corrected fit parameter A∗+ is correlated with the simulated primary energy squared. The
distribution as well as the resolution is shown in figure 5.4. In contrast to earlier work, the
spread around the fit is asymmetric for the given simulation set and the resolution shows a
rather long tail on one side. This discrepancy is most likely due to a shortcoming of the usage
of A∗+ as an energy estimator. Depending on the geometry, e.g., the distance to the emission
region, the emission pattern can be spread over a larger or smaller area due to projection effects.
Therefore, two air showers with the same primary energy and the same arrival direction will
show a spread in A∗+ because of the fluctuations in Xmax which correspond to fluctuations
of the distance to the emission region. An extreme example is shown in figure 5.5 where a
proton and an iron simulation, for the same primary energy and incoming direction give almost
a factor of 10 difference in the maximum energy-density A+ . Depending on the simulated
shower geometries, primaries, and energies, this effect has more or less impact on the energy
resolution. To improve the energy estimation it is necessary to take the amplitude of the energy
density as well as the width of the pattern into account. Therefore, an improved estimator is
5.2. Relation between LDF Parameters and Air-Shower Properties 105

90°
1100
135° 45°
1000
60 ◦
40 ◦ 900

Xmax (g/cm2 )
20 ◦
800
180° 0°

700

600
225° 315°
500
17.0 17.2 17.4 17.6 17.8 18.0 18.2 18.4 18.6
270° log10(energy (eV))

Figure 5.3: Distribution of parameters in the simulated dataset. Left: Skyplot of the incoming direction
of the simulated showers. The star denotes the direction of the geomagnetic field. Right: Xmax values
as function of the simulated cosmic-ray energy in the set.
number of simulated showers
log(A /eV)

4
350
*

3.5
300
3
250
2.5
200
2
150
1.5

1 100

0.5 50

0 0
17 17.2 17.4 17.6 17.8 18 18.2 18.4 18.6 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
log(E /eV) (E -ERD)/E
sim sim sim

Figure 5.4: Correlation between A∗+ and the simulated primary energy. Left: Linear fit to the distribu-
tion with a slope of 2.14±0.02. Right: Relative difference between the reconstructed energy based on
A∗+ and the simulated primary energy. The distribution has a RMS of 30.2% and a mean of −3.1%.

introduced which is the spatial integral over the distribution of the energy density u(~r),
Z
Erad = u(~r) d2~r . (5.5)

The integrated energy density is also a well motivated quantity as it is the total energy radiated
by the air-shower particles. This quantity, when corrected for the geomagnetic angle, is ex-
pected to be proportional to the primary energy of the cosmic ray as Erad is expected to scale
quadratically with the number of electrons and positrons in the shower due to the coherent
emission. For a parametrization-independent determination of the total radiated energy, the
interpolation of the simulated energy density on the star-shape antenna pattern is used. The
interpolated profile is numerically integrated on a circle with a radius of 500 m which corre-
sponds to the extend of the star-shape pattern. This energy in the radiation is taken as the
106 Chapter 5. The Radio Lateral Distribution Function

80
9.0
400 400 70
 )-direction (m)

 )-direction (m)
7.5
60

energy density (eV/m2 )

energy density (eV/m2 )

200 200
6.0 50
position in v ×(v ×B

position in v ×(v ×B
0 0 40
4.5

30
−200 −200
3.0
20

−400 1.5 −400

10

−400 −200 0 200 400 0.0 −400 −200 0 200 400 0

 -direction (m)
position in v ×B  -direction (m)
position in v ×B

Figure 5.5: Simulations of the energy-density pattern of two extensive air showers with the same
primary energy (0.44 EeV) and arrival direction (zenith angle: 55.7◦ azimuth angle: 5.9◦ ). Left: Sim-
ulation with an Xmax value of 592 g/cm2 (iron). Right: Simulation with an Xmax value of 1221 g/cm2
(proton). The color scales differ by almost a factor of 10.

reference radiation energy of the following studies. When corrected for the geomagnetic an-
gle (Sradio = Erad / sin2 (α)), the correlation with the primary energy shows a small spread of
9.1% taken as the width of the distribution of relative differences for the used simulation set
as shown in figure 5.6 (top row). The resolution has an intrinsic limit due to shower-to-shower
fluctuations and second-order geometric effects. Determining the resolution only for showers
initiated by iron nuclei yields a value of 6.0%. As air showers initiated by iron nuclei show
in general significantly less intrinsic shower-to-shower fluctuations the contributions of these
fluctuations is estimated to be comparable to the ones from geometric effects.
For the application to experimental data, the resolution of the cosmic-ray energy determination
procedure using the integral over the energy-density distribution also depends on the accuracy
with which the real energy density profile can be described by the LDF. Fitting the original
equation 5.1 and the AERA specific parametrization in form of equation 5.2 to the star-shape
simulations shows the impact of the simplified description on the resolution. For the original
LDF description, the correlation between Sradio and the primary energy for the same shower
simulations as used before is displayed in figure 5.6 (center row). Fitting the energy density
distribution with the empirical parametrization in equation 5.1 introduces an additional scat-
ter of 4% on the energy resolution. The reduction of fit parameters in the binned approach
does not lead to a decrease in resolution compared to the full parameter fit. Even though the
shower-by-shower determination of Erad reveals differences arising from the two parametriza-
tions, the overall resolution in terms of the energy reconstruction is almost the same in both
cases as shown in figure 5.6 (bottom row) for the case of using the AERA parametrization.

In the next step, the influence of a sparse sampling of the energy-density profile on the ground
5.2. Relation between LDF Parameters and Air-Shower Properties 107

number of simulated showers

/eV)

700
8.5
radio

600
log(S

8
7.5 500
7
400
6.5
6 300

5.5 200
5
100
4.5

4 0
17 17.2 17.4 17.6 17.8 18 18.2 18.4 18.6 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
log(E /eV) (E -ERD)/E
sim sim sim

9 500

number of simulated showers

/eV)

8.5
radio
log(S

8 400
7.5

7 300
6.5
6 200
5.5
5 100
4.5
4 0
17 17.2 17.4 17.6 17.8 18 18.2 18.4 18.6 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
log(E /eV) (E -ERD)/E
sim sim sim

9
number of simulated showers
/eV)

450
8.5
radio

400
log(S

8
7.5 350

7 300

6.5 250

6 200

5.5 150

5 100

4.5 50

4 0
17 17.2 17.4 17.6 17.8 18 18.2 18.4 18.6 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
log(E /eV) (E -ERD)/E
sim sim sim

Figure 5.6: Correlations between Sradio based on the integral over the different versions of the LDF and
the simulated primary energy. Left: Linear fit to the distribution. Right: Relative difference between the
reconstructed energy based on Sradio and the simulated primary-particle energy. Top row: Numerical
integration over the interpolated energy-densities of the star-shape simulations. Center row: Integration
over the full parameter LDF given in equation 5.1. Bottom row: Integration over the LDF as adapted
for AERA given in equation 5.2. The fit results as well as the obtained resolutions are listed in table 5.2.

as it is the case for an experimental setup is analysed. The sampling of the distribution depends
on the grid of antenna stations as well as the core position and the arrival direction of the air
shower. As it is computationally extremely expensive to re-simulate every air shower in the set
with different core positions, a generic method is employed using the star-shape simulations.
For every simulated air shower in the set, the AERA124 (and also AERA24 only) antenna-
station positions are transformed into the ~v × B ~ vs. ~v × (~v × B)
~ plane. In this coordinate
108 Chapter 5. The Radio Lateral Distribution Function

1000
generated core positions generated core positions
800 AERA stations 1500 AERA stations
 )-direction (m)

 )-direction (m)
simulated antenna positions simulated antenna positions
600 1000

400 500
position in v ×(v ×B

position in v ×(v ×B
200 0

0 −500

−200 −1000

−400 −1500

−600
−1000 −500 0 500 −3000 −2000 −1000 0 1000 2000
 -direction (m)
position in v ×B  -direction (m)
position in v ×B

Figure 5.7: Example of the generated core positions for which a star-shape simulation is evaluated
at the AERA antenna positions to test the influence of sampling on the LDF fit. Left: Core positions
generated around the AERA24 RDS positions. Right: Core positions generated around the AERA124
RDS positions.

Table 5.2: Fitted slopes and corresponding resolutions of the correlations between the integrated energy
density patterns and the primary particle energy using different parametrizations as shown in figure 5.6.
In addition the values for the correlation for the usage of the corrected amplitude parameter A∗+ (see
figure 5.4) are added.

used parametrization slope resolution (%)

A∗+ parameter 2.14 ± 0.02 30.2 (RMS)
interpolation 2.02 ± 0.01 9.1 ± 0.1
LDF original (equation 5.1) 2.00 ± 0.01 13.2 ± 0.2
LDF AERA (equation 5.2) 2.00 ± 0.01 13.3 ± 0.2

system, a convex hull with a distance of 500 m is defined around the array. Within the hull,
core positions are generated on a regular grid with a spacing of 50 m (20 m for the AERA24
array) and transformed back to the East-North-vertical-coordinate system. An example of the
generated shower-core positions for a given simulation is displayed in figure 5.7. For every
core position, the energy density pattern taken from the interpolation of the simulated antenna
positions is evaluated at the station positions corresponding to the chosen core position in the
shower plane. In addition, also the simulated maximum amplitudes of the electric field at
each simulated antenna station are interpolated and evaluated at the station positions of the
AERA array. The interpolated amplitudes are used to define a lower limit below which the
station would not have measured a signal in an experimental setup. This minimum ampli-
tude is defined as 100 µV/m [158]. If the evaluation of the interpolation of less than three
station positions on the AERA grid reveals an amplitude above this threshold, the procedure
is aborted. As the parametrization of the LDF for AERA has 4 free parameters, at least five
antenna stations with signal are needed to constrain the fit. However, in case of the experi-
5.2. Relation between LDF Parameters and Air-Shower Properties 109

Table 5.3: Values for mean and width of the Gaussian fits shown in figure 5.8. Additionally, the values
for the total distributions are given. The fit uncertainties are of the order of 10−4 and thereby smaller
than the stated precision.

array # stations contained mean (Gaussian) σ (Gaussian) mean std. dev.

configuration
AERA24
3−4 yes -0.159 0.110 -0.127 0.160
3−4 no -0.038 0.127 -0.174 0.240
5+ yes -0.042 0.068 -0.039 0.125
5+ no -0.052 0.123 -0.032 0.218
all all -0.051 0.080 -0.071 0.212
AERA124
3−4 yes -0.054 0.087 -0.106 0.203
3−4 no -0.009 0.063 -0.051 0.178
5+ yes -0.008 0.046 -0.022 0.119
5+ no -0.035 0.087 -0.029 0.136
all all -0.020 0.060 -0.051 0.161

mental hybrid detection of air showers the two core-position parameters can be fixed to the
core position as reconstructed externally, e.g., by the SD. For this analysis, the core position
is fixed to the simulated core position in these cases. Stations below the amplitude threshold
are included in the procedure as it is also the case for externally triggered data, where all sta-
tions are readout. The energy densities for stations further than 500 m away from the shower
axis are set to 0 eV/m2 . The generated values for the energy density are fitted with the AERA
parametrization of the LDF. The resulting radiation energy Erad can then be compared to the
one yielded from the numerical integral over the interpolation of the original simulation. The
relative differences are distinguished for the cases of fixed and free core-position parameters
(for 3 − 4 or five and more signal stations) as well as for the core position being contained
and not contained within the antenna array. In total, this method yields 4223690 reconstructed
LDFs on the AERA24 array and 4686092 on the AERA124 array. For both arrays, the relative
differences in radiation energy are shown in figure 5.8. The distributions in the varying cases
are fitted with Gaussian functions to approximate the mean and the width of the distributions.
The results are summarized in table 5.3. The widths of the distributions are of the order of 10%.
Combined with the intrinsic energy resolution of Erad taken from the integrated interpolation,
the resulting resolution is about 14% which is in good agreement with the earlier findings for
fitting the AERA parametrization directly to the energy density on the star shape pattern. The
distributions especially for the cases of 3 and 4 signal stations are not Gaussian shaped in the
tails. This has two reasons, one is that the shown distributions are superpositions of Gaussian
110 Chapter 5. The Radio Lateral Distribution Function

×10 ×10
3 6
number of simulated core positions

number of simulated core positions

70 3-4 signal stations 5+ signal stations
0.12
core contained core contained
60
core not contained 0.1 core not contained
50
0.08
40
0.06
30

20 0.04

10 0.02

0 0
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
(E -E30-80MHz)/E (E -E30-80MHz)/E
30-80MHz,AERA 30-80MHz 30-80MHz,AERA 30-80MHz

×10 ×10
3 6
number of simulated core positions

number of simulated core positions

100 3-4 signal stations 5+ signal stations
0.35
core contained core contained
0.3
80 core not contained core not contained
0.25
60
0.2

40 0.15

0.1
20
0.05

0 0
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
(E -E30-80MHz)/E (E -E30-80MHz)/E
30-80MHz,AERA 30-80MHz 30-80MHz,AERA 30-80MHz

Figure 5.8: Comparison between the radiation energy Erad taken from the interpolated energy density
pattern and the one taken from the integration over the fit of the AERA LDF (equation 5.2) to the energy
density evaluated at the AERA station positions (top row AERA24, bottom row AERA124). See text
for further details on the procedure. Left: Relative difference in Erad for three or four signal stations
(evaluated amplitude > 100 µV/m). The parameters of the core position in the LDF are fixed to the
true core position. Right: Relative difference in Erad for 5 and more signal stations where the core
position is a free fit parameter. The parameters of Gaussian fits to all distributions in the plotted range
are given in table 5.3 together with the mean and the standard deviation of the whole distribution.

distributions of the relative differences in radiated energy of one air shower simulation under
variation of the core position. Therefore, the total distribution is by construction not Gaussian
in the tails. The second reason is the lower degree to which the LDF is constrained for certain
geometries, e.g., when the signal station positions are very close together in the shower plane
or the sampling is only on one site of the actual energy-density profile giving only very rough
indications for the total amplitude or width. This effect gets bigger in the case of only 3 or
4 signal-stations despite the fixed core position. To also take the tail of the distributions into
account the mean and the standard deviation are also calculated and given in table 5.3. From
the standard deviations of the order of 15 − 20% a total energy resolution of 17 − 22% is
determined which is in good agreement with experimental results discussed in chapter 6.
All distributions show a small negative offset of the order of 5% which corresponds to an un-
derestimation of the radiation energy. The exact value however depends on the geometry of
the air showers in the dataset. Taking subsets for only one type of primary or a small zenith
5.2. Relation between LDF Parameters and Air-Shower Properties 111

angle range, changes the absolute numbers. Therefore, a general correction for a bias intro-
duced by the simplified description is not indicated. For the case of the AERA24 grid and 3
or 4 signal-station showers, the offset is significantly larger. For contained core positions, this
constellation is suppressed as the AERA24 grid is rather dense and therefore only air showers
with a very small footprint lead to the small number of signal stations. Mainly air showers
with small zenith angles are in this subset which are known to be described by the LDF less
accurately. On a large data sample, the influence of these effects on the overall resolution is
however negligible as the fraction of these cases is suppressed since the radio footprint is small
compared to the distance between the antennas. For not contained core-positions, the sampling
on only one side of the axis is not sufficient to constrain the fit in all cases. This leads to an
underestimation of the radiation energy. A possible bias in the reconstructed radiation energy
from measured data is in general not expected to exceed 10% but the exact number can only
be determined using simulations explicitly re-sampling the measured showers in geometry, en-
ergy, and shower development. Compared to the typical experimental systematic uncertainties
in the radio detection of cosmic rays of the order of 30%, the possible contribution arising
from the LDF parametrisation is in any case small.

5.2.3 Core Position

For air showers that were detected in five or more antenna stations, the LDF is used to re-
construct the core position. For this purpose, the resulting ~rcore parameters are transformed
from the shower plane to the ground plane. For the case where the AERA parametrization
is fitted directly to the simulated energy densities on the star shaped grid, the spread of the
~
reconstructed core position is of the order of 10 m in the ~v × B-direction and about 5 m in
~
the ~v × (~v × B)-direction. The asymmetry in the spread arises from the construction of the
LDF as the sum of a positive and a negative Gaussian which are displaced against each other
~
in the ~v × B-direction. After the transformation to the ground coordinated system, the tail
of the distribution gets fanned out. The distributions in both coordinate systems are shown in
figure 5.9 (top row). Most of the positions on the ground are also distributed along the hori-
zontal axis which is due to the geomagnetic field pointing almost northwards at the location of
AERA. From the generated energy density dataset with varying core positions and five or more
signal stations, the core resolution of fitting the AERA LDF to data sampled on a realistically
sparse grid can be evaluated. For this purpose, the two datasets generated for the AERA24
and the AERA124 arrays are re-used. The containment requirement of the shower core which
means that the core position is surrounded by antenna stations on all sides is expected to have
a major influence on the core position reconstruction. Therefore, the sets are split up in two
subsets each with contained and not contained core positions. The initial value of the core
parameter ~rcore in the fit is set to (50 m, 50 m) in the shower plane. The reconstructed core po-
sitions for the AERA24 (center row) and the AERA124 (bottom row) array simulation datasets
112 Chapter 5. The Radio Lateral Distribution Function

40 102 40 102

core position in y-direction (m)

core position in vx(vxB)-direction (m)

number of simulated showers

number of simulated showers
30 30

20 20
10 10
10 10

0 0

-10 -10
1 1
-20 -20

-30 -30

-40 10-1 -40 10-1

-40 -30 -20 -10 0 10 20 30 40 -40 -30 -20 -10 0 10 20 30 40
core position in vxB-direction (m) core position in x-direction (m)

40 40
core position in vx(vxB)-direction (m)

core position in y-direction (m)

number of simulated showers

number of simulated showers

3
30 3 30 10
10
20 20

10 10
102 102
0 0

-10 -10
10 10
-20 -20

-30 -30

-40 1 -40 1
-40 -30 -20 -10 0 10 20 30 40 -40 -30 -20 -10 0 10 20 30 40
core position in vxB-direction (m) core position in x-direction (m)

40 40
core position in vx(vxB)-direction (m)

core position in y-direction (m)

number of simulated showers

number of simulated showers

30 30 3
3 10
10
20 20

10 10
102
102
0 0

-10 -10

10 10
-20 -20

-30 -30

-40 1 -40 1
-40 -30 -20 -10 0 10 20 30 40 -40 -30 -20 -10 0 10 20 30 40
core position in vxB-direction (m) core position in x-direction (m)

Figure 5.9: Reconstructed shower-core positions by fitting the AERA parametrization of the LDF
(equation 5.2) to simulated data on the star-shape grid (top row), to simulations evaluated on the
AERA24 array (center row), and to simulations evaluated on the AERA124 array (bottom row). The
~ vs. ~v ×(~v ×B)
true core-position is set to (0 m, 0 m). Left: Distribution of the core positions in the ~v ×B ~
plane. Right: Projection of the reconstructed core-position onto the ground plane (y-axis pointing north-
wards).

are shown in figure 5.9 in the shower as well as in the ground plane. Compared to the case
of fitting the whole star shaped antenna position pattern, the distributions extend further in
~
the ~v × B-direction ~
and also show a spread along the ~v × (~v × B)-axis. The distribution
is not radially symmetric due to the spread introduced from shifting the two Gaussian func-
tions also seen in figure 5.9. After the transformation into the ground plane, the spread is
more uniformly distributed around the true core position extending slightly further out. This
5.2. Relation between LDF Parameters and Air-Shower Properties 113

Table 5.4: Core resolution taken as the width of the distributions shown in figure 5.9. The means of
the distributions are less than 1 m offset for all distributions. The fit uncertainties are of the order of
10−3 m and thereby smaller than the stated precision.

array contained σ~v×B~ (m) ~ (m) σx (m) σy (m)

σ~v×(~v×B)
AERA24
yes 6.6 4.1 7.2 5.2
no 9.7 11.3 13.1 13.0
all 8.6 7.3 10.8 9.4
AERA124
yes 6.8 4.2 7.7 5.8
no 7.1 11.3 9.5 15.1
all 7.1 4.6 7.9 6.1

broadening is however only visible due to the chosen log-scale and has no major impact on
the overall resolution which is approximated by fitting a two-dimensional Gaussian function
to the distributions for the contained, the not contained, and the combined cases as given in
table 5.4. The resolution is highest for the contained core positions inside the AERA24 array
as it allows the densest sampling. For cores outside of the instrumented area, the resolution
decreases and is even worse than for the case of the AERA124 array. This is due to the dense
grid as it allows to reach five or more signal stations distributed on a smaller area and therefore
also sampling a smaller part of the energy density distribution. It is worth noticing that the
total resolution on the core position is not easily comparable between the two distributions
as the number of contained and not contained showers differs by construction of the method.
The fraction of contained showers is significantly lower for the AERA24 array and thereby the
total resolution is strongly influenced towards the not-contained cases. This however should
represent the situation as it is in the actual experimental dataset. In addition to the theoretical
studies presented here, the influence of background noise on the fitting procedure needs to be
analysed separately. It can however be concluded, that the principle resolution power from
the method itself is comparable to the one of particle detector arrays with a similar detector
spacing e.g., KASCADE-Grande with 5 m core resolution [82].

5.2.4 Depth of the Shower Maximum

The correlation of the width of the LDF σ+ of the original LDF parametrization (equation 5.1)
with the atmospheric distance (column density [g/cm2 ]) to the shower maximum Dmax atm
has
been studied on star-shape antenna-array simulations produced for the AERA site in [156]. It
was found that the distribution can be described reasonably well by a third-degree polynomial
function with a combined overall resolution of 23 g/cm2 . The 1572 simulations of the anal-
ysed set had values of Dmax
atm
ranging up to 1200 g/cm2 . With the simulation set presented in
114 Chapter 5. The Radio Lateral Distribution Function

section 5.2.1, the same range in Dmax atm

is covered whereas the number of simulations is dou-
bled. Furthermore, the set also includes showers induced by iron nuclei. Using a third degree
polynomial leads to an unsatisfying description of the obtained distribution for the presented
set of simulations. Therefore, a fifth degree polynomial is used to describe the distribution
as shown in figure 5.10 (top left). The resolution obtained over the whole range of σ+ is
13.9 ± 0.3 g/cm2 and thereby is greatly improved compared to the earlier findings. As the
slope of the distribution changes with increasing width, the resolution is not constant for all
σ+ . Therefore, the resolution as function of the footprint width is calculated for bins with
a width of 5 m and shown in figure 5.10 (top right). For large atmospheric distances to the
shower maximum, mainly corresponding to inclined air showers, the spread of the distribution
increases which results in a decrease in resolution. If Xmax is close to ground and the footprint
is very small, the resolution is also decreased.
In the next step, it is tested whether the simplified AERA parametrization has an impact on the
resolution. This is expected to some extent due to the binning of function parameters in zenith
angle bins. The distribution of the σ parameter, using equation 5.2 for the LDF fit and the
resulting binned resolution using again a fifth-degree polynomial fit are shown in figure 5.10
(bottom row). The resolution over the whole range of σ is 17.8 ± 0.3 g/cm2 in this case.
The resolution is decreased and the effect of the binning becomes visible in terms of line-like
structures in the distribution for larger values of Dmaxatm
. The spread as function of σ however
decreases for the largest bins, as the distribution is extended to larger σ values. Correlating
the σ(+) parameter to the atmospheric distance to Xmax as shown here and also used in earlier
work, reveals a very competitive limit on the resolution compared to other methods such as the
air fluorescence technique.
In a simple theoretical model, the relation between σ(+) and the point of emission can be de-
scribed with the intercept theorem, see figure 5.2. Therefore the width of the LDF should be
correlated with the geometric rather than the atmospheric distance to the point of emission.
The geometric distance to the shower maximum is introduced here as Dmax geo
calculated from
equation 5.4 using geometrical distances for Xtotal and Xmax instead of column densities. The
correlation between Dmax geo
and σ+ (equation 5.1) is shown in figure 5.11 (top row). The spread
in this case is slightly smaller compared to Dmaxatm
. The resolution is given in units of kilometers
and needs to be transformed into the atmospheric depth of the shower maximum per shower
using the individual zenith angle and the respective atmospheric profile. The resolution over
the whole range of σ+ is 0.167 ± 0.003 km. Using the AERA parametrization a similar ef-
fect is visible in the distribution compared to what is seen when looking at Dmax atm
. Using the
AERA parametrization of the LDF with parameters binned in zenith angles limits the freedom
of the function. This limitation results in a larger σ for larger zenith angles and thereby gives
a smaller spread in this region. The overall spread however increases for σ bins up to about
180 m. The corresponding distributions are shown in figure 5.11 (bottom row). The resolution
over the whole range of σ slightly increases to 0.217 ± 0.004 km. The parameters for the poly-
5.2. Relation between LDF Parameters and Air-Shower Properties 115

75
Dmax (g/cm2)

1200
60

1000 45
atm

(g/cm2 )
30
800
15

600 0

atm
max
f(σ )−D
−15
400

+
−30
200
−45

0 −60

50 100 150 200 250 −75

50 100 150 200
σ+ (m)
σ + (m)

75
Dmax (g/cm2)

1200
60

1000 45
atm

(g/cm2 )
30
800
15

600 0
atm
max
f(σ)−D
−15
400
−30
200
−45

0 −60

50 100 150 200 250 300 −75

50 100 150 200 250
σ (m) σ (m)

Figure 5.10: Correlation between the width of the footprint σ(+) and the atmospheric distance to the
shower maximum Dmax atm for the original LDF parametrization (equation 5.1) in the top row and the

AERA LDF parametrization (equation 5.2) in the bottom row. Left: Fifth degree polynomial fit f
to the distribution. Right: Resolution calculated for different bins in σ(+) with a width of 5 m. The
geo
points indicate the mean of the differences between f (σ) and the true Dmax , the errorbar represents the
standard deviation.

nomial fits of all shown distributions are given in the appendix A.5.
The resulting resolution on Xmax can in this case not be taken from the distribution of residuals
due to the ambiguity between the geometrical distance to Xmax and the atmospheric density
around that point. Therefore, the reconstructed values of Dmax geo
are transformed to Xmax ac-
cording to the zenith angles of the air showers and then compared to the Xmax value as taken
from the simulations. The resulting resolutions are 12.2 g/cm2 for the full fit of the original
LDF parametrization and 15.4 g/cm2 for the AERA LDF parametrization which is an im-
provement of 1 − 2 g/cm2 compared to using Dmax atm
. As before, the resolution in Xmax is not
constant over the range of σ(+) -values. The corresponding profiles for the resolution as func-
tion of bins in σ(+) are shown in figure 5.12. The resolution of Xmax achievable in this way is
very competitive. It can be seen as a limit for the proposed method and will be influenced by
experimental effects depending on the exact condition of the measurement.
The relation between the footprint width and the distance to the shower maximum is mainly
dominated by the zenith angle and in second order by Xmax . The dependency on the zenith
116 Chapter 5. The Radio Lateral Distribution Function

1.0
Dmax (km)

16 0.8
geo

14 0.6

12 0.4

(km)
10 0.2

f(σ +)−Dmax
geo
8 0.0

6 −0.2

4 −0.4

2 −0.6

0 −0.8

50 100 150 200 250 −1.0

50 100 150 200
σ+ (m)
σ + (m)

1.0
Dmax (km)

16 0.8
geo

14 0.6

12 0.4

10 f(σ)−Dmax
geo
(km) 0.2

8 0.0

6 −0.2

4 −0.4

2 −0.6

0 −0.8

50 100 150 200 250 300 −1.0

50 100 150 200 250
σ (m) σ (m)

Figure 5.11: Correlation between the width of the footprint σ(+) and the geometric distance to the
geo
shower maximum Dmax for the original parametrization (equation 5.1) in the top row and the AERA
parametrization (equation 5.2) in the bottom row. Left: Fifth degree polynomial fit f to the distribution.
Right: Resolution calculated for different bins in σ(+) with a width of 5 m. The points indicate the mean
of the differences between f (σ) and the true Dmax
atm and the errorbar represents the standard deviation.

angle is however not constant over the range of covered angles and the used polynomial func-
tion is a compromise for the description of the entire set of simulations. Figure 5.13 shows the
different shapes of the correlation between Dmax geo
and σ+ for various zenith angle bins. From
the separated distributions, it is possible to see how the shape of the overall distribution is built
up. It can be seen that the shape of the individual distributions is not completely re-sampled by
the polynomial fit. For the application to state-of-the-art experiments like AERA or LOFAR,
the detector characteristics introduce uncertainties which outrange these second order effects.
For a next-generation radio experiment such as the Square Kilometer Array (SKA) with more
than 60 000 antennas on a square kilometer, such effects will play a role and should be con-
sidered when developing new reconstruction tools. The resolution e.g., can be improved by
fitting distinct ranges of zenith angles separately. For the given simulation set and the single
polynomial fit function, the resolution can also be expressed as a function of the zenith angle
of the air shower. The resulting profiles are shown in figure 5.14.
Analogue to the resolution of the energy and the core position reconstruction, also the res-
olution achievable for reconstructing Xmax using the AERA LDF parametrization has been
5.2. Relation between LDF Parameters and Air-Shower Properties 117

75 75

60 60

45 45
−Xmax (g/cm2 )

−Xmax (g/cm2 )
30 30

15 15

0 0

−15 −15
Reco

Reco
Xmax

Xmax
−30 −30

−45 −45

−60 −60

−75 −75
50 100 150 200 50 100 150 200 250
σ + (m) σ (m)

Figure 5.12: Resolution of the reconstructed value of Xmax using the fitted relation between the width
geo
of the footprint σ(+) and the geometric distance to the shower maximum Dmax transformed to atmo-
spheric depth according to the individual zenith angles of the showers. The resolution is given for 5 m
bins of σ(+) . Left: Resolution when using the original LDF parametrization (equation 5.1). Right:
Resolution obtained when fitting with the AERA parametrization of the LDF (equation 5.2).

evaluated for the station positions of the AERA array. Therefore, the same two datasets (one
for AERA24 and one for AERA124) as described before with the generated core positions
are used. In this case, the situation for the AERA124 array is especially interesting as for
AERA124 the statistics on multi-hybrid data containing radio and FD detections of the same
air showers are increasing rapidly which brings the experimental extraction of the correlation
between the footprint width and the distance to Xmax into reach. The experimentally deter-
mined relation can then be compared to the simulation results. Therefore, the results presented
in the following are focussing on the AERA124 simulation set. As shown above, the resolution
improves when using Dmax geo
which will be employed for the intermediate step to reconstruct
Xmax in all following analyses.
As air showers with an arrival direction close to the zenith only show a small footprint. With
a sparse antenna grid, the sampling for these showers is sometimes not sufficient to recon-
struct the exact shape of the density pattern and the fitted LDF is too broad and has a smaller
amplitude. The effects of smaller amplitude and larger width mainly compensate each other
when looking at the integral over the LDF. Concerning the footprint width however, these air
showers have a significant impact on the spread, but also lead to a systematic shift in terms
of σ. The correlation between σ and Dmax geo
for the AERA124 array is shown in figure 5.15.
For small values of σ, the distribution shows a shift and therefore the data cannot be described
with the polynomial function taken from the fit in figure 5.11 (bottom row) for values below
σ = 150 m. The distribution is re-fitted by polynomials of different degrees, however the
differences in resolution compared to a linear fit are negligible. Thus, only the linear fit is dis-
cussed in the following which shows an Xmax resolution of 32.4 ± 0.1 g/cm2 . The resolution
as function of σ using the fitted relation for the sub-range σ = [95 m, 255 m] is displayed in
118 Chapter 5. The Radio Lateral Distribution Function

60
offset (m): 0 50 100 150 200 250 300 350 400 450 500 550
15 54

48

42

zenith angle ( ◦ )
10 36
(km)

30
Dmax
geo

24
5
18

12

6
0
0
100 200 300 400 500 600 700 800
σ + (m)

geo
Figure 5.13: Distribution of Dmax as function of σ+ for zenith angle bins of 5◦ width. Every zenith
angle bin is offset by 50 m in σ+ with respect to the precursor bin. The color coding is continuously.
The lines represent the polynomial function obtained from the fit to the distribution in figure 5.11 using
equation 5.1 for the LDF fit. All lines represent the same function but are also offset for every zenith
angle bin.

figure 5.16. In addition, the resolution obtained from applying the same method to a subset of
showers for which 5 or more stations were evaluated with a signal amplitude above the detec-
tion threshold is also shown. In this case, the overall resolution improves to 27.4±0.1 g/cm2
for the linear relation. Taking all together, the resolution is limited for small footprint widths
and improves with growing width. This is an indication that the method requires a sufficiently
dense array to sample the energy density distribution with high precision. The situation can
also be improved when the detector array is at low altitude as Dmax geo
gets larger for all air
showers in this case. But also for AERA, with increasing size of σ the in principle achievable
resolution rapidly approaches a level comparable to the resolution of about 20 − 30 g/cm2
achieved with established techniques like the fluorescence [29] or the air Cherenkov detection
[90]. The obtained resolutions for all studied constellations of LDF parametrizations and an-
tenna arrays are summarized in table 5.5. The offset of the reconstructed values of Xmax is
smaller than 3 g/cm2 for all constellations except using the AERA LDF on the AERA124 array
including 3 and 4 signal-station showers where an offset of 7.7 ± 0.1 g/cm2 is observed.
For the reconstruction of Xmax with AERA, a direct calibration against the FD measurements
in multi-hybrid air showers is the goal. The number of multi-hybrid RD-FD showers increases
but is still limited for the frame of this work, therefore the determination of the calibration
function on simulations is important to be able to reconstruct information about Xmax for all
air showers detected with AERA.
The general analyses presented in this chapter have shown the feasibility of using the method
5.2. Relation between LDF Parameters and Air-Shower Properties 119

75 75

60 60

45 45
−Xmax (g/cm2 )

−Xmax (g/cm2 )
30 30

15 15

0 0

−15 −15
Reco

Reco
Xmax

Xmax
−30 −30

−45 −45

−60 −60

−75 −75
0 10 20 30 40 50 60 0 10 20 30 40 50 60
zenith angle ( ◦ ) zenith angle ( ◦ )

Figure 5.14: Resolution of the Xmax determination analogue to figure 5.12 but for 2.5◦ bins in zenith
angle. Left: Resolution for the application of the original LDF parametrization (equation 5.1). Right:
Resolution for the application of the AERA LDF parametrization (equation 5.2).

Table 5.5: Resolutions of the reconstruction of the depth of the shower maximum. Two different LDF
parametrizations, the original one from equation 5.1 and the one modified for AERA (equation 5.2) are
used and applied to data sampled with the star-shape array and the AERA124 array. See text for further
details on the method.

LDF array atm

σ(Dmax ) (g/cm2 ) σ(Dmax
geo
) (km) σ(Xmax ) (g/cm2 )
original star-shape 13.9 ± 0.3 - 13.9 ± 0.3
AERA star-shape 17.8 ± 0.3 - 17.8 ± 0.3
original star-shape - 0.167 ± 0.003 12.2 ± 0.2
AERA star-shape - 0.217 ± 0.004 15.4 ± 0.3
AERA AERA124 0.509 ± 0.001 32.4 ± 0.1
AERA AERA124 - 0.447 ± 0.001 27.4 ± 0.1
5+ stations

for the AERA124 array. The influence of the characteristics of the specific dataset under study
(i.e., energy and direction ranges, geometry of the antenna array) on the correlation and reso-
lution is however not negligible. For the reconstruction of Xmax from the AERA124 data as
presented in chapter 7, an additional simulation set which is exactly re-sampling the measured
air showers and also includes measured noise is analysed for further studies of the influence of
geometry and background noise. Before that, the determination of the energy of extensive air
showers measured by AERA24 in coincidence with the Surface Detector array based on the
AERA LDF is presented in the next chapter.
120 Chapter 5. The Radio Lateral Distribution Function

16 star shape simulation fit

linear function
14

12

10
(km)

8
Dmax
geo

0
50 100 150 200 250
σ (m)
Figure 5.15: Correlation between the width of the LDF σ (equation 5.2) and the geometric distance
geo
to Xmax Dmax for the AERA124 array. The gray scale gives the number of entries in the background
histogram from white to black. The solid line indicates the polynomial fit to the distribution obtained
by fitting the AERA parametrization to the energy density at the star-shape antenna positions (see
figure 5.11 (bottom row)). The dotted line represents a linear fit to the distribution.

100 100

80 80

60 60
−Xmax (g/cm2 )

−Xmax (g/cm2 )

40 40

20 20

0 0

−20 −20
Reco

Reco
Xmax

Xmax

−40 −40

−60 −60

−80 −80

−100 −100
100 120 140 160 180 200 220 240 100 120 140 160 180 200 220 240
σ (m) σ (m)

Figure 5.16: Resolution of the reconstructed Xmax when sampling the energy density pattern with the
array of AERA124 and fitting the parametrization of the LDF (equation 5.2). Left: Binned resolution
for all showers. Right: Subset with 5 or more stations evaluated to be above the signal amplitude
threshold.
 Radio Energy Reconstruction
6
This chapter has been published as Auger Full Author List Publications:

The Pierre Auger Collaboration, A. Aab et al.,

Energy Estimation of Cosmic Rays with the Engineering Radio Array of the
Pierre Auger Observatory
Submitted to Physical Review D, arXiv:1508.04267

The Pierre Auger Collaboration, A. Aab et al.,

The energy in the radio signal of extensive air showers
Submitted to Physical Review Letters

In this analysis we present an absolute calibration of cosmic-ray energies around 1 EeV using
the energy radiated between 30 and 80 MHz as an estimator. The radiated energy is determined
based on the integral over the 2D lateral distribution of the energy density as discussed in the
preceding chapter.

6.1 Data Selection and Event Reconstruction

For this study we are using RD and SD data recorded between April 2011 and March 2013
when AERA was operating in its first commissioning phase. For this analysis, both self-
triggered and externally triggered events are used.

6.1.1 Preselection of Cosmic Ray Candidates

In the case of the self-triggered events, a preselection is performed offline by searching for
coincidences with the SD events. A radio event has to agree in time and location with an SD
event to be considered as cosmic-ray candidate. The radio-trigger time and the time when the
air shower core hits the ground have to agree within ±20 µs. Such a conservative coincidence
window also accounts for horizontal events, for which the time difference is expected to be
larger. For both trigger types, only events with a clear radio pulse in at least three stations
are considered, to allow for a reconstruction of the incoming direction of the signal. For
externally triggered events the requirement is a signal-to-noise ratio (SNR) greater than ten.
122 Chapter 6. Radio Energy Reconstruction

pulse maximum east-west

1000 north-south

electric field [µV/m]

vertical
Hilbert envelope
500

−500

−1000
420 440 460 480 500
time [ns]

Figure 6.1: Reconstructed electric-field trace of one of the measured air showers. An upsampling by a
factor of five was applied. The shown Hilbert envelope (dashed line) is the square root of the quadratic
sum of the Hilbert envelopes of the three polarization components.

For self-triggered events the signal threshold is dynamically adjusted to the noise level to keep
the trigger rate at a constant level of 100 Hz. We require that the reconstructed incoming
directions from the radio and the Surface Detectors array agree within 20◦ to be accepted as a
cosmic-ray candidate. The 20◦ cut does not reflect the angular resolution of the SD nor that of
the radio detector. This preselection cut retains the maximum number of cosmic-ray signals
and significantly reduces the number of anthropogenic noise pulses, which originate mainly
from the horizon. In addition, we apply quality cuts on the data of the Surface Detector array
as described in [67]. The most important cuts are that the core position be surrounded by a
hexagon of active stations and that the zenith angle of the incoming direction be less than 55◦ .
A total of 181 cosmic-ray candidates with energies above 1017 eV remain.

6.1.2 Reconstruction of Radio Data

We use a similar reconstruction scheme for the hybrid data as described in chapter 3. For
this analysis we also apply the RFI cleaning procedure discussed in section 3.5.3. The energy
density u of the incoming electromagnetic radio pulse at each radio station is determined by
calculating the time integral over the absolute value of the Poynting vector as described in
section 3.5.4. As the radio detector effects are corrected for in the data reconstruction, the
obtained energy density can be directly compared to air-shower simulations. We also calculate
the direction of the electric-field vector, i.e., the polarization direction of the signal. In the full
width half maximum (FWHM) interval around the pulse maximum of the Hilbert envelope
we observe that the reconstructed electric-field vectors are aligned approximately along the
same direction for every time bin. To accurately determine the mean direction of the electric-
field vector, we average over all vectors in the FWHM interval of the Hilbert envelope (see
figure 6.1).
6.1. Data Selection and Event Reconstruction 123

6.1.3 Selection of Radio Signals Induced by Cosmic Rays

Given the presents of pulsed background noise at the AERA site, the preselected events are
likely to contain non cosmic ray signals that mimic cosmic-ray pulses. There are two scenarios
possible: Signals in one or more stations are not caused by the air shower or an event contains
only noise pulses that by chance led to a reconstructed incoming direction similar to that of
the SD. In order to reject background signals, we take advantage of the expected polarization
of the radio signal. The polarization of the radio pulse is only used for this purpose and not
considered for the energy estimation. In the frequency range of AERA geomagnetic emis-
sion is dominant compared to the charge-excess emission process. The expected direction of
the electric-field vector is therefore calculated from the geomagnetic and the charge-excess
contributions
~ exp ∝ sin α ~egeo + a ~eCE ,
E (6.1)
where α is the angle between shower axis and magnetic field of the Earth, and a is the average
relative charge-excess strength (0.14 ± 0.02 at AERA). Details on the emission mechanisms
and the polarization of the radio signal are given in section 1.2.4. In this approach, the di-
rection of the geomagnetic contribution depends only on the incoming direction of the air
shower whereas the charge-excess contribution depends in addition on the position of the ra-
dio station relative to the shower axis. In figure 6.2, all stations with signal of a cosmic ray
candidate are shown, and the measured polarization is compared with the expectations of the
two radio-emission mechanisms. The overall agreement between measured and expected field
polarizations is quantified using the angular difference

~ meas,i , E
βi = ∠(E ~ exp,i ) (6.2)

at each station i. For each event, the average deviation β̄ of the individual deviations βi of
the stations with signal is calculated and will be used as criterion for a quality cut. Relevant
uncertainties are taken into account as follows:

• The relative strength a of the charge-excess can vary due to shower-to-shower fluctu-
ations, and additional dependencies on the geometry of the air shower. Therefore, for
each possible value of a between 0 and 0.5 the average deviation β̄ is calculated and
only the smallest value of β̄ is considered.

• The uncertainty of the SD shower core position is taken into account by variation of
the core within its estimated uncertainties. In our dataset the uncertainty varies between
10 m and 80 m depending on the energy and zenith angle. For each trial of the core
position β̄ is calculated. Again, only the smallest value of β̄ is considered.

• Interference of the cosmic-ray radio signal with noise pulses can alter the polarization.
Simulation studies showed that for a single radio station the uncertainty in β due to
noise is below 8◦ at detection threshold, and decreases to 1◦ at high signal-to-noise
124 Chapter 6. Radio Energy Reconstruction

300 measurement
geomagnetic
charge-excess

position in ~v × (~v × ~B) [m]

200 geomagnetic + charge-excess
β = 0.5◦

β = 3.5◦
100
β = 1.9◦

β = 0.6◦
β = 1.9◦
0

β = 2.9◦
−100
−200 −100 0 100 200 300
position in ~v × ~B [m]

Figure 6.2: Polarization map of a single air shower. The SD shower core is at the coordinate origin.
The measured polarizations are shown as black arrows. The gray arrows are the model expectations,
and the red and blue arrows are the geomagnetic and the charge-excess components, respectively. The
definition of β is described in the text. The properties of this air shower are: energy: 0.9 EeV, zenith
angle: 36◦ , azimuth angle: 207◦ . For the emission model of equation 6.1, the optimal value of the
relative charge-excess strength is a = 0.18.

ratios. To obtain the average value of β for all radio stations in the event we compute a
weighted mean with weights wi = 1/σβ2i with σβi being the expected uncertainty from
the simulation.

We impose a limit on the average deviation β̄ of the polarization direction. This maximum de-
viation is fixed at a value of 3◦ . This value is slightly above the combination of the following
effects. The incoming direction of an air shower reconstructed with the Surface Detector array
has an uncertainty between 1.3◦ and 0.7◦ depending on the cosmic ray energy and the zenith
angle [159]. Hence, the expected direction of the electric-field vector will have the same
uncertainty. All antennas are aligned to the magnetic north (or perpendicularly to the mag-
netic north in case of the other polarization direction) with a precision of better than 1◦ [160].
All antennas are uniformly constructed and the two antennas of a radio station are identical.
Asymmetries in the ground conditions have only negligible influence as the LPDA antenna is
mostly insensitive towards the ground. A measurement at AERA has shown that the responses
of all antennas differ by less than 0.3% [161]. A difference in the amplification of the signal
chain of the north-south and east-west polarized antenna will influence the polarization mea-
surement. From an individual measurement of the signal chain of all antennas the uncertainty
is estimated to be 2.5% which results in a polarization uncertainty below 0.7◦ . In addition, we
neglect the dependence of the relative strength a of the charge-excess on the distance between
observer position and shower axis. For a single station this effect is relevant. However, in our
approach we only use the average deviation of all stations with signal also taking into account
the uncertainty in the core position. Therefore the distance dependence will mostly average
6.1. Data Selection and Event Reconstruction 125

Table 6.1: Overview of selection cuts and the number of showers surviving these cuts. Preselection
means: ESD ≥ 0.1 EeV, standard SD quality cuts, ≥ 3 radio stations with signal, SD and RD recon-
structed incoming directions agree within 20◦ . See text for details.

cut number of showers after cut

preselection (section 6.1.1) 181
polarization cut (β̄ < 3◦ , section 6.1.3) 136
no thunderstorm conditions (section 6.1.3) 134
LDF fit converged (σ < 300 m, section 6.2) 126
≥ 5 stations with signal
(only high quality dataset, section 6.3) 47

out. We estimate that the remaining additional scatter is 1.5◦ . We account for individual radio
stations being contaminated with substantial noise signals by iterating through all configura-
tions with only one and then more stations removed, down to the minimum of three stations.
An event where the weighted average deviation β̄ is greater than 3◦ for all station combinations
is rejected. If β̄ is less than 3◦ for any station combination and the fraction of selected stations
is larger than 50% of the total number of stations with signal, the shower candidate is consid-
ered a cosmic-ray event and only the stations from this particular combination are used. After
this cut 136 showers remain. The number of excluded single stations and complete events is
compatible with the measured rate of noise pulses. Most of the events recorded during thun-
derstorm conditions appear to be rejected by this selection procedure as the strong atmospheric
electric-fields of a thunderstorm influence the radio emission and alter the polarization of the
radio signals [162, 163]. For two thirds of the showers, a measurement of the atmospheric
electric field is available. These showers are checked for thunderstorm conditions using an
algorithm described in [164]. Based on this check, two additional showers were rejected. All
cuts are summarized in table 6.1.

6.1.4 Uncertainties on the Energy Density in a Single Radio Station

In addition to the uncertainties on the amplification of the signal chain of 2.5% discussed
above, no further uncertainties are expected that would result in a different response of stations
within one event. To first order, the frequency content and the incoming direction of the radio
pulse are similar at all observer positions. Therefore, an uncertainty of the antenna-response
pattern has a negligible influence as it is evaluated for the same direction at all stations. Possi-
ble different ground conditions at different station positions that result in a different reflectivity
of the soil are negligible due to the insensitivity of the antenna towards the ground. The 2.5%
amplification uncertainty results in 5% uncertainty on the energy density u, as u scales quadrat-
ically with the electric-field amplitude. This uncertainty is added in quadrature to the signal
uncertainty resulting from noise.
126 Chapter 6. Radio Energy Reconstruction

400

energy density u [eV/m2]

data 240
250 fit
300 sub-threshold
data
210
position in ~v × (~v × ~B) [m]

flagged

energy density u [eV/m2]

200
200 core (SD) sub-threshold
180
150
100 150
100
0 120
50
−100 90
0
−200 60 1

u−fit
−300 0

σu
30

−400 0 −1
−400 −200 0 200 400
0 50 100 150 200 250 300 350 400
position in ~v × ~B [m] distance to RD shower axis [m]

Figure 6.3: Lateral signal distribution of an air shower. The properties of this air shower are: energy:
0.75 EeV, zenith angle: 37◦ , azimuth angle: 44◦ . Left: The energy density in the shower plane. The
measurements are indicated as circles where the color shows the energy density. Grey squares are sta-
tions with signal below threshold and the red cross marks a station that is rejected due to a mismatch in
the signal polarization. The background map shows the LDF parametrization (equation 5.2). The co-
ordinate origin is the reconstructed core position of the radio LDF fit. Right: The radio energy density
is shown as a function of the perpendicular distance to the shower axis. Blue squares are the measure-
ments and the circular symbols indicate the value of the LDF parametrization at the position of the
measurement. Gray curves show the projection of the two-dimensional LDF onto a line connecting the
radio-core position with each station position. Also shown are the residuals in units of the uncertainty
of the measurement.

6.2 Energy Estimator

To obtain an absolute energy estimator from the signals at the different distances to the shower
axis (energy density u in units of eV/m2 ) the AERA LDF introduced in equation 5.2 in sec-
tion 5.1 is used. The LDF is fitted to the data using a chi-square minimization. An example of
one air shower within our dataset is shown in figure 6.3. Some shower data do not contain suf-
ficient information to fit the LDF, such as when only three stations with signal are present that
have roughly the same signal strength. This results in an unphysically broad LDF. To reject
these events we impose the quality cut σ < 300 m (table 6.1). An analysis of simulations for
the AERA geometry showed that the σ parameter of the LDF is never larger than 300 m (see
section 5.2.4). In the following, only the 126 events that pass the quality cuts are considered
and will be referred to as the full dataset. To derive the accuracy of the energy estimation
method, the dataset will be further divided in a high-quality dataset containing only events
with at least five stations with signal where the core position is reconstructed in the LDF fit.

6.2.1 Definition
The spatial integral of the lateral distribution function gives the amount of energy that is trans-
ferred from the primary cosmic ray into radio emission in the AERA frequency band during
6.2. Energy Estimator 127

90°

135° 45°
60◦

40◦

20◦

180° 0°

225° 315°

270°

Figure 6.4: Skymap of the 126 selected cosmic rays. Green filled circles denote air showers with at
least five stations with signal and open circles denote air showers with less than five stations with signal.
The red star denotes the direction of the magnetic-field axis at AERA. All measured cosmic rays are at
least 20◦ away from the magnetic-field axis. Therefore, the geomagnetic emission gives the dominant
contribution to the radiation energy for all showers.

the air-shower development. We use the energy estimator Sradio as introduced in section 5.2.2
which is the radiation energy divided by sin2 (α) to account for different emission strengths at
different angles between shower axis and magnetic field,
1
Z
Sradio = u(~r) d2~r
sin2 α
R2 (6.3)
Aπ
σ 2 − C0 C32 e2C4 σ ,


=
sin2 (α)
where R2 denotes the shower plane. The positive σ 2 term dominates by far over the negative
second term resulting in a positive value of Sradio . The sin2 (α) correction only holds if the
geomagnetic emission is the dominant contribution which is the case for α > 10◦ at AERA.
Due to the reduced emission strength the number of detections for arrival directions within
10◦ of the geomagnetic field axis is suppressed. The angular distribution of the air showers is
shown in figure 6.4.

6.2.2 Shower-by-Shower Uncertainties

The following uncertainties are relevant for the energy estimator due to shower-by-shower
fluctuations and summarized in table 6.2:

• The gains of the low-noise amplifiers and filter amplifiers exhibit a temperature depen-
dence. The effect has been measured and amounts to −42 mdB/K. Each air shower
128 Chapter 6. Radio Energy Reconstruction

Table 6.2: Overview of uncertainties of the electric field amplitude σ|E|

~ and the energy estimator Sradio .
"⊕" denotes a quadratic sum. The average fit uncertainty of Sradio is 46% and 24% for the high quality
subset of showers with at least five stations with signal.

source of uncertainty σ|E|

~ σSradio
shower-by-shower
temperature dependence 4% 8%
angular dependence of
antenna response pattern 5% 10%
reconstructed direction negligible negligible
LDF fit uncertainty - error propagation of
fit parameters
total shower-by-shower uncertainty 6.4% 12.8% ⊕ fit uncertainty
absolute scale
absolute scale of antenna
response pattern 12.5% 25%
analogue signal chain 6% 12%
total absolute scale uncertainty 14% 28%

is measured under specific environmental conditions. In particular this implies that we

have a random distribution of ambient temperatures which exhibit a Gaussian distribu-
tion with a standard deviation of 8.3◦ C. This corresponds to a fluctuation of the gain of
4%.

• An uncertainty of the simulated antenna response that depends on the incoming direction
of the radio signal will lead to an shower-by-shower uncertainty as each shower has a
different incoming direction. The effect is determined to be 5% by comparison of the
simulated antenna response with a measurement at AERA [111].

• The reconstructed direction of the air shower obtained with the SD has an uncertainty
of less than 1.3◦ . This has negligible influence on the antenna response pattern, since it
can be considered uniform over such a small change of angle.

As the different uncertainties are independent, the total uncertainty of the electric field am-
√
plitude is 4%2 + 5%2 ≈ 6.4% and therefore 12.8% on Sradio . The uncertainty of α can be
neglected. The fit uncertainties of A and σ including their correlation are propagated into Sradio
using Gaussian error propagation. In the case of showers with less than five stations with sig-
nal, the core position of the SD reconstruction is used and its uncertainty is propagated into the
fit uncertainty of Sradio . This fit uncertainty is added in quadrature to the statistical uncertainty
of 12.8% of the energy estimator. The average fit uncertainty of Sradio is 46%. For showers
with at least five stations with signal the average uncertainty reduces to 24%.
6.3. Energy Calibration 129

6.2.3 Absolute Scale Uncertainties

The dominant systematic uncertainties of the reconstructed electric-field amplitudes are the
calibration of the analog signal chain and the antenna response pattern. The analog signal chain
consists of the low-noise amplifier, the filter amplifier and all cables between the antenna and
the analog-to-digital converter. The analog signal chain has been measured for each channel
of each radio station separately in the field and differences are corrected for. The systematic
uncertainty of the analog chain amounts to 6%. The simulated antenna response pattern has
been confirmed by measurements at an overall level of 4%. The systematic uncertainty of
the measurement is 12.5% in the vector effective length [165]. Conservatively, the systematic
uncertainty of the antenna-response pattern is therefore estimated as 12.5%. Combining both
uncertainties in quadrature, the systematic uncertainty of the electric-field amplitude is 14%.
The radio energy density and the energy estimator scale with the amplitude squared. Therefore,
the systematic uncertainty of the absolute scale of the radiation energy is 28%. We note that,
as the cosmic ray energy is proportional to the square root of the radiation energy (see next
section), the systematic uncertainty of a radio cosmic-ray energy scale would remain at 14%.

6.3 Energy Calibration

The radio-energy estimator Sradio is shown as a function of the cosmic ray energy measured
with the Surface Detector array in figure 6.5 (top). A clear correlation is observed. For the
calibration function we follow the same method as used for the calibration of the Surface
Detector array with the Fluorescence Detector (see section 3.4.3). The calibration function

Sradio = A × 107 eV (ESD /1018 eV)B (6.4)

is obtained by maximizing a likelihood function that takes into account all measurement un-
certainties, detector efficiencies and the steeply falling energy spectrum (the functional form
of the likelihood function can be found in appendix A.6). The result of the calibration fit is
A = 1.58 ± 0.07 and B = 1.98 ± 0.04. The resulting slope is compatible with an exponent
of B = 2 implying that the energy deposited in radio emission increases quadratically with
the cosmic ray energy. We can infer from equation 6.4 that, for a 1 EeV air shower perpen-
dicular to the magnetic field axis, 15.8 MeV are deposited on average in radio emission in
the frequency range of 30 to 80 MHz. In the lower left panel of figure 6.5 the scatter around
the calibration curve for all air showers in our dataset is shown. This amounts to 29%. We
also tested a high-quality dataset containing only air showers with at least five stations with
signal, where a determination of the core position in the radio LDF fit is possible. These air
showers are marked by green filled circles in figure 6.5. The fit of the calibration curve gives
a compatible result (A = 1.60 ± 0.08, B = 1.99 ± 0.05) and the scatter around the calibration
curve reduces to 24% (lower right panel of figure 6.5). To obtain a goodness-of-fit estimator,
130 Chapter 6. Radio Energy Reconstruction

the measured distribution is compared to the expected distribution which is computed from
the likelihood function, i.e., from the probability model that describes the fluctuations. The
comparison yields a reduced chi-square value of χ2 /ndf = 13.8/12 for the full dataset and
χ2 /ndf = 8.43/6 for the high-quality dataset. In particular, it shows that the estimated un-
certainties of the energy estimator in section 6.2.2 are compatible with the observed scatter
around the calibration curve.

6.3.1 Uncertainties of the Reconstructed Cosmic-Ray Energy

To determine the energy resolution of the radio detector, the known resolution of the Surface
Detector array needs to be subtracted from the combined scatter. The average (statistical)
SD energy resolution for all air showers in our dataset is 18%. To obtain an estimate of the
radio-energy resolution we use a Monte Carlo study which takes into account the energy and
zenith angle dependence of the SD energy resolution. The combined scatter is simulated for
different radio-energy resolutions, according to the number of air showers and the energy and
zenith distribution of the dataset. We find that the energy resolution of the radio detector is
22% for the full dataset and 17% for the air showers where the core position could be deter-
mined in the radio LDF fit, when five or more radio stations have a significant signal. In the
above calculation we assumed that the energy estimates from the SD and radio reconstruc-
tion are uncorrelated for a fixed energy. However, an anti-correlation is expected as radio
emission originates from the electromagnetic part of the air shower whereas the SD signal is
mostly due to muons resulting from the hadronic shower component [166] and which are anti-
correlated shower parameters for a fixed cosmic ray energy. In case of an anti-correlation, the
estimated radio-energy resolution would be even smaller making the above values conserva-
tive estimates. Furthermore, we studied the effect of a possible bias in the SD reconstructed
energy for different primaries where the detector is not fully efficient (0.1 − 0.3 EeV) and has
a slightly different efficiency curve for the two extreme scenarios of proton and iron primaries
[159]. We found that the effect is negligible for our dataset. The uncertainty on the absolute
scale of the energy estimator as discussed in section 6.2.3 is calibrated out by correlating Sradio
with ESD . The method, however, inherits the uncertainties of the SD energy scale. This scale
uncertainty is dominated by the FD scale uncertainty, which is used to calibrate the SD. It is
16% at 1017.5 eV and decreases to 14% at energies ≥ 1018 eV [103].

6.3.2 Precision and Possible Improvements of the Reconstruction

We have found that the instrumental noise and the environmental influences are not the dom-
inant contributions to our energy resolution. Applying the method described to a CoREAS
Monte Carlo dataset, including a representative set of shower geometries as well as shower-to-
shower fluctuations, but no instrumental or environmental uncertainties, a similar energy reso-
lution is obtained for the same detector layout. The intrinsic limitation in the energy resolution
6.3. Energy Calibration 131

due to shower-to-shower fluctuations of the electromagnetic part of the shower is predicted to

be smaller than 10% [93, 167] and we expect that the current energy resolution can be further
improved. Under the condition that the LDF samples the relevant part of the signal distribution
on the ground correctly for all geometries, the energy estimator should only be affected by the
shower-to-shower fluctuations in the electromagnetic part of the shower. The dominant addi-
tional geometric dependence is due to the fact that the air shower might not be fully developed
when reaching the ground, i.e., some part of the shower is clipped away. As the atmospheric
depth increases with the secant of the zenith angle, clipping mostly affects high-energy verti-
cal showers. Hence, we expect an additional dependence on the zenith angle. In the future,
with larger statistics, this effect will be parametrized from data and will further improve the
energy resolution. Also, a better understanding of the detector and the environmental effects,
such as temperature dependencies, will help to improve the energy reconstruction. Combined
measurements, such as they are possible at the Pierre Auger Observatory, hold great poten-
tial for future improvements of the energy resolution due to the anti-correlation of the energy
reconstructed with the radio and Surface Detector array.

6.3.3 The Energy Content of Extensive Air Showers in MHz Radiation

So far, the energy content of extensive air showers in the radio frequency range of 30 to 80 MHz
has only been measured at the Pierre Auger Observatory in Argentina. However, our findings
can be generalized by the following consideration. To obtain a prediction that is independent
of the location of the experiment, i.e., a universal formula to calculate the radiation energy
from the cosmic-ray energy, the calibration function in equation 6.4 can be normalized to the
local magnetic field. We found that it is sufficient to correct only for the dominant geomagnetic
part of the radio emission. This is because the increase of radiation energy due to the charge-
excess emission is small, as constructive and destructive interference with the geomagnetic
emission mostly cancel out in the integration of the energy densities over the shower plane,
see equation 6.3. For the average relative charge-excess strength of 14% at AERA the increase
in radiation energy is only 2%. As most locations on Earth have a stronger magnetic field than
the AERA site the effect of the charge-excess emission on the radiation energy will be even
smaller. Within the statistical accuracy of the calibration function this effect can be neglected
which leads to the universal prediction of the radiation energy
 2
E BEarth
E30−80 MHz = (15.8 ± 0.7(stat) ± 6.7(sys)) MeV × sin α 18 , (6.5)
10 eV 0.24 G

where E is the cosmic ray energy, BEarth denotes the local magnetic-field strength and 0.24 G
is the magnetic-field strength at the AERA site. The systematic uncertainty quoted here is the
combined uncertainty of Sradio (28%) and the SD energy scale (16% at 1017.5 eV). This formula
will become invalid for radio detectors at high altitudes because the amount of radiation energy
132 Chapter 6. Radio Energy Reconstruction

decreases as – depending on the zenith angle – a significant part of the air shower is clipped
away at the ground.

6.4 Relevance for the Absolute Energy Scale Determination

The energy in the radio signal can be used for a cross-calibration of different experiments and
detection techniques. Alternatively, it can be used for an independent determination of the
absolute energy scale of a cosmic-ray observatory. Such an absolute calibration requires input
from the experimental and theoretical side, where the procedure for measuring the radiation
energy has already been outlined above. On the theoretical side, the radiation energy emit-
ted by an extensive air shower with a given energy in the electromagnetic component can be
predicted from first principles using classical electrodynamics. Uncertainties in the propaga-
tion of the radio signal are insignificant since the atmosphere is transparent to radio waves
in this frequency range. Using the sophisticated simulation codes discussed in section 1.2.4
the radio emission from extensive air showers can be calculated on the basis of Monte Carlo
simulations. Electromagnetic radiation from the acceleration of charged particles in the elec-
tromagnetic component of the air shower, coherence effects due to the spatial particle distribu-
tion, and propagation effects in the refractive index gradient of the atmosphere fully determine
the radio emission. These simulations thus involve no free parameters. There are no scaling
uncertainties or “yield” factors. Work is still needed to compare the predictions of existing
simulation codes in more detail, but no insurmountable problems have been identified. To re-
late the radiation energy emitted by the electromagnetic component of an extensive air shower
to the energy of the primary cosmic ray, a correction for the invisible energy, i.e., the energy
carried away by high-energy muons and neutrinos, needs to be applied. This correction can be
obtained by a data-driven method exploiting the correlation between the invisible energy and
the muon number at the ground [168], and only adds a small contribution to the overall energy
scale uncertainty [103].
Combining accurate measurements of the energy in the radio signal and predictions from first-
principle calculations, an independent determination of the absolute energy scale using radio
detection of cosmic rays becomes possible.
6.4. Relevance for the Absolute Energy Scale Determination 133

109

108
Sradio [eV]

107

106

105 17
10 1018 1019
ESD [eV]
35 number of signal stations ≥5
N = 126 data
20 N = 47 data
30 µ = 0.04 ± 0.03 model
σ = 0.29 µ = 0.02 ± 0.04 model
σ = 0.24
25
15
20
entries

entries

15 10

10
5
5

0 0
−1.5 −1.0 −0.5 0.0 0.5 1.0 1.5 −1.5 −1.0 −0.5 0.0 0.5 1.0 1.5
2( ESD − ERD )/( ESD + ERD ) 2( ESD − ERD )/( ESD + ERD )

Figure 6.5: Top: The radio energy estimator Sradio as a function of the cosmic-ray energy measured
with the Surface Detector array. A power law is fitted to the data using a likelihood approach which
takes all uncertainties and detection efficiencies into account. Green filled circles denote air showers
where the core position has been determined in the radio LDF fit, i.e., all air showers with at least five
stations with signal. Open circles denote events with less than five stations with signal and use the SD
core position. Bottom: Relative energy resolution: The energy of the radio detector is obtained using
the fit in the top figure. The left histogram contains all air showers, and the right histogram contains
the air showers with at least five stations with signal (green filled circles). The expected distribution
is shown as a gray shaded area which is computed from the fitted probability model that describes the
fluctuations.
 Radio Xmax Reconstruction
7
For the reconstruction of the depth of the shower maximum Xmax , two different methods are
presented in this chapter and applied to AERA124 multi-hybrid data, performing a compari-
son to the measurements of the Fluorescence Detector. Both methods are based on the radio
energy-density pattern measured on the ground. In one approach, the width of the 2D LDF as
discussed in chapter 5 is used. In the other approach, sets of simulations for different shower
developments are generated and Xmax is determined from the agreement between the indi-
vidual simulations and the measured data. Both methods are also compared to each other.
Finally, the width of the LDF is reconstructed for the thousands of showers in the RD-SD hy-
brid dataset and the average value of Xmax as function of energy is determined and compared
to other experimental data and predictions from different hadronic interaction models.

7.1 Xmax from the 2D Radio LDF

The radio lateral distribution function can be used to reconstruct Xmax with AERA124, as
discussed in chapter 5. To apply the method to measured data, the correlation found for the
generic simulations is redetermined for the real detector geometry and also including measured
noise. This correlation is verified using the RD-SD-FD multi-hybrid showers.

7.1.1 Calibration with Full Detector Simulations

The correlation between the width of the LDF σ and the geometric distance to Xmax Dmax geo

depends on the layout of the antenna array (see section 5.2.4). Furthermore, the influence of
real background noise on the reconstruction of the width parameter needs to be investigated.
This however cannot be easily determined from the generic simulation set for the real AERA
antenna grid. Therefore the specific simulations purpose-made for the AERA layout and the
properties (energy and arrival direction) of the reconstructed showers in the AERA dataset are
used (see section 3.7).
In the presence of noise and considering the uncertainties introduced by the detector setup,
the reconstruction of the radio signals and the determination of the LDF becomes more chal-
lenging. The uncertainty associated with the signal chain has a significant influence on the
largest signals whereas the noise mainly impacts stations with signals around or below the
signal-to-noise threshold. Both effects together influence the reconstruction quality and result
136 Chapter 7. Radio Xmax Reconstruction

12

10

8
(km)

6
Dmax
geo

fit to generic simulations

0
100 200 300 400 500
σ (m)
geo
Figure 7.1: Correlation between the geometric distance to Xmax Dmax and the width of the LDF σ
obtained from all successfully reconstructed simulations (with zenith angle < 55◦ ) which are purpose-
made for AERA124. The simulations are re-sampling the measured air showers and include measured
background noise. No quality cuts are applied to the shown dataset. The line indicates the linear
correlations obtained from the generic simulations in figure 5.15.

in cases for which the reconstruction is not successful. Altogether, 5832 simulated showers
were successfully reconstructed. The distributions of σ and Dmax geo
, for the 4479 reconstructed
air showers for which the zenith angle of the arrival directions is smaller than 55◦ , are shown
in figure 7.1. In addition, the linear dependency obtained from the fit shown in figure 5.15 is
plotted as an indicator of the expected correlation. The bulk of the data points is located around
the line predicted from the generic simulations. The distribution however has a rather broad
spread and a number of points are located in a separated region vertically oriented around
σ = 400 m. For further analysis, quality cuts are enforced on the reconstructed data as sum-
marized in table 7.1. Due to the relatively large spread in the distribution, rather strict quality
cuts are imposed on the events for this Xmax analysis to obtain a high-quality sample. These
cuts can be loosened or otherwise optimized depending on the target of the specific analysis. A
cut on the number of signal stations (above SNR threshold) is not necessary in this case, as for
these simulations, the core position parameters are fixed to the real core position for showers
with less than 5 signal stations. For real data this however is not the case and a cut might
be needed. The distribution is fitted with a linear function revealing a χ2 /ndf of 7.68. This
indicates that the uncertainty on σ taken directly from the LDF fit are either underestimated
or that there are additional influences playing a role, e.g., the background noise. To correct
for this, the uncertainties on σ are increased by quadratically adding 11.15 g/cm2 resulting
in a χ2 /ndf of 1.0. The modified uncertainties can then be propagated into the uncertainty
of Xmax . The distribution is shown together with the linear fit to the data and the linear re-
7.1. Xmax from the 2D Radio LDF 137

Table 7.1: Overview of selection cuts and the number of showers surviving these cuts starting with 5832
simulated showers. For the robustness against value variation, the values of the measured or simulated
energy densities are varied according to their uncertainties and the LDF is re-fitted 1000 times. If less
than half of these fit attempts are successful, the shower is rejected.

cut number of showers after cut

zenith angle > 55 ◦
4479
geomagnetic angle α > 10◦ 4479
σ < 270 m 4004
χ2 /ndf < 2 2662
uncertainty of σ < 10% 957
robustness against value variation 950

lation obtained from the generic simulations in figure 7.2. The width of the distribution of
σ = 53.5 ± 1.7 g/cm2 is taken as the resolution on Xmax . The similarity between the relation
found in the two independent simulation sets indicates that the added background noise has no
significant systematic influence on the reconstructed σ. The spread however increases as the
noise varies the reconstructed energy density at the individual station positions and also the
uncertainties of the energy density introduce more freedom for the minimization process.

7.1.2 Comparison of Simulations and Multi-Hybrid Data

For the reconstruction of Xmax using the lateral distribution function, the Pierre Auger Ob-
servatory provides a unique ensemble of detectors. On the one hand multi-hybrid RD-SD-FD
showers can be used to perform a calibration between radio and the FD independently of sim-
ulation results and on the other hand the large RD-SD hybrid dataset allows to reconstruct the
values of Xmax of measured air showers for which no FD information is available. To compare
the results obtained from simulations to hybrid data, a selection is made from the dataset of
multi-hybrid RD-SD-FD showers described in chapter 3.
For simulated air-showers, the value of Xmax as well as the value of Dmax geo
is based on a
parametrized atmospheric model. During the reconstruction of measured data, the atmospheric
overburden at the time of the shower is extracted from GDAS (see section 2.2). This informa-
tion is used when fitting the Gaisser-Hillas function to the FD data and thereby determining
Xmax . It is also used for its transformation into a geometric distance using
(hGDAS (Xmax / cos θ) − hAuger )
geo
Dmax (Xmax , θ) = . (7.1)
cos θ
Here, hGDAS (Xmax ) is the vertical geometrical height above sea level for a given GDAS profile
and a given shower with a certain Xmax and zenith angle θ. hAuger is thereby the elevation of
the AERA site above sea level.
138 Chapter 7. Radio Xmax Reconstruction

180

number of showers
14 linear fit
fit to generic sim 160
12
140
10 120
(km)

8 100
Dmax
geo

80
6
60
4
40
2
20

0 0
0 50 100 150 200 250 300 -300 -200 -100 0 100 200 300
σ (m)
reco
Xmax -Xmax (g/cm2)

Figure 7.2: Xmax determination from the purpose-made AERA124 simulations including measured
geo
noise. Left: Correlation between the geometric distance to Xmax Dmax and the width of the LDF σ
together with the best linear fit (solid line) and the linear relation obtained from the generic simula-
tions in figure 5.15 (dashed line). Right: Absolute differences between the reconstructed and the true
value of Xmax . The distribution is fitted with a Gaussian function with µ = −0.3 ± 1.7 g/cm2 and
σ = 53.5 ± 1.7 g/cm2 .

Table 7.2: Overview of selection cuts and the number of showers surviving these cuts starting with 105
RD-SD-FD multi-hybrid showers.

cut number of showers after cut

LDF fit not successful 102
≥ 5 stations with signal 30
geomagnetic angle α > 10◦ 30
σ < 270 m 30
uncertainty of σ < 20% 28

The RD-SD-FD multi-hybrid dataset contains 105 showers of which 52 have 3 signal stations
(above SNR threshold), 23 showers have 4 signal stations and 30 showers have 5 or more sig-
nal stations. Three showers with 3 signal stations do not have a successful LDF fit. To ensure
a high reconstruction quality and to stay independent of influences of the SD reconstruction in
terms of the shower core position, only showers with 5 or more signal stations are used for the
following analysis. From this subset, showers are selected based on the radio quality cuts in-
troduced in table 7.2. These cuts are less strict than the ones applied to the simulation dataset
in order to increase the number of showers. In addition it is checked, that the atmospheric
electric fields are not enhanced by thunderstorm conditions. A detailed description on this cri-
terion can be found in [164]. Nineteen of the selected showers are found to be not associated
with thunderstorm conditions, whereas for 9 of the selected showers no atmospheric electric
field data are available. The distribution of the arrival directions of the selected cosmic rays
as well as the distributions of the energy and Xmax as measured with the Surface Detector and
7.1. Xmax from the 2D Radio LDF 139

90° 5 6

135° 45° 4 5

number of showers

number of showers
4
50 ◦ 3
30 ◦

10 ◦ 3
180° 0° 2
2
1
1
225° 315° 0 0
17.4 17.6 17.8 18.0 18.2 18.4 18.6 600 700 800 900 1000
270° log10(energy (eV)) Xmax (g/cm2 )

Figure 7.3: Parameters of the shower set for the comparison of the reconstructed values of Xmax using
the width of the LDF and the FD measurement. Left: Arrival directions taken from the SD. Center:
Energy distribution taken from the SD. Right: Xmax distribution taken from the FD.

the Fluorescence Detector, respectively, are shown in figure 7.3. The selected air showers are
mainly arriving from the south-west. This is related to three aspects. First, the domination of
the geomagnetic emission mechanism leads to a suppression of showers being detected from
the North i.e., parallel to the geomagnetic field. Second, the Fluorescence Detector quality cri-
teria do not allow showers coming directly towards the telescopes which suppresses showers
from the East. And third, the antenna spacing of AERA favours inclined showers with larger
footprints to be detected by multiple stations. The distributions of energy and Xmax resample
those for the whole dataset (see figure 3.10) except for a higher minimum energy of 1017.2 eV
caused by the applied selection criteria.
The selected dataset is used to experimentally verify the results of the simulation studies.
Therefore, the correlation between σ and Dmax geo
is determined for this set analogue to the pro-
cedure for the simulations. The uncertainties of Xmax based on the radio LDF are propagated
from the modified uncertainties of the LDF fit. The uncertainties of Xmax measured with the
FD are taken as discussed in section 3.6. The corresponding correlation is shown in figure 7.4
on top of the results from the generic and the purpose-made AERA124 simulation studies. The
measured showers show a similar correlation compared to the AERA124 simulations and the
generic ones. The exact parameters of the linear fits are given in table 7.3 also indicating the
agreement of the results obtained from both simulation sets compared to the measured data.
The much larger number of showers in the purpose-made AERA124 simulation dataset results
in smaller uncertainties of the fit parameters. Therefore, the calibration obtained from this
simulation set is used to reconstruct the value of Xmax from the width of the LDF. The reso-
lution on Xmax which can be obtained from the measured air shower set is 59.3 ± 15.7 g/cm2
and thereby increased compared to the simulations. This can be attributed to the less strict
cuts applied to the set. Applying the same cuts to the AERA simulation set results in a com-
parable resolution. The reconstructed Xmax values as function of the Xmax values measured
with the FD are shown in figure 7.5 together with the corresponding residuals. The correla-
140 Chapter 7. Radio Xmax Reconstruction

12000

10000

8000
Dmax
geo
(m)

6000

4000 fit to sim

fit to data
2000
fit to generic sim
AERA sim
AERA data
0
0 50 100 150 200 250 300 350
σ (m)
geo
Figure 7.4: The distance to the shower maximum Dmax as a function of the with of the radio footprint
σ, based on RD-SD-FD multi-hybrid showers (red squares and red dashed line). In the background the
data obtained from simulations as shown in figure 7.2 are plotted.

geo
Table 7.3: Fit parameters obtained for the linear correlation Dmax = p0 σ+p1 for the different simulated
and measured datasets.

dataset p0 p1 (m)
generic Sim 64.6 ± 0.1 −3297 ± 2
AERA124 Sim 65.0 ± 0.8 −3129 ± 121
AERA124 measurement 67.5 ± 9.6 −3755 ± 1217

tion of the data points is 0.76. The values of Xmax based on the radio LDF show an offset
of −12.0 ± 13.4 g/cm2 compared to the FD measurements. Even though the value is still
compatible with 0, investigating possible systematic effects is subject of ongoing studies.

7.2 Xmax from Simulated Energy-Density Profiles

In addition to the reconstruction of Xmax based on the fit of the parametrized LDF description,
a more fundamental approach was developed by the LOFAR-collaboration by fitting detailed
Monte Carlo simulations to the measured data [71]. For this method a typical resolution for
the determination of Xmax of about 17 g/cm2 is stated for LOFAR with its high antenna density
7.2. Xmax from Simulated Energy-Density Profiles 141

1200
8 σ =59.3 ± 15.7 g/cm2

number of showers
7
1100 6
5
4
3
1000 2
1
0
900 −200 0 200
Xmax -Xmax
(g/cm2 )

RD FD
(g/cm2 )

800
Xmax
RD

700

600

500

400
σRD + σFD
2
(∆Xmax)2

4
3
2
2

1
0
sgn(∆Xmax)

−1
−2
−3
−4
400 500 600 700 800 900 1000 1100 1200
X FD
max (g/cm )
2

Figure 7.5: Comparison between the reconstructed depth of the shower maximum Xmax based on the
width of the radio LDF and the measurements of the Fluorescence Detector. The dashed line represents
the bisecting line. In the bottom panel, the residuals are shown. The uncertainties do not include the
systematic uncertainties of the standard FD measurement. In the inset, the distribution of the differences
of the reconstructed values is shown. The solid line represents a fit with a Gaussian function with
µ = −12.0 ± 13.4 g/cm2 and σ = 59.3 ± 15.7 g/cm2 .

with a theoretical duty cycle of almost 100%. The application of the method to AERA data
allows high-resolution Xmax measurements and the results are compared to FD data.

7.2.1 Basic Principle of the Method

The basic idea behind the method is to produce a set of air shower simulations resampling
the measured air-shower. The simulations are produced with the generic star-shape antenna
142 Chapter 7. Radio Xmax Reconstruction

pattern, as discussed in section 3.8. For these simulations, the calculated energy density in
the shower plane is interpolated on a two-dimensional map and fitted to the measured energy
density by varying the core positions and the scaling. For better accuracy, information acquired
with the Surface Detector array is included and compared to the simulated particle distribution
at ground level in a similar way, using a one-dimensional particle LDF. The best value for
Xmax is then taken from the fitted minimum of the reduced chi-square distribution from the
entire simulation set. A detailed description of the individual steps of the method, modified
for AERA is outlined in the following.

7.2.2 Individual Shower Simulations

For every measured air shower, 20 simulations for showers induced by protons and 10 simula-
tions for showers induced by iron nuclei are produced with CORSIKA and CoREAS, based on
the energy and direction reconstructed from the SD data. In the simulations, the height of the
first interaction of the cosmic ray in the atmosphere is determined by the hadronic interaction
model using different reproducible random seeds for different showers. Thereby, a broad but
realistic (under the assumption that the interaction models are correct) range of heights of first
interactions and further shower developments is achieved. The Xmax values from a simulation
set of 30 showers cover the value of Xmax measured by the FD in most cases. If this is not
the case, CONEX (see section 1.2.3) is used to find the random seeds for which the simulated
showers are likely to have a Xmax value in the region measured by the FD. These seeds are used
as input for the production of additional simulations. For this analysis, the CONEX option is
only used in case the original distribution of simulated Xmax values is not covering the value
of Xmax measured by the FD. All showers are simulated using the US standard atmosphere
parametrized according to Lindsey [53].

Simulated Electric Field and Energy Density

The simulated electric-field traces are processed to be comparable to measured data. There-
fore, the traces are down sampled to the experimental time resolution and filtered between 30
and 80 MHz. The peak position is defined as the maximum of the Hilbert envelope of the
magnitude trace of the electric field. Around this peak position, a 100 ns broad window is
used to determine the total energy density of the three-dimensional electric field as well as the
energy density in the individual polarizations. The total energy density is then interpolated in
~ vs. ~v × (~v × B)
the ~v × B ~ plane.

Simulated Lateral Particle Distribution

In addition to the radio information, particle information can optionally be included in the pro-
cedure if available. The method is in principle similar to the one for the radio data, but uses
a different reconstruction procedure. To be able to compare particle simulations to measured
7.2. Xmax from Simulated Energy-Density Profiles 143

data, a common quantity (such as the energy density for the radio emission) needs to be cho-
sen and reconstructed for simulations and measurements. For the SD, the standard quantity for
particle signals is the number of vertical equivalent muons (VEM) as it is discussed in chap-
ter 4. To determine the number of VEM from simulations, the detector response is applied
to the simulated particles on the ground. Due to the thinning mechanism used during the air-
shower simulation, a de-thinning procedure [169] is performed before applying the detector
simulation for the SD stations [170]. The full GEANT4 simulation for a large number of water
Cherenkov detector stations is very computational expensive and therefore look up tables are
used for the detector simulation1 .
The classical particle LDF has a one-dimensional form (see NKG function 1.23) and does not
account for geometrical effects for inclined showers where on one side of the axis the shower
develops further in the atmosphere before reaching the ground than on the other side. This
breaks the rotational symmetry of the LDF and leads to a biased reconstruction of the core
position for inclined showers. To lower the impact of the asymmetry, a ground pattern of sim-
ulated detector stations is used that forms a regular 18-arm star in the shower plane. The arms
have an angular spacing of 20◦ and an increasing radial distance. As the number of particles
decreases with the distance to the shower axis, the area which is used to resample the particles
has to be increased. Here, a value of dR = ±10% of the radial distance of the stations to
the shower axis defines the resampling area together with a central angle of ±10◦ . The radial
spacing between two neighbouring stations with R1 < R2 is then set by R2 = 1.23R1 . After
the simulation of the response of all stations in this dense array to the simulated particles, a
trigger simulation is made and sub-threshold stations are rejected. The generated data for the
remaining stations are used to reconstruct an one-dimensional LDF. This is done by using the
standard SD reconstruction module sequence (see section 3.4.1), where for the special case of a
dense array individual reconstruction modules need to be modified to not reject these stations.
An example LDF including all stations from the dense star-shape grid is shown in figure 7.6.
The LDFs of all simulations can then be included in a combined fit mechanism as explained in
the following. It should be noted that the reconstructed standard LDF for the Surface Detector
array is only valid for air showers with zenith angles smaller than 55◦ . Therefore, the particle
information will only be included in the fitting procedure for showers with zenith angles below
this limit.

7.2.3 Data Selection

From the whole RD-SD-FD multi-hybrid dataset presented in section 3.6, 80 air showers are
reconstructed. The showers are required to have more than 4 radio signal stations, a successful
radio LDF fit (without applying the additional quality cuts), and an uncertainty of the value
of Xmax obtained from the FD of less than 40 g/cm2 . In addition to the standard RdObserver
1
TabulatedTankSimulator module of the Off line framework
144 Chapter 7. Radio Xmax Reconstruction

Signal [VEM]
3
10 signal station
saturated station

102

10

0 200 400 600 800 1000 1200 1400 1600

r [m]

Figure 7.6: Lateral particle distribution reconstructed from a CORSIKA air shower simulation pro-
cessed in Off line. The simulated signal is plotted in units of vertical equivalent muons (VEM) as
a function of the distance to the shower axis. As simulated detector array, an idealized 18-arm star
grid with increasing radial distances between the stations is used. The saturated stations are recovered
during the reconstruction. See text for details on the simulations.

reconstruction, the RFI cleaning method as discussed in section 3.5.3 is applied. Only a part
of the selected showers are actual high-quality multi-hybrid events, the other air showers are
additionally selected for test purposes. The direction and energy needed as input for the sim-
ulation is taken from the SD reconstruction to test the method already for the case that no FD
information is available. Due to the low number of candidate showers, no further selection
on the zenith angle is made here, even though for 13 inclined showers in the set the particle
information cannot be included in the fit as the zenith angle exceeds 55◦ .
For the fit procedure itself, a station selection is applied on a shower-by-shower basis. Hereby,
all stations with a signal pulse (SNR of ≥ 10, as defined in chapter 3), are selected together
with a number of sub-threshold stations that were up and running but registered no signal
above the threshold. The sub-threshold stations are only considered if their position is less
than 500 m away from the shower axis and included for the same reason as in chapter 6 for
the topic of the 2D LDF fit which becomes unreasonably flat and broad for showers, where
all signal stations measured roughly the same energy density. For the simulation based Xmax
reconstruction, this case leads to misleading best fits of the broadest simulated energy density
profile. Furthermore, the possible shift of the core position to optimize the fit is limited in
case the signal stations are surrounded by sub-threshold stations. The uncertainty of the mea-
sured energy density is calculated from the energy density of the noise in a correspondingly
long time window added in quadrature to the uncertainty introduced from the signal chain of
5% (see also chapter 6). Saturated radio detection stations are excluded from the analysis.
Together with further quality cuts summarized in table 7.4, an event rejection based on the
7.2. Xmax from Simulated Energy-Density Profiles 145

atmospheric electric field conditions is applied at a later stage. For the data from the SD, the
signal stations are taken as they are reconstructed in Off line for the Infill array, in addition sta-
tions marked as accidental are discarded and for saturated stations the recovered signal values
are used.

7.2.4 Fit Procedure for the Reconstruction of Xmax

Following the procedure in [71], the two-dimensional energy-density profile and the particle
LDF are fitted simultaneously to the RD and the SD data in a least χ2 fit, using:
X  uRDS − Sr usim (~rRDS − ~rcore ) 2
2
χ =
RD stations
σRDS
2 (7.2)
pSDS − Sp psim (~rSDS − ~rcore )
X 
+ .
SD stations
σ SDS

Here, uRDS is the measured energy-density of a radio detection station (as defined in equa-
tion 3.9) at ground position ~rRDS with its uncertainty σRDS as discussed above and a scaling
parameter Sr to account for the uncertainty of the input energy, possible offsets in the absolute
antenna calibration, and the absolute scale of the simulations. Analogue, pSDS is the measured
particle signal of a Surface Detector station at ground position ~rSDS with uncertainty σSDS and
a scaling factor Sp to account for the uncertainty of the input energy and the overall offset
of the reconstructed LDF from simulation due to missing muons. For the fit procedure, the
core position ~rcore is varied on the ground plane in every step and the system is then trans-
formed to the ~v × B ~ vs. ~v × (~v × B)
~ frame to calculate the χ2 . In case of inclined showers
with zenith angles above 55◦ , the fit is performed based on the radio data only which reduces
equation 7.2 to its first summand. For every simulation the procedure is repeated resulting
in a χ2 /ndf-distribution as function of the Monte Carlo Xmax value. To obtain the final Xmax
value from this distribution, a parabola is fitted to the data points within ±200 g/cm2 around
the simulation with the lowest χ2 /ndf value. If the fit does not converge or returns a negative
parabola, the range is gradually extended in steps of ±50 g/cm2 until the fit is successful or a
total range of 500 g/cm2 is reached where the procedure is aborted. The minimum of the fit is
taken as the best Xmax estimate from this method.

7.2.5 Application to Multi-Hybrid Data

The outlined method is applied to the selected set of 80 air showers and exemplary, the fit
results for two air showers are discussed in detail in the following. The measured radio data
together with the best fitting simulation for these two showers are displayed in figure 7.7. The
footprints as spatially sampled with AERA are indicated by the large coloured circles together
with the interpolation (background colour) of the best fitting (lowest χ2 /ndf) simulation (small
circles). For one of the shown air showers the initial cosmic ray has an arrival direction from
146 Chapter 7. Radio Xmax Reconstruction

600
50
sub-thres data sim 40 sub-thres
CoREAS sim
position in v ×(v ×B)-direction [m]

400 35 40 AERA data

energy density (eV/m2 )

energy density (eV/m2 )
30
200
25
30
0
20
20
200 15

10 10
400
5
0
600 0
600 400 200 0 200 400 600 0 100 200 300 400 500
position in v ×B-direction [m] distance to shower axis (m)
600
25
sub-thres data sim 32 sub-thres
CoREAS sim
position in v ×(v ×B)-direction [m]

400 28
20 AERA data
energy density (eV/m2 )
energy density (eV/m2 )

24
200
20 15
0 16
10
200 12

8 5
400
4
0
600 0
600 400 200 0 200 400 600 0 100 200 300 400 500
position in v ×B-direction [m] distance to shower axis (m)

Figure 7.7: Reconstructed radio lateral distribution function for two example showers. Left: The best
fitting CoREAS simulation (small circles) interpolated in the shower plane (background color map) as
fitted to the measured energy densities by the AERA stations (large circle), sub-threshold stations are
additionally marked by a "+". The reconstructed position of the shower axis is shown by a small "×"
at (0,0). Right: One-dimensional representation of the simulated and measured LDF. The sub-threshold
stations are marked by crosses.

the South (zenith angle: 49.6◦ , azimuth angle: 292.6◦ ) and for the other one from the North
(zenith angle: 42.5◦ , azimuth angle: 95.4◦ ). Thereby, the second shower developed almost
parallel to the geomagnetic field which results in the very distinct footprint shape. The one-
dimensional representation of the LDF has in this case the maximum of the energy-density
pattern about 100 m away from the shower axis and the station closest to the shower axis has
measured a significantly lower energy density. Using simulated energy-density patterns to fit
the data automatically takes these effects into account and no further restrictions on the shower
geometry are necessary for this method. Only when using the method in hybrid mode, a re-
striction might originate from e.g., the particle detector reconstruction.
In addition to the radio measurement, also the particle data are used in the fit procedure for pre-
sented events. In figure 7.8 (left), the particle counterparts to the radio measurements shown
7.2. Xmax from Simulated Energy-Density Profiles 147

1.2
103 SD LDF
LDF CORSIKA sim 1.0
SD data

dE/dX (PeV/(g/cm2 ))
with optimized core 0.8
signal (VEM)

2
10
0.6

0.4
101 0.2

0.0

100 0 0.2
100 200 300 400 500 600 700 800 900 200 400 600 800 1000 1200
distance to shower axis (m) atmospheric depth (g/cm2 )
103 2.5
SD LDF
LDF CORSIKA sim 2.0
SD data dE/dX (PeV/(g/cm2 ))
102
with optimized core 1.5
signal (VEM)

1.0

101 0.5

0.0

100 0 0.5
200 400 600 800 1000 1200 1400 1600 200 400 600 800 1000 1200
distance to shower axis (m) atmospheric depth (g/cm2 )

Figure 7.8: Auger baseline measurements of the two air showers shown in figure 7.7. Left: Particle
signal in vertical equivalent muons (VEM) as function of the measured (circles) and fitted (squares)
distance to the shower axis together with the original reconstructed LDF (red solid) and the LDF based
on the best fitting simulation (blue dashed). Right: Longitudinal shower-profile as measured with
the Fluorescence Detector (see section 1.2.5 for details on the technique). The energy deposition in
the atmosphere is depicted as function of the atmospheric depth. The vertical band represents the
reconstructed value of Xmax together with the corresponding uncertainty.

in figure 7.7 are plotted including the result from the combined fit regarding the core position
and the particle LDF. Depending on the size of the particle footprint and thereby the number
of triggered SD stations, the influence of the particle distribution on the sensitivity to Xmax is
in general limited. In some cases the fit procedure is however dominated by the influence of
the particle signal. For these cases the fit can be performed in a radio-only mode. In general it
is however beneficial to include the information in a combined fit as it helps to constrain the
geometry e.g., when the shower core position is outside of the radio array. As the particle SD
array is larger than AERA, the core in this cases is always confined in the particle array. In
figure 7.8 (right), the longitudinal profiles of the corresponding air showers as measured with
the Fluorescence Detector are shown. The atmospheric depth of the maximum shower de-
148 Chapter 7. Radio Xmax Reconstruction

10 10
SD Xmax(AERA) p
RD Xmax(FD) Fe
8 8

6 6
2 /ndf

2 /ndf
4 4

2 2

0 0
550 600 650 700 750 800 850 900 500 550 600 650 700 750 800 850 900
Xmax (g/cm2 ) Xmax (g/cm2 )
9 7
8
SD Xmax(AERA) p
RD 6 Xmax(FD) Fe
7
5
6
5 4
2 /ndf

2 /ndf

4 3
3
2
2
1 1

0 0
600 650 700 750 800 850 900 950 600 650 700 750 800 850 900 950
Xmax (g/cm2 ) Xmax (g/cm2 )

Figure 7.9: Distribution of the fit quality for the simulated showers against the depth of the shower
maximum Xmax as taken from the simulations. Left: Fit qualities of the radio (triangles) and particle
(squares) parts separated for the two showers presented in 7.8. Right: Reconstructed value of Xmax
from the combined radio and particle fit. The dots (proton primaries) and squares (primary iron nuclei)
represent the simulations. The solid line is a parabolic fit to the data points around the simulation
which yields the lowest χ2 /ndf. The dashed vertical line represents Xmax taken as the minimum of the
parabola together with the one sigma uncertainty on this value. The solid vertical line represents Xmax
as measured with the FD together with its uncertainty.

velopment is obtained by fitting the Gaisser-Hillas function (see equation 1.24) to the shower
profiles.
The resulting χ2 /ndf distributions for all simulations for the two example showers are plot-
ted against the true Xmax value and shown in figure 7.9. Here the results for the particle as
well as the radio χ2 /ndf are shown together with the combined distribution including the fitted
parabola and the determined value for Xmax . The uncertainty is calculated as described in
section 7.2.6. In both cases the reconstructed value of Xmax agrees with the FD measurement
within the uncertainties.
7.2. Xmax from Simulated Energy-Density Profiles 149

90° 6 12

135° 45° 5 10

number of showers
number of showers
4 8
50 ◦
30 ◦

10 ◦ 3 6
180° 0°
2 4

1 2

225° 315° 0
0 500 600 700 800 900 1000 1100
17.2 17.4 17.6 17.8 18.0 18.2 18.4
270° log10(energy (eV)) Xmax (g/cm2 )

Figure 7.10: Parameters of the shower set for the comparison of the reconstructed values of Xmax ,
using the simulated radio energy-density profiles and the FD measurement. Left: Arrival directions
taken from the SD. Center: Energy distribution taken from the SD. Right: Xmax distribution taken
from the FD.

Table 7.4: Overview of the quality cuts for the comparison between the reconstructed Xmax values
based on the fit of simulated radio energy-density profiles and the FD measurement. The selection
starts with a set of 80 showers.

cut number of showers after cut

SD-FD energy calibration quality cuts 32
minimum in χ2 /ndf-distribution 26
more than 75% of the simulations converge in the fitting 25

For the comparison of the reconstructed values of Xmax with the Fluorescence Detector data,
additional quality cuts targeting the baseline reconstruction need to be applied to the set. For
the 80 showers for which simulations were run, 32 fulfil the quality criteria for the SD-FD
energy calibration as described in section 3.6. For these 32 showers, the χ2 /ndf-distribution of
6 of them does not show a minimum but keeps continuously de- or increasing over the whole
covered Xmax range. One additional shower is deselected as in the fitting procedure for more
than half of the simulated profiles no minimum could be found. Table 7.4 summarizes the
selection criteria of the considered showers. The distribution of the arrival directions of the
25 cosmic rays in the final selection as well as the distributions of the energy, and Xmax as
measured with the Surface Detector and the Fluorescence Detector, respectively, are shown in
figure 7.10. The asymmetry in the distribution of the arrival directions of the selected showers
is similar to the one discussed in section 7.1.2.
Using the presented dataset, the individually reconstructed values of Xmax obtained from the
FD and the radio measurements are compared. The resulting scatter plot of Xmax RD
vs. Xmax
FD
is
shown together with the corresponding residuals in figure 7.11. A correlation with a corre-
lation coefficient of 0.83 is found and the values from the radio reconstruction are consistent
with the measurements of the Fluorescence Detector. The resolution taken as the width of
150 Chapter 7. Radio Xmax Reconstruction

the distribution of Xmax

RD FD
− Xmax is σ = 47.0 ± 16.9 g/cm2 . The mean of the distribution is
µ = 4.4 ± 13.9 g/cm2 and thereby compatible with 0.
The method should be finally verified as soon as more multi-hybrid data that fulfil the qual-
ity cuts of the various detector components are available. Especially showers were the whole
longitudinal shower-profile is observed by the Fluorescence Detector are rare. This might also
be a geometrical effect similar as for the arrival directions of the selected shower set. Even
though all of the selected showers fulfil the SD-FD energy calibration quality cuts, for some
showers only a fraction of the longitudinal shower-profile was observed by the Fluorescence
Detectors. An example for such a shower is shown in figure 7.12 which also represents the
shower with the largest residual in the used set. In addition, the shower with the second largest
residual is also shown for which the profile is measured with a higher quality.

7.2.6 Uncertainty of the Reconstructed Value

The uncertainty on the reconstructed value of Xmax is not taken from the χ2 /ndf-distribution
but determined by applying the method to shower data generated from the simulations. For
each simulated air shower, artificial radio data are produced by evaluating the interpolated en-
ergy density map at the selected AERA station positions with respect to the measured shower
core positions. Therefore, the signals are varied according to Gaussian fluctuations as given
by the original signal uncertainties. Similarly, particle data are produced by evaluating the
simulation-based lateral distribution function (reconstructed as described in section 7.2.2) at
the participating SD station positions. The generated SD signals are also varied according to
Gaussian fluctuations as given by the original signal uncertainties. The produced shower data
are then reconstructed analogue to real data by fitting all simulated energy-density profiles ex-
cept for the one used to produce the artificial shower data. As the Xmax value is known for
the simulated data, it can be compared to the reconstructed Xmax . Repeating the procedure for
all simulations, a distribution of the differences is obtained and the uncertainty on the recon-
structed value of Xmax for measured data is taken as the 68% quantile of this distribution.
An overall correction needs to be applied for the difference between the atmospheric profile
used in the simulation and the actual profile of the atmosphere in which the air shower devel-
oped. The atmospheric conditions at the time of measurement are obtained from the GDAS.
Figure 7.13 gives an example for the differences in the atmospheric profile between GDAS
and the US standard parametrization for one of the showers in the dataset. The difference of
Xmax is then determined as
1
∆X(h) = XGDAS (h), (7.3)
cos θ
using the zenith angle of the air shower, the atmospheric depth of a given altitude extracted
from GDAS XGDAS (h), and the geometrical hight h calculated from combining the Monte
Carlo Xmax with the atmospheric profile used in the simulation. This correction is possible
because the overall size of the energy density pattern on the ground is determined by the geo-
7.2. Xmax from Simulated Energy-Density Profiles 151

1200
6
σ =47.0 ± 16.9 g/cm2

number of showers
5
1100 4
3

1000 2
1
0
900 −200 −100 0 100 200
Xmax -Xmax
(g/cm2 )

RD FD
(g/cm2 )

800
Xmax
RD

700

600

500

400
σRD + σFD
2
(∆Xmax)2

4
3
2
2

1
0
sgn(∆Xmax)

−1
−2
−3
−4
400 500 600 700 800 900 1000 1100 1200
X RD
max (g/cm )
2

Figure 7.11: Comparison between the reconstructed Xmax based on the simulated energy-density pro-
files and the measurements of the Fluorescence Detector. The dashed line represents the bisecting line.
In the bottom panel, the residuals are shown. The uncertainties do not include systematic uncertainties
of the FD measurement. In the inset, the distribution of the differences of the reconstructed values
is shown. The solid line represents a fit with a Gaussian function with µ = 4.4 ± 13.9 g/cm2 and
σ = 46.7 ± 16.9 g/cm2 .

metrical distance to Xmax (see section 5.2.4). The typical correction is of the order of 20 g/cm2 .
The uncertainty associated with this correction is less than 1 g/cm2 . However, a difference in
the atmospheric profile also effects the whole shower development. From re-simulating a
subset of showers with an adapted atmosphere and thereby different deviations compared to
the GDAS profile, the reconstructed value of Xmax is affected by less than 3 g/cm2 . For the
152 Chapter 7. Radio Xmax Reconstruction

0.8 4.0
Xmax(AERA) Xmax(AERA) p
0.7 Xmax(FD) 3.5 Xmax(FD) Fe
0.6
dE/dX (PeV/(g/cm2 ))

3.0
0.5
2.5

2 /ndf
0.4
2.0
0.3
1.5
0.2
0.1 1.0

0.0 0.5
300 400 500 600 700 800 900 1000 1100 550 600 650 700 750 800 850 900
atmospheric depth (g/cm2 ) Xmax (g/cm2 )
2.5 18
Xmax(AERA) Xmax(AERA) p
Xmax(FD) 16 Xmax(FD) Fe
2.0
14
dE/dX (PeV/(g/cm2 ))

1.5
12
2 /ndf

1.0 10
8
0.5
6
0.0
4
0.5 2
400 600 800 1000 1200 500 600 700 800 900 1000 1100 1200
atmospheric depth (g/cm2 ) Xmax (g/cm2 )

Figure 7.12: χ2 /ndf-distributions and longitudinal profiles for the two showers with the largest residuals
in the subset used for the direct Xmax comparison.

correction of the atmospheric density profile, the arrival direction is used which also intro-
duces an uncertainty that depends on the value of Xmax itself and due to the secant relation
given in equation 7.3 also on the shower geometry. The contribution of this uncertainty de-
pends on the angular resolution and is determined during reconstruction for every individual
shower. The determination of the uncertainty originating from simulating the shower only for
the reconstructed arrival direction not taking the angular resolution into account is pending.
Uncertainties for the FD measurements are taken as presented in section 3.6. A summary of
the discussed uncertainties is given in table 7.5. In addition, the used simulation codes and
therein especially the hadronic interaction models introduce uncertainties as they are based
on extrapolations from values for lower energies. These uncertainties are not included in the
presented analyses. It is up to future work to extend the simulation sets to also include and
compare the different interaction models. The analysis is also only performed with simula-
tions based on CoREAS. Efforts to compare the results with a similar but not identical kind
of analysis based on ZHAires simulations are ongoing and no principle problems have been
7.2. Xmax from Simulated Energy-Density Profiles 153

16
14
12
10

∆X (g/cm2 )
8
6
4
2
0
20 5 10 15 20 25 30 35 40
height a.s.l. (km)

Figure 7.13: Difference in atmospheric depth as a function of the height above sea level between GDAS
and the US standard parametrization of the atmosphere. The GDAS profile is evaluated for March 6th,
2014 at 5:34 UTC.

Table 7.5: Overview of uncertainties on the reconstructed value of Xmax for the radio as well as for the
Fluorescence Detector reconstruction.

source of uncertainty σXmax

RD
GDAS profile < 1 g/cm2
correction for the simulated atmosphere < 3 g/cm2
angular resolution propagated from the SD
reconstruction
minimum in χ2 /ndf-distribution determined from
sim set
FD
missing Mie data (for a subset of the data) 6.2 g/cm2
photonstatistics, geometry, taken from FD
atmosphere, timing reconstruction
FD systematic uncertainty 8 − 11 g/cm2

identified so far. Also no indication for large discrepancies due to different simulation codes is
given by the authors in [71] describing the original application of the method to LOFAR data.
The average uncertainty of Xmax taken from all 55 successfully reconstructed showers is
34 g/cm2 . The increased average uncertainty compared to the results from LOFAR can mainly
be attributed to the sparser antenna station grid which results in only a handful of signal sta-
tions per event compared to hundreds of antennas measuring the radio signal per shower for
the LOFAR experiment. Nevertheless, the technique improves the achievable resolution for
Xmax measurements with AERA compared to other methods developed so far.
154 Chapter 7. Radio Xmax Reconstruction

1200

1100

1000
(g/cm )
2

900

800
LDF
max
X

700

600

500
500 600 700 800 900 1000 1100 1200

X sim
max
(g/cm )
2

Figure 7.14: Comparison of the reconstructed Xmax from the simulated energy-density profiles and
from the 2D LDF fit. The black dashed line represents the bisecting line, the red dotted line is the best
fit to the data (Xmax
LDF = (1.05 ± 0.20) X sim − (37 ± 137) (g/cm2 )).
max

7.2.7 Direct Comparison between both Methods

The two presented methods to determine Xmax can be cross checked against each other. It
should be noted, that the methods are not completely independent, as the parametrization of
the 2D LDF is based in the same type of simulations as directly fitted to the data and both
approaches use the same measured quantity, the energy density, for the reconstruction. For the
check, 40 showers from the set which also survive the quality criteria associated with the LDF
fit given in table 7.6 are selected. For the other showers either the simulation-based reconstruc-
tion does not succeed or the LDF reconstruction quality cuts are not fulfilled. The correlation
is shown in figure 7.14. Within the uncertainties, a linear fit to the data is in agreement with the
bisecting line. The scatter around the line has a mean of µ = 14.7 ± 6.0 g/cm2 and a width of
σ = 32 ± 8 g/cm2 . The seen offset is thereby compatible with the combination of the offsets
of the two radio reconstructions when compared to the FD measurement (see figure 7.5 and
figure 7.11). The width has a comparable value as seen in a similar comparison for LOFAR
data where the resolution is 38 g/cm2 [39].

7.3 Xmax Reconstruction using RD-SD Hybrid Data

Based on the calibration presented in section 7.1, Xmax is reconstructed for the whole RD-
SD hybrid dataset for which mainly no FD observations are available. All events observed
with three or more RDSs are potentially usable for the reconstruction of Xmax . For showers
detected with three and four RDS however, the core position needs to be taken from the SD
7.3. Xmax Reconstruction using RD-SD Hybrid Data 155

Table 7.6: Overview of selection cuts and the number of showers surviving these cuts starting with
6569 RD-SD hybrid showers.

cut number of showers after cut

successful 2D LDF fit (equation 5.2) 6192
zenith angle < 55◦ 4708
geomagnetic angle α > 10◦ 4701
≥ 5 stations with signal 1543
σ < 270 m 1410
uncertainty of σ < 20% 1203
no thunderstorm conditions 1074
(no data available for 230 events)

reconstruction which in many cases introduces additional scatter due to the uncertainties in
the SD core position reconstruction. Therefore, these events are not taken into account for the
following analysis. This hard cut has an impact on the energy distribution as well as the dis-
tribution of arrival directions as it requires the lateral distribution of the electric field with an
amplitude above the SNR to be wide enough. Thereby, the low-energy showers are cut away
and at the same time showers with a small zenith angles are disfavoured as their footprints are
small due to the small distances to Xmax . As before, a cut on the maximum zenith angle is
set to 55◦ . Furthermore, the same radio quality cuts as applied to the multi-hybrid dataset are
used for this set. The cut statistics are shown in table 7.6. The distributions of energy and
arrival direction in the resulting dataset are shown in figure 7.15. Remarkable is the horse-
shoe like population in the skyplot. This is due to the combination of the suppression of radio
emission for air showers arriving parallel to the geomagnetic field and the fact that for vertical
air showers, the distance to the emission region is small and therefore detections with high
radio station multiplicity are suppressed. The energy distribution of the selected showers has
a higher energy threshold compared to the total set. This is also due to the effects mentioned
before in combination with the minimum number of signal stations set to five.
Besides these influences on the distributions of energy and arrival direction, there is also a
possible influence on the measured distribution of Xmax . Deeply penetrating vertical (proton)
primaries induce an extensive air shower at lower altitude and thereby the air shower is not
fully developed before reaching ground. This on one hand leads to less energy in the radiation
as discussed in chapters 5 and 6 and on the other hand gives a lower detection probability with
a radio array due to the smaller footprint and the lower amplitudes. The full efficiency for
the radio detector depends on multiple air-shower parameters which are energy, zenith angle,
geomagnetic angle and Xmax , and also on which part of AERA the air shower was detected
in as the antenna spacing is not homogeneous. On top, it also needs to be kept track of the
uptime of the individual RDSs. Taking all these influences together further data reduction and
156 Chapter 7. Radio Xmax Reconstruction

103 90°

135° 45°
60 ◦
number of events

102 40 ◦
20 ◦

180° 0°
1
10

225° 315°
100
16.5 17.0 17.5 18.0 18.5 19.0 19.5
log10(energy (eV)) 270°

Figure 7.15: Distributions of energy (left) and arrival direction (right) for all air showers detected with
a signal above SNR threshold in five or more RDSs. The star in the skyplot indicates the direction of
the geomagnetic field at the site of AERA.

extended simulation studies are necessary to determine a fully unbiased dataset. For now, only
a plain representation of the reconstructed Xmax distribution is shown in the following as a
preliminary result, not including bias corrections and systematic uncertainties.
The comparison between the measured average value of the depth of the shower maximum
hXmax i and the predictions from simulation studies for different hadronic interaction models
gives an indication of the composition of the cosmic rays. To extract this information, the
reconstructed and selected air showers in the hybrid dataset presented above are used to deter-
mine hXmax i and the elongation rate for the AERA data. A histogram of the Xmax distribution
as well as a scatter plot of the reconstructed values of Xmax using AERA as function of the
cosmic-ray energy as measured with the SD for all selected events is shown in figure 7.16. The
overall distribution of Xmax is similar to the one obtained from the RD-SD-FD multi-hybrid
events (see figure 3.10). The shown uncertainties are determined by propagating the modi-
fied fit uncertainties of the LDF width parameter. The general shape indicates the increase in
Xmax with increasing energy and a decreasing overall spread. To compare the obtained distri-
bution with other experimental results and model predictions, the reconstructed Xmax values
are binned in energy and summarized in a profile histogram. The chosen energy range spans
over one decade from 1017.2 eV to 1018.2 eV and is split up in 5 bins which are equally spaced
in the common logarithm of the energy. The obtained values are directly compared to results
from experiments using optical detectors to determine Xmax like Tunka, Yakutsk, or the Auger
Fluorescence Detector. The corresponding plot is shown in figure 7.17 and also contains sim-
ulation results for proton and iron nuclei based on the hadronic interaction models EPOS-LHC
and QGSJET-II-04 (see section 1.2.3). The simulated lines should be interpreted as indicators
7.3. Xmax Reconstruction using RD-SD Hybrid Data 157

180 1600

160 1400
140 1200
number of showers

120
1000

Xmax (g/cm2 )
100
800
80
600
60
40 400
20 200
0 0
200 400 600 800 1000 1200 1400 1017 1018 1019
Xmax (g/cm2 ) energy (eV)

Figure 7.16: Reconstructed values of Xmax using the width of the LDF and the calibration presented
in section 7.1. Left: Distribution of the reconstructed Xmax values. Right: Scatter plot of Xmax as
function of the cosmic ray energy.

for the two extreme cases in terms of single cosmic-ray species. For the investigated energy
range, the hXmax i as determined from the AERA data is in agreement with the measurements
of the other experiments shown and also indicates the trend from a heavier towards a lighter
composition with increasing energy. The values determined with AERA are smaller than the
ones obtained with the Fluorescence Detector (Auger) which is in line with the offset seen in
figure 7.5. Correcting for this offset would shift the values upwards very close to the ones
obtained from the Fluorescence Detector. The relation between the average depth of the maxi-
mum shower development hXmax i for a given primary type and cosmic-ray energy is described
by the elongation rate (see section 1.2.2). It has been determined by extensive simulation stud-
ies for different hadronic interaction models and has a literature value of 50 − 60 g/cm2 per
decade of energy for a single cosmic-ray species in the energy range investigated by AERA
(e.g., [175]). Using a linear fit to the presented average values results in an elongation rate of
89 ± 14 g/cm2 per decade of energy which is very similar to the value obtained for example
from the Auger Fluorescence Detector of 85.0 g/cm2 per decade of energy [171] for the en-
ergy range from 1017 eV to 1018.3 eV.
These results are to be seen as a first effort to reconstruct Xmax and from their the composition
of the measured cosmic rays using the lateral distribution of the radio emission with AERA.
Further studies of the detection efficiencies and the systematic uncertainties are pending. The
obtained result is nevertheless a promising step on the way to composition measurements with
radio detectors.
158 Chapter 7. Radio Xmax Reconstruction

900
Auger
850 Yakutsk
Tunka
800 AERA

750
Xmax (g/cm2 )

700

650

600 QGSJET-II, p
QGSJET-II, Fe
550 EPOS-LHC, p
EPOS-LHC, Fe
500 16
10 1017 1018 1019 1020
energy (eV)

Figure 7.17: Average depth of the shower maximum hXmax i as function of the energy as measured
with AERA. In addition, the data from the Auger Fluorescence Detector [171], Tunka [172], and the
Yakutsk array [173, 174] are shown. The lines indicate hXmax i values obtained from simulations for
protons and for iron nuclei using the hadronic interaction models EPOS-LHC [49] and QGSJET-II-04
[148]. It should be noted here that the AERA dataset is not corrected for possible efficiency biases and
other systematic effects.
 Conclusions and Outlook
8
Rapid developments of the experimental as well as the theoretical aspects of the radio detec-
tion of extensive air-showers have drawn the attention from the cosmic-ray research commu-
nity towards the recent results of experiments like AERA, LOFAR, and Tunka-Rex. The radio
technique not only offers a competitive sensitivity to the three observables, direction, energy,
and particle type but also a duty cycle of 100%. Thereby, it is suited to become the key tool
to unriddle the nature and the origin of ultra high-energy cosmic rays. With an improved un-
derstanding of the microscopic and the macroscopic effects responsible for the radio emission
and propagation taking place in extensive air showers, big improvements became possible in
the theoretical prediction and also in the experimental reconstruction of the measured signals
and their interpretation (chapter 1). The high-quality multi-hybrid data from AERA and the
baseline detectors of the Pierre Auger Observatory thereby allow a precise calibration of prop-
erties reconstructed from the radio signals with complementary detection techniques.

The work presented in this thesis is contributing to the field of radio detection by targeting the
three observables, especially the particle type with AERA. On the way to this goal, hardware
and field work were carried out to extent the existing engineering array becoming the world’s
largest radio cosmic ray detector which AERA is today (chapter 2). A revised station design
and the enlarged array required testing the solar power system as well as the electromagnetic
compatibility of the electronics to be suitable for the planed application. Additionally, the
communication of the radio detection stations has been extended from an optical fibre to a
wireless network to bridge the large distances. All together the system integration efforts have
contributed to two efficient deployment campaigns and a smooth operation of AERA in its
extended stages with 124 and later 153 radio detection stations. The gained experience is also
valuable input for future hardware development. The small internal particle detectors deployed
with a number of radio detection stations are a reliable trigger source. Based on the measure-
ments with these detectors, information about the shower is obtained and can be included in
the data reconstruction (chapter 4). The size of the detectors is however small and the trigger
efficiency as well as the sensitivity to shower parameters would benefit from larger units.
Since 2011 and especially after the extension of AERA in 2013, thousands of air showers have
been measured in coincidence with the Surface Detector array and partly also with the Fluores-
cence Detector. The acquired dataset represents the largest number of cosmic rays measured
through there radio emission and simultaneously with a classical detection technique so far.
160 Chapter 8. Conclusions and Outlook

The development of a standardized reconstruction pipeline for these data is a major effort of
the whole AERA group and contributions were made in this thesis (chapter 3). The resulting
dataset also contains several hundred reconstructed multi-hybrid measurements of air showers
which is unique in its kind and is used for various analyses. From a basic comparison between
the arrival directions as reconstructed from the radio signal times and the arrival directions as
measured with the Surface Detector array, a resolution of better than 0.5◦ is obtained.
On the simulation side, great progress has been made by generating two large sets of simula-
tions for AERA. One set is based on the properties of the measured air showers using realistic
antenna positions on the ground. Combined with noise data from the time of the measurement,
this set is used for realistic studies and further development of the reconstruction of air shower
parameters. The other set is using a generic antenna array forming a regular star in the shower
plane. The special alignment of the simulated antenna positions allows to resample the 2D
energy-density pattern in great detail. Based on this set, observables are tested for their corre-
lation with shower parameters and second order effects in the radio emission processes have
been investigated.
The lateral distribution of the radio signal cannot be described in one dimension only. There-
fore, the parametrization of the lateral distribution was a long-standing challenge. With the
recent developments on the simulation site e.g., the star-shape antenna grid, and the detailed
measurements from the LOFAR telescope, this challenge has been tackled and an empirical
two-dimensional description is now available. A detailed simulation study reveals an unseen
sensitivity to the energy, the position of the shower axis, and the depth of the shower maximum
(chapter 5). A new observable which is the total energy in the radio signal taken as the integral
over the 2D lateral distribution function (LDF) is strongly correlated with the total energy in
the shower. Using this quantity yields a better resolution than using the maximum intensity
of the lateral distribution. Based on the general description of the LDF, a revised version has
been formulated for AERA, optimized for a less densely populated antenna station array. The
influences of this simplified model on the reconstructed parameters are investigated and found
to be small compared to the influence of the number of antennas and their spacing. The AERA
spacing is however dense enough to use the 2D LDF as a basis for the reconstruction of shower
parameters with a higher resolution than other methods available so far.
Employing the 2D LDF to reconstruct the energy radiated in the MHz frequency band for
measured hybrid data has been used by the AERA group for a calibration with the energy
measured with the Surface Detector array which in turn is calibrated with the Fluorescence
Detector (chapter 6). The radio energy shows a quadratic dependency on the shower energy
when corrected for geometric influences on the strength of the emission processes. This rela-
tion is expected for the coherent emission of the radiation. The reached energy resolution is
22% for the used dataset and is improved to 17% by using a high-quality subset. The energy
in radiation is only a small fraction of the total energy in the shower and a value of 15.8 MeV
for a shower with 1 EeV is found when the shower arrives perpendicular to the geomagnetic
 161

field. The energy in the radiation is not only a very good energy indicator, but also a quantity
that can be determined from first principles based on the laws of electro-magnetism. Thereby
it can be used to determine the energy scale of cosmic rays and to calibrate other detection
techniques.
Based on the results of simulation studies using the generic and the purpose-made simulations
for AERA, a calibration of the width of the lateral distribution as function of the distance to the
shower maximum has been performed for AERA for the first time (chapter 7). This calibra-
tion allows for the unique opportunity to compare the results obtained from radio data to the
measurements of the Fluorescence Detector. The amount of high quality RD-SD-FD multi-
hybrid data is at the moment unfortunately still too limited for a purely data driven calibration,
the simulation-based calibration is however in agreement with the available measurements of
the Fluorescence Detector. The confirmation of the method is important for AERA as well
as for other radio experiments which can not be cross checked with an independent detection
technique. The resolution on the depth of the shower maximum obtained from the calibration
is 53.5 ± 1.7 g/cm2 and limited by the relatively sparse antenna array. The application of the
method to the whole AERA RD-SD hybrid dataset has been used to measure the average depth
of the shower maximum for 5 bins in the energy range from 1017.2 − 1018.2 eV. Thereby, an
independently confirmed method to measure the depth of the shower maximum is used for
the first time to extract information about the cosmic ray composition from the radio signals.
The measurements exhibit a trend towards a lighter composition as a function of energy. The
obtained results are in perfect agreement with results from other experiments using classical
techniques operating in the same energy range. It is up to future work to incorporate the effects
on the measured distribution by means of efficiency, preferred shower geometries, and other
aspects which can influence the resulting average values.
A second technique to reconstruct the depth of the maximum shower development proposed
by the LOFAR collaboration is based on the agreement between simulations for the measured
energy and arrival direction of the cosmic rays and the measured data. Sets of simulations
covering the whole range of possible depths of the shower maximum are compared to the
measured values and the interpolated depth for which the best agreement is achieved, is taken
as the reconstructed value. The procedure has been adapted and developed further for the ap-
plication to data acquired by AERA and also the Surface Detector array. The resulting values
of the depth of the shower maximum are compared to the measurements with the Fluorescence
Detector (chapter 7). This comparison is of great importance to validate the method also for
other experiments. It has been shown here that the method works for AERA despite the limited
number of signal stations and provides a measurement of the depth of shower maximum with
an average uncertainty of 34 g/cm2 . A subset has been compared to data from the Fluorescence
Detector and the reconstructed values are in agreement within their uncertainties. Comparing
both method the reconstruct the depth of maximum shower development with AERA, a linear
correlation is found compatible with a slope of 1 and a spread of 32 ± 8 g/cm2 .
162 Chapter 8. Conclusions and Outlook

Altogether, the presented work contains answers to the questions of how to reconstruct air
shower parameters with a radio cosmic-ray observatory. Competitive resolution is achieved
for the measured arrival direction and the energy. The central scientific results of this thesis
are the two methods to determine the depth of the maximum shower development. The
methods are discussed in detail and the obtained values from AERA are validated by the di-
rect comparison to the measurements of the Fluorescence Detector. In addition, the results are
compared on a statistical basis to other experiments for the first time.

The next generation of radio cosmic-ray observatories will benefit from the achieved techno-
logical and methodical developments. Cost efficient, robust, and self sufficient radio detection
stations are essential for the construction of a large-scale detector-array targeting the highest
energies. For these stations, thorough system integration is of extreme importance and the
hardware developers will need to go new ways in the station design.
With the already very precise energy reconstruction of AERA, the focus of related work shifts
towards the determination of second order effects. This is a necessary step on the way towards
the determination of the energy scale of cosmic rays from first principles. Future work will
also target improvements on the parametrization of the lateral distribution function. Here, a
description based on the now well understood emission mechanisms and their interplay build-
ing up the observed density patterns could improve the agreement between data and model
even further. First efforts in these directions have already been carried out and early results
will be available soon.
On the shorter time scale, the focus of research at radio cosmic ray observatories is heading
towards the precise determination of the depth of the maximum shower development. Mainly
the combination of various shower development sensitive parameters offers great potential to
improve the sensitivity even further. The timing information and also the frequency content
of the pulses were found to be sensitive to the depth of the maximum shower development.
Combined with the lateral distribution function, these observables will be included in joined
analyses. The same holds true for the simulation-based reconstruction of the depth of the
shower maximum for which these information is also available from the simulations and can
be included in the fitting procedure.
The ongoing efforts in the radio detection of cosmic rays will allow to better understand the
origin of ultra high-energy cosmic rays. Right now, the foundations are laid but further work
is needed to pin point the objects responsible for these fascinating particles. As a final future
prospect of the contribution of the radio detection of cosmic rays to this challenge, the average
values of the depth of the shower maximum measured with AERA (figure 7.17) are converted
to the mean logarithmic mass number and shown together with compiled data from various ex-
periments in figure 8.1. The AERA measurements are in good agreement with the world data.
In addition, predictions of the average mass number from a small selection of proposed astro-
physical models are shown. The ankle model for example is based on the assumption that the
 163

5
published data GRB ankle
AERA data SNR & AGN WR-CRs
4 Fe

3
N
ln A

2

He
1

0 p

−1 16
10 1017 1018 1019
energy (eV)

Figure 8.1: Values of the mean logarithmic mass number hln Ai from the experimental data and as-
trophysical models as a function of energy. The bands of published experimental data as well as the
AERA data are computed using the two hadronic interaction models QGSJET-II-04 [148] and EPOS-
LHC [49]. The collection of published data is taken from [25]. The values for the different models
(ankle, GRB, SNR & AGN, and WR-CRs) are based on the work presented in [32, 176, 177, 178] and
taken partly from [25].

composition and the origin of the cosmic rays changes around the ankle at 4 × 1018 eV from
being mainly galactic iron nuclei towards extragalactic protons [32]. Also focussing on the
transition from galactic to extragalactic cosmic rays is the SNR & AGN model which assumes
that the lower energetic component of the cosmic radiation (E < 1017 eV) is accelerated by
supernova remnants and the highest energetic cosmic rays are of extragalactic origin, acceler-
ated by active galactic nuclei [176]. The GRB model in contrast to that assumes cosmic rays
are accelerated in highly relativistic jets originating from gamma ray bursts [177]. The last
model shown is the WR-CRs model which proposes the sources of high energy cosmic rays
to be Wolf-Rayet star explosions [178]. Further details on the shown models are given in the
corresponding references. The WR-CRs model shows the best agreement with the AERA data
and is also compatible with the compiled data of various experiments (taken from [25]) over
the whole energy range presented. In general also the other models are in agreement with this
compiled data but less favourable concerning the results from AERA. The presented AERA
data are preliminary and investigations of systematic effects are not yet considered. Neverthe-
less, the presented result shows the great potential of the radio technique to help answering the
fundamental questions about the nature of ultra high-energy cosmic rays.
 Appendix
A
A.1 Wireless Communication Hardware and Settings
Information about the installed hardware and the used settings for frequency, channel width,
and firmware for the AERA wireless communication network presented in chapter 2.

Table A.1: RDS with Ubiquiti COMs

Hardware Bullet M5 HP
Channel width 40 MHz
Polarization vertical
Firmware XM.v5.5.4
Antenna L-COM HG5827EG

Table A.2: Ubiquiti point-to-point CRS-Coihueco I

Hardware Bullet M5 HP (Titanium @ CO)

SSID AERACoihueco-CRS
Frequency 5.660 GHz
Channel width 40 MHz
Polarization vertical
Firmware XM.v5.5.4
Antenna Poynting Antennas K-GRID-003-06
166 Appendix

Table A.3: Ubiquiti point-to-point CRS-Coihueco II

Hardware Bullet M5 HP (Titanium @ CO)

SSID AERACoihueco-CRS#2
Frequency 5.600 GHz
Channel width 40 MHz
Polarization vertical
Firmware XM.v5.5.6
Antenna Poynting Antennas K-GRID-003-06

Table A.4: Access Points CRS

AP CRS-East CRS-West
Hardware Rocket M5 GPS Rocket M5
SSID AERAsectorE_CRS AERAsectorW_CRS
Frequency 5.785 GHz 5.805 GHz
Channel width 40 MHz 40 MHz
Polarization vertical vertical
Firmware XM.v5.5.4 XM.v5.5.4
Antenna Airmax Sector 5G-90-20 Airmax Sector 5G-90-20

Table A.5: Access Points Coihueco

Hardware Rocket M5 GPS Titanium Rocket M5 GPS

SSID AERAsector_Coihueco AERAsector_Coihueco2
Frequency 5.745 GHz 5.560 GHz
Channel width 40 MHz 40 MHz
Polarization vertical vertical
Firmware XM.v5.5.4 XM.v5.5.4
Antenna Airmax Sector 5G-90-20 Airmax Sector 5G-90-20
Appendix 167

A.2 Module Sequence RdObserver

Module sequence of the RdObserver which is used to reconstruct the RD-SD hybrid and RD-
SD-FD multi-hybrid datasets presented in chapter 3 and used in the analyses in chapter 7.

<loop numTimes="unbounded">
<module> EventFileReaderOG </module>
<module> RdEventPreSelector </module>
<module> EventCheckerOG </module>
<module> SdQualityCutTaggerOG </module>
<module> SdPMTQualityCheckerKG </module>
<module> TriggerTimeCorrection </module>
<module> SdCalibratorOG </module>
<module> SdBadStationRejectorKG </module>
<module> SdSignalRecoveryKLT </module>
<module> SdEventSelectorOG </module>
<module> SdPlaneFitOG </module>
<module> LDFFinderKG </module>
<try>
<module> SdHorizontalReconstruction </module>
</try>
<module> RdEventInitializer </module>
<module> RdStationPositionCorrection </module>
<module> RdStationRejector </module>
<module> RdChannelADCToVoltageConverter </module>
<module> RdChannelSelector </module>
<module> RdChannelPedestalRemover </module>
<module> RdChannelResponseIncorporator </module>
<module> RdChannelBeaconTimingCalibrator </module>
<module> RdChannelBeaconSuppressor </module>
<module> RdStationTimingCalibrator </module>
<module> RdStationTimeWindowConsolidator </module>
<module> RdChannelTimeSeriesTaperer </module>
<module> RdChannelBandstopFilter </module>
<module> RdChannelUpsampler </module>
<module> RdChannelRiseTimeCalculator </module>
<module> RdAntennaChannelToStationConverter </module>
<module> RdStationSignalReconstructor </module>
<module> RdStationEFieldVectorCalculator </module>
<loop numTimes="unbounded">
168 Appendix

<module> RdTopDownStationSelector </module>

<module> RdPlaneFit </module>
</loop>
<module> RdClusterFinder </module>
<module> RdPlaneFit </module>
<module> RdStationRiseTimeCalculator </module>
<module> RdEventPostSelector </module>
<module> RdLDFMultiFitter </module>
<module> Rd2dLDFFitter </module>
<try>
<module> FdCalibratorOG </module>
<module> FdEyeMergerKG </module>
<module> FdPulseFinderOG </module>
<module> FdSDPFinderOG </module>
<module> FdAxisFinderOG </module>
<module> HybridGeometryFinderOG </module>
<module> HybridGeometryFinderWG </module>
<module> FdApertureLightKG </module>
<module> FdEnergyDepositFinderKG </module>
<module> FdProfileReconstructorKG </module>
</try>
<module> RdStationTimeSeriesWindowCutter </module>
<module> RdStationTimeSeriesTaperer </module>
<module> RdREASSimPreparator </module>
<module> EventFileExporterOG </module>
<module> RecDataWriterNG </module>
</loop>
Appendix 169

A.3 Quality Cuts for the RD-SD-FD Multi-Hybrid Dataset

Quality cuts used for the energy calibration of the Surface Detector Infill array using SD-FD
hybrid events. These cuts are applied to the RD-SD-FD multi-hybrid data to extract high
quality Xmax measurements. The selected showers are used in the analysis in chapter 7.

Table A.6: SD-FD energy calibration cuts for the Infill array.

cut value
!lightning
minRecLevel 3
maxZenithSD 55◦
T4Trigger 2
T5Trigger 2
heatOrientationUp
eyeCut 101111
minLgEnergyFD 1e-20
skipSaturated
badFDPeriodRejection
maxVAOD 0.1
LidarCloudRemoval 25
MinCloudDepthDistance params: -50 50
MaxCloudThickness 100
!badPixels 1
xMaxObsInExpectedFOV params: 40 20
xMaxError 40.0
energyTotError 0.18
profileChi2Sigma 2.5 -1.1
maxDepthHole 20.
maxCoreTankDist 750m
170 Appendix

A.4 Scintillator Calibration

Values of the upper limits of the particle density for sub-threshold scintillator detectors and
calibration values for the deposited energy in the scintillator detectors by single muons. Both
quantities are calculated as discussed in chapter 4.

Table A.7: Upper limits (UL) for the integrated signal in sub-threshold stations.

No. UL Top UL Bottom No. UL Top UL Bottom

(VEM) (VEM) (VEM) (VEM)
34 0.60 0.60 110 0.64 0.96
35 0.66 0.84 111 0.66 0.58
36 0.62 0.68 112 0.82 0.74
43 0.60 0.62 113 0.62 0.60
44 0.52 0.56 122 0.46 0.62
49 0.70 0.60 123 0.50 0.60
58 0.64 0.64 124 0.58 0.64
59 0.54 0.64 125 0.58 0.84
60 0.52 0.58 135 0.52 0.52
69 0.62 0.92 136 0.50 0.48
70 0.42 0.50 137 0.60 0.64
83 0.56 0.66 144 0.54 0.56
84 0.50 0.66 145 0.40 0.70
85 0.62 1.14 151 0.50 1.52
92 0.62 0.64 152 0.56 0.58
93 0.60 0.80 88 0.58 -
94 0.64 0.66 103 0.58 -
100 0.58 0.78 126 0.58 -
101 0.62 0.82 146 0.58 -
102 0.64 0.84 157 0.58 -
Appendix 171

Table A.8: Values in units of fC extracted from the single muon calibration for the individual scintilla-
tors. The values for a combined distribution of all scintillators are displayed under <average>.

No. widthL, T peakT widthG, T widthL, B peakB widthG, B

<average> 0.08±0.01 0.91±0.01 0.35±0.01 0.09±0.01 0.76±0.01 0.34±0.01
34 0.07±0.01 0.87±0.01 0.30±0.02 0.07±0.01 0.87±0.02 0.34±0.02
35 0.06±0.01 0.81±0.02 0.30±0.02 0.08±0.01 0.68±0.01 0.26±0.02
36 0.07±0.01 0.85±0.01 0.28±0.02 0.08±0.01 0.80±0.01 0.29±0.02
43 0.09±0.01 0.93±0.01 0.29±0.02 0.07±0.01 0.84±0.01 0.31±0.02
44 0.06±0.01 0.99±0.02 0.44±0.02 0.08±0.01 0.92±0.01 0.30±0.02
49 0.05±0.01 0.98±0.02 0.49±0.03 0.07±0.01 0.84±0.02 0.31±0.02
58 0.06±0.01 0.94±0.02 0.30±0.02 0.08±0.01 0.78±0.01 0.28±0.02
59 0.10±0.01 0.99±0.02 0.33±0.02 0.06±0.01 0.76±0.01 0.28±0.01
60 0.07±0.01 0.99±0.01 0.30±0.01 0.06±0.01 0.84±0.01 0.30±0.02
69 0.06±0.01 0.88±0.01 0.30±0.02 0.06±0.01 0.55±0.03 0.45±0.04
70 0.08±0.01 1.25±0.02 0.41±0.02 0.07±0.01 1.13±0.02 0.43±0.02
83 0.07±0.01 0.92±0.02 0.34±0.02 0.09±0.01 0.79±0.02 0.43±0.03
84 0.08±0.01 1.06±0.01 0.33±0.02 0.08±0.01 0.76±0.01 0.30±0.02
85 0.08±0.01 0.99±0.02 0.35±0.02 0.09±0.01 0.63±0.02 0.28±0.02
92 0.08±0.01 0.84±0.01 0.29±0.02 0.09±0.01 0.82±0.01 0.31±0.02
93 0.08±0.01 0.85±0.01 0.27±0.02 0.07±0.01 0.73±0.01 0.31±0.02
94 0.05±0.01 0.80±0.02 0.34±0.02 0.09±0.01 0.74±0.01 0.23±0.02
100 0.08±0.01 0.93±0.02 0.31±0.02 0.09±0.01 0.79±0.01 0.27±0.02
101 0.05±0.01 0.88±0.02 0.34±0.02 0.09±0.01 0.67±0.01 0.26±0.02
102 0.07±0.01 0.84±0.01 0.27±0.01 0.06±0.01 0.58±0.01 0.25±0.02
110 0.06±0.01 0.90±0.02 0.30±0.02 0.08±0.01 0.56±0.01 0.27±0.02
111 0.05±0.01 0.78±0.02 0.36±0.02 0.07±0.01 0.85±0.02 0.33±0.02
112 0.08±0.01 0.65±0.01 0.24±0.02 0.13±0.01 0.73±0.01 0.25±0.03
113 0.07±0.01 0.78±0.01 0.27±0.02 0.05±0.01 0.86±0.02 0.34±0.02
122 0.07±0.01 1.11±0.02 0.42±0.02 0.07±0.01 0.80±0.01 0.28±0.02
123 0.06±0.01 1.04±0.01 0.31±0.01 0.09±0.01 0.86±0.01 0.29±0.02
124 0.07±0.01 0.90±0.02 0.34±0.02 0.08±0.01 0.76±0.01 0.26±0.02
125 0.06±0.01 0.88±0.02 0.32±0.02 0.06±0.01 0.71±0.01 0.31±0.02
135 0.05±0.01 0.99±0.02 0.36±0.02 0.06±0.01 0.96±0.02 0.39±0.02
136 0.08±0.01 1.04±0.02 0.39±0.02 0.15±0.02 1.05±0.02 0.39±0.03
137 0.07±0.01 0.83±0.01 0.26±0.01 0.07±0.01 0.79±0.01 0.27±0.02
144 0.06±0.01 0.99±0.02 0.35±0.02 0.10±0.01 0.89±0.01 0.31±0.02
145 0.13±0.02 1.28±0.03 0.49±0.04 0.09±0.01 0.72±0.01 0.25±0.02
151 0.14±0.02 1.03±0.02 0.33±0.04 0.09±0.01 0.26±0.01 0.00±0.03
152 0.07±0.01 0.95±0.02 0.32±0.02 0.08±0.01 0.87±0.02 0.33±0.02
172 Appendix

A.5 LDF Width Parameter Correlation

Parameters for the fifth-degree polynomial fit to the distribution of the atmospheric and the
geometric distance to Xmax as function of the width of the LDF σ(+) as discussed in chapter 5.
The fit parameters are obtained by either using the original LDF function (equation 5.1) or the
AERA parametrization (equation 5.2).

Table A.9: Atmospheric distance to Xmax - Dmax

atm

parameter full fit AERA

p0 -516.27 -571.323
p1 20.5449 23.8009
p2 -0.16958 -0.248084
p3 0.000545931 0.00143211
p4 4.06132e-07 -3.99593e-06
p5 -3.05614e-09 4.46421e-09

geo
Table A.10: Geometric distance to Xmax - Dmax

parameter full fit AERA

p0 -4.53748 -5.74906
p1 0.159968 0.229873
p2 -0.000627602 -0.00219756
p3 -4.21959e-06 1.22506e-05
p4 4.4592e-08 -3.28206e-08
p5 -9.05486e-11 3.67051e-11
Appendix 173

A.6 Likelihood Function of the Energy Calibration

Likelihood function as used in chapter 6 for the energy calibration of AERA. The likelihood
function (for one pair of radio signal Sradio and SD cosmic-ray energy estimate ESD ) has the
following form

1 X εSD (ESD , Θi ) εRD (ESD , Θi , Φi )

f (Sradio , ESD ) = ×
N i εSD (ESD,i , Θi ) εRD (ESD,i , Θi , Φi )
gRD (Sradio |S(ESD,i ), ...) ×
gSD−sh (ESD |ESD,i , Θi ) . (A.1)

The summation is performed over all events in the selected dataset. gRD (Sradio |S, ...) and
gSD−sh (ESD |E, Θ) are the conditional probability density functions, which describe the proba-
bility to measure a radio signal Sradio or energy ESD if the true radio signal, energy and zenith
angle are S, E and Θ. Φ denotes the azimuth angle. gRD (Sradio ) is obtained for each event in
a Monte Carlo simulation where all reconstructed parameters that influence the radio-energy
estimator are varied within their uncertainties. εSD (ESD , Θ) and εRD (ESD , Θ, Φ) are the effi-
ciencies of the surface and the radio detector. The radio efficiency has been determined with
Monte Carlo air-shower simulations and a full-detector simulation and depends on the energy,
the zenith and the azimuth angle. N is the normalization of the function to an integral of one.
Bibliography
[1] B. A BELEV et al., ALICE C OLLABORATION, Performance of the ALICE experiment
at the CERN LHC, International Journal of Modern Physics A, 29 (2014), p. 30044.

[2] J. L INSLEY, Evidence for a Primary Cosmic-Ray Particle with Energy 1020 eV, Physical
Review Letters, 10 (1963), pp. 146–148.

[3] V. F. H ESS, Über Beobachtungen der durchdringenden Strahlung bei sieben Freibal-
lonfahrten, Physikalische Zeitschrift, 13 (1912), pp. 1084–1091.

[4] P. AUGER et al., Extensive Cosmic-Ray Showers, Reviews of Modern Physics, 11

(1939), pp. 288–291.

[5] W. KOLHÖRSTER et al., Gekoppelte Höhenstrahlen, Naturwissenschaften, 26 (1938),

pp. 576–576.

[6] V. B ONVICINI et al., PAMELA C OLLABORATION, The PAMELA experiment in space,

Nuclear Instruments and Methods in Physics Research A, 461 (2001), pp. 262–268.

[7] M. AGUILAR et al., AMS C OLLABORATION, First Result from the Alpha Mag-
netic Spectrometer on the International Space Station: Precision Measurement of the
Positron Fraction in Primary Cosmic Rays of 0.5-350 GeV, Physical Review Letters,
110 (2013), p. 141102.

[8] T. S ANUKI et al., BESS C OLLABORATION, Precise Measurement of Cosmic-Ray Pro-

ton and Helium Spectra with the BESS Spectrometer, Astrophysical Journal, 545 (2000),
pp. 1135–1142.

[9] M. AVE et al., The TRACER instrument: A balloon-borne cosmic-ray detector, Nuclear
Instruments and Methods in Physics Research A, 654 (2011), pp. 140–156.

[10] H. S. A HN et al., CREAM C OLLABORATION, The Cosmic Ray Energetics And Mass
(CREAM) instrument, Nuclear Instruments and Methods in Physics Research A, 579
(2007), pp. 1034–1053.

[11] P. AUGER et al., Les grandes gerbes cosmiques de l’atmosphère, Proceedings of the
Academy of Sciences, 207 (1938), pp. 228–230.

[12] J. A. S IMPSON, Elemental and isotopic composition of the galactic cosmic rays, Annual
Review of Nuclear and Particle Science, 33 (1983), pp. 323–382.
176 Bibliography

[13] W. D. A PEL et al., KASCADE-G RANDE C OLLABORATION, Kneelike Structure in the

Spectrum of the Heavy Component of Cosmic Rays Observed with KASCADE-Grande,
Physical Review Letters, 107 (2011), p. 171104.

[14] J. R. H ÖRANDEL, On the knee in the energy spectrum of cosmic rays, Astroparticle
Physics, 19 (2003), pp. 193–220.

[15] J. R. H ÖRANDEL et al., The Knee in the Energy Spectrum of Cosmic Rays in the Frame-
work of the Poly-Gonato and Diffusion Models, in Proceedings of the 28th International
Cosmic Ray Conference, Tsukuba, Japan, 2003.

[16] Y. S. YOON et al., CREAM C OLLABORATION, Cosmic-ray Proton and Helium Spec-
tra from the First CREAM Flight, Astrophysical Journal, 728 (2011), p. 122.

[17] A. D. PANOV et al., Elemental energy spectra of cosmic rays from the data of the
ATIC-2 experiment, Bulletin of the Russian Academy of Science, Physics, 71 (2007),
pp. 494–497.

[18] O. A DRIANI et al., PAMELA C OLLABORATION, PAMELA Measurements of Cosmic-

Ray Proton and Helium Spectra, Science, 332 (2011), p. 69.

[19] R. U. A BBASI et al., H IGH R ESOLUTION F LY ’ S E YE C OLLABORATION, First Ob-

servation of the Greisen-Zatsepin-Kuzmin Suppression, Physical Review Letters, 100
(2008), p. 101101.

[20] W. D. A PEL et al., KASCADE-G RANDE C OLLABORATION, The spectrum of high-

energy cosmic rays measured with KASCADE-Grande, Astroparticle Physics, 36
(2012), pp. 183–194.

[21] M. A MENOMORI et al., T IBET ASγ C OLLABORATION, The All-Particle Spectrum of

Primary Cosmic Rays in the Wide Energy Range from 1014 to 1017 eV Observed with
the Tibet-III Air-Shower Array, Astrophysical Journal, 678 (2008), pp. 1165–1179.

[22] A. S CHULZ FOR THE P IERRE AUGER C OLLABORATION, The measurement of the
energy spectrum of cosmic rays above 3 × 1017 eV with the Pierre Auger Observatory,
in Proceedings of the 33rd International Cosmic Ray Conference, Rio de Janeiro, Brazil,
2013.

[23] P. S CHELLART, Measuring Radio Emission from air showers with LOFAR, PhD thesis,
Radboud University Nijmegen, 2015.

[24] J. B LÜMER et al., Cosmic rays from the knee to the highest energies, Progress in Particle
and Nuclear Physics, 63 (2009), pp. 293–338.
Bibliography 177

[25] K.-H. K AMPERT et al., Measurements of the cosmic ray composition with air shower
experiments, Astroparticle Physics, 35 (2012), pp. 660–678.

[26] T. A NTONI et al., KASCADE C OLLABORATION, KASCADE measurements of energy

spectra for elemental groups of cosmic rays: Results and open problems, Astroparticle
Physics, 24 (2005), pp. 1–25.

[27] D. D ’E NTERRIA et al., Constraints from the first LHC data on hadronic event genera-
tors for ultra-high energy cosmic-ray physics, Astroparticle Physics, 35 (2011), pp. 98–
113.

[28] T. K. G AISSER, The Cosmic-ray Spectrum: from the knee to the ankle, Journal of
Physics Conference Series, 47 (2006), pp. 15–20.

[29] A. A AB et al., P IERRE AUGER C OLLABORATION, The Pierre Auger Cosmic Ray Ob-
servatory, Nuclear Instruments and Methods, A798 (2015), pp. 172–213.

[30] W. H. BAUMGARTNER et al., The 70 Month Swift-BAT All-sky Hard X-Ray Survey,
Astrophysical Journal Supplement, 207 (2013), p. 19.

[31] A. A AB et al., P IERRE AUGER C OLLABORATION, Searches for Anisotropies in the

Arrival Directions of the Highest Energy Cosmic Rays Detected by the Pierre Auger
Observatory, Astrophysical Journal, 804 (2015), p. 15.

[32] V. B EREZINSKY et al., On astrophysical solution to ultrahigh energy cosmic rays,

Physical Review D, 74 (2006), p. 043005.

[33] K. G REISEN, End to the Cosmic-Ray Spectrum?, Physical Review Letters, 16 (1966),
pp. 748–750.

[34] G. T. Z ATSEPIN et al., Upper Limit of the Spectrum of Cosmic Rays, Soviet Journal of
Experimental and Theoretical Physics Letters, 4 (1966), p. 78.

[35] R. A LVES BATISTA et al., Effects of uncertainties in simulations of extragalactic

UHECR propagation, using CRPropa and SimProp, ArXiv e-prints, (2015).

[36] E. F ERMI, On the Origin of the Cosmic Radiation, Physical Review, 75 (1949),
pp. 1169–1174.

[37] A. M. H ILLAS, The Origin of Ultra-High-Energy Cosmic Rays, Annual Review of

Astronomy and Astrophysics, 22 (1984), pp. 425–444.

[38] J. C ANDIA et al., Turbulent diffusion and drift in galactic magnetic fields and the ex-
planation of the knee in the cosmic ray spectrum, Journal of High Energy Physics, 12
(2002), p. 33.
178 Bibliography

[39] A. N ELLES, Radio emission from air showers, PhD thesis, Radboud University Ni-
jmegen, 2014.

[40] A. M. H ILLAS, TOPICAL REVIEW: Can diffusive shock acceleration in supernova

remnants account for high-energy galactic cosmic rays?, Journal of Physics G Nuclear
Physics, 31 (2005), p. 95.

[41] A. A BRAMOWSKI et al., HESS C OLLABORATION, HESS J1640-465 - an exception-

ally luminous TeV gamma-ray supernova remnant, Monthly Notices of the Royal As-
tronomical Society, 439 (2014), pp. 2828–2836.

[42] M. ACKERMANN et al., F ERMIN -LAT C OLLABORATION, Detection of the character-

istic pion-decay signature in supernova remnants, Science, 339 (2013), pp. 807–811.

[43] M. G. A ARTSEN et al., I CE C UBE C OLLABORATION, Evidence for High-Energy Ex-

traterrestrial Neutrinos at the IceCube Detector, Science, 342 (2013), p. 1.

[44] M. G. A ARTSEN et al., I CE C UBE C OLLABORATION, First Observation of PeV-Energy

Neutrinos with IceCube, Physical Review Letters, 111 (2013), p. 021103.

[45] W. H EITLER, Quantum theory of radiation, 1954.

[46] J. M ATTHEWS, A Heitler model of extensive air showers, Astroparticle Physics, 22

(2005), pp. 387–397.

[47] J. R. H ÖRANDEL, Cosmic Rays from the Knee to the Second Knee:. 1014 to 1018 eV,
Modern Physics Letters A, 22 (2007), pp. 1533–1551.

[48] J. E NGEL et al., Nucleus-nucleus collisions and interpretation of cosmic-ray cascades,

Physical Review D, 46 (1992), pp. 5013–5025.

[49] T. P IEROG et al., EPOS LHC : test of collective hadronization with LHC data, ArXiv
e-prints, (2013).

[50] G. BATTISTONI et al., The FLUKA code: description and benchmarking, in Hadronic
Shower Simulation Workshop, M. Albrow et al., eds., vol. 896 of American Institute of
Physics Conference Series, 2007, pp. 31–49.

[51] S. O STAPCHENKO, QGSJET-II: towards reliable description of very high energy

hadronic interactions, Nuclear Physics B Proceedings Supplements, 151 (2006),
pp. 143–146.

[52] E.-J. A HN et al., Cosmic ray interaction event generator SIBYLL 2.1, Physical Review
D, 80 (2009), p. 094003.
Bibliography 179

[53] D. H ECK et al., CORSIKA: A Monte Carlo Code to Simulate Extensive Air Showers,
FZKA–6019, FZKA, 1998.

[54] S. J. S CIUTTO, AIRES: A system for air shower simulations (Version 2.2.0), ArXiv
e-prints, (1999).

[55] T. P IEROG et al., First results of fast one-dimensional hybrid simulation of EAS using
CONEX, Nuclear Physics B Proceedings Supplements, 151 (2006), pp. 159–162.

[56] L. G. D EDENKO, A new method of solving the nuclear cascade equation, in Proceed-
ings of the 9th International Cosmic Ray Conference, London, UK, 1965.

[57] A. M. H ILLAS, Calculations on the propagation of mesons in extensive air showers, in

Proceedings of the 9th International Cosmic Ray Conference, London, UK, 1965.

[58] G. B OSSARD et al., Cosmic ray air shower characteristics in the framework of the
parton-based Gribov-Regge model NEXUS, PRD, 63 (2001), p. 054030.

[59] G. A. A SKARYAN, Excess Negative Charge of the Electron-Photon Shower and Coher-
ent Radiation Originating from It. Radio Recording of Showers under the Ground and
on the Moon, Journal of the Physical Society of Japan Supplement, 17 (1962), p. C257.

[60] F. D. K AHN et al., Radiation from Cosmic Ray Air Showers, Royal Society of London
Proceedings Series A, 289 (1966), pp. 206–213.

[61] S. G REBE, Finger on the pulse of cosmic rays - dependence of the radio pulse shape on
the air shower geometry, PhD thesis, Radboud University Nijmegen, 2013.

[62] J. V. J ELLEY et al., Radio Pulses from Extensive Cosmic-Ray Air Showers, Nature, 205
(1965), pp. 327–328.

[63] F. G. S MITH et al., The detection of radio pulses of wavelength 6. 8 m, in coincidence

with extensive air showers, in the energy region 1016 -1017 eV, in Proceedings of the 9th
International Cosmic Ray Conference, London, UK, 1965.

[64] H. R. A LLAN, Radio Emission From Extensive Air Showers, Progress in Elementary
Particle and Cosmic Ray Physics, 10 (1971), pp. 171–302.

[65] O. R AVEL FOR THE CODALEMA C OLLABORATION, The CODALEMA experiment,

Nuclear Instruments and Methods in Physics Research A, 662 (2012), p. 89.

[66] P. L AUTRIDOU et al., Some possible interpretations from data of the CODALEMA ex-
periment, in American Institute of Physics Conference Series, R. Lahmann et al., eds.,
vol. 1535 of American Institute of Physics Conference Series, 2013, pp. 99–104.
180 Bibliography

[67] J. S CHULZ FOR THE P IERRE AUGER C OLLABORATION, Status and prospects of the
Auger Engineering Radio Array, in Proceedings of the 34th International Cosmic Ray
Conference, The Hague, The Netherlands, 2015.

[68] A. A AB et al., P IERRE AUGER C OLLABORATION, Probing the radio emission from air
showers with polarization measurements, Physical Review D, 89 (2014), p. 052002.

[69] P. S CHELLART et al., LOFAR C OLLABORATION, Detecting cosmic rays with the LO-
FAR radio telescope, Astronomy & Astrophysics, 560 (2013), p. A98.

[70] P. S CHELLART et al., Polarized radio emission from extensive air showers measured
with LOFAR, Journal of Cosmology and Astroparticle Physics, 10 (2014), p. 14.

[71] S. B UITINK et al., Method for high precision reconstruction of air shower Xmax using
two-dimensional radio intensity profiles, Physical Review D, 90 (2014), p. 082003.

[72] O. S CHOLTEN et al., A Macroscopic Description of Coherent Geo-Magnetic Radiation

from Cosmic Rays, in Proceedings of the 30th International Cosmic Ray Conference,
Merida, Mexico, 2008.

[73] K. W ERNER et al., A realistic treatment of geomagnetic Cherenkov radiation from cos-
mic ray air showers, Astroparticle Physics, 37 (2012), pp. 5–16.

[74] J. A LVAREZ -M UÑIZ et al., Monte Carlo simulations of radio pulses in atmospheric
showers using ZHAireS, Astroparticle Physics, 35 (2012), pp. 325–341.

[75] T. H UEGE et al., Simulating radio emission from air showers with CoREAS, in Ameri-
can Institute of Physics Conference Series, R. Lahmann et al., eds., vol. 1535 of Amer-
ican Institute of Physics Conference Series, 2013, pp. 128–132.

[76] C. W. JAMES et al., General description of electromagnetic radiation processes based

on instantaneous charge acceleration in endpoints, Physical Review E, 84 (2011),
p. 056602.

[77] J. A LVAREZ -M UÑIZ et al., Radio pulses from ultra-high energy atmospheric showers
as the superposition of Askaryan and geomagnetic mechanisms, Astroparticle Physics,
59 (2014), pp. 29–38.

[78] K. K AMATA et al., The Lateral and the Angular Structure Functions of Electron Show-
ers, Progress of Theoretical Physics Supplement, 6 (1958), pp. 93–155.

[79] K. G REISEN, Cosmic Ray Showers, Annual Review of Nuclear and Particle Science, 10
(1960), pp. 63–108.
Bibliography 181

[80] A. H AUNGS et al., Energy spectrum and mass composition of high-energy cosmic rays,
Reports on Progress in Physics, 66 (2003), pp. 1145–1206.

[81] T. A NTONI et al., KASCADE C OLLABORATION, The cosmic-ray experiment KAS-

CADE, Nuclear Instruments and Methods in Physics Research A, 513 (2003), pp. 490–
510.

[82] W. D. A PEL et al., KASCADE-G RANDE C OLLABORATION, The KASCADE-Grande

experiment, Nuclear Instruments and Methods in Physics Research A, 620 (2010),
pp. 202–216.

[83] G. VAN A AR et al., hxmax i measured with the surface detector. Auger Internal Publi-
cation (GAP2015_032), April 2015.

[84] M. AVE et al., AIRFLY C OLLABORATION, Measurement of the pressure dependence

of air fluorescence emission induced by electrons, Astroparticle Physics, 28 (2007),
pp. 41–57.

[85] T. K. G AISSER et al., Reliability of the method of constant intensity cuts for recon-
structing the average development of vertical showers, in Proceedings of the 17th Inter-
national Cosmic Ray Conference, Plovdiv, Bulgaria, 1977.

[86] M. U NGER et al., Reconstruction of longitudinal profiles of ultra-high energy cosmic

ray showers from fluorescence and Cherenkov light measurements, Nuclear Instruments
and Methods in Physics Research A, 588 (2008), pp. 433–441.

[87] T. A BU -Z AYYAD et al., The prototype high-resolution Fly’s Eye cosmic ray detector,
Nuclear Instruments and Methods in Physics Research A, 450 (2000), pp. 253–269.

[88] H. K AWAI et al., T ELESCOPE A RRAY C OLLABORATION, Telescope Array Experiment,

Nuclear Physics B - Proceedings Supplements, 175-176 (2008), pp. 221–226.

[89] J. A BRAHAM et al., P IERRE AUGER C OLLABORATION, The fluorescence detector of

the Pierre Auger Observatory, Nuclear Instruments and Methods in Physics Research
A, 620 (2010), pp. 227–251.

[90] N. B UDNEV et al., Tunka-25 Air Shower Cherenkov array: The main results, Astropar-
ticle Physics, 50 (2013), pp. 18–25.

[91] A. H ORNEFFER FOR THE LOPES C OLLABORATION, LOPES Detecting Radio Emis-
sion from Cosmic Ray Air Showers, ArXiv Astrophysics e-prints, (2004).

[92] F. G. S CHRÖDER et al., Tunka-Rex: A radio antenna array for the Tunka experiment,
in American Institute of Physics Conference Series, R. Lahmann et al., eds., vol. 1535
of American Institute of Physics Conference Series, 2013, pp. 111–115.
182 Bibliography

[93] W. D. A PEL et al., LOPES C OLLABORATION, Reconstruction of the energy and depth
of maximum of cosmic-ray air showers from LOPES radio measurements, Physical Re-
view D, 90 (2014), p. 062001.

[94] A. C ORSTANJE et al., LOFAR C OLLABORATION, The shape of the radio wavefront
of extensive air showers as measured with LOFAR, Astroparticle Physics, 61 (2015),
pp. 22–31.

[95] Q. D OROSTI H ASANKIADEH FOR THE P IERRE AUGER C OLLABORATION, Advanced

Reconstruction Strategies for the Auger Engineering Radio Array, 6th Conference on
Acoustic and Radio EeV Neutrino Detection, Annapolis, Maryland, USA, (2014).

[96] A. N ELLES et al., A parameterization for the radio emission of air showers as predicted
by CoREAS simulations and applied to LOFAR measurements, Astroparticle Physics, 60
(2015), pp. 13–24.

[97] J. S CHULZ et al., EMC-test of the AERA124 reference station. Auger Internal Publica-
tion (GAP2013_063), November 2013.

[98] R. M. T ENNENT, The Haverah Park extensive air shower array, Proceedings of the
Physical Society, 92 (1967), pp. 622–631.

[99] I. A LLEKOTTE et al., P IERRE AUGER C OLLABORATION, The surface detector system
of the Pierre Auger Observatory, Nuclear Instruments and Methods in Physics Research
A, 586 (2008), pp. 409–420.

[100] J. A BRAHAM et al., P IERRE AUGER C OLLABORATION, Properties and performance

of the prototype instrument for the Pierre Auger Observatory, Nuclear Instruments and
Methods in Physics Research A, 523 (2004), pp. 50–95.

[101] J. A BRAHAM et al., P IERRE AUGER C OLLABORATION, Trigger and aperture of the
surface detector array of the Pierre Auger Observatory, Nuclear Instruments and Meth-
ods in Physics Research A, 613 (2010), pp. 29–39.

[102] S. A RGIRÒ et al., P IERRE AUGER C OLLABORATION, The offline software framework
of the Pierre Auger Observatory, Nuclear Instruments and Methods in Physics Research
A, 580 (2007), pp. 1485–1496.

[103] V. V ERZI FOR THE P IERRE AUGER C OLLABORATION, The Energy Scale of the Pierre
Auger Observatory, in Proceedings of the 33rd International Cosmic Ray Conference,
Rio de Janeiro, Brazil, July 2013.

[104] J. T. B RACK et al., Absolute calibration of a large-diameter light source, Journal of

Instrumentation, 8 (2013), p. 5014P.
Bibliography 183

[105] A. C. ROVERO et al., Multi-wavelength calibration procedure for the pierre Auger
Observatory Fluorescence Detectors, Astroparticle Physics, 31 (2009), pp. 305–311.

[106] J. A BRAHAM et al., P IERRE AUGER C OLLABORATION, A study of the effect of molec-
ular and aerosol conditions in the atmosphere on air fluorescence measurements at the
Pierre Auger Observatory, Astroparticle Physics, 33 (2010), pp. 108–129.

[107] B. K EILHAUER FOR THE P IERRE AUGER C OLLABORATION, The Balloon-the-Shower

programme of the Pierre Auger Observatory, Astrophysics and Space Sciences Trans-
actions, 6 (2010), pp. 27–30.

[108] P. A BREU et al., P IERRE AUGER C OLLABORATION, Description of atmospheric con-

ditions at the Pierre Auger Observatory using the Global Data Assimilation System
(GDAS), Astroparticle Physics, 35 (2012), pp. 591–607.

[109] F. S ANCHEZ FOR THE P IERRE AUGER C OLLABORATION, The AMIGA detector of the
Pierre Auger Observatory: overview, in Proceedings of the 32nd International Cosmic
Ray Conference, Beijing, China, 2011.

[110] H. K LAGES FOR THE P IERRE AUGER C OLLABORATION, Enhancements to the South-
ern Pierre Auger Observatory, Journal of Physics Conference Series, 375 (2012),
p. 052006.

[111] P. A BREU et al., P IERRE AUGER C OLLABORATION, Antennas for the detection of
radio emission pulses from cosmic-ray induced air showers at the Pierre Auger Obser-
vatory, Journal of Instrumentation, 7 (2012), p. 11P.

[112] M. E RDMANN et al., Antenna Alignment for the first 24 Stations of AERA. Auger
Internal Publication (GAP2010_083), July 2010.

[113] A. N ELLES et al., A Survey of Narrowband and Broadband Radio-frequency Interfer-

ence at AERA. Auger Internal Publication (GAP2011_062), 2011.

[114] A. VAN DEN B ERG et al., Fiber Communication System for the Auger Engineer-
ing Radio Array at the Southern Auger Observatory. Auger Internal Publication
(GAP2011_035), March 2011.

[115] D. C HARRIER, C ODALEMA C OLLABORATION, Antenna development for astroparti-

cle and radioastronomy experiments, Nuclear Instruments and Methods in Physics Re-
search A, 662 (2012), p. 142.

[116] B. R EVENU FOR THE P IERRE AUGER C OLLABORATION AND THE CODALEMA
C OLLABORATION, Radio detection of cosmic ray air showers by the RAuger experi-
ment, a fully autonomous and self-triggered system installed at the Pierre Auger Obser-
vatory, Nuclear Instruments and Methods in Physics Research A, 662 (2012), p. 130.
184 Bibliography

[117] C. T IMMERMANS et al., Description of the scintillator triggered AERA-II stations.

Auger Internal Publication (GAP2013_074), August 2013.

[118] F. G. S CHRÖDER et al., New method for the time calibration of an interferometric radio
antenna array, Nuclear Instruments and Methods in Physics Research A, 615 (2010),
pp. 277–284.

[119] Y. Z HU FOR THE P IERRE AUGER C OLLABORATION, A flexible FPGA-based mod-

ule for wireless communications in astroparticle physics experiments, in Proc. of the
9th Internationl Symposium on Telecommunications, Piscataway, N.J., October 2012,
Behlilovic, N.

[120] J. K ELLEY et al., Design and Testing of a 5 GHz Commercial Wireless Network for
AERA. Auger Internal Publication (GAP2012_054), April 2012.

[121] M. H EVINGA et al., Photo-Voltaic System for the Auger Engineering Radio Array.
Auger Internal Publication (GAP2011_034), March 2011.

[122] PVSYST SOFTWARE PACKAGE, http://www.pvsyst.com.

[123] A. VAN DEN B ERG, private communication.

[124] HTTPS :// EOSWEB . LARC . NASA . GOV /, Atmospheric Science Data Center.

[125] VICTRON ENERGY , http://www.victronenergy.nl/batteries/Gel-and-AGM-batteries,

http://www.victronenergy.nl/batteries/Gel-and-AGM-batteries.

[126] J. K ELLEY et al., AERA EMC Tests at ASTRON. February 2010.

[127] J. L. K ELLEY FOR THE P IERRE AUGER C OLLABORATION, Data acquisition, trig-
gering, and filtering at the Auger Engineering Radio Array, Nuclear Instruments and
Methods in Physics Research A, 725 (2013), pp. 133–136.

[128] B. F. R EVENU, AERA Central Trigger. Auger Internal Publication (GAP2012_115),

2012.

[129] J. M ALLER FOR THE P IERRE AUGER C OLLABORATION, Radio detection of exten-
sive air showers at the Pierre Auger Observatory, Nuclear Instruments and Methods in
Physics Research A, 742 (2014), pp. 232–236.

[130] J. M ALLER et al., Two methods for rejecting background radio traces in RAuger data
at the level of a single station (T1 or T2). Auger Internal Publication (GAP2012_087),
November 2013.

[131] S. JANSEN et al., Towards standardizes selection criteria for AERA data. Auger Internal
Publication (GAP2015_007), January 2015.
Bibliography 185

[132] F EDERAL AVIATION A DMINISTRATION, Automatic Dependent Surveillance Braod-

cast (ADS-B) Out Performance Requirements to Support Air Traffic Control (ATC) Ser-
vice, Federal Register, 75 (2010).

[133] R. B RUN et al., ROOT - An object oriented data analysis framework, Nuclear Instru-
ments and Methods in Physics Research A, 389 (1997), pp. 81–86.

[134] S. M ATHYS et al., Development and applications of a ROOT based IO library for
AERA. Auger Internal Publication (GAP2015_01), January 2015.

[135] P. A BREU et al., P IERRE AUGER C OLLABORATION, Advanced functionality for radio
analysis in the Offline software framework of the Pierre Auger Observatory, Nuclear
Instruments and Methods in Physics Research A, 635 (2011), pp. 92–102.

[136] D. V EBERIC et al., SD Reconstruction - Offline Reference Manual. Auger Internal

Publication (GAP2005_035), 2005.

[137] M. S ETTIMO, Measurement of the cosmic ray energy spectrum using hybrid events of
the Pierre Auger Observatory, European Physical Journal Plus, 127 (2012), p. 87.

[138] A. P ORCELLI, Measurement of the Depth of Shower Maximum in the Transition Region
between Galactic and Extragalactic Cosmic Rays with the Pierre Auger Observatory,
PhD thesis, Karlsruher Insitute of Technology, 2014.

[139] A. A AB et al., P IERRE AUGER C OLLABORATION, Depth of maximum of air-shower

profiles at the Pierre Auger Observatory. I. Measurements at energies above 1017.8 eV,
Physical Review D, 90 (2014), p. 122005.

[140] I. C. M ARIS FOR THE P IERRE AUGER C OLLABORATION, Measurement of the Energy
Spectrum of Cosmic Rays above 3 × 1017 eV at the Pierre Auger Observatory, vol. EPS-
HEP2013, 2013, p. 405.

[141] S. G REBE et al., Suppression of self-introduced narrowband RFI in the time domain.
Auger Internal Publication (GAP2013_012), November 2013.

[142] R. C. J ONES, A new calculus for the treatment of optical systems, Journal of the Optical
Society of America, 31 (1941), pp. 488–493.

[143] G. B URKE et al., Numerical electromagnetics code (nec) method of moments, part ii
and iii. tech.rep., 1981.

[144] D. C HARRIER, Calculating the vector equivalent length of the butterfly antenna from
nec2 by simulating the antenna in transmitting mode. Auger Internal Publication
(GAP2014_025), 2014.
186 Bibliography

[145] S. JANSEN, private communication.

[146] K. L INK et al., LOPES C OLLABORATION, Revised absolute amplitude calibration of

the LOPES experiment, ArXiv e-prints, (2015).

[147] A. N ELLES et al., Calibrating the absolute amplitude scale for air showers measured
at lofar, Journal of Instrumentation, 10 (2015), p. P11005.

[148] S. O STAPCHENKO, QGSJET-II: physics, recent improvements, and results for air show-
ers, in European Physical Journal Web of Conferences, vol. 52 of European Physical
Journal Web of Conferences, 2013, p. 2001.

[149] M. KOBAL FOR THE P IERRE AUGER C OLLABORATION, A thinning method using
weight limitation for air-shower simulations, Astroparticle Physics, 15 (2001), pp. 259–
273.

[150] A. A AB et al., P IERRE AUGER C OLLABORATION, Reconstruction of inclined air show-

ers detected with the Pierre Auger Observatory, Journal of Cosmology and Astroparti-
cle Physics, 8 (2014), p. 19.

[151] S. AGOSTINELLI et al., GEANT4 – a simulation toolkit, Nuclear Instruments and Meth-
ods in Physics Research A, 506 (2003), pp. 250–303.

[152] J. B ERINGER et al., PARTICLE DATA G ROUP, Review of Particle Physics, Physical
Review D, 86 (2012), p. 010001.

[153] T. A NTONI et al., KASCADE C OLLABORATION, Electron, muon, and hadron lat-
eral distributions measured in air showers by the KASCADE experiment, Astroparticle
Physics, 14 (2001), pp. 245–260.

[154] S. T HOUDAM et al., LORA: A scintillator array for LOFAR to measure extensive air
showers, Nuclear Instruments and Methods in Physics Research A, 767 (2014), pp. 339–
346.

[155] P. K. F. G RIEDER, Extensive Air Showers, 2010.

[156] A. N ELLES et al., A parameterization for the radio signal measured with AERA. Auger
Internal Publication (GAP2014_073), September 2014.

[157] A. N ELLES et al., The radio emission pattern of air showers as measured with LO-
FAR - a tool for the reconstruction of the energy and the shower maximum, Journal of
Cosmology and Astroparticle Physics, 5 (2015), p. 18.

[158] T. H UEGE, private communication.

Bibliography 187

[159] I. M ARI Ş FOR THE P IERRE AUGER C OLLABORATION, The amiga infill detector of
the Pierre Auger Observatory: performance and first data, in Proceedings of the 32nd
International Cosmic Ray Conference, Beijing, China, 2011.

[160] S. F LIESCHER, Antenna Devices and Measurement of Radio Emission from Cosmic
Ray induced Air Showers at the Pierre Auger Observatory, PhD thesis, RWTH Aachen
University, 2011.

[161] M. S TEPHAN FOR THE P IERRE AUGER C OLLABORATION, Antennas, Filters and
Preamplifiers designed for the Radio Detection of Ultra-High-Energy Cosmic Rays,
Proceedings of the Asia Pacific Microwave Conference, TH3G-49 (2010).

[162] M. E NDER et al., Radio Emission of Extensive Air Showers during Thunderstorms, in
Proceedings of the 31st International Cosmic Ray Conference, Lodz, Poland, 2009.

[163] P. S CHELLART et al., Probing atmospheric electric fields in thunderstorms through

radio emission from cosmic-ray-induced air showers, Physical Review Letters, 114
(2015), p. 165001.

[164] S. N EHLS, Calibrated Measurements of the Radio Emission of Cosmic Ray Air Show-
ers, PhD thesis, Institut für Kernphysik, Universität Karlsruhe, 2008.

[165] K. W EIDENHAUPT, Antenna Calibration and Energy Measurement of Ultra-High En-

ergy Cosmic Rays with the Auger Engineering Radio Array, PhD thesis, RWTH Aachen
University, 2014.

[166] G. FARRAR FOR THE P IERRE AUGER C OLLABORATION, The muon content of hybrid
events recorded at the Pierre Auger Observatory, in Proceedings of the 33rd Interna-
tional Cosmic Ray Conference, Rio de Janeiro, Brazil, July 2013.

[167] T. H UEGE et al., Dependence of geosynchrotron radio emission on the energy and depth
of maximum of cosmic ray showers, Astroparticle Physics, 30 (2008), pp. 96–104.

[168] M. T UEROS FOR THE P IERRE AUGER C OLLABORATION, Estimate of the non-
calorimetric energy of showers observed with the fluorescence and surface detectors
of the Pierre Auger Observatory, in Proceedings of the 33rd International Cosmic Ray
Conference, Rio de Janeiro, Brazil, 2013.

[169] P. B ILLOIR, A sampling procedure to regenerate particles in a ground detector from a

thinned air shower simulation output, Astroparticle Physics, 30 (2008), pp. 270–285.

[170] T. M C C AULEY et al., GEANT4 simulation of the surface detectors. Auger Internal
Publication (GAP2000_055), 2000.
188 Bibliography

[171] A. P ORCELLI FOR THE P IERRE AUGER C OLLABORATION, Measurements of Xmax

above 101 7 eV with the fluorescence detector of the Pierre Auger Observatory, in Pro-
ceedings of the 34th International Cosmic Ray Conference, The Hague, The Nether-
lands, 2015.

[172] S. E PIMAKHOV FOR THE T UNKA C OLLABORATION, Elemental Composition of Cos-

mic Rays above the Knee from Xmax measurements of the Tunka Array, in Proceedings
of the 33rd International Cosmic Ray Conference, Rio de Janeiro, Brazil, 2013.

[173] S. P. K NURENKO et al., The depth of maximum shower development and its fluctua-
tions: cosmic ray mass composition at E0 >1017 eV, Astrophysics and Space Sciences
Transactions, 7 (2011), pp. 251–255.

[174] S. P. K NURENKO et al., Study of cosmic rays at the Yakutsk EAS array: energy spec-
trum and mass composition, Nuclear Physics B Proceedings Supplements, 212 (2011),
pp. 241–251.

[175] M. U NGER et al., P IERRE AUGER C OLLABORATION, Study of the Cosmic Ray Compo-
sition above 0.4 EeV using the Longitudinal Profiles of Showers observed at the Pierre
Auger Observatory, Astronomische Nachrichten, 328 (2007), p. 614.

[176] E. G. B EREZHKO, Composition of Cosmic Rays Accelerated in Active Galactic Nuclei,

Astrophysical Journal, Letters, 698 (2009), pp. L138–L141.

[177] A. DAR et al., A theory of cosmic rays, Physics Reports, 466 (2008), pp. 179–241.

[178] S. T HOUDAM et al., Study of the energy spectrum and composition of cosmic rays up
to the highest energies, in preperation.
 Index
access point (AP) Auger Muon and Infill radio simulations, 111
AERA, 41, 166 Ground Array scintillators, 96
SD, 29 (AMIGA), 33 SD, 62
active bowtie antenna, 39 SD-FD hybrid, 32
backbone link, 29 CORSIKA, 15
active galactic nuclei
beacon, 41 cosmic microwave
(AGN), 6
bean shape, 17 background
ADST, see Advanced Data
Bullet (Ubiquity), 41 (CMB), 7
Summary Tree
Butterfly antenna, 39 cosmic rays, 3
Advanced Data Summary
Tree (ADST), 70 ultra high-energy
calorimetric energy, 22
AERA, see Auger (UHECR), 6
cascade
Engineering Radio critical energy, 13
electromagnetic, 12
Array CRS, see Central Radio
hadronic, 12
AERA124, 38 Station
CDAS, see central data
AERA153, 39 cut-off, 4
acquisition system
AERA24, 36 central data acquisition data merging, 58
air Cherenkov detectors, 23 system (CDAS), 29 data set
air shower, 11 central radio data AERA, 71
simulations, 15 acquisition system, AERA24, 125
AMIGA, see Auger Muon 37 multi-hybrid, 72
and Infill Ground Central Radio Station scintillators, 92
Array (CRS), 38 depth of shower maximum,
analysis modules, 59 characteristic radius, 20, 94 see Xmax
angular resolution charge excess detector description, 59
AERA, 72 emission, 15 digitizer, 37
scintillators, 97 fraction, 17 dipole moment, 16
SD, 31 Coihueco, 28 direction reconstruction,
SD-FD hybrid, 32 communication system 60, 69
ankle, 4 AERA, 41 distance to Xmax , 103
antenna channel level, 66 AERA tests, 50 atmospheric, 113
antenna response, 67 SD, 29 geometric, 114
AP, see access point composition, 4
arrival direction, 6 CONEX, 15 electric field, 17
Auger Engineering Radio CoREAS, 18 electromagnetic
Array (AERA), 35 core position compatibility, 46
190 Index

electromagnetic FOV, see field of view lateral distribution function

component, 11 (LDF)
Gaisser-Hillas function, 22,
elongation rate, 13 particles, 20
63
end point formalism, 19 radio, 24, 101
gamma rays, 11
energy calibration RD, 101
GDAS, see Global Data
RD, 129 scintillators, 90, 94
Assimilation
SD, 64 SD, 61
System
energy density, 69 SD simulation, 142
geomagnetic angle, 17, 106
energy estimator, 126 LDF, see lateral distribution
geomagnetic emission, 16
energy reconstruction function
Global Data Assimilation
FD, 22, 63 LNA, see low-noise
System (GDAS),
radio simulations, 104 amplifier
33
RD, 121 ln A, see logarithmic mass
golden events, 64
SD, 61, 64 log-periodic dipole antenna
GZK-effect, 8
energy resolution (LPDA), 36
FD, 65 hadronic component, 11 logarithmic mass, 162
radio simulations, 106 HEAT, see High Elevation Loma Amarilla, 28
RD, 130 Auger Telescopes, longitudinal shower
SD, 65 see High Elevation development, 11
energy spectrum, 3 Auger Telescopes Los Leones, 28
event data, 59 Heitler model, 12 Los Morados, 28
event level, 69 High Elevation Auger low-noise amplifier (LNA),
event rate, 72 Telescopes 36
extensive air shower, see air (HEAT), 28, 33 LPDA, see log-periodic
shower Hilbert-envelop, 69 dipole antenna
Hillas criterion, 9
FD, see Fluorescence
hybrid shower macroscopic model, 18
Detector
RD+SD, 71 mean logarithmic mass,
Fermi acceleration, 9
RD+SD+FD multi-, 72 162
field of view (FOV), 31
AERA, 163
filter-amplifier, 37 integrated energy density, microscopic model, 18
fluorescence 105 multi hybrid detector
detectors, 22 interaction models, 15, 75 setups, 24
light, 22 interactions, 7 muonic component, 11
light yield, 22
Fluorescence Detector Jones matrix, 67
neutrinos, 11
(FD), 31 NKG-function, 20
knee, 4
fluorescence telescope, 31 noise, 55
calibration, 33 laser facilities (XLF, CLF),
footprint width, 114 33 Off line, 59
Index 191

on-sky coordinate system, radio frequency star shape, 76

67 interference (RFI), skymap, 6
optical fibre, 42 36 SNR, see signal-to-noise
RD, see Radio Detector ration
particle detector arrays, 20 array, AERA solar power system
photomultiplier tube (PMT) RdObserver, 70, 167 AERA124, 39, 45
scintillators, 80 RDS, see radio detection AERA24, 36
SD, 30 stations SD, 28
Pierre Auger Observatory, RFI, see radio frequency sources, 8
27 interference spectral breaks, 4
PMT, see photomultiplier RFI suppression, 70 station level, 67
tube rigidity, 8 superposition principle, 14
point-to-multipoint link, 41 ring buffer, 37 Surface Detector array
point-to-point link, 41 Rocket (Ubiquity), 41 (SD), 27
polarization, 123 system integration, 44
sampling rate, 37
propagation, 7 scattering, 63 TDMA, see Time Division
scintillator, 79 Multiple Access
quality cuts Bottom-, 79 thunderstorm conditions,
FD, 64, 169 calibration, 84, 87, 170 125
RD energy, 125 module, 80 Time Division Multiple
RD LDF, 136 Top-, 79 Access (TDMA),
RdObserver, 71 SD, see Surface Detector 29
RD Xmax LDF, 138, array trigger (RD)
155 SD Infill array, 34 external-, 56
RD Xmax profile, 149 second knee, 4 internal particle-, 57
scintillators, 93 sector antenna, 41 non cosmic ray-, 58
SD, 60, 169 selection criteria, see self-, 56
SD-FD hybrid, 73, 169 quality cuts trigger conditions
shower age, 20 FD, 32
radiated energy, 105, 131 shower front, 12 RD, 55, 81
radiation length, 13 shower reconstruction scintillators, 81
radio detection stations FD, 62 SD, 30
(RDS), 35 RD, 65, 122
Radio Detector array (RD), scintillators, 92 UHECR, see cosmic rays,
see AERA SD, 59 ultra high-energy
radio detector arrays, 23 signal-to-noise ration uncertainty
radio detector description, (SNR), 69 energy density, 125
66 simulation set energy estimator, 127
radio emission, 15 AERA, 75 energy scale, 129
192 Index

Xmax FD, 74 wireless communication Xmax resolution

Xmax RD LDF, 139 system, see FD, 64
Xmax RD profiles, 150 communication radio simulations, 115
system RD LDF, 139
vector effective length
AERA, 165 RD profiles, 150
(VEL), 67
VEL, see vector effective Xmax reconstruction RD simulations, 137
length FD, 64 Xmax , 4
VEM, see vertical radio simulations, 113, average-, 5
equivalent muon 172 average- AERA, 157
vertical equivalent muon RD LDF, 135
(VEM), 30, 86 RD profiles, 140 ZHAires, 18
 Summary

The Earth is continuously exposed to a flux of particles coming from outer space. These
particles are called cosmic rays and their existence has been known for more than a century.
The beginning of the life-cycle of the most energetic ones among them is however still un-
explained. Especially the sources able to accelerate particles to the observed energies have
not been identified yet. One of the reasons for the still open questions about the nature of the
highest energetic cosmic rays is their low flux. Over long ranges of the known spectrum of
cosmic rays, the flux decreases according to a power law (∝ E −2.7 ). At the end of the mea-
sured spectrum at about 1020 eV, only one cosmic ray is observed per square-kilometer per
century. Therefore, air- or space-borne experiments cannot be employed for their direct detec-
tion. Experiments that measure enough of these rare particles to infer information about their
origin must be large and can only be built on the ground. There however, only secondary par-
ticles such as muons, photons, electrons, positrons, and other products of the collisions of the
cosmic rays with nuclei in the Earth’s atmosphere can be measured. During their propagation
through the atmosphere, these particles are likely to also interact, creating new particles which
subsequently also interact and create new particles and so on. Thereby, the initial energy of
the cosmic ray is transferred to millions of secondary particles in a particle cascade which is
called extensive air shower. From the observation of the extensive air shower, the properties of
the cosmic ray which are the energy, the arrival direction, and the type of the original particle
need to be reconstructed to learn about its origin and propagation towards the Earth.
The first two properties are in general reconstructed from the secondary particles measured
on the ground. This technique has been used since the first description of the phenomenon of
extensive air showers and yields high resolution for both observables. The measurement of
the particle type of the cosmic rays is still challenging. Without the direct measurement of the
primary particle, information about its kind is only indirectly available via the development
of the air shower which differs for different primary particles on a statistical basis. At the
highest energies, mainly optical methods are employed to directly or indirectly detected light
induced by the secondary particles in the air shower. These techniques are however limited to
a duty cycle of about 10% by environmental conditions as they can only operate during clear
moonless nights.
Already proposed in the 1960’s, the detection of radio signals originating from the charged
particles in the air showers is an alternative to the established techniques. The radio emission
originates from two mechanisms which are both based on the separation of positive and nega-
tive charges during the air shower development. One mechanism is the charge excess mecha-
nism. It is based on the effect that during the air shower development electrons are knocked out
194 Summary

of atmospheric molecules and join the front of the air shower whereas the positively charged
ions are left behind. The other mechanism is the geomagnetic emission mechanism which is
based on the separation of electrons and positrons in the shower front due to the Lorentz-force
in the geomagnetic field. On the ground, the emission can be detected in the MHz-regime
which shows a characteristic interference pattern.
After the experimental investigation of the emission mechanisms, and also improvements of
the theoretical understanding and modelling of the underlying processes, the sensitivity of the
radio technique to air-shower and thereby cosmic-ray parameters is in the focus of today’s re-
search. A reliable way is the comparison to established detection techniques when measuring
the same air showers simultaneously which is the subject of this thesis and carried out at the
Pierre Auger Observatory in Argentina. There, the Auger Engineering Radio Array (AERA)
is operated together with a fluorescence light detector and a particle detector array. The pre-
sented work focuses on the simultaneous detection of extensive air-showers with the different
detector types and the comparisons of the radio measurements to the other techniques. The
basic idea of multi-hybrid air shower detection is sketched on page 195. During the air shower
development, fluorescence light is emitted from excited nitrogen molecules in the atmosphere.
This light is used to image the shower development. Furthermore, MHz radiation is emit-
ted and detected on the ground. The distribution of the energy density in the radiation has a
complex shape as indicated by the color coding in the sketch. Together with the radio pulses,
the particles of the air shower are detected with particle detectors on the surface and some-
times also with additional buried muon counters. The reconstruction of the arrival direction
of a cosmic ray is based on the arrival times of the pulses or particles. The resolution of
the measurement is similar for both, particle and radio detectors. For the reconstruction of
the cosmic ray energy, the radio energy yield, which is the amount of energy transferred in
MHz-radiation (the integral over the sketched color map) is introduced and calibrated with the
particle detector measurement in this work and the resolution is shown to be competitive to
the other techniques. Finally, the sensitivity of the radio technique to the shower development
is investigated in simulation studies and demonstrated on data of simultaneous observations
of the radio emission and the fluorescence light. A cross calibration is performed between
the two methods, based on the measured energy density of the radio signals on the ground.
One method is based on sets of Monte Carlo simulations performed for protons and iron nu-
clei using the measured arrival direction and energy as input. The sets cover a broad range of
possible shower developments and the better the prediction of the energy density pattern fits
the measurement, the more likely the simulated shower development corresponds to the one of
the actual shower. The other method employs an empiric parametrization of the lateral energy
density distribution which is fitted to the data. The width of the distribution is sensitive to
the shower development. After the calibration with the fluorescence light measurements, this
method is applied to a much larger dataset for which no fluorescence light data are available to
also infer information about the shower development for these showers. This results in the first
Summary 195

measurement of the depth of the shower maximum from radio data that has been calibrated
with an independent detection technique and thus the mass (type) of cosmic rays as a function
of energy.

cosmic ray!
kosmische straling!
kosmische Strahlung

atmospheric nucleus!
atmosferische atoomkern!
atmosphärischer Atomkern

air shower!
deeltjeslawine!
Luftschauer

fluorescence telescope!
fluorescentietelescoop!
coherent MHz radiation! Fluoreszenzteleskop
coherente MHz straling!
kohärente MHz Strahlung
radio antenna!
radioantenne!
Radioantenne

particle detector!
energy density distribution on the ground! deeltjesdetector!
muon counter! Teilchendetektor
energiedichtheidsverdeling op de grond! muonteller!
Energiedichteverteilung am Erdboden Myonenzähler

Sketch of the multi-hybrid air shower detection at the Pierre Auger Observatory.
Illustratie van de waarneming van een deeltjes lawine bij het Pierre Auger Observatorium.
Schematische Darstellung der Hybridmessungen von Luftschauern am Pierre Auger Observatorium.
 Samenvatting

De aarde is blootgesteld aan een voortdurend bombardement van subatomaire deeltjes van-
uit de ruimte. Deze deeltjes worden kosmische straling genoemd, en hun bestaan is al meer
dan een eeuw bekend. De oorsprong van de meest energetische van deze deeltjes is echter nog
steeds niet verklaard. Om meer specifiek te zijn, zijn de bronnen die in staat zijn om de deeltjes
tot de waargenomen hoge energieën te versnellen nog niet geïdentificeerd. Een van de redenen
waarom vragen over de oorsprong van hoogenergetische kosmische straling nog steeds open
staan is de zeer lage flux. Over een groot deel van het bekende spectrum van kosmische straling
neemt de flux af met toenemende energie volgens een machtsfunctie (∝ E −2.7 ). Aan het einde
van het gemeten spectrum, op ongeveer 1020 eV, wordt slechts één kosmisch deeltje waargeno-
men per vierkante kilometer per eeuw. Daarom kunnen ballon- en satellietexperimenten niet
gebruikt worden voor detectie van kosmische straling met deze energieën. Experimenten die
genoeg van deze zeldzame deeltjes kunnen meten om informatie over hun oorsprong te verza-
melen zijn daarom noodzakelijkerwijs groot, en worden op de grond gebouwd. Op de grond
kunnen echter alleen secundaire deeltjes zoals muonen, fotonen, elektronen, positronen en an-
dere producten van de botsingen tussen kosmische deeltjes en atomen in de atmosfeer van de
aarde worden gemeten. Tijdens de propagatie van kosmische straling door de atmosfeer van
de aarde is er een grote kans op interactie. De daarbij ontstane deeltjes gaan op hun beurt
ook weer interacties aan, enzovoort. Hierdoor wordt de initiële energie van het kosmische
deeltje verdeeld over miljoenen secundaire deeltjes in een deeltjeslawine. Door middel van
waarnemingen van deeltjeslawines moeten eigenschappen van het oorspronkelijke kosmische
deeltje, zoals de energie, de aankomstrichting en het type deeltje, worden gereconstrueerd om
kennis te verkrijgen over hun bronnen en de processen die plaats vinden tijdens hun reis door
de ruimte op weg naar de aarde.
De eerste twee van de bovengenoemde eigenschappen worden typisch gereconstrueerd van-
uit metingen van de secundaire deeltjes op de grond. Deze techniek wordt gebruikt sinds de
eerste beschrijving van deeltjeslawines en geeft een hoge resolutie in zowel energie als aan-
komstrichting. Het reconstrueren van het type deeltje is nog steeds een grote uitdaging. Zonder
een directe meting van het primaire deeltje is informatie over de aard van het deeltje alleen in-
direct beschikbaar, via statistische verschillen tussen de ontwikkeling van deeltjeslawines voor
verschillende soorten primaire deeltjes. Op de hoogste energieën worden voornamelijk opti-
sche methoden gebruikt om licht op te vangen dat direct of indirect wordt geproduceerd door
de secundaire deeltjes in de lawine. Deze technieken zijn echter beperkt tot een lage “duty
cycle” van ongeveer 10% doordat deze waarnemingen alleen kunnen plaatsvinden gedurende
heldere, maanloze nachten.
De detectie van radiostraling afkomstig van de geladen deeltjes in de deeltjeslawine is al in
198 Samenvatting

de jaren ‘60 voorgesteld als alternatief voor de bovengenoemde methoden. Deze radiostraling
wordt geproduceerd door twee mechanismen, die beide gebaseerd zijn op ladingsscheiding
van de positieve en negatieve lading in de deeltjeslawine. Een van deze twee mechanismen
is het ladingsoverschot-mechanisme. Hierbij worden elektronen losgeslagen uit atmosferi-
sche moleculen, waarna ze meebewegen met de deeltjeslawine en waarbij de positief geladen
ionen achtergelaten worden. Daarnaast is er het geomagnetische mechanisme waarbij de la-
dingsscheiding tussen elektronen en positronen in de deeltjeslawine veroorzaakt wordt door
de Lorentz-kracht in het magnetisch veld van de aarde. Bij beide mechanismen verandert
tevens het aantal vrije deeltjes in de lawine als functie van de tijd, hetgeen ook weer resul-
teert in radiostraling. Interferentie tussen de radiostraling van deze mechanismen geeft een
karakteristiek intensiteitspatroon, dat op de grond gedetecteerd kan worden in het megahertz-
frequentiebereik.
Volgend op experimentele verificatie van de emissiemechanismen, en verbeteringen in het the-
oretisch begrip van de onderliggende processen, is het huidig onderzoek gericht op het bepalen
van de gevoeligheid van radiostralingsmetingen van eigenschappen van de deeltjeslawine en de
kosmische straling zelf. Een betrouwbare methode hiervoor is een rechtstreekse vergelijking
met bestaande meettechnieken wanneer dezelfde deeltjeslawine tegelijkertijd met meerdere
meettechnieken wordt waargenomen. Deze methode is het onderwerp van dit proefschrift. Dit
onderzoek is uitgevoerd bij het Pierre Auger Observatorium in Argentinië. Daar opereert de
Auger Engineering Radio Array (AERA) samen met optische fluorescentietelescopen en deel-
tjesdetectoren. Het werk dat in dit proefschrift gepresenteerd wordt richt zich op de gelijktij-
dige detectie van deeltjeslawines met de verschillende meetinstrumenten en een vergelijking
van de meettechnieken. Het concept van een dergelijke multi-hybride waarneming is geschetst
op pagina 195. Gedurende de ontwikkeling van een deeltjeslawine wordt fluorescentielicht
uitgestraald door aangeslagen stikstofmoleculen in de atmosfeer. Dit licht wordt gebruikt om
met de fluorescentietelescopen een foto te maken van de ontwikkeling van de deeltjesregen.
Daarnaast komt er radiostraling vrij in het megahertz-frequentiebereik die met AERA gede-
tecteerd wordt op de grond. De distributie van de energiedichtheid in deze radiostraling heeft
een complexe vorm, zoals aangegeven door de kleur in de schets. Naast de radiopulsen wor-
den ook de secundaire deeltjes gedetecteerd met deeltjesdetectoren op de grond, en soms met
ingegraven muonendetectoren. De reconstructie van de aankomstrichting van het kosmische
deeltje is gebaseerd op metingen van de aankomsttijd van de radiopulsen of van de secundaire
deeltjes. De resolutie van beide meettechnieken is vergelijkbaar. Voor de reconstructie van
de energie van het kosmische deeltje wordt de radio-opbrengst, de hoeveelheid energie aan-
wezig in de megahertz-radiostraling (ofwel de integraal over het geschetste patroon), bepaald.
Deze wordt gemeten, en gekalibreerd met de waarnemingen van de deeltjesdetectoren. De
resolutie van deze nieuwe methode is vergelijkbaar met die van eerder ontwikkelde technie-
ken. Als laatste wordt de nauwkeurigheid van de radiodetectietechniek voor de ontwikkeling
van de deeltjesregen onderzocht. Aanvankelijk aan de hand van simulatiestudies, en vervol-
Samenvatting 199

gens door directe vergelijkingen tussen radio- en fluorescentiewaarnemingen. Een kalibratie

is uitgevoerd tussen deze twee methodes, gebaseerd op de gemeten energiedichtheid van het
radiosignaal op de grond. Vervolgens zijn er twee methoden gebruikt voor de reconstructie van
de ontwikkeling van de deeltjeslawine. Voor de eerste methode zijn Monte Carlo-simulaties
uitgevoerd met protonen en ijzerkernen als primaire deeltjes, met de gemeten aankomstrich-
ting en energie als invoerparameters. Deze simulaties beslaan een groot bereik aan mogelijke
deeltjeslawine-ontwikkelingen. Hoe beter het gesimuleerde energiedichtheidspatroon van de
radiostraling overeenkomt met het gemeten patroon, hoe groter de kans dat het gesimuleerde
ontwikkelingspatroon van de deeltjeslawine overeenkomt met het ontwikkelingspatroon van
de gemeten deeltjeslawine. De tweede methode gebruikt een empirische parametrisatie van
de laterale energiedichtheidsdistributie, die vervolgens zo gekozen wordt dat hij overeen komt
met de gemeten waarden. De breedte van de distributie is afhankelijk van de ontwikkeling van
de deeltjeslawine. Na een calibratie met fluorescentiewaarnemingen is deze methode toegepast
op een veel grotere set waarnemingen, waarvoor geen fluorescentiewaarnemingen beschikbaar
zijn. Dit resulteert in de eerste meting van de atmosferische diepte van het maximum van de
deeltjeslawine, en dus de massa (ofwel het deeltjestype) van kosmische straling, als functie
van de energie op behulp van radiometingen die zijn gekalibreerd met een onafhankelijke de-
tectiemethode.
 Zusammenfassung

Die Erde wird kontinuierlich von kleinen Teilchen und größeren Objekten aus dem Weltall
getroffen. Das Licht der Sonne oder der Sterne und die kleinen Meteoren, die als Sternschnup-
pen in unserer Atmosphäre glühen sind zum Beispiel allgemein bekannte Vertreter. Neben den
masselosen Lichtteilchen und den teils massiven Meteoren befinden sich auch viele masse-
behaftete Elementarteilchen und Atomkerne unter den eintreffenden Teilchen. Diese werden
als kosmische Strahlung bezeichnet und ihre Existenz fasziniert die Wissenschaft seit nun-
mehr über 100 Jahren. Seit der ersten Beschreibung dieses Phänomens durch Victor Hess in
der er seine Entdeckung Höhenstrahlung nennt, konnte das Verständnis dieser Teilchen wei-
ter und weiter verbessert werden. Die gesammelten Erkenntnisse basieren dabei hauptsächlich
auf Messungen der Energie, der Ankunftsrichtung und der Art der Teilchen in der kosmischen
Strahlung. Aus diesen Observablen können Rückschlüsse auf die Natur der individuellen Teil-
chen gezogen werden. Besonderes Interesse gilt dabei den höchstenergetischen Teilchen und
der Frage nach den Quellen, die in der Lage sind Teilchen auf diese Energien zu beschleuni-
gen. Heutzutage wissen wir, dass die Anzahl der Teilchen in der kosmischen Strahlung die die
Erde erreichen stark abhänging von ihrer Energie ist. Grundsätzlich gilt, je höher die Energie
desto weniger Teilchen werden beobachtet. Dies hat zur Folge, dass direkte Messungen ober-
halb der Erdatmosphäre an Bord von Satelliten oder Ballonen auf Grund der begrenzten Größe
der Detektoren nur bis zu einer gewissen Energie genügend Teilchen nachweisen können um
sichere Rückschlüsse über deren Charakteristik zu ziehen. Diese maximale Energie für direkte
Messungen der Kosmischen Strahlung ist etwa 100 TeV, was ungefähr dem hundertfachen der
Energie einer fliegenden Mücke entspricht. Für höhere Energien und damit kleinere Teilchen-
zahlen sind größere Detektorflächen nötig wie sie hauptsächlich auf der Erdoberfläche gebaut
werden können. Messungen mit Detektoren auf der Erdoberfläche können jedoch nur indirekt
Aufschluss über das Primärteilchen geben, da dies mit der Erdatmosphäre wechselwirkt und
auf dem Erdboden nicht mehr nachweisbar ist. Die Wechselwirkung der Teilchen der kosmi-
schen Strahlung mit Atomkernen der Atmosphärenmoleküle ähnelt den Ereignissen in einem
Teilchenbeschleuniger wie dem Large Hadron Collider am CERN. Durch die während der
Kollision freiwerdende Energie werden unzählige neue Teilchen erzeugt, die ihrerseits wieder
wechselwirken und auch zerfallen können und dabei jeweils neue Teilchen erzeugen. Die-
se Entwicklung führt zu einer Teilchenkaskade und einer Zunahme der Teilchenzahl bis die
primäre Energie auf so viele Teilchen aufgeteilt ist, dass diese im Durchschnitt zu klein ist
um neue Teilchen zu erzeugen und die Kaskade langsam wieder ausstirbt. Die Teilchenkas-
kade in der Erdatmosphäre wird als (ausgedehnter) Luftschauer bezeichnet und wurde zu erst
von Pierre Auger, Bruno Rossi und Werner Kolhörster beschrieben. Nur ein kleiner Teil der
erzeugten Teilchen erreicht dabei den Erdboden und kann dort nachgewiesen werden. Zum
202 Zusammenfassung

Nachweis werden häufig regelmäßig angeordnete Teilchendetektoren benutzt. Basierend auf

der Messung der sekundären Teilchen am Boden können Rückschlüsse über die Bewegungs-
richtung und die Energie des ursprünglichen primären Teilchens gewonnen werden. Aktuell
entwickelte Analysemethoden versprechen außerdem Rückschlüsse auf die (durchschnittliche)
Massenzahl über die Rekonstruktion der Luftschauerentwicklung. Je größer die Massenzahl
des primären Teilchens, desto höher in der Atmosphäre beginnt der Luftschauer im Durch-
schnitt. Die Bestimmung der Massenzahl ist ein essenzieller Aspekt für die Beantwortung der
Frage nach den Quellen der höchstenergetischen kosmischen Strahlung. Eine weitere bewähr-
te, wenn auch herausfordernde Messmethode ist die Detektion von Fluoreszenzlicht, welches
von atmosphärischen Stickstoffmolekülen emittiert wird, nachdem diese durch die Teilchen in
der Luftschauerkaskade angeregt wurden. Die Menge an Licht ist dabei abhängig von der Teil-
chenzahl, die wiederum mit der Energie des primären Teilchens zusammenhängt. Zusätzlich
lässt sich die Entwicklung des Luftschauers mit dieser Technik sehr genau verfolgen und es
können damit Rückschlüsse auf die primäre Massenzahl gezogen werden. Da die betreffende
Lichtmenge jedoch extrem gering ist und nur bei klaren und mondlosen Nächten detektiert
werden kann, ergeben sich kurze Betriebszyklen von nur etwa 10% der gesamten Zeit. Auf
Grund der erreichten Präzision ist die Fluoreszenztechnik trotzdem nach wie vor das Zugpferd
im Bereich der Vermessung der höchstenergetischen kosmischen Strahlung und wird meist
in einem hybriden Ansatz in Kombination mit einem Teilchendetektor eingesetzt. Die beiden
Techniken können dann gegeneinander kalibriert werden. Somit kann der Teilchendetektor,
der einen Betriebszyklus von 100% aufweist auch während der Zeiten zu denen keine Fluores-
zenzlichtmessungen möglich sind, hochwertige Informationen über die Eigenschaften der kos-
mische Strahlung liefern. Diese Kombination von Detektionstechniken wird unter anderem am
Pierre Auger Observatorium in Argentinien eingesetzt. Auf einer Fläche von über 3000 km2
sind hier 1600 Teilchendetektoren in einem regelmäßigen Muster angeordnet und werden von
27 Fluoreszenzteleskopen komplementiert. Damit bildet das Pierre Auger Observatorium das
weltgrößte Instrument zur Untersuchung der kosmischen Strahlung und liefert gleichzeitig
einzigartige Voraussetzungen um neue Technologien zu testen und mit den beiden Hauptde-
tektoren zu kalibrieren. Eine dieser Technologien ist die Detektion von extrem kurzen Radio-
pulsen die während der Luftschauerentwicklung entstehen und am Boden mit Radioantennen
detektiert werden können. Die Detektion solcher Pulse und deren Interpretation ist Thema die-
ser Arbeit. Erste Versuche auf dem Gebiet wurden bereits in den 1960er Jahren durchgeführt.
Wenn auch vielversprechend, wurde die Methode auf Grund der technischen Limitationen der
damaligen Zeit nicht weiter verfolgt. Seit Beginn des neuen Jahrtausends sind diese Limitatio-
nen nun mit heutiger Elektronik überwunden und die Radiodetektion von kosmischer Strah-
lung ist mittlerweile etabliert. Durch die modernen Radiodetektoren war es möglich Theorien
zu den Emissionsmechanismen der Radiostrahlung zu überprüfen und das Verständnis dieses
Phänomens weiter zu vertiefen. Es hat sich dabei gezeigt, dass die Radioemission auf zwei
verschiedenen Mechanismen beruht, die in Kombination zu den auf dem Erdboden beobach-
Zusammenfassung 203

teten Pulsen führen. Die Mechanismen basieren auf der zeitlich veränderlichen Trennung von
Ladungen während der Luftschauerentwicklung. Einerseits werden positive und negative La-
dungen durch die Lorenzkraft im Erdmagnetfeld getrennt. Andererseits tritt eine Trennung
von positiven Ladungen ein, wenn bewegliche negative Elektronen aus Luftmolekülen aus-
gelöst werden und ein positives Ion hinter sich zurück lassen. Die Ladungstrennung und die
sich gleichzeitig durch die Luftschauerentwicklung ändernde Teilchenzahl führen zu einem
sich ändernden Dipolmoment welches die Emission von elektromagnetischer Strahlung verur-
sacht. Da dieses Dipolmoment räumlich begrenzt ist, ergibt sich eine kohärente Überlagerung
der einzelnen Beiträge die zu den beobachteten Nanosekundenpulsen führt. Neben den experi-
mentellen Erkenntnissen ist damit heutzutage auch das theoretische Verständnis und damit die
Möglichkeit für Vorhersagen zur Radioemssion von ausgedehnten Luftschauern vorhanden.
Die Vorhersagen stimmen mittlerweile gut mit den Beobachtungen überein und haben dadurch
ihrerseits zur besseren Interpretation der Messdaten beitragen können. Die Radiodetektion von
ausgedehnten Luftschauern bietet wie auch die Detektion der Teilchen, die Möglichkeit eines
kontinuierlichen Betriebszyklus und ist nur eingeschränkt durch das Auftreten atmosphäri-
scher elektrischer Felder, wie sie z.B. in Gewitterwolken vorkommen. Dadurch bietet die Ra-
diotechnik bei entsprechender Sensitivität für die entscheidenden Eigenschaften Energie, An-
kunftsrichtung und Massenzahl (Luftschauerentwicklung) der gemessenen Teilchen ein sehr
mächtiges Werkzeug zur Erforschung der kosmischen Strahlung. Die Bestimmung der Sensi-
tivität im Hinblick auf diese drei Eigenschaften ist Hauptthema dieser Arbeit. Hierzu wurden
Daten des größten Radiodetektors für kosmische Strahlung, dem Auger Engineering Radio Ar-
ray, kurz AERA, analysiert. AERA ist ein Bestandteil des Pierre Auger Observatoriums und
wird in Kombination mit den dort etablierten schon beschriebenen Detektoren betrieben. Um
die Anzahl der detektierten Luftschauer zu vergrößern wurde AERA 2014 von 0.5 km2 auf
12 km2 und 2015 nochmals auf 17 km2 erweitert und umfasst nun 153 Radiodetektorstationen.
Die technischen Vorbereitungen und die Durchführung zur Erweiterung des Antennenfeldes
sind ebenso Teil dieser Arbeit. Auf Grund der gleichzeitigen Messung der ausgedehnten Luft-
schauer mit sowohl den Radio-, als auch den Fluoreszenz- und den Teilchendetektoren können
die gewonnen Ergebnisse untereinander kalibriert und verglichen werden. Dadurch ist es mög-
lich die Sensitivität der Radiomessungen mit den etablierten Detektortypen zu vergleichen,
die Technik weiter zu entwickeln und für zukünftige Forschungsprojekte als Haupttechnik zu
positionieren. Eine illustrierte Darstellung der hybriden Detektion von Luftschauern am Pier-
re Auger Observatorium ist auf Seite 195 gezeigt in der die wichtigsten Bestandteile sowohl
des Luftschauers als auch der Detektoren zusammengefasst sind. Der gezeigten Energiedich-
teverteilung des Radiosignals am Erdboden (farbig gekennzeichnet) kommt für die Analysen
im Rahmen dieser Arbeit eine besondere Bedeutung zu. Anders als die Rekonstruktion der
Ankunftsrichtung des primären Teilchens die auf den Signalankunftszeiten beruht, wird für
die Energiemessung und die Rekonstruktion der Luftschauerentwicklung die Verteilung der
Energiedichte herangezogen. Auf Grund der Interferenz der Radiostrahlung der zwei Emis-
204 Zusammenfassung

sionsmechanismen kann diese nicht durch eine rotationssymmetrische Funktion beschrieben

werden und eine analytische Beschreibung konnte erst kürzlich gefunden werden. Das Integral
über die Energiedichtefunktion entspricht der totalen Energie in der Radiostrahlung. Diese ist
proportional zur Energie des primären Teilchens, wie durch eine Kalibrierung mit dem Teil-
chendetektor des Observatoriums gezeigt werden kann. Die resultierende Energieauflösung ist
dabei vergleichbar mit der Energieauflösung von etablierten Detektionsmethoden. Auch für
die Rekonstruktion der Luftschauerentwicklung kann die Energiedichtefunktion herangezo-
gen werden. Dazu werden in dieser Arbeit zwei Methoden vorgestellt und auf AERA Daten
angewendet. Einerseits kann die Breite der Energiedichtefunktion als Indikator für die Entfer-
nung zum Ort der Radioemission, der wiederum von der Luftschauerentwicklung abhängt, ge-
nutzt werden. Andererseits ermöglichen aktuelle Simulationsprogramme für gegebene Energie
und Ankunftsrichtung (z.B. durch Messung mit den Teilchendetektoren) eine Vorhersage der
Energiedichteverteilung für verschiedene Luftschauerentwicklungen. Durch den Vergleich der
Vorhersagen mit den Messungen kann dann das wahrscheinlichste Szenario bestimmt werden.
Beide Methoden wurden im Rahmen dieser Arbeit zum ersten Mal überhaupt mit Ergebnissen
einer anderen, unabhängigen Detektionsmethode, in diesem Fall der Messung des Fluores-
zenzlichtes verglichen. Die Ergebnisse sind daher von Bedeutung für die aktuelle Forschung
und gleichzeitig auch für die Etablierung von Radiodetektoren als Hauptbestandteil zukünfti-
ger Observatorien für höchstenergetische kosmische Strahlung.
 Acknowledgments
During the work for this thesis, I was supported by many people, directly and indirectly, sci-
entifically and privately and I would like to thank these people.

I will start here with my supervisor Jörg Hörandel whom I met in the middle of nowhere and
who offered me to join his group for this PhD project. Jörg, thank you for having me work in
Nijmegen and for all the support and discussions during the last four years. I learned so many
new things and got in touch with many interesting people. Thanks for all the opportunities you
created during this time.
I would also like to thank my Promotor Heino Falcke for his comments on tricky scientific
questions and his general advices.
In addition to these two senior scientists from the department of astrophysics, I want to thank
Sijbrand de Jong and Charles Timmermans from the department of high energy physics for
their commitment to the AERA group in Nijmegen and their feedback and ideas related to my
work. Many thanks also go to John Kelley for introducing me to the subject of radio detection
of cosmic rays and to Stijn Buitink for all the discussions and help along the way of my the-
sis. I learned a lot about electrical engineering from Peter Dolron with whom I worked on the
subject of system integration.

Within the radio cosmic ray group in Nijmegen, I benefited a lot from the younger generation.
Anna Nelles and Pim Schellart often made my day either from the scientific or the social point
of view. With the two of them, our office became the hot spot for a special kind of humour.
I also enjoyed travelling with Anna to South and North America and having many fruitful
brainstorming occasions with Pim e.g., about selection criteria. I would further like to thank
all the PhD students from the Auger group at EHEF, Harm Schoorlemmer, Stefan Grebe, Ste-
fan Jansen, Guus van Aar, Giuseppe De Mauro and Fabrizia Canfora for their collaboration
and teamwork. A special thanks goes to Stefan Jansen for forcing me to speak Dutch in the
middle of the Argentinian Pampa. A big thank you also to my colleagues Laura Rossetto and
Antonio Bonardi for scientific discussions and an introduction to Italian gestures. Arjen van
Vliet deserves my thanks for answering all my in-between questions and also for making our
office vital again.

My office mates Marianne Heida and Thomas Kupfer created a warm and pleasant atmosphere
and introduced me to many new things on various subjects. I enjoyed the discussions about
soccer with Thomas despite his Bavarian opinion and fed Gianni under Marianne’s supervision
on a daily basis. Unfortunately only visiting irregularly, I enjoyed the company of Jan Kuijpers
206 Acknowledgments

in our office and benefited from his experience and profound knowledge of astrophysics.
Already during my first visit to the department of astrophysics I felt at home and this feeling
never left me again. Therefore, I would like to thank all members and ex-members of the
department for generating this great spirit making the department such a nice place to work. It
was often interesting to see how different the approaches between astrophysics and astroparti-
cle physics sometimes are and what one can learn from each other.

Working within the Pierre Auger Collaboration set up the framework for many fruitful dis-
cussions with colleagues from all over the world. Thanks to everyone how made this project
possible! Especially within AERA, I got in contact with many interesting people and I want
to thank them all for their collaborativeness. My closest foreign collaborator Christian Glaser
I would like to thank for all the inspiring discussions and teamwork and for becoming a friend
during many hours of scientific conversations and especially common travels to meetings, con-
ferences and the AERA site.

Before starting my PhD-project, numerous people from my former institute in Münster en-
couraged me to take this step forward. Thank you!

Despite the many great people from my scientific environment, I got the most important sup-
port from my family and friends. Thanks to all of those who helped me to be able to make it
to this point. Thanks for the incredible support during the most difficult times and for never
doubting that there is sunshine after the rain.
Without the great people I met and became friends with over the years I wouldn’t feel the same
way I do today. Thank you!

Ganz besonders möchte ich meiner Familie danken, dass sie mich immer bedingungslos un-
terstützt hat. Mama und Papa, ohne euch wäre es nie so weit gekommen. Danke für alles!

Was würde ich ohne dich machen? Du bist immer für mich da, danke Jenni.
 About the Author
Johannes was born in Lemgo (Germany) in October 1986. He attended the Christian-Dietrich-
Grabbe Gymnasium and specialized in physics and mathematics in the final years. During his
school time, he was active in the local swim team both as a swimmer and as a coach. In 2006,
he graduated and went to the University of Münster to study Physics. As an undergraduate
student he got in contact with the various facets of the subject of physics. In the following he
decided to specialize in the two fields of "nuclear and astroparticle physics" as well as "medical
physics and biophysics". For his Diploma thesis, he got the opportunity to work in the group
of Prof. Dr. C. Weinheimer on the topic of direct dark matter searches within the XENON-
project. He designed, constructed, and tested a 2-phase time projection chamber for electron
drift length measurements in liquid Xenon. Throughout this project, he worked in close col-
laboration with physicists, engineers and technicians further strengthening his interest in the
combination of science and engineering.
To broaden his scientific as well as his personal horizon, Johannes switched subjects and
moved to the Netherlands to work with Dr. J. Hörandel on the radio detection of cosmic rays
at the Radboud University in Nijmegen. The scientific results of his work are discussed in this
manuscript. Besides the presented topics, he was also actively involved in the deployment of
the second stage of the Auger Engineering Radio Array.
During his Diploma and PhD project, Johannes worked in an international collaboration and
presented his work at various conferences. He was also actively involved in teaching problem
sessions and lab classes for Bachelor and Master students.