Observation of Air Showers with the Iceact 7 Pixel Demonstrator in Coincidence with Icecube and Icetop

Observation of air showers with the IceAct 7 pixel demonstrator in coincidence with IceCube and IceTop

von

Erik Ganster

Masterarbeit in Physik

Vorgelegt der Fakultät für Mathematik, Informatik und Naturwissenschaften der Rheinisch-Westfälischen Technischen Hochschule Aachen

im Oktober 2018

angefertigt am III. Physikalischen Institut B Univ.-Prof. Dr. Christopher Wiebusch

revised version, January 2019

Betreuer Dr. Jan Auenberg III. Physikalisches Institut B RWTH Aachen

Erstgutachter Univ.-Prof. Dr. Christopher Wiebusch III. Physikalisches Institut B RWTH Aachen

Zweitgutachter Dr. Andreas Haungs Institute for Nuclear Physics Karlsruhe Institute of Technology

iii iv Contents

1 Introduction1

2 Cosmic Rays3 2.1 Energy Spectrum ...... 3 2.2 Air Showers ...... 5 2.3 Cherenkov Light ...... 7

3 The IceCube Neutrino Observatory 11 3.1 The IceCube In-Ice Array ...... 12 3.2 The surface detector IceTop ...... 16 3.3 IceCube-Gen2 ...... 17

4 The IceAct Telescope 19 4.1 Mechanical structure/design ...... 21 4.2 Camera and Optics ...... 22 4.3 Silicon Photomultiplier (SiPM) ...... 25 4.4 The IceAct demonstrator at South Pole ...... 29 4.4.1 Camera ...... 30 4.4.2 Data Acquisition System ...... 32 4.4.3 Telescope Operation ...... 34

5 Coincident Data Taking with IceAct and IceCube 37 5.1 Connection between IceAct and IceCube ...... 37 5.2 Method of Time Synchronization ...... 39 5.3 Time Synchronization Results ...... 41

6 Analysis of Coincident Events 45 6.1 Analysis methods ...... 45 6.1.1 Software trigger ...... 45 6.1.2 Signal Extraction ...... 47

v Contents

6.2 Event Selection ...... 48 6.2.1 IceAct Event Selection ...... 48 6.2.2 IceTop Event Selection ...... 54 6.2.3 IceCube Event Selection ...... 56 6.3 IceAct Telescope Performance ...... 59 6.3.1 Observing Cosmic Rays ...... 64 6.3.2 Energy threshold of the IceAct Telescope ...... 72 6.4 Correlation between IceAct and IceCube/IceTop ...... 74

7 Summary and Outlook 77

A Appendix 81

Bibliography 83

List of Figures 91

List of Tables 95

vi 1 Introduction

For more than a decade scientists all over the world have searched for the origin of cosmic rays and tried to explain their acceleration mechanisms leading to energies of 1020 eV. The IceCube Neutrino Observatory is a 1 km3 lage neutrino telescope melted deep in the glacial ice at the geographic South Pole. It was built to search for very high energetic neutrinos of astrophysical origin. The identication of their sources could answer the question for the sources of high energy cosmic rays, since they are believed to be the same. An air shower surface detector called IceTop is located above the IceCube in- ice array. It measures cosmic rays and can be used as a veto for cosmic ray induced muons and neutrinos that are the main background in IceCube's search for astrophysical neutrinos. A larger air shower detector on the surface is part of the Gen-2 extension of Ice- Cube. These surface detector should improve the veto capabilities of IceTop as well as the investigation of cosmic rays. One idea is the use of an array of small imaging air Cherenkov telescopes called IceAct. By observing an independent component of air showers, Cherenkov light, with a low energy threshold they could improve IceTop's composition measurements and veto cosmic rays by using the whole atmosphere in their eld of view as an active volume. In order to demonstrate the capabilities of such an air Cherenkov telescope even in the harsh environments at South Pole, an IceAct telescope demonstrator was installed in December 2015. After a brief introduction to cosmic rays and the principle of Cherenkov radiation an introduction to the IceCube Neutrino Observatory is given. The concept of Ice- Act telescopes is introduced followed by a chapter on how data is taken in coincidence with IceAct, IceCube and IceTop. The focus of this thesis is the analysis of coincident events recorded with all three detectors in 2016 proong a successful operation of an air Cherenkov telescope at South Pole and its dierent capabilities: vetoing cosmic ray air showers, composition measurements together with IceTop and energy calibration for IceTop. Parts of this thesis will be published in a paper on the design and performance of the rst IceAct demonstrator at South Pole.

1 1 Introduction

2 2 Cosmic Rays

Cosmic ray is a collective term for charged particles or nuclei that traverse through the interstellar medium, reach Earth and enter atmosphere. During balloon ights in 1912 Victor Hess observed a faster discharge of electroscopes with increasing altitude [1]. He correctly concluded that the ionization of air must increase with altitude causing the electroscopes to lose their charge faster. This ionization is caused by cosmic rays hitting the atmosphere and producing secondary charged particles in extensive air showers. The origin of cosmic rays reaching energies of more than 1020 eV (gure 2.1) is still unknown, as well as mechanisms accelerating particles to these energies or their sources. This chapter will give an introduction to cosmic rays and their well measured energy spectrum. It describes how air showers evolve in the atmosphere and what eect is used to measure these air showers at highest energies.

2.1 Energy Spectrum

The dierential energy spectrum of cosmic rays is shown in gure 2.1. Measurements of dierent experiments are combined to cover an energy range from 1013 eV to 1020 eV. The dierential ux, or number of particles, of cosmic rays is given in units of 1/(GeV m2 s sr). The spectrum follows a power law of the form:

dN ∝ Eγ . (2.1) dE To point out three important features of the spectrum it is often multiplied by a factor of Ex. In gure 2.1, it is multiplied with a factor of E2.6 and its main features (knee, 2nd knee and ankle) are marked. Direct measurements of cosmic rays are feasible up to energies of about 100 TeV. For higher energies large detection areas are required to collect a suitable number of events in a reasonable time. The power law behavior causes the rate to heavily decrease from more than 1000 particles per second and square meter at GeV energies to about one particle per square meter and year at PeV.[2]

3 2 Cosmic Rays

Figure 2.1: Energy spectrum of cosmic rays. Flux measurements of dierent experiments are shown. The spectrum is multiplied by E2.6 to point out its features: knee, 2nd knee and ankle. Taken from [3].

A steepening of the spectrum occurring between 1015 eV to 1016 eV is called knee (gure 2.1). At the knee the spectral index γ of equation (2.1) changes from ≈ −2.7 to ≈ −3.1, a softening of the spectrum. Energy and ux measurements of cosmic rays at the knee and beyond rely on indirect measurements of secondary particles produced in air showers (section 2.2) by ground based instruments.[2, 4] A second, further softening of the spectrum at about 4 × 1017 eV is called 2nd knee. At about 4 × 1018 eV the spectrum recovers to a spectral index of ≈ −2.7 again as can easily be seen in gure 2.1. This transition is called ankle. Event rates lower than one particle per square kilometer and century above 100 EeV requires huge detectors for measurements. The worlds largest detector for cosmic rays is the Pierre Auger Cosmic Ray Observatory in the Province of Mendoza, Argentina, covering an area of ≈ 3000 km2 [5].[2]

4 2.2 Air Showers

Dierent models have been evolved to explain the knee and the ankle in the cosmic rays' energy spectrum. Common explanations uses dierent acceleration mechanisms and origins to describe these features. Knee and 2nd is associated with an upper limit of acceleration in galactic sources like supernova remnants and dierent compositions from light particles to heavy nuclei. The ankle, on the other hand, is the onset of an extra galactic component with a harder spectrum compared to galactic sources.[2, 6] An interaction of very high energetic cosmic rays with photons from the cosmic mi- crowave background (CMB) called GZK-cuto would lead to a hard cuto in the energy spectrum at about 6 × 1019 eV. The GZK-cuto is driven by production of a ∆+- resonance:

p + γ → ∆+ → p + π0 → n + π+ .

It was introduced and rst calculated by K. Greisen, G. T. Zatsepin and V. A. Kuzmin in 1966 giving it the name GZK-cuto shortly after the CMB was discovered by A. Penzias and R. Wilson [79].

2.2 Air Showers

Primaries of the cosmic rays will produce so called air showers when they hit Earth's atmosphere and interact with air nuclei. Such air showers can be divided into two categories (or parts): electromagnetic and hadronic showers. A simple model for the description of processes in electromagnetic showers was introduced by W. Heitler in 1954 and can be expanded to hadronic cascades. The model cannot replace fully detailed simulations of the processes inside an air shower, especially for hadronic ones, but it describes the most important features fairly well. The Heitler model for electromagnetic and its adapted version for hadronic showers is shown in gure 2.2 and shall be discussed in more detail.[10, 11] The left side of gure 2.2 shows an electromagnetic shower induced by a primary high energy photon. In a rst step the photon produces an electron-positron pair. Each electron and positron will radiate a single photon by bremsstrahlung after traveling a certain distance in the atmosphere. In both cases the energy is assumed to be equally divided between both outgoing particles. Both processes, pair production and bremsstrahlung repeats several times and form an electromagnetic cascade of high energetic particles. Using the assumption of electrons, positrons and photons losing half of their energy be-

5 2 Cosmic Rays fore next interaction, the distance d between each step n of the shower (gure 2.2) can easily be calculated:

d = λr · ln(2) , with being the electromagnetic radiation length in the medium ( air 2).[11] λr λr ≈ 36 g/cm

Figure 2.2: Schematic of the Heitler model for air shower description. Left side shows an electromagnetic cascade induced by a primary photon while the right side shows an adaption of the Heitler model for the development of hadronic showers. Taken from [11].

The total number of particles after n steps and a total distance x is simply given by:

n x/λr x = nλr ln(2) ,N = 2 = e . (2.2)

The cascade ceases when the energy of the particles is to low for further pair production or bremsstrahlung. In the Heitler model this energy is called critical energy EC , and in air it is ≈ 85 MeV. If the energy of the primary particle is E0, the maximum number of particles and the depth when it is reached can be calculated using (2.2):

E0 Xmax/λr Nmax = = e (2.3) EC E0 ⇔ Xmax = λr · ln (2.4) EC

The total number of particles Nmax is described fairly well by the Heitler model and the depth of the shower maximum is predicted in very good accordance with a full simulation

6 2.3 Cherenkov Light for an electromagnetic shower. More details on the Heitler together with this simulation is given in [11]. For pure electromagnetic cascades the Heitler model holds well but it cannot describe showers induced by hadrons, e.g. protons (2.2 right). Nevertheless, models use a similar approach with an hadronic interaction length λI instead of radiation length λr. This 2 length is assumed to be constant over energy, e.g. λI ≈ 120 g/cm for pions in air.

Instead of a simple pair production or bremsstrahlung after one interaction length, Nch charged pions ( ±) and 1 neutral pions ( 0) are produced. Because of their short π 2 Nch π lifetime π0 immediately decay into two photons each initiating an electromagnetic sub- shower as described above. Charged pions again produce charged and neutral pions until their energy falls below critical energy π± . Remaining pions then will decay yielding EC muons and neutrinos. Assuming an equal energy division in each generation, charged pions carry a total energy of 2 n , where is the energy of the cosmic ray primary.[11] 3 E0 E0 For other cosmic ray primaries, e.g. iron nuclei, a superposition model is used. A nucleus with mass number A and energy E0 is taken as A individual single nucleons

(protons) each with energy E0/A. The resulting shower is an overlay of A individual proton showers all starting at the same point. This superposition predicts a higher Xmax for iron showers compared to single protons. In addition, an iron shower will produce ≈ 1.8 times more muons than a proton induced shower with same primary energy. Both quantities are important parameters to determine the composition of cosmic rays.[11]

2.3 Cherenkov Light

Cherenkov light is named after the Soviet physicist Pavel Alekseyevich Cherenkov who rst discovered "Visible Radiation Produced by Electrons Moving in a Medium with Velocities Exceeding that of Light" in 1934. The eect was later interpreted by Ilya Mikhailovich Frank and Igor Yevgenyevich Tamm who were, together with Cherenkov, awarded the Nobel Prize in physics for "the discovery and the interpretation of the Cherenkov eect" in 1958.[1214]

For a medium with refractive index n the velocity of light is given by c0/n < c0 where c0 denotes to the speed of light in vacuum. A particle, with a sucient amount of energy, can reach a velocity faster than the speed of light in that medium. Charged particles traversing a dielectric medium polarizes surrounding atoms/molecules along its path as illustrated in gure 2.3 and two cases can be distinguished based on their velocity:

7 2 Cosmic Rays

(a) For particle velocities v < c0/n polariza- (b) If the velocity of the particle exceeds the

tion no overall polarization occurs, hence speed of light (v > c0/n) a net polarization there is no Cherenkov light emission. occurs along its path resulting in the emission of Cherenkov radiation.

Figure 2.3: Illustration of polarization yielding to Cherenkov light in a dielectric medium. Taken from [15] .

c0 In this case there is no net polarization in the dielectric medium. The po- v < n : larization is spherical symmetric around the particle traversing the medium and hence superimposing destructively.

c0 When the velocity of the charged particle exceeds the speed of light, an v > n : asymmetric polarization appears. Polarization only propagates with c0/n in the medium which yields to a cone of polarized atoms/molecules moving along the particle's track. This time dependent polarization causes the directed emission of light as observed by Cherenkov.

I. M. Frank and I. E. Tamm interpreted the eect observed by Cherenkov and developed the famous Frank-Tamm-formula describing number of photons emitted per unit length

8 2.3 Cherenkov Light and wavelength interval [13]:

d2N 1 1 = 2παq2 · 1 − , where dxdλ λ2 n2(λ) · β2 α = ne structure constant, q = particle charge, λ = wavelength, n = refractive index of the medium, wavelength dependent.

The shorter the wavelength the more photons are generated. The fact that the refractive index gets < 1 for x-rays sets a natural lower limit on the generated wavelength spectrum [15].

(a) Geometrical sketch of the Cherenkov light (b) Spectrum of Cherenkov light by a 1 TeV emission angle. Taken from [15]. vertical air shower in a height of 2200 m in arbitrary units. Taken from [16].

Figure 2.4: Cherenkov spectrum and illustration of emission angle.

The direction of emission of Cherenkov light is characteristic for each medium and is based solely on its refractive index:

c/n 1 cos(θ) = = , v βn

9 2 Cosmic Rays with v as can be easily seen by comparison with gure 2.4a. For air with β = c nair ≈ 1.0003 this yields to an angle of ≈ 1.4◦. The minimum energy of a charged particle to emit Cherenkov light is given by:

1 m c2 β > ⇔ E > 0 0 , n q 1 1 − n2 with m0 being the rest-mass of the particle. Electrons (m0 = 511 keV) need to have an energy exceeding 20.9 MeV in order to emit Cherenkov light in air. Air showers, as described in section 2.2, produce high energy charged particles, especially electrons and positrons. There energies exceed the Cherenkov threshold of ≈ 20 MeV which results in Cherenkov light emitted by air showers in the atmosphere. Air Cherenkov telescopes like VERITAS [17], H.E.S.S. [18] and MAGIC [19] detect this light and reach eective collection areas of more than 105 m2 by using Earth's atmosphere as their detection volume. These huge collection ares make them ideal instruments to investigate the high energy cosmic rays (but mainly γ-rays) as their rate falls with a steep power law spectrum (section 2.1). The Cherenkov spectrum peaks at about 350 nm as shown in gure 2.4b. A typical light pool on the surface has a radius of about 130 m with ≈ 100 photons/m2 arriving in a short ash (O(ns)). It has no sharp cuto but is smeared out as can be seen in gure 2.5. It shows the radial light pool brightness in a.u. normalized to a distance of r = 100 m as measured by the HEGRA telescope in a height of 2200 m on La Palma [20].[3, 16]

Figure 2.5: Light pool of a γ-ray induced air shower as measured on ground. The light pool brightness in a.u. is normalized to r = 100 m and measured for γ-rays in the energy range of 3.3 TeV to 6.2 TeV. Taken from [20].

10 3 The IceCube Neutrino Observatory

Located at the Amundsen-Scott South Pole station in Antarctica the IceCube Observa- tory is currently the largest neutrino telescope on Earth. It consists of a huge cubic- kilometer in-ice array and a square-kilometer surface detector array as shown in the sketch in gure 3.2. The greatest successes include the observation of high energy extraterrestrial neutrinos in 2013 and the detection of a neutrino coincident in time and direction with a gamma-ray are of a blazar [2123].

Figure 3.1: Artistic view of cosmological particle propagation. Neutrinos are neither absorbed like photons nor deected by magnetic elds like charged particles. They directly point to their source. Taken from [24].

Since neutrinos are neither absorbed (like photons) nor deected by magnetic elds (like charged particles) they are ideal for point source searches as they point directly back to their source (gure 3.1). The following sections will give an overview of the design and detection principle for the in-ice array and the surface detector of the IceCube Neutrino Observatory. Afterwards an overview of future extension plans is given.

11 3 The IceCube Neutrino Observatory

3.1 The IceCube In-Ice Array

The in-ice array (IceCube Array) shown in gure 3.2 consists of 5160 optical sensors called Digital Optical Modules (DOM, see gure 3.4) that are deployed on 86 strings deep in the ice. The strings reach a depth of ≈ 2450 m below surface whereby only depths between 1450 m and 2450 m are instrumented with DOMs. In total, they instrument a volume of ≈ 1 km3. The 78 strings of the primary IceCube Array follow a hexagonal footprint with a spacing of 125 m. Eight strings build the DeepCore inll-array that has a lower string spacing of 41 m to 105 m and a DOM-to-DOM spacing of 7 m compared to 17 m. This design was chosen to have a high sensitivity in the energy range of O(TeV − PeV) in the array and O(10 − 100 GeV) in the DeepCore inll.[25]

Figure 3.2: Sketch of the IceCube Neutrino Observatory. The true to scale sketch of the Eiel Tower in the bottom right corner gives a feeling for the size of IceCube.

The deep glacial ice at South Pole has ideal optical properties for the detection of Cherenkov light (2.3) emitted by charged particles traversing the in-ice array. It is very clear with almost no impurities and bubbles in depths below 1400 m. The depth of IceCube also reduces the background of atmospheric muons for the neutrino detection in

12 3.1 The IceCube In-Ice Array

IceCube signicantly due to absorption in the ≈ 1.5 km ice above the array.[26]

Charged particles traversing the ice with a higher speed than light in ice (nice = 1.356 at 400 nm [27]) emit Cherenkov light as described in section 2.3. They can originate from cosmic ray air showers (section 2.2) or from charged current interactions of atmospheric and astrophysical neutrinos. The later one, neutrinos of astrophysical origin, are events IceCube is hunting for. All events in IceCube can be roughly divided into two dierent event topologies:

(a) Example of a shower in the IceCube array. (b) Example of a track-like event. In addi- Light emission is spherically around the pri- tion to spherical light emission from the ini- mary vertex. tial vertex the track of an outgoing muon is celarly visible.

Figure 3.3: Examples of the two event topologies in the IceCube array. Shown are the strings together with their DOMs in gray, DOMs that detected photons are colored corresponding to their arrival time. The size of the color bubbles is proportional to the amount of light that was detected. Taken from [21]

Track-like events A high energy muon neutrino produces an outgoing muon along with a hadronic shower at the vertex of a charged-current interaction with a nucleus from the ice. The outgoing muon traverses the ice and emits a Cherenkov light cone along its track (compare with 3.3b). Above a critical energy of about 1 TeV muon energy losses are dominated by radiative processes with large uctuations which makes

13 3 The IceCube Neutrino Observatory

a conversion from deposited energy in the detector to muon or neutrino energy dicult. In contrast, the angular reconstruction reaches a resolution of about 0.6◦ due to the track's large lever.[25, 28, 29] Electromagnetic or hadronic showers Showers can be induced by neutrinos of all avors through neutral-current interactions with surrounding nuclei Electron (anti)neutrinos will also produce a shower in charged-current reactions. Light emission of such an electromagnetic or hadronic shower is almost spherically around the primary neutrino interaction vertex (compare with 3.3a)). The light output is directly proportional to its energy which makes the conversion from deposited light to neutrino energy of such events much more precise than for track-like events. The average resolution of deposited energy is about 15 % in all channels.[30]

(a) Picture of a DOM. The upper hemisphere (b) Sketch of the inner layout of a DOM. Com- houses all electronics while the lower one munication, readout and digitization elec- contains the 10 inch PMT. Taken from [31]. tronic is located on the annular main board. Taken from [25]

Figure 3.4: Picture and a sketch of an IceCube DOM.

Light emitted in the detector gets detected by the so called Digital Optical Modules, short DOMs. A picture of an IceCube DOM and a sketch of its inside is shown in g- ure 3.4. Each DOM houses a 10 inch PMT facing down, data acquisition, control and communication electronics on an annular circuit board (main board) in the upper hemisphere. For in situ calibration purposes each DOM is equipped with light-emitting diodes (LEDs). All 5160 DOMs of IceCube (and those of IceTop, section 3.2) are connected to

14 3.1 The IceCube In-Ice Array the IceCube Lab (ICL) in the middle of the array on the surface via cables (gure 3.2). As the name "Digital Optical Module" already suggests, the analog recorded data of the PMT gets digitized by the DOM itself before it is sent to the ICL. Together with charge information a timestamp for the PMT signal is sent which allows for an angular reconstruction of the primary particle direction. The reconstruction resolution depends directly on the time accuracy of each DOM. For this reason clocks and signal travel times to the ICL are calibrated every second with an accuracy of < 1 ns for all DOMs.[25] In addition to that, the PMT hit rate is recorded in time intervals of 1.6384 ms for each DOM. Low energy neutrinos (O(10 MeV)), e.g. from a supernova in our galaxy, would cause an overall increase of the PMT rate but be to dim (in terms of light yield) to trigger the standard DAQ. This design expands the range of applications from the search for astrophysical neutrinos and their sources to the detection of supernovae in our galaxy.[32]

15 3 The IceCube Neutrino Observatory

3.2 The surface detector IceTop

IceTop is a cosmic ray air shower array located on top of the IceCube in-ice array at the South Pole at an altitude of 2835 m. It consists of 162 Cherenkov tanks lled with clear ice and arranged in 81 stations. The stations follow the same hexagonal pattern as the strings of the IceCube in-ice array (section 3.1) with a spacing of approximately 125 m between stations (gure 3.5a). The two tanks of each station are about 10 m apart. A denser inll array is situated on top of the in-ice inll array DeepCore next to the ICL in the center.[33]

(a) Layout of the IceTop surface array. Shown (b) Sketch of an IceTop tank. It consists of are the locations of IceCube strings (section two DOMs that are embedded into ice inside 3.1) as well as IceTop stations. tank coated with a reective liner.

Figure 3.5: Array layout of the IceTop surface detector and a detailed sketch of one of its stations. Taken from [33].

A cross section sketch of an IceTop tank is shown in gure 3.5b. The tank itself is made out of polyethylene coated with a reective liner on the inside and closed by a wooden lid. Up to a height of 0.9 m it is lled with clear ice in which two standard IceCube DOMs are embedded. To increase the dynamic range of air shower detection, the DOMs are operated at dierent gains: one at low gain (105) and one at high gain (5 × 106).[25, 33] The design of IceTop and its location in ≈ 2800 m height denes an energy threshold of about 300 TeV (lowered to about 100 TeV for the denser inll) for an ecient air shower reconstruction that requires at least 3 station hits. The area of ≈ 1 km2 limits cosmic ray measurements to about 1 EeV. Hence, it covers the region around the knee of the

16 3.3 IceCube-Gen2 cosmic ray energy spectrum completely. The energy resolution is strongly dependent on the air shower energy, reaching from 25 % at 2 PeV to 12 % above 10 PeV.[34] Besides the measurement of the cosmic ray spectrum IceTop serves as a partial veto for down-going muons in IceCube and for direction calibration. In a hybrid operation together with the in-ice array it is capable of analyzing the cosmic ray composition and the consistency of surface muon measurements with dierent hadronic interaction models.[33, 35]

3.3 IceCube-Gen2

IceCube-Gen2 is a planned extension of the IceCube Neutrino Observatory. It will enlarge the existing IceCube array by nearly a factor of 10 to expand IceCubes capabilities in the measurement of very high energy neutrinos. Very long absorption lengths of photons in the deep ice at South Pole allow string spacing of 250 m reducing costs signicantly. In addition, a new dense inll array inside the existing DeepCore array would lower the energy threshold down to several GeV to improve atmospheric oscillation measurements as well as indirect searches for dark matter. A view of a possible layout is shown in gure 3.6.[36, 37]

(a) Top view. (b) Side view.

Figure 3.6: Top and side view of proposed Gen2 extensions of the IceCube Neutrino observatory. The red area is the footprint of the IceCube Neutrino Observatory as of completion in 2010. The proposed high energy in-ice extension is shown (light blue) together with even bigger cosmic ray arrays surface detectors (green and yellow). Taken from [38].

Besides high energy in-ice extension with further strings and improved DOMs, a new surface detector is proposed to expands the existing IceTop surface detector. One purpose

17 3 The IceCube Neutrino Observatory of this new and much bigger surface detector is a veto for cosmic ray air showers. High energy muons produced in air showers easily reach the in-ice array in a depth of about 1.5 km forming the main background for astrophysical neutrino searches. By detecting air showers in the atmosphere above the in-ice detector such muons can eectively be vetoed. IceTop is already used as a veto detector in several analyses.[39, 40] For a future surface array, acting as a veto for the in-ice array, three dierent approaches are currently discussed: scintillator detectors (IceScint), radio antennas, and imaging air Cherenkov telescopes (IceAct). In addition to veto capabilities all three approaches will improve the measurement of cosmic rays composition by combining dierent observation channels.[36, 37, 4144] In order to demonstrate a successful operation of an air Cherenkov telescope at South Pole, an IceAct telescope demonstrator has been deployed in December 2015. This thesis focus on the analysis of data taken with this demonstrator in coincidence with IceCube and IceTop.

18 4 The IceAct Telescope

Imaging air Cherenkov telescopes are ideal instruments to observe cosmic- and especially γ-ray air showers. First observations of Cherenkov light ashes from by cosmic rays were performed by W. Galbraith and J.V. Jelly in 1952 [45]. Modern Cherenkov telescopes like VERITAS [17], H.E.S.S. [18] and MAGIC [19] using mirrors with diameters up to 17 m observe high energetic γ-rays in the energy regime of about 50 GeV to 10 TeV. They reach an angular resolution of 0.07◦ allowing them to draw detailed maps of γ-ray sources in the night sky.

Figure 4.1: Photograph of the IceAct imaging air Cherenkov telescope installed at South Pole in 2017. This picture shows the successor of a 7-pixel demonstrator that was deployed at South Pole in 2015 (gure 4.10). Provided by [46].

Additionally, air Cherenkov telescopes can be used to improve measurements of ground based particle detectors as recent studies show [47]. By detecting Cherenkov light emitted mainly by the electromagnetic part an air shower, they draw an image of the shower during its maximum development in the atmosphere. Cherenkov light yield is directly

19 4 The IceAct Telescope proportional to energy of the primary particle, thus Cherenkov telescopes, even small ones, could improve energy measurements of classical air shower arrays signicantly. They could also improve the composition measurement of cosmic ray primaries by projection the light yield of a shower onto its axis reconstructed by an air shower array.

The maximum Cherenkov light yield is very close to Xmax which directly depends on the primary particle type. In combination with an air shower array a Xmax measurement is possible and thus e.g. a separation of proton and iron induced showers. Studies on the ability of composition measurements of small air Cherenkov telescopes are currently ongoing.[47] The IceAct telescope is such a small air Cherenkov telescope. Its design is adapted from the 61-pixel uorescence telescope FAMOUS developed for the Pierre Auger Observatory that uses silicon photomultipliers (SiPM ) for photon detection instead of classical photo multiplier tubes (PMT ) [48]. Due to its small size and its setup as a fully enclosed system, an array of IceAct imaging air Cherenkov telescopes is proposed as part of the high energy Gen2-extension of the IceCube Neutrino Observatory. Besides expanding veto capabilities for muons in the in-ice array it can be used to improve IceTop's energy and composition measurement of cosmic ray air showers. A further modication of the FAMOUS telescope, called HAWC's Eye was successfully operated together with the HAWC gamma ray observatory. A rst analysis of this hybrid setup shows and proofs that even small air Cherenkov telescopes with 61-pixels can improve energy measurements signicantly [15].[5, 44, 49] In order to demonstrate the capabilities and a successful operation in the harsh environment at South Pole, an even smaller telescope demonstrator with a 7-pixel camera was deployed in December 2015. This chapter will discuss the mechanical setup of a (full) 61-pixel IceAct telescope, its optical design, its camera and data acquisition system as well as its smaller 7-pixel demonstrator version. A successor of the demonstrator telescope was deployed in exchange in January 2017 and is shown in gure 4.1.

20 4.1 Mechanical structure/design

4.1 Mechanical structure/design

The basic design of IceAct is adapted from the FAMOUS uorescence telescope [48]. It features an enclosed system design approach, that is very well suited for operation at South Pole. A mechanical sketch of the telescope is shown in gure 4.2.

glass plate 549.7 mm Fresnel lens 23.5 mm

134.8 mm 61 pixel mm 502.1 camera 134.8 mm camera with Winston cones

Data acquisition and storage

Figure 4.2: Mechanical sketch of the IceAct imaging air Cherenkov telescope. A glass plate that is covering a Fresnel lens is used in combination with a carbon ber tube to build an enclosed system to protect the telescope from the harsh environment at South Pole. An insulated box beneath the tube houses the data acquisition (DAQ) system while ensuring a safe stand of the telescope. The inlay (upper left) shows a more detailed sketch of the 61-pixel camera with attached Winston cones to each pixel. Adapted from [44, 50].

A carbon-ber tube with a length and diameter of about 0.55 m is covered by a 4 mm BOROFLOAT 33 borosilicate glass plate that protects the entire system from snow, dust and very low temperatures. To prevent snow accumulation on the glass plate itself, it is ush mounted with the tube allowing wind to simple blow o snow. To prevent reections inside the tube it is lined with black fabric. The tube is mounted on a wooden box housing the data acquisition system and ensuring a safe stand of the telescope while allow tilting of the whole telescope. The camera with a diameter of ≈ 13 cm and a total

21 4 The IceAct Telescope height of ≈ 2.5 cm) is embedded in the base plate of the tube in the focal plane of a Fresnel lens. It is connected to the data acquisition system via cables.[44, 51] Details on the optics consisting of a Fresnel lens and light collecting cones attached on the SiPMs are given in section 4.2. The working principle of SiPMs is described in section 4.3.

4.2 Camera and Optics

An Orafol SC943 Fresnel lens is used as the main optic of the telescope. It has a diameter of D = 549.7 mm, a focal length of f = 502.1 mm (for 546 nm [52]) and is installed directly beneath the glass plate protecting it from scratches. The lens has a thickness of about 2 mm. Detailed measurements of optical properties of the Fresnel lens, point spread function (PSF ), focal length, etc. can be found in [53]. The camera uses 6 mm × 6 mm SensL J-Series silicon photomultipliers as single pixels. More details on the working principle of SiPMs can be found in section 4.3. A hexagonal arrangement of all 61-pixels leads to same distance of 15 mm between all pixel centers as can be seen in gure 4.3. In addition, three "blind" pixels are located outside the main hexagonal shaped (a) (b) camera.[44] Figure 4.3: Picture of IceAct's camera cir- To improve light collection of the cuit board with (left) and without (right) SiPMs, Winston cones are glued on all 61- Winston cones glued on the SiPMs pixels. Figure 4.3 shows the circuit board housing all SiPMs with (left) and without (right) Winston cones glued on. Winston cones are non imaging light collectors (or concentrators) using reection on a parabolic shaped surface to redirect rays. An ideal Winston cone have a circular entrance and exit window and gathers light that enters the cone up to a maximum acceptance angle θM (compare with gure 4.4a). The light concentration ratio of such a cone is given by [54]:

2 2 Aen πren n = 2 = 2 . Aex πrex sin(θM ) ren and rex corresponds to the radii of entrance and exit window while n denotes the refractive index of the cone material. By choosing both radii, the maximum acceptance

22 4.2 Camera and Optics

angle θM is xed and thereby the parabolic shape and the length L of the cone. A close- up picture of cones used for the IceAct telescope is shown in gure 4.5. More details on Winston cones can be found in [54].

Ω Ω

D = 549.7mm

Ω 502.1mm = f

γ δ

rfp = 60mm

(a) Skecth of the prole of a Winston cone (b) Sketch of the light path through the IceAct with circular entrance and exit window. telescope dening the eld of view. Taken Adapted from [50]. from [44].

Figure 4.4: Sketch of a Winston cone and of the telescope optic.

Since circular entrance and exit windows are neither ideal for a hexagonal camera layout nor for squared SiPMs, a special Winston cone design was developed in [55]. This design combines a hexagonal shaped entrance window for a maximal ll factor in the focal plane and a squared exit window tting exactly the SiPM. The nal design of such a "Hex-to-Square" cone is shown in gure 4.5 (left side). These cones are made of PMMA and use the principle of internal total reection at the transition PMMA-air. The distance between camera and Fresnel lens is selected so that the cone entrance windows lie in the focal plane.

23 4 The IceAct Telescope

The eld of view (FOV ) of the 61-pixel camera can be calculated using simple geo- metric as shown in gure 4.4b:

r FOV = 2Ω = 2 arctan fp f 60 mm = 2 arctan 502.1 mm ≈ 2 · 6.8◦ = 13.6◦ , where rfp = 60 mm gives the distance from the optical axis to the outermost Winston cone. The additional angles γ and δ dene the Winston cones needed acceptance angle and can be calculated in the same way, yielding γ ≈ 28.7◦ for the central cone and δ ≈ 33.7◦ for the outermost.

Hex-to-Square Cone Former Circular Al Cone 14.8mm

6mm 6mm

13.42mm

Figure 4.5: Close-up photography of Winston Cones used for IceAct telescopes. Left one is the newly developed "Hex-to-Square" PMMA cone while the right cone is a standard circular cone made of aluminum and used in the 7-pixel demonstrator telescope (section 4.4). Taken from [55].

24 4.3 Silicon Photomultiplier (SiPM)

4.3 Silicon Photomultiplier (SiPM)

Silicon photomultiplier (SiPM ) are devices used for photon detection in dierent applications. In contrast to photo-multiplier tubes (PMT ) that are based on a system of dynodes accelerating and multiplicating a primary electron emitted by a photo-cathode, SiPMs consists of an array of Geiger-mode avalanche photo diodes (GAPD) that feature a higher photon detection eciency (PDE). GAPDs are based on p-n junctions shown in gure 4.6. By doping a semiconductor with impurities free moving electrons and holes can be inserted. In case of silicon commonly phosphor is used for n-doping since it adds one electron to the lattice while aluminum has one electron less than silicon and is used for p-doping. When an n-doped material is brought into contact with a p-doped material, free electrons of the n-doped region will ll the holes in the p-doped region, this process is called electron-hole annihilation. It can be visualized by electrons moving to the p-doped side of the p-n junction while positive charged holes move opposed. A region with no free charges (neither electrons or holes) is created and called depletion zone. Due to separated electrons and holes it is charged with an electrical eld pointing from n-doped to the p-doped side. As more electrons moves towards the p-doped region, the electrical eld gets stronger until it compensates the energy that is gained by electron-hole annihilation and an equilibrium is reached.[56]

n-doped Depletion zone p-doped + + __ Electron Hole + + + __ _ Photon + + __

Figure 4.6: Sketch of a p-n junction. The left side is n-doped while the right side is p-doped. When brought into contact free electrons from the n-doped region ll holes in the p-doped region yielding a so called depletion zone with an electric eld pointing from n- to p-doped side. By applying a voltage this depletion zone can be increased. Adapted from [55].

By applying a reverse voltage with the anode at the n-doped side of the p-n junction the depletion zone can be increased. An incoming photon can generate a new electron- hole pair by the photoelectric eect in the depletion zone. This newly created pair is accelerated in the electric eld resulting in a current ow in the p-n junction. Further

25 4 The IceAct Telescope increase of the applied reverse voltage will directly increase energy gain of electron-hole pairs in the depletion region. If this gain is sucient to generate more electron-hole pairs, an avalanche is formed resulting in a breakdown of the applied voltage.

If multiple layers of n and p doped mate- rials are combined with an applied voltage greater than the breakdown voltage of the p-n junction, it is possible to maximize energy gain in the depletion zone and detect single photons. These cells are then called GAPD - Geiger-mode avalanche photodi- ode. The phrase "Geiger-mode" indicates, that the output of such a cell is independent of the actual number of photons detected. They are designed in a way that even single photons initiate an avalanche Figure 4.7: Structure of a single GAPD cell that charge depends on the applied voltage used in SiPMs. It shows a -on- type p n and the used p-n junction only. A quench- diode structure optimized for high photon ing resistor is used to stop the avalanche detection eciency in the blue and UV. once it is initiated and allow the cell to re- Taken from [57]. cover quickly. An example of the layout of a GAPD cell used in the SensL J-Series SiPMs used in the IceAct telescope is shown in gure 4.7. A p-on-n diode structure is used that features a high photon detection eciency in the blue and UV region, the peak region of Cherenkov light (compare with gure 2.4b).[15, 5759]

A SiPM consists of several hundreds to thousands GAPDs arranged in an array. A close up photography of a SensL J-Series SiPM is shown in gure 4.8 (left) along with a simplied representative schematic (right). The SensL J-Series SiPM used in the camera of the IceAct telescopes consists of 22 929 GAPDs, each 35 µm × 35 µm in size, parallel connected and arranged in a 6 mm × 6 mm grid. Each GAPD has its own quenching resistor. The special Fast Output of SensL's J-Series is not used by the camera of the IceAct telescope.[59]

26 4.3 Silicon Photomultiplier (SiPM)

(a) Image of the front (left) and back (right) of an SensL J-Sereis (b) Simplied schematic SiPM as used in the IceAct telescope. It consits of 22 292 of an SiPM showing 12 35 µm × 35 µm GAPDs that arranged in a 6 mm × 6 mm ar- GAPDs connected in ray. The ne structure of this array is barely visible. parallel.

Figure 4.8: Image and simplied schematic of an SiPM. Taken from [59]

The signal of an SiPM is given by a multiple of the output of a single GAPD cell - corresponding to the number of cells that break down. This results in a discrete pulse spectrum. An overlay of SiPM waveforms recorded with an oscilloscope is shown in gure 4.9.

Figure 4.9: Overlay of measured SiPM waveforms. Peaks labeled 1, 2, 3 p.e. correspond to 1, 2 and 3 GAPD cell breakdowns. Since GAPD detect single photons their signal is called photo electron (PE) equivalent. Taken from [58].

27 4 The IceAct Telescope

The shape of an SiPM pulse can be described by an overlay of two exponential functions [60]:

! 1 t−t0 A(t, N) = cN · 1 − · e λ . t−t0 1 + e τ

The rst exponential with time constant τ describes the steep rising edge caused by the avalanche in a GAPD cell while the second one with time constant λ describes the tail where the GAPD cells recover after a breakdown. Later one is very similar to the voltage measured when charging or discharging a capacitor. N corresponds to the number of cell breakdowns while c denotes to a calibration factor giving the amplitude of a single GAPD breakdown (1 PE). Typical values are 0.9 ns to 1.1 ns for τ and ≈ 20 ns for λ.[60] As mentioned before, the gain and hence photon detection eciency of a GAPD depends directly on the voltage applied to them. This bias voltage Vbias is given by the sum of the breakdown voltage Vbd of the cell, marking the beginning of Geiger-mode operation, and the over-voltage Vov which is directly proportional to the gain of the GAPD:

Vbias = Vbd + Vov .

However, the breakdown voltage of GAPDs strongly depends on temperature but can be described linear in a wide range of temperatures [61]:

Vbd(T ) = Vbd(T0) + β · (T − T0) .

SensL C/J-Series SiPMs that are used in the IceAct telescope have a temperature co- ecient β of 21.5 mV/K [59, 62]. For an analysis it is crucial to operate SiPMs with a constant gain and hence a constant photon detection eciency. Thus, the over-voltage has to kept constant what can be achieved by precisely regulating the bias voltage de- pending on the actual temperature of the SiPMs.

28 4.4 The IceAct demonstrator at South Pole

4.4 The IceAct demonstrator at South Pole

To demonstrate the capabilities of an air Cherenkov telescope even in such harsh environments as at South Pole, a 7-pixel demonstrator telescope was installed in December 2015. A picture of the deployment at South Pole is shown in gure 4.10. It shows the telescope itself (right side) and the deployment crew with Leif Rädel (second from right) from RWTH Aachen University. The telescope is closed with a light proof lid during astronomical day at South Pole since sunlight entering the telescope would get focused by the Fresnel lens and could seriously damage the telescope. In contrast to the IceAct telescope described in sections 4.1 and 4.2, this technology demonstrator has a 7-pixel camera only. The pixels are arranged in a 2 two-ring hexagonal pattern. In addition to that, the acrylic (PMMA) Winston cones were developed during operation of the 7-pixel demonstrator in 2016/17 that was equipped with circular hollow aluminum cones instead. Fresnel lens and the glass plate covering it, on the other side, are the same. More details on the setup and the operation of this demonstrator are given in the following sections.

Figure 4.10: Deployment of the IceAct telescope demonstrator at South Pole in De- cember 2015. The picture shows John Kelley, John Felde, Leif Rädel (RWTH Aachen) and Aongus O'Murchadha (from left to right) who deployed the telescope on the roof of the ICL. Taken from [63].

29 4 The IceAct Telescope

4.4.1 Camera

A 7-pixel camera is much cheaper than a full 64-pixel camera and sucient for a technology demonstration. Each pixel of this camera is a SensL C-Series SiPM with a size of 6 mm × 6 mm consisting of 18 980 35 µm × 35 µm GAPD micro-cells The pixels are arranged in a hexagonal pattern to ensure an equal distance between all pixels as can be seen in gure 4.11. Figure 4.13 shows the camera being installed in the telescope tube with a human hand for size comparison. Both gures (4.11 and 4.13) also show the Winston cones that are used as light collectors to concentrate the Cherenkov light to the SiPM. These Winston cones are "round" or circular cones, which means that they have a circular entrance and exit window. Nev- ertheless, the working principle is the same as for the "hex-to-square" cones described in section 4.2 (see also gure 4.5). Further they were built of aluminum and not of PMMA and thus, are hollow. PMMA cones works due to total internal reection on the boundary PMMA-air. Aluminum cones, in contrast, use simple reection on its surface to collect light (further details in section 4.2).

Figure 4.11: Picture of the 7-pixel camera showing the SiPMs and Winston cones. The picture shows all seven aluminum Winston cones on the left that concentrate the light to the SiPMs of the camera shown on the right. The circuit board housing the SiPMs (green) gets screwed to the plate holding the Winston cones once the camera is fully assembled. Cables (visible in the background) are used to connect the camera to its readout electronics.Taken from [46].

30 4.4 The IceAct demonstrator at South Pole

A sketch of the used aluminum Winston cones is shown in gure 4.12. Due to man- ufacturing their outer diameter is slightly

bigger (dout = 17.4 mm) as their entrance

diameter of den = 13.42 mm. The diam-

eter of the circular exit window is dex = 6 mm to maximize the usable area of the 6 mm × 6 mm sized SiPM:

π · (3 mm)2 π = ≈ 78.5 % . 6 mm · 6 mm 4 Between the light collecting Winston cones and the SiPM a Schott UG11 UV- Pass lter is installed [65]. It is mainly Figure 4.12: Sketch of the single pixel of transparent at 300 nm to 350 nm reducing the 7-pixel camera with Winston cone. the night sky background while Cherenkov Between exit aperture of the Winston photons peaking at ≈ 320 nm (gure 2.4b cone a SCHOTT UG11 UV-Pass lter was in section 2.3) can pass easily. installed. Taken from [64]. The eld of view for a single pixel of the 7-pixel camera can be calculated in the same ways as for the 61-pixel camera, yielding:

r 6.71 mm FOV/pixel = 2 arctan en = 2 arctan f 502.1 mm ≈ 1.5◦ .

For the whole camera with three pixels in a row, that results in a FOV of ≈ 2.25◦ since there is some space between the cones due to their wall thickness.

31 4 The IceAct Telescope

Figure 4.13: The demonstrator's 7-pixel camera gets installed inside the tube (left). The SiPMs are situated below the Winston cones that are good to see. Cables in the background connect the SiPMs to the data acquisition system in a box below the telescope (not shown).Taken from [46].

4.4.2 Data Acquisition System

Two DRS4 Evaluation boards designed by the Paul-Scherrer-Institute are used as the data acquisition system of the 7-pixel telescope demonstrator. The boards have a ring capacitor array with 1024 cells each, used to store a waveform. They oer possible sampling rates in the range of 0.7 GSa/s to 5 GSa/s. In this case, a sampling rate of 1 GHz was used which, in combination with the 1024 cells of the ring capacitor array, yields to a 1 µs waveform with a time resolution of 1 ns. The picture in gure 4.14 shows both DRS4 boards before they were installed in the telescope demonstrator. Dynamic range of the DRS4 boards was set to ±500 mV and SiPMs were connected to show up as negative signals.[66, 67] Each DRS4 board oers 4 analog input channels giving the opportunity to double readout one of the seven SiPMs of the camera. This can be used for cross calibration purposes, since the signal of that specic SiPM must look the same in both DRS4 boards. Compared to the signals of the other 6 SiPMs, its amplitude is reduced by a factor 2 . 3 The mapping of input channels to SiPMs is shown in gure 5.1 (channels A1, B1, C1 and

32 4.4 The IceAct demonstrator at South Pole

Figure 4.14: The picture shows the DRS4 Evaluation boards used in the IceAct telescope demonstrator. Both boards are running a custom, self-written rmware. Also shown are the cables used to connect the DRS4 boards to the SiPMs. Taken from [46].

D1 belong to the rst DRS board and channels A2, B2, C2 and D2 to the second). It was planned to connect the central pixel to both DRS boards but by mistake, an outer pixel (top left in gure 5.1) was chosen. Whenever one of the boards trigger, either by a xed rate trigger or a discriminator crossing of one ore more signals, it sends a trigger signal from its external trigger output to the others trigger input. This ensures that one always obtain a full camera picture of all 7 SiPMs. Both DRS4 boards are connected to a computer via USB that controls data acquisition and sets xed rate trigger frequency, discriminator threshold, etc. A feature introduced by the custom, self developed rmware that was used is a pixel multiplicity trigger. It allows requiring at least N-channels crossing the discriminator before readout is initiated. To ensure that even a two pixel coincidence is triggered correctly when the corresponding pixels are connected to dierent boards, a further digital in and output of the DRS4 board is used. It was originally designed to allow for a clock synchronization between two or more DRS boards but reprogrammed in the rmware to allow two pixel coincidence trigger across both boards. Whenever there is a pixel above the discriminator threshold on one board, a signal is sent to the other board via this connection (called Sync in gure 5.1). Both boards now treat this signal as a

33 4 The IceAct Telescope further discriminator crossing. This way a two or three pixel coincidence trigger across both boards is implemented. Of course this works for two or three pixel coincidences only, as the clock in/output allows for digital signals only and cannot be used to tell how many pixels above threshold are observed needed to trigger e.g. on a 4 pixel coincidence with 2 pixels above threshold on each board.[46, 68]

4.4.3 Telescope Operation

The telescope had to be operated remotely since direct access at South Pole is not possible during astronomical night. A slow control featuring a web interface is running on a small computer and accessible from all over the world. It was also developed for the FAMOUS telescope and controls data acquisition of the IceAct telescope demonstrator [69]. For voltage regulation a power supply developed especially for silicon photomultipliers was used. It is capable of regulating the bias voltage of the SiPMs in steps of 1 mV.[70]

The telescope was operated in a two pixel coincidence mode combined with a low frequency xed rate trigger during most runs. An overview over all runs taken during the 2016 pole season is given in gure 4.15. The lid of the telescope was removed just before the beginning of astronomical twilight on May 12th that lasts to August 1st. In total there were 83 runs taken before it was too bright to take physics runs anymore. In order to investigate the inuence of the sun slowly rising, the telescope was kept open until October 4th.

After gaining experience with remote operation at the South Pole and some bug-xing, telescope runs got longer as operation got more stable. Longer periods without data runs in June and July are due to extreme baseline variations and trigger issues prohibiting a physics data run. The run this analysis focus on, was taken at the end of July from 20th to 27th and the most stable telescope run of that season.[46]

34 4.4 The IceAct demonstrator at South Pole

IceAct runs 2016

May

June

July

August

September IceAct season IceAct runs October analyzed run

03 06 09 12 15 18 21 24 27 30 day of the month

Figure 4.15: Overview of data runs taken with the IceAct telescope demonstrator in 2016. Gray shows the IceAct season were the telescope could be operated. This analysis focus on the run shown in orange in July 2016. Other telescope data runs are shown in blue.

35 4 The IceAct Telescope

36 5 Coincident Data Taking with IceAct and IceCube

The IceAct telescope demonstrator was installed to demonstrate a successful operation of an air Cherenkov telescope in the harsh environment at South Pole. Further it should demonstrate that such a telescope is capable to improve the energy calibration of IceTop and the angular resolution of IceCube but also to measure the composition of cosmic rays. In order to take data in coincidence with all three detectors (IceCube, IceTop and IceAct) the telescope needs to be integrated into the data acquisition system of IceCube and the clocks of all detectors have to be synchronized. Since IceCube and IceTop were designed together, they can be operated in a coincidence (or veto) mode by default. The following sections give details on how IceAct was connected to and integrated into IceCube and how the time synchronization works. A rst time synchronization was done in [71]. The method described here is adapted from this thesis and improved.

5.1 Connection between IceAct and IceCube

A schematic of the connection between the telescope and IceCube is shown in gure 5.1. It is a simple one way connection where only the telescope is able to send information to IceCube and not vice versa. Every time the telescope triggers, both xed rate and physics (see section 4.4.2), it sent a pulse to a main board of a standard IceCube DOM. Each of IceCube's 5160 DOMs have such a main board that houses all the readout electronics for the PMT and digitization (see section 3.1). Instead of digitizing the analog signal of a connected PMT this main board (that is connected to IceAct) digitizes the TTL pulse sent by the telescope. This main board is itself integrated into the IceCube data acquisition system, just as the other 5160 DOMs.

37 5 Coincident Data Taking with IceAct and IceCube

D1/2 A1 Data Data DRS PC Board USB LAN IceCube Sync B2 C2 B1 Trigger DOM Network Flag Mainboard DRS Board ITFilterbank 22 Datastream A2 C1 DAQ ITHub08 Telescope ICL Figure 5.1: Schematic of the trigger connection between IceAct and IceCube. The external trigger output of the telescope is connected to a DOM main board that is integrated into the IceCube data acquisition system (IceCube Network). Apart from a telescope trigger ag no data can be sent to IceCube's data stream. The data stream of the telescope, however, is recorded independently.

A priori IceCube and IceAct take their data independently: neither IceCube nor IceAct is capable to trigger the other.1 If an event is recorded by IceAct a pulse is sent to the DOM main board that gets digitized. The advantage of this connection is that the DOM main board of IceAct can be (and is) fully integrated into IceCube's data acquisition system and readout is identical to all other DOMs. Each trigger of IceCube has a certain time window in which hits of DOMs are assumed to belong to the initial trigger condition [25]. If IceAct triggers, either a xed rate or physics (2-pixel coincidence), during such a trigger time window the hit of the DOM main board gets stored in the data stream. Such a coincidence can be caused by chance by a xed rate trigger of the telescope or due to the fact, that IceTop has observed the same air shower too (or IceCube detects muons from that air shower at the same time). As only a simple TTL pulse is sent from the telescope to IceCube, only timing information is passed from the telescope to IceCube and not any further like charge or pixels above threshold. By using the DOM main board a precise time stamp is assigned to the IceAct ag. This one way communication ensures that all events that are observed in coincidence by IceCube and/or IceTop and IceAct are agged in the data stream of IceCube. The data stream of IceAct, in contrast, contains all events recorded by the telescope: coincident ones but also not coincident ones. Therefore, a synchronization of both data streams is needed.

1At least for the demonstrator telescope in 2016. For a successor telescope it is planned to trigger an IceCube detector readout by telescope triggers.

38 5.2 Method of Time Synchronization

5.2 Method of Time Synchronization

Event times of the telescope are given in UTC from its internal computer and thus, only achieve a resolution in the order of O(ms). IceCube however uses a GPS synchronized clock that achieves a resolution in the order of O(ns). This fact makes it impossible to tell which events are coincident from simply comparing both time stamps. One must therefore search for the unique pattern of IceCube events (with the IceAct DOM hit) in the data stream of IceAct. Coincidences caused by physic triggers (observing an air shower) occur randomly distributed over time which makes this pattern unique for each observation run.

Visualization of synchronization algorithm

IceAct physic trigger IceAct FRT coincident IceCube event time difference

total time (shifted) data streams difference

+0 time shifts [a.u.]

-1

-2

0 20 40 60 80 90 95 100 time [a.u.]

Figure 5.2: Visualization of the time synchronization algorithm. Shown are events recorded by the telescope (blue) and events marked as coincident by IceCube (orange). The pattern is shown for dierent time shifts between both data streams (y-axis) where event times are given in arbitrary units (x-axis). The time dierence between IceCube events and their nearest counterpart event of IceAct is indicated in green. The total time dierence (sum of absolute values) is shown on the right side of the plot.

39 5 Coincident Data Taking with IceAct and IceCube

Exemplary data streams of IceCube and IceAct are shown in gure 5.2. Shown are all events recorded by the telescope, FRT (Fixed Rate Trigger) (light blue) as well as physics trigger (blue), and IceCube events having an IceAct DOM hit (orange). The gure also shows the working principle of the synchronization algorithm as described below. Looking at the row showing the event times for a shift on 0 a.u. (the correct one in this example) one can easily spot, that all events of IceCube have a counterpart in IceActs event stream but not vice versa. Fixed rate triggers are coincident only by chance, namely if IceCube is triggered by an independent trigger at the same time (IceCube's over all trigger rate is ≈ 3 kHz [25]). Telescope triggers below the threshold of IceCube or IceTop can not have any counterpart since they will not trigger a detector readout. In order to analyze coincident events and to conclude on the performance of the telescope demonstrator, a time synchronization (or matching) between both data stream is needed. Afterwards it is possible to remove xed rate triggers of IceAct, that are not marked in IceCubes data stream before synchronization, and keeping only real physic coincidences. The individual steps of the algorithm used for synchronization are described below (compare with gure 5.2):

1. Select the rst IceCube event with an IceAct DOM hit and use its accurate time to search for its nearest counterpart in IceAct's data stream.

2. Calculate the time dierence between these two events and mark them as a possible event pair.

3. Repeat the previous steps for all events that are marked coincident by the telescope and sum up their absolute (or squared) time dierences.

4. Calculate the mean time dierence between both data streams by dividing the summed time dierence by the total number of coincident events and store this together with the matched event pairs.

5. Shift the whole data stream of IceCube by a xed time interval, e.g. 0.5 ms and repeat the previous steps.

6. Repeat the time shift multiple times to scan a certain range (e.g. −5 s to 5 s).

7. Create pairs of coincident events by using the time shift with the minimal mean time dierence between the data streams.

40 5.3 Time Synchronization Results

Instead of scanning possible time shifts in steps of e.g. 1 ms it can be modied to use a minimizer algorithm to search for the shift yielding to a minimal mean time dierence between both data streams. The output of step 4 is used as an input for a minimizer that determines the time shift between both data streams. This leads to a more precise result since the shift is not xed to multiples of a xed time interval (in this case multiples of 1 ms). On the other side, this "t" like approach does not show the impact of xed rate triggers compared to physics trigger only as described in the next section (5.3). The main dierence to the method used in [71] is the use of a free time shift between both arrays. Not forcing a perfect alignment for the rst pair (as done in [71]) leads to a better result in the mean time dierence of all event pairs since it is more natural to allow a time oset for all events instead of xing one pair to be exactly matching.

5.3 Time Synchronization Results

Results of the time synchronization for the selected IceAct telescope run (section 4.4.3 are shown in gure 5.3. For this particular synchronization it was assumed that the timestamps of IceCube and IceAct do not dier by more than 80 s. The mean time deviation in gure 5.3 is the mean time dierence of event pairs as determined by the algorithm described in section 5.2. The synchronization scan yields to a mean time dierence of 0.53 ms for a time shift of 2 ms. This indicates that the UTC times of IceAct events are much more accurate than expected. This result can be improved further by using a minimizer algorithm instead of a simple scan as described above (section 5.2). In this case the standard minimizer of scipy, an open-source software library for python, was used. The minimizers used the mean of the squared time dierence between IceAct and IceCube as an input to determine the time shift between both data streams. For minimization the algorithm was congured to use the so called Nelder-Mead-method/approach. The t lead to more accurate values for the time shift: 2.5 ms with a corresponding mean time dierence of 0.32 ms. Besides the global minimum (indicated with the red dot), further minima are visible in gure 5.3. These minima are caused by xed rate triggers of the telescope that have, by chance, a counterpart in IceCube's data stream. If the time shift between IceCube and IceAct is equal to a multiple of the xed rate trigger frequency of IceAct, all xed rate triggers will be aligned perfectly. The mean time dierence of about 1.5 s for these shifts is much higher than for the correct time shift between both data streams, since physics trigger will be aligned only for this correct shift and not for multiples of the

41 5 Coincident Data Taking with IceAct and IceCube

xed rate trigger frequency. This frequency can be determined directly by the time intervals between those minima. In this case it yields to a xed rate trigger frequency of TFRT = 17.49 s ⇔ fFRT = 0.057 Hz which is in good agreement with the set FRT frequency.

3.5

3.0

2.5

2.0

1.5

mean time deviation [s] 1.0

0.5 Minimum: shift: 0.0025s 0.0 err: 0.00032s

80 60 40 20 0 20 40 60 80 timeshift [s]

Figure 5.3: Time synchronization results using all coincident events. For each time shift (as described in section 5.2) on the x-axis the corresponding mean time dierence between event pairs of IceAct and IceCube is shown on the y-axis. Clearly visible are the global minimum (also indicated by the red dot) and local minima that occur repeatedly all ≈ 17 s caused by xed rate triggers of the telescope. This frequency can be determined directly by this scan.

Once the synchronization is done, one can distinguish between xed rate triggers and physics trigger of IceAct in the IceCube data stream as. The time shift scan can then be repeated taking only physics trigger of the telescope into account. This is shown in gure 5.4. In comparison with gure 5.3, the local minima in intervals of ≈ 17 s disappeared while the global minima remains the same, as expected. The physics trigger will align only for

42 5.3 Time Synchronization Results one time shift while for other ones the mean time dierence is dominated by randomly selected event pairs. Figure 5.4 shows a signal-to-noise ratio of about 3 s to 0.33 ms. If there were any xed rate triggers remaining, they would be clearly visible in this plot which makes it an excellent cross-check.

3.0

2.5

2.0

1.5

1.0 mean time deviation [s]

0.5 Minimum: shift: 0.0025s 0.0 err: 0.00033s

80 60 40 20 0 20 40 60 80 timeshift [s]

Figure 5.4: Time synchronization results using only coincident physic events. Each time-shift between the data streams of IceCube and IceAct (x-axis) yields to a certain mean time dierence between event pairs. In contrast to gure 5.3 only physics trigger of the telescope were used. Hence, no further minima besides the global one (indicated by the red dot) can be seen.

Figure 5.5 shows the event rate of all coincident events. Shown are all 1284 events, split into 669 xed rate triggers (orange) and 615 physics trigger (blue) of IceAct. The approximately constant rate of coincident xed rate triggers proves a stable operation of IceCube and IceAct during the run (07/20 - 07/27/2016) otherwise there would be gaps in the rate of coincident FRT trigger. It also proves, that the connection between IceAct and IceCube was stable and working ne. On the other side, the rate of coincident physics trigger is not constant at all. How-

43 5 Coincident Data Taking with IceAct and IceCube ever, a constant rate is not expected either. The trigger rate of the telescope depends, for example, strongly on weather conditions such as a clear sky with no clouds in the telescope's eld of view. Figure 5.5 shows these variations in the rate of coincident physic events with only a few events from July 25th on.

80 all coincidences: 1284 70 FRT coincidences: 669

30 number of events [#] 20

07/20/16 07/21/16 07/22/16 07/23/16 07/24/16 07/25/16 07/26/16 07/27/16

Figure 5.5: Event rate of IceAct - IceCube coincidences. The number of events is shown as a stacked histogram composed of coincidences of xed rate trigger (FRT, orange, bottom) and physic trigger (blue, top) with a bin width of 4 h. The rate of FRT is rather constant whereas the physic trigger rate strongly uctuates during the run.

44 6 Analysis of Coincident Events

After coincident events of IceAct, IceCube and IceTop are identied as described in section 5.2 and 5.3 a more detailed look on these events is possible. For the search of coincident events IceAct raw data and IceCube Level2 data was used. In a rst step quality cuts based on simple quantities measured by IceAct, IceCube and IceTop are dened and applied yielding to dierent data sets with higher analysis level. The rst part of this section will give an overview on these data sets, the quality cuts applied and the methods used to do so. Analysis performed on these data sets is presented in the second part of this chapter, including a conclusion on the performance of the telescope. Correlation of IceAct data and IceCube/IceTop reconstructions is shown afterwards.

6.1 Analysis methods

In this section an overview of the used analysis methods including the software trigger for IceAct and SiPM pulse signal extraction will be given.

6.1.1 Software trigger

The algorithm used for software re-triggering is based on the software trigger used in the very rst look at data from the demonstrator telescope in [72] (2016). The algorithm searches for a signal crossing the threshold in a xed time interval of the recorded voltage trace. The threshold for the software trigger was set to −40 mV, same as the DAQ, and the time interval in which the pulse is expected to 550 ns to 700 ns (compare with gure 6.1a). This time interval can be xed, since the DRS boards have a trigger delay time setting used to shift the trigger time to the same position for all self triggered traces.

45 6 Analysis of Coincident Events

waveform channel 1 10 10 smoothed waveform trigger threshold trigger index 20 20

30 30 voltage [mV] 40 voltage [mV] 40

50 waveform channel 1 50 smoothed waveform trigger threshold 60 trigger search window 60 0 200 400 600 800 1000 600 610 620 630 640 650 time [ns] time [ns] (a) Raw (blue) and smoothed (orange) traced (b) Zoom into the SiPM pulse and its of an example event. The orange area gives smoothed trace. Shown are raw (blue) the expected signal start interval. The hor- and smoothed (orange) trace and the trig- izontal black dashed line corresponds to the ger time (vertical line) for a given trigger trigger threshold. threshold (horizontal line).

Figure 6.1: Example signal trace for software trigger algorithm.

In a rst step the software trigger smooths the waveform using a running mean with a window of 5 samples. The smooth window is intentionally chosen rather small as smooth- ing with a small window will compensate for high frequency noise while not distorting the pulse shape, especially the rising edge, much. An example SiPM pulse, its raw (blue) and smoothed (orange) trace, is shown in gure 6.1a. A zoom into the region of interest is shown directly besides in gure 6.1b. When the trace is smoothed, the software trigger iterates through samples in the given range stopping at the rst one that crosses the threshold. In this case it stops at the rst sample below −40 mV (horizontal black dashed line) due to negative SiPM pulses of the telescope demonstrator. In addition to threshold crossing the software trigger requires at least one sample before and after crossing to be below/above the threshold. This ensures a stable software trigger even for signals with just a few samples above the threshold. An example is shown in gure 6.1b. The vertical black line is the found trigger position for a threshold of −40 mV indicated by the horizontal black dashed line. The algorithm can be congured to search for positive pulses as well. Thus, it is independent of the DAQ and can be used in dierent applications.

46 6.1 Analysis methods

6.1.2 Signal Extraction

The method (or algorithm) used to extract a signal out of the recorded voltage trace of one SiPM is also adapted from the method used in [72]. The software trigger is used as described in section 6.1.1 to identify traces with an SiPM pulse. The further signal extraction process for these traces is described in the following. Figure 6.2 shows the same trace that was used in section 6.2.1 to illustrate the working principle of the software trigger (gures 6.1a and 6.1b). The trigger threshold and the found trigger position in the trace is indicated with a dashed horizontal, respectively a solid vertical black line. The signal (or pulse) start time is set to 5 ns before trigger time for all pulses (vertical red line).

waveform channel 1 10 smoothed waveform trigger threshold trigger index signal/pulse start 20 pulse minimum baseline baseline window 30 integration window

voltage [mV] 40

60 400 450 500 550 600 650 700 750 800 time [ns]

Figure 6.2: Example trace for signal extraction. The recorded trace (blue) gets smoothed with a running mean of 31 samples. The baseline is dened as the median of samples in the blue area while the green area gives the samples used to calculate the pulse integral.

For the actual signal extraction, the rst step is to correct the baseline for each pulse since it uctuates due to photons from night sky background and temperature variations of the DAQ system. In a window with a size of 128 ns samples before the pulse the median of all samples is calculated and used to correct the baseline shift. To ensure that this window does not contain the SiPM pulse itself, it starts at 16 ns before the pulse (so 16 + 5 = 21 ns before the trigger index). The blue area in gure 6.2 shows the position of

47 6 Analysis of Coincident Events this baseline window. In addition to the baseline median, the RMS (root mean square) is calculated for the same samples, giving a reference for baseline variations. After baseline correction the pulse integral is calculated by summing up 64 samples starting at the signal start index (green area). The integral of the pulse is a measurement of the released charge of the SiPM due to photons hitting the SiPM. It is hence a measurement of the number of photons (or amount of light) seen by the SiPM. The pulse minimum is obtained from a smoothed trace using again a running mean. This time the window is chosen to 31 samples, much larger than for software triggering (5 samples). The obtained smoothed trace is shown as an orange overlay in gure 6.2. The minimum of the waveform is searched in a window of 64 ns starting at the trigger index. It is identical to the window used for calculating the pulse integral (green). Same applies for the baseline window (blue) that is used for correcting the smoothed trace by the baseline median. The pulse amplitude, in this case the minimum with respect to the baseline, is proportional to the number of detected photons (compare to section 4.3) like the pulse integral.

6.2 Event Selection

6.2.1 IceAct Event Selection

The analyzed 7-day run from July 20th to 27th was taken with a 2-pixel coincidence trigger in combination with a ≈ 0.06 Hz xed rate trigger (FRT). Obviously FRT events cannot be used to conclude on the performance of the telescope and thus, are removed in a very rst step. As already shown in section 5.3, 615 physics triggered events out of 1284 total coincident events remain. Besides removing FRT coincidences, additional quality cuts are applied. First, they ensure a normal operation of the telescope during that time and second they take care of the special channel D of the demonstrator's camera. Channel D, the top left pixel of the camera, is connected to both DRS4 boards. That makes any signal of this SiPM showing up twice in the DAQ resulting in a false two pixel coincidence for this channel. Also the signal of this SiPM is reduced by a factor of 2 compared to the other SiPMs 3 that are connected to one channel only. To account for this, the software trigger threshold is reduced accordingly for channel D on both DRS boards and the obtained pulse amplitude/integral is multiplied by 3 . 2 All applied quality cuts are described in the following:

48 6.2 Event Selection

Trigger Level/Threshold

First events with an incorrect trigger setting are removed. The trigger threshold for the discriminator was set to −40 mV during most of the run. However, for the rst few minutes it was set to −25 mV. Since trigger rate and energy threshold depends directly on the threshold it has to kept constant. If the telescope and all pixel would be perfectly calibrated, it would be possible to correct for dierent trigger settings during a run. But since there is no gain calibration for neither one of the 7 SiPMs, this is not possible for the IceAct demonstrator telescope.

event frequency [#] 200 trigger threshold cut all events: N = 615 kept events: N = 567 25

35 set trigger threshold [mV] 40

45 0 200 400 event frequency [#] 07/20/16 07/21/16 07/22/16 07/23/16 07/24/16 07/25/16 07/26/16 07/27/16

Figure 6.3: Trigger threshold setting for the analyzed IceAct run. The y-axis gives the trigger threshold setting in mV as a function of the event time. Each blue point is one of 615 events in total while the smaller orange ones indicate events passing the trigger threshold cut. Projected to the top is the event rate of all (blue) and passing events (orange, dashed) while the projection to the right shows a histogram of the trigger thresholds. The black dashed line corresponds to the cut applied (−30 mV).

Figure 6.3 shows the trigger level of all 615 coincident events as a function of run-time It also shows the resulting event rate as a projection to the top and a histogram of used trigger threshold on the right. Events with a trigger threshold setting above −30 mV are discarded. In total 567 events pass this rst quality cut while 48 events have to be discarded.

49 6 Analysis of Coincident Events

IceAct Trigger Rate

Another cut removes all events where the overall readout rate of the IceAct telescope, respectively of the DRS4 evaluations boards, is too high. Paul Scherrer Institute states a possible readout rate of ≈ 500 Hz a single DRS4 Evaluation Board. [66]. However due to the use of a custom rmware, a DAQ consisting of two boards and the use of a two channel coincidence trigger across them the maximum possible readout rate is smaller. It was estimated that a readout rate of less than 1 can be handled safely by 3 ms ≈ 333 Hz both boards in 2-pixel coincidence mode [46, 68]. For a higher readout rate it cannot be guaranteed that both boards are operating in the same mode: they have to be activated again in sequence after a readout. A higher event rate may result in one board triggering before the other one was fully congured for a new readout. The cut forces all IceAct events (not only coincidences) to be separated by more than 0.003 s. There is only one coincident event not passing.

Trigger/Pixel Multiplicity

Third simple quality cut is the actual pixel multiplicity per event. The data acquisition was set to require a 2-pixel coincidence across both boards. However, one pixel (channel D, top left as seen in gure 5.1) is connected to both DRS4 boards. This will make any signal exceeding the trigger threshold of −40 mV2 to show up as a 2-pixel coincidence for the data acquisition system. In order to check for a real 2-pixel coincidence of two dierent SiPMs, a software trigger is used. The software trigger algorithm is described in detail in section 6.1.1. The obtained software trigger (or pixel) multiplicity for all coincident events is shown in gure 6.4. The green one shows the trigger multiplicity obtained by the software trigger without correcting for the double readout of one SiPM (channel D). A signal in this SiPM will always show up as two triggered pixels for the data acquisition and trigger an IceAct event. By requiring a trigger in channel D in both boards simultaneously and counting this as a single pixel above threshold, the blue curve is obtained. The maximum pixel multiplicity is consequently reduced from 8 to 7. Very few events with only one software trigger in channel D (so not for both boards) are discarded as well. These triggers originate from independent baseline uctuations on both boards resulting in pulses exceeding −40 mV only in one board. If only the trigger multiplicity cut (at least two triggered pixels) is applied to all 615 events, 521 events pass (orange dashed

2exceeding is here meant as "falling below" since the IceAct demonstrator telescope uses negative SiPM pulses

50 6.2 Event Selection line).

all events: N = 615 160 kept events: N = 521 not channel D corrected 140

120

100

80 frequency [#] 60

0 0 1 2 3 4 5 6 7 8 9 pixel with trigger per event

Figure 6.4: Pixel multiplicity of all coincident physics trigger. The blue curve shows the multiplicity before the cut, orange dashed shows that all events with less than 2 pixels above threshold are removed. The green dashed curve shows the trigger multiplicity not corrected for the double readout of one SiPM.

Time Synchronization

A further quality cut is applied to ensure a good time synchronization between IceAct and IceCube. For each coincident event pair of IceAct and IceCube the time dierence between both is calculated using the synchronization result of section 5.3 (∆¯t = 0.3 ms). The time dierence for single event pairs after synchronization is shown in gure 6.5. In order to pass this quality cut, event pairs of IceAct and IceCube are required to have a time dierence (after synchronization) of less than 10 · ∆¯t = 3 ms. In total 601 of 615 events pass this cut.

51 6 Analysis of Coincident Events

event frequency [#] 8 time difference cut 7 all events: N = 615 kept events: N = 601

3 time difference [ms] 2

0 0 100 event frequency [#] 07/20/16 07/21/16 07/22/16 07/23/16 07/24/16 07/25/16 07/26/16 07/27/16

Figure 6.5: Time dierence for IceAct-IceCube event pairs after time synchronization. Time dierence is shown in seconds on y-axis for all events with their time on the x- axis. Projection above shows the event rate with (dashed orange) and without (blue) applied time dierence cut. The histogram of time dierences (projection right) shows the same.

Baseline RMS

In addition to the quality cuts described above, this cut is applied after signal extraction of the recorded traces. Strong baseline uctuations can be caused by night sky background or other noise caught up by the cables or SiPMs directly. A histogram of the baseline RMS determined by the signal extraction algorithm (section 6.1.2) is shown in gure 6.6 (left) along with an example event (right) that does pass the applied baseline RMS cut but all other cuts. The cut is chosen conservatively again: RMS ≤ 8 mV. 606 events pass this cut.

52 6.2 Event Selection

40 120 channel 0, N = 302 channel 1, N = 253 channel 2, N = 360 20 100 channel 3, N = 304 channel 4, N = 273 channel 5, N = 280 0 80 channel 6, N = 278 channel 7, N = 311 20 60 voltage [mV] frequency [#] 40 40

60 Ch 0, RMS = 0.0mV Ch 4, RMS = 17.4mV 20 Ch 1, RMS = 0.0mV Ch 5, RMS = 12.4mV Ch 2, RMS = 9.4mV Ch 6, RMS = 11.8mV 80 Ch 3, RMS = 17.6mV Ch 7, RMS = 16.5mV 0 0 2 4 6 8 10 0 200 400 600 800 1000 baseline RMS [mV] time [ns] (a) Histogram of baseline RMS of all coinci- (b) Example traces of an event not passing the dent events for each channel. baseline RMS cut.

Figure 6.6: Histogram of the baseline RMS for all channels and example of noise. The event (right) passes all quality cuts except for the baseline RMS cut. A quoted RMS of 0.0 mV (only right) means that there was no software trigger found for this channel.

Summary

A summary of all applied IceAct quality cuts is given in table 6.1. It also shows the number of events passing each cut when applied for its own as well as for all cuts together. All cuts are chosen to be as conservative as possible and there are 501 events passing all cuts.

cut setting events passing Trigger threshold < −30 mV 567 Trigger rate < 333 Hz 614 Pixel multiplicity ≥ 2 521 Time synchronization < 3 ms 601 Baseline RMS < 8 mV 606 combined 501 /615

Table 6.1: Summary of applied IceAct quality cuts.

53 6 Analysis of Coincident Events

6.2.2 IceTop Event Selection

In order to ensure a good air shower reconstruction by IceTop, some quality cuts have to be applied. To understand what the cuts do and how IceTop data is used later on, the air shower reconstruction algorithm is briey described before the applied cuts are explained. Secondary particles of air showers will be detected by IceTops 81 stations, consisting of 2 ice Cherenkov tanks each (see section 3.2). The charge or signal expectation S for a tank in the distance R to the shower axis can be described as [33]:

R −β−κ log10(R/Rref ) S(R) = Sref · . Rref This function describes the lateral distribution of an air shower and is thus called Lateral Distribution Function (LDF). β and κ denote to the slope and curvature in a logarithmic representation of S(R). Simulations have shown that κ can be set to a constant κ = 0.303 for all air showers. Sref gives the air shower signal in a tank at a distance of Rref to shower axis. The reference is for IceTop chosen to 125 m (so it is S125 at Rref = 125 m).[33] Besides the LDF, the time information of each tank can be used to describe the shower front. The expected signal time of a signal in a tank at position ~a is given by

1 t(~x) = t + (~x − ~x ) · ~n + ∆t(R) , 0 c core R2 ∆t(R) = aR2 + b 1 − exp − . 2σ2

T0 is the the shower impact time on ground at position core~x . ∆t(R) gives the deviation from a plane perpendicular to shower that moves in the direction given by the unit vector ~n. The constants are xed to a = 4.823 × 10−4 ns/m2, b = 19.41 ns and σ = 83.5 m.[33]

All free parameters of the air shower reconstruction, shower core position core~x =

(xcore, ycore), shower direction θ, φ, shower size S125, slope parameter β, and time at ground t0 are tted using a maximum likelihood method. This method accounts for hits not passing trigger threshold or resulting in a saturated signal [33, 73]. The t also account for snow accumulation on IceTop tanks by reducing their signal S according to:

h Scorrected = S · exp − . λs cos θ

The snow height on each IceTop tank is given by h, θ is the air shower zenith and

λs = 2.1 m an eective absorption length in snow [73]. It has to be measured every year

54 6.2 Event Selection and individually for each IceTop tank.

In principle, it is possible to convert S125 to air shower energy using Monte-Carlo simulations of air showers. Due to dierent snow heights above each IceTop tank and accumulation of new snow every year, a lot of new Monte-Carlo is needed to nd the proper conversion for each year. For the data of 2016 there is no conversion from S125 to air shower or cosmic ray primary energy yet available [74]. For this analysis, following further quality cuts based on this t and other IceTop observables are applied to ensure a good air shower reconstruction (as suggested by [74]):

1. At least 5 triggered stations (needed to run the reconstruction),

2. Successful air shower reconstruction (convergence especially for β),

3. Reconstructed shower core must lie within the IceTop footprint on surface and the tank with maximum signal is not an outermost tank,

4. There is at least one signal above 6 VEM (vertical muon equivalent3),

5. The neighbor tank of the tank with the maximum signal has a signal > 4 VEM.

Table 6.2 summarizes the applied cuts and gives the number of events passing each cut. In total there are 338 passing these cuts when only these IceTop quality cuts are applied. If also IceAct cuts, as described in section 6.2.1, are applied, 320 events are remaining.

cut events passing > 5 stations 424 t convergence 393 containment 423 Smax > 6 VEM 408 neighbor Smax > 4 VEM 355 combined 338 /615 with IceAct cuts 320 /615

Table 6.2: Summary of applied IceTop quality cuts.

3The signal generated by a single vertical down-going muon inside one IceTop tank

55 6 Analysis of Coincident Events

6.2.3 IceCube Event Selection

In addition to the air showers being directly detected by IceTop, muons originating from air showers can reach the IceCube in-ice array when their energy is sucient to travel ≈ 1.5 km in ice. The track of a muon, its direction an energy loss is then reconstructed using a maximum Likelihood method. The number of photons detected by each DOM (section 3.1) as well as their arrival time is used to maximize a multi-photo-electron (MPE) Likelihood that compares measured data (arrival time of rst out of N photons) to a muon track hypothesis [75, 76]. It is assumed, that the muon traverses the detector (or a part of it) on a straight line with the speed of light emitting Cherenkov light [77]. This hypothesis is fully dened by ve parameters: vertex position ~x = (x, y, z) and direction of the muon, given by θ, φ. For the angular reconstructions this Likelihood became time dependent to describe the expected arrival time of photons at given position (DOM). For all 615 coincident events, there are 240 successful reconstructed muon tracks using the MPE Likelihood reconstruction. In the following, a short description of applied quality cuts is given:

Timing: Due to long readout windows of the IceCube detector, it is possible that there are multiple muon tracks in the IceCube in-ice array for an IceAct coincident event. Since muons originating from cosmic ray air showers seen by IceAct have to travel at least 1.5 km from surface to the in-ice array, a time cut can be used to select only muons of interest. The time of the muon in the in-ice array is thus required to be within a window of 5 µs to 8 µs after the IceAct trigger time. This time dierence is shown in gure 6.7a. In total 194 out of 240 muons pass this cut.

Fit quality: For a maximum Likelihood reconstruction method a quality parameter,

called Lreduced, can be inferred directly by using the maximum value L and the number of degrees of freedom ndof (here ve):

L Lreduced = −ndof .

This parameter corresponds to the reduced χ2 for a Gaussian probability up to a constant [77]. Here a reduced Likelihood < 10 is chosen, resulting in 223 passing events (compare to gure 6.7b).

Energy reconstruction: The muon energy reconstruction based on the charge deposited in DOMs along its track is independent of the angular reconstruction of the track.

56 6.2 Event Selection

Therefore, coincident events are required to have a successful track and energy reconstruction. There are 20 events with an angular track reconstruction but no energy estimation.

Direct hits: Based on the track hypothesis DOM hits by photons are called direct when they occur in a time window chosen to −15 ns to 125 ns relative to the expected photon arrival time. Obviously the t gets better the more direct hit photons are detected compared to scattered photons arriving later. For IceAct coincident events at least 36 direct DOM hits are required on two (or more) dierent strings (compare with gure 6.8). This ensures a good reconstruction quality since it is dicult to reconstruct muons precisely that are vertically down-going along one string only. This cut is rather strong and removes 104 reconstructed muons, leaving 136 for further analysis.

All applied cuts are summarized with event numbers in table 6.3 and visualized in gures 6.7a, 6.7b and 6.8.

cut events passing timing 194 reduced Likelihood 223 energy reconstruction 220 direct hits 136 combined 134 with IceAct cuts 130

Table 6.3: Summary of applied InIce quality cuts.

57 6 Analysis of Coincident Events

all N = 240 all N = 239 25 not in hist: 41 cut N = 223 cut N = 194 100 not in hist: 0 20 80

# # 60

10 40

5 20

0 0 0 2000 4000 6000 8000 10000 0 5 10 15 20 25 30 35 40 time [ns] rlogl

(a) Histogram of the time dierence between (b) Reduced logarithmic likelihood of recon- the in-ice muon and the IceAct trigger strucet muons in the in-ice detector (blue). (blue). Only muons in a range of 5 µs to 8 µs To ensure a good reconstruction, events after the IceAct trigger are used for further with a value of more than 10 are removed analysis (orange). (remaining events in orange).

Figure 6.7

N = 240 14 N = 240

60 12

50 10

40 8

30 counts [#] counts [#] 6

20 4

10 2

0 0 0 2 4 6 8 10 12 14 0 20 40 60 80 100 120 140 n_strings n_doms

Figure 6.8: Histogram of direct DOM (right) and string hits (left). The minimum required value to pass the quality cut is indicated with a vertical black dashed line. The minimum in the distribution of direct DOM hits at about 35 (the chosen cut) indicates a separation between signal (> 35) and background (≤ 35).

58 6.3 IceAct Telescope Performance

6.3 IceAct Telescope Performance

After event selection 501 coincident events remain for further analysis. The rate of these events is shown in gure 6.9 stacked with all coincident events (left) and divided into trigger multiplicity (right). The event shows variations than can be caused by e.g. clouds in the FOV absorbing the Cherenkov light of air showers with only very few coincidences for the last day of the run. Data of a Sky-cam tracking stars and a LIDAR (light detection and ranging, measuring the amount of light back scattered by clouds in the atmosphere) indicate more clouds towards the end of the run [46]. The peak of events in the beginning of the run vanishes when applying the IceAct quality cuts. It is mainly caused by a lower trigger threshold as shown in gure 6.3.

60 50 all events: 615 2 pixel trigger kept N = 501 3 pixel trigger 50 4 pixel trigger 40 5 pixel trigger 6 pixel trigger 40 7 pixel trigger 30

20 20 number of events [#] number of events [#]

10 10

0 0

07/20/16 07/21/16 07/22/16 07/23/16 07/24/16 07/25/16 07/26/16 07/27/16 07/20/16 07/21/16 07/22/16 07/23/16 07/24/16 07/25/16 07/26/16 07/27/16 (a) Stacked event rate for coincident events (b) Event rate for dierent trigger mutliplici- befor (blue) and after apllying the IceAct ties for events selected as described in 6.2.1. quality cuts of section 6.2.1.

Figure 6.9: Event rate of IceAct-IceCube/IceTop coincident events.

Figure 6.10 shows a 2d-histogram of all waveforms of all events passing the event selection for channel 0 (left) and channel 3 (right) of board 0. The traces are baseline corrected and aligned by shifting them with respect to the trigger time of the software trigger. Since channel 3 of both boards is connected to the same SiPM (double readout), its recorded amplitudes are smaller compared the other channels. In fact, they are reduced by a factor of 2 . The saturation level for pulses appears to be dierent for 3 some pulses, however, this is due the baseline correction done for each pulse (waveform) individually. Histograms of the waveforms for the other channels are shown in A.1 in appendix A.

59 6 Analysis of Coincident Events

(a) 2d-histogram of waveforms of channel 0 (b) 2d-histogram of waveforms of channel 3 board 1. board 1.

Figure 6.10: 2d-histogram of waveforms of channel 0 and 3 of board 1.

The number of triggers for each pixel is shown in the left of gure 6.11. The trigger √ rate is distributed uniformly across the camera when a Poissonian error of N for each pixel is taken into account. The mean charge of each pixel can be calculated by summing up all pulse integrals/amplitudes and dividing that sum by the number of triggers. It is shown in the right of gure 6.11.

7 600

6 1 0 1 5 0 500 3.51 4.43 5 =

310 298 t = o

t t

o 5 N t

, N ]

, 400 s ]

V #

[

s 4 r 0 e 1 g [

4.12 4.00 3.43 l 274 272 252 300 i r a t r

3 g e r e t a n i w

t f 200 e o s l s

2 u r p e

b n a 269 358 m 4.11 3.45 e 100 u n 1 m

0 0

(a) Trigger multiplicity of each camera pixel. (b) Mean charge of each camera pixel.

Figure 6.11: Trigger multiplicity (left) and mean charge (right) of each camera pixel. Both distributions are uniformly across the camera when allowing a 10 % variation √ in gain and taking the Poissonian error ( N) for the trigger multiplicity (left) into account.

The mean charge for the pixels in the lower left, middle and top right of the camera is

60 6.3 IceAct Telescope Performance slightly higher compared to the other pixels. Since there is no gain calibration available the values are allowed to deviate from pixel to pixel by a few percent. The mean charge is consistent with a uniform distribution when allowing a deviation of 10 % between the pixels due to dierent gains.

Two example event displays of IceAct are shown in gure 6.12. The pulse amplitude of each pixel is indicated by a horizontal black line and color coded. A horizontal line at 0 mV means that there is no software trigger found for this pixel. Figure 6.13 shows both events as seen by IceTop and the IceCube in-ice array. The size of the bubbles corresponds to the amount of light detected by the DOM at this position and the red line is the reconstructed muon direction. The number of pixel with a trigger of IceAct and the pulse amplitudes suggests that the left event (6961) originate from a lower energetic cosmic ray than the right one (40686). The number of DOM hits in IceTop as seen below (gure 6.12) is also much higher for the right one. IceTop reconstructs the right event to a and the left one to . The in-ice muon energy is log10(S125) = 1.81 log10(S125) = −0.09 measured with 1.78 PeV (right) and 2.6 TeV (left) respectively.

IceAct Event 6961 IceAct Event 40686

0 0

50 50

100 100

150 150

200 200

250 250 50 50 0 Voltage Minimum [mV] 0 Voltage Minimum [mV] 50 300 50 300 100 100 150 150 200 350 200 350

voltage [mV] 250 voltage [mV] 250 300 300 350 350 0 256 512 768 1024 400 0 256 512 768 1024 400 time [ns] time [ns] (a) IceAct event display. (b) IceAct event display.

Figure 6.12: IceAct event display of two coincident events.

61 6 Analysis of Coincident Events

(a) IceCube event display. (b) IceCube event display.

Figure 6.13: IceCube event display of coincident events shown in gure 6.12. The left event (6961) is reconstructed with to and while the log10(S125) = −0.09 Eµ = 2.6 TeV right one (40686) is reconstructed to and log10(S125) = 1.81 Eµ = 1.78 PeV

The integral and minima of all pulses (as determined by the method described in section 6.1.2) are summed up for each coincident event and shown as a histogram in 6.14a and 6.14b. Both quantities are proportional to the amount of light seen by the whole camera and thus to the total charge. The Cherenkov light yield of an air shower is proportional to the energy of the cosmic ray primary. As described in section 2.1 the dierential energy spectrum follows a power law with a spectral index of ≈ −2.7. In a double logarithmic plot a power law of the is a straight line that slope corresponds to the spectral index. A t to the falling spectrum is performed (in gures 6.14a and 6.14b) and yields −1.7 for the integral and −1.6 for the amplitude measurement of total charge. These values are not compatible with a spectral index of γ = −2.7 for cosmic rays. However, the event selection is biased since only coincidences between IceAct, IceTop and IceCube are taken into account. Bad weather, especially clouds, could also bias the spectrum since lower energetic showers do not penetrate deep into the atmosphere and thus emitting no light below the clouds that could be detected by the telescope.

62 6.3 IceAct Telescope Performance

N = 309 N = 309 power law fit: power law fit: = -1.7 = -1.6

101 101 frequency [#] frequency [#]

0 10 100 0 1 10 2 10 10 2 3 /ndf = 1.75 10 10 2 pulse integral [10 9 V s] /ndf = 2.24 0 pulse iminimum [mV] 0 residuals 10 residuals 10 0 1 10 10 102 103 pulse integral [10 9 V s] pulse iminimum [mV] (a) Histogram of IceAct pulse integral for all (b) Histogram of IceAct pulse minimum for all events. events.

Figure 6.14: Histogram of IceAct pulse integral and minimum for all events. A power law ((2.1)) t is applied to the data points marked in orange.

63 6 Analysis of Coincident Events

6.3.1 Observing Cosmic Rays

The performance of the IceAct telescope demonstrator can be directly inferred from measurements of the IceTop surface detector and the IceCube in-ice array. The following focuses on coincident events of IceAct, IceTop and IceCube. They were identied using the time synchronization described in sections 5.2 and 5.3 for events selected as described in section 6.2.1 for IceAct, 6.2.2 for IceTop and section 6.2.3 for IceCube.

Air shower/muon impact point

Figure 6.15 shows a two-dimensional histogram of the air shower impact point (left) as measured by IceTop and the in-ice reconstructed muon backtracked to the surface (right). Both distributions are centered around the position of the IceAct telescope. A 2-dimensional Gaussian KDE (kernel density estimator) is used to approximate the distribution and to draw the 1σ level in gure 6.15.

300 300 Stations Strings 8 IceAct IceAct COG COG 3 d = 6.3±10.7m d = 13.5±11.0m 200 7 200 1 1 r = 79m r = 105m

100 100 2 5

0 4 0 y [m] y [m]

3 1 100 100

2 frequency [#], N = 320, binsize 10x10m frequency [#], N = 130, binsize 10x10m

200 1 200 0

300 300 400 300 200 100 0 100 200 300 400 400 300 200 100 0 100 200 300 400 x [m] x [m] (a) Air shower impact point as reconstructed (b) Position of the in-ice muon track extrapo- by IceTop. lated back to surface.

Figure 6.15: Shower core position of coincident events. Positions are given in (x, y) of the IceCube coordinate system and IceTop tanks respectively string positions are marked with blue circles. The center of gravity of both distributions is indicated by a golden star while the green diamond shows the position of the IceAct telescope. The black dashed line corresponds to the 1σ level.

Due to the small eld of view of the IceAct demonstrator of ≈ 2.5◦ and its vertical alignment, pointing straight up4, observed air showers are expected to be very close to the telescope. The 1σ radius of 79 m for IceTop air showers and 104 m for backtracked muons as well as the distance between the COG (center of gravity) and IceAct of 6.3 m

4more or less straight up as discussed later one

64 6.3 IceAct Telescope Performance

(13.5 m) shows this expectation in measured data. It is a rst evidence for coincident cosmic ray air showers observed by the IceAct demonstrator. Figure 6.16 shows the surface position of coincident events as already shown in gure 6.15 but as a zoomed in scatter plot instead of a two-dimensional histogram. The size of each point is proportional to the air shower/muon energy that is also color coded.

For IceTop coincidences (left) the tted S125 (section 6.2.2) and for the in-ice muon the truncated energy (section 6.2.3) is used. Both observables are proportional to the energy of the air shower. The signal in IceTop is strongly dependent on the snow accumulation/height on the Cherenkov tanks, making a conversion from S125 to energy of the cosmic ray primary dicult. Due to a lack of Monte-Carlo events there is no energy conversion from log(S125) to cosmic ray primary energy for 2016 (yet). Nevertheless,

S125 is proportional to the cosmic ray primary energy. Figure 6.16 shows, that outer air showers tend to have higher energies (respectively higher S125 and truncated energy) compared to showers closer to the telescope.

200 3.0 200 COG COG Strings IceTop Stations d = 6.3±10.7m d = 13.5±11.0m IceTop reconstruction muon tracks 150 1 2.5 150 1 IceAct IceAct r = 104m 0

r = 79m 3 1

6 10 =

100 t

100 2.0 o t N

, ] 0 v 2 e

3 50

50 1.5 G

[

y t o g t r N e

, 0 n 0 1.0 e )

y [m] 5 y [m] n 2 1 o S u ( 5 0

10 m 1 50 0.5 50 g d o e l t c u r 100 t 100 0.0 s n o c e r 150 0.5 150

200 1.0 200 104 200 100 0 100 200 200 100 0 100 200 x [m] x [m] (a) Air shower impact point and (b) Position of the backtracked in-ice muon log10(S125) as reconstructed by IceTop. with its truncated energy.

Figure 6.16: Air shower surface positions with color coded energy. The footprint of IceTop/IceCube is shown as black circles while the scatter points indicate the air shower position on the surface. The air shower energy, or its proportional observable, is color coded and visualized by the size of each scatter point.

Figure 6.17 shows this correlation in more detail. Air shower energy estimator of IceTop ( ) and IceCube (truncated muon energy) is shown as a function of distance log10(S125) between air shower and IceAct telescope. Especially for IceTop data (left) it is clearly visible that air showers up to a distance of about 100 m are observed independently of their energy while for larger distances only higher energetic air showers show up in IceAct coincident events. For in-ice data it is basically the same but due to less statistic (130

65 6 Analysis of Coincident Events to 319 events) it is harder to see.

2.5 median 14 median 4 2.0 12 106

1.5 10 )

5 1.0 2 8 1

S 5 ( 10 2 0 1 g

o 0.5 6 l frequency [#], N = 130 4 frequency [#], N = 317 0.0 InIce muon energy [GeV]

104 2 0.5 0 0 1.0 0 25 50 75 100 125 150 175 200 0 25 50 75 100 125 150 175 200 IceAct shower distance [m] IceAct shower distance [m] (a) of IceTop versus distance be- (b) Truncated muon energy versus distance log10(S125) tween IceAct and shower impact point. between backtracked muon and IceAct.

Figure 6.17: Air shower energy estimator versus distance. For IceTop is used log10(S125) as en energy estimator for the air shower energy while for in-ice muons the truncated energy is shown as a function of the air shower distance to IceAct. The green bars correspond to the median and the 25 % and 75 % level of each distance bin.

This is expected as the Cherenkov light pool on surface has a limited spatial extension. The amount of light reaching the ground of an air shower is approximately constant up to ≈ 100 m, given by the Cherenkov cone projected to surface. Higher energetic showers generate more secondary particles with a high pT that emit Cherenkov light "outside" the mean light cone centered on the shower axis (compare with section 2.3). These showers can thus be observed for larger distances to the telescope.

Angular reconstruction

In addition to the shower/muon impact point the direction of coincident air showers is interesting. The camera of the 7-pixel demonstrator has a limited eld of view (FOV) of about 1.5◦ per pixel (see 4.4) resulting in a FOV of ≈ 4.5◦ for the whole camera. Taking the Cherenkov light cone with an opening angle of ≈ 1.4◦ (see 2.3) into account the telescope should be able to observe air showers up to a zenith of ≈ 3.5◦. Figure 6.18 shows the angular reconstruction of IceTop air showers (left) and in-ice muons (right).

◦ The zenith (each on the left) is binned in steps of 1 and each bin count Ni is weighted

66 6.3 IceAct Telescope Performance to account for the cos(θ) dependency of the solid angle:

N N N˜ = c · i , c = tot . (6.1) i cos(θ ) − cos(θ ) P ˜ i i+1 i Ni √ The error is given by Ni (also for the histogram of the azimuth) and weighted the same way, resulting in larger errors for small zenith values as can be seen in the right plot gure 6.18. Given uncertainties in the angular reconstruction and high energetic air showers to be slightly outside the eld of view but still visible due to high pT particles, the distribution is compatible with expectations. There are no coincident air showers or muons with a zenith > 14◦. However, the rst zenith bin (0◦ to 1◦) has less entries than the second one (1◦ to 2◦) for IceTop reconstructions (gure 6.18 left). Moreover, the distribution of the azimuth is not at as one would expect from a telescope pointing vertically up, neither for air showers nor muons.

140 50 N = 320 N = 320 40 N = 130 N = 130 binsize = 1.0 binsize = 60.0 binsize = 1.0 binsize = 60.0 120

120 35 40 100

100 30

80 25 30 80 ) ) ( ( N N s s # # d d o o

c c 20

d 60 d 60 20 15

40 40 10 10

20 20 5

0 0 0 0 0 2 4 6 8 10 12 14 0 50 100 150 200 250 300 350 0 2 4 6 8 10 12 14 0 50 100 150 200 250 300 350 zenith [ ] azimuth [ ] zenith [ ] azimuth [ ] (a) IceTop air shower direction reconstruction. (b) IceCube muon direction reconstruction.

Figure 6.18: Histogram of zenith and azimuth of coincident air showers for IceTop and IceCube. In both cases the zenith histogram is weighted as given by equation (6.1). √ The error is given by a Poisson distribution: σ = Ni, and weighted accordingly (zenith only).

Figure 6.19 shows the angular reconstruction of air showers/muons in a polar projection. Distance r to center corresponds to the zenith while φ equivalent to the azimuth. The mean direction of showers/muons is indicated by the yellow star and given by the direction of all vectors lined up together. In both cases, the mean direction is not pointing straightly up (θ = 0◦) but into φ ≈ 220◦ with a zenith of ≈ 1.13◦. It is notable that both data sets, air showers reconstructed by IceTop and muons in the in-ice detector, lead to the same zenith of 1.13◦ and nearly the same azimuth. This supports the hypothesis, that the telescope is slightly tilted into that direction instead of being vertically aligned.

67 6 Analysis of Coincident Events

For better visualization of the limited FOV of the telescope, it is shown in gure 6.19 as a red area centered on the mean direction.

0° 0° 14 2.25 FOV 14 2.25 tilted FOV 12 N = 320 315° 45° 12 N = 130 COG: 315° 45° 10 10 COG: =53.2 =0.00 =222.3 =1.13 8 8 6 6 4 4 2 2

270° 90° 270° 90°

225° 135° 225° 135°

180° 180° (a) IceTop reconstructed air shower direction. (b) IceCube reconstructed muon direction.

Figure 6.19: Polar scatter plot of zenith and azimuth of coincident air showers. The radial distance corresponds to the zenith while φ is equivalent to the azimuth. The FOV of the telescope is shown as a red area centered on the mean direction indicated by the yellow star.

To test this hypothesis all individual directions are tilted with respect to their mean direction, separately for IceTop and IceCube. A histogram of the resulting directions is shown in gure 6.20. The azimuth is clearly more uniformly distributed compared to gure 6.18 and the zenith is steeply falling. The resulting distributions for zenith and azimuth strongly supports the hypothesis of a tilted telescope. In order to test this hypothesis further, the maximum Likelihood method is used to determine the telescope tilt. The measured number of events in each azimuth bin (Ni) is compared to the expected number for a tilt obtained from a simple simulation ( sim): Ni Many (O(106)) particles are generated uniformly in the small FOV of the telescope. These particles are tilted into one direction given by θ and φ and put into a histogram using the same bins as 6.18. Re-weighting this distribution to the number of coincident events yields to the number of expected events for each bin used in the Likelihood sim. Ni For computational reasons the Likelihood function that needs to be maximized is inverted and minimized:

X sim sim −2LLH = −2 · Ni log(Ni ) − Ni . i

68 6.3 IceAct Telescope Performance

The t is performed using IceTop's air shower reconstruction due to more statistic: 319 events compared to 130 events with an in-ice muon reconstruction. The maximum Likelihood is obtained for a tilt of θ = 0.53◦ into a direction of φ = 230◦. A LLH scan is shown in a polar projection in gure 6.21 (left) along with the obtained azimuth histogram (right) for the best t. It shows that the telescope is indeed tilted by about 0.5◦ and that this small tilt is sucient to cause the observed non uniformity in direction (azimuth) of coincident cosmic rays. The scan (left of gure 6.21) shows, that the direction can be strongly constraint by only 319 coincident events.

140 40 N = 320.0 N = 320 N = 130.0 N = 130 binsize = 1.0 binsize = 60.0 binsize = 1.0 binsize = 60.0 50

120 35 120

30 40 100 100

80 80 30 ) ) ( ( N N s s

# # 20 d d o o c c d d 60 60

20 15

40 40 10

10 20 20 5

0 0 0 0 0 2 4 6 8 10 12 14 0 50 100 150 200 250 300 350 0 2 4 6 8 10 12 14 0 50 100 150 200 250 300 350 zenith [ ] azimuth [ ] zenith [ ] azimuth [ ] (a) IceTop reconstructed and tilted air shower (b) IceCube reconstructed and tilted muon di- direction. rection.

Figure 6.20: Histogram of tilted zenith and azimuth of coincident air showers. All showers/muons are tilted with respect to their mean direction (compare with gure 6.19). The azimuthal distribution is at while the zenith is steeply falling towards higher values.

The expected zenith distribution is also shown in the right of gure 6.21. However, it is not used in the t. The hard cuto at 3◦ is caused by the uniformly simulation in the eld of view of the telescope instead of a Gaussian distribution with some events outside the actual eld of view. Also there are no reconstruction uncertainties taken into account that cause an additional smearing of the zenith distribution up to larger angles.

69 6 Analysis of Coincident Events

IceAct best fit: =230.2 =0.53

measurement measurement N = 317 N = 317 140 60 simulation: simulation: N = 1000000 N = 1000000

120 50

100 40

80 N d d 30

60 relative frequency

20 40

10 20

0 0 0 4 8 12 16 20 0 80 160 240 320 zenith [ ] azimuth [ ]

(a) LLH scan of the telescope tilt. (b) LLH best t for the telescope tilt.

Figure 6.21: LLH t of telescope tilt using coincident air showers. The reconstructed air showers (IceTop) are shown in blue and the expected number of events for the best t in orange for zenith and azimuth (right side). The left side shows the LLH scan of dierent θ , φ along with the best t (yellow plus).

2.5 median 8 median 34 32 2.0 30 106 28 26 6 1.5 24 22 )

5 1.0 2 20 1

S 5 ( 18 10 4 0 1

g 16

o 0.5 l 14 12 frequency [#], N = 317 0.0 10 frequency [#], N = 130 InIce muon energy [GeV] 2 8 4 6 10 0.5 4 2 0 1.0 0 0.975 0.980 0.985 0.990 0.995 1.000 cos( ) 0.975 0.980 0.985 0.990 0.995 1.000 cos( ) (a) IceTop energy estimator vs. log10(S125) (b) IceCube truncated muon energy vs cos(θ). cos(θ).

Figure 6.22: Zenith of coincident air showers versus energy estimator. The zenith (θ) is binned uniformly in cos(θ). The green bars are median and 25 % and 75 % level of each zenith bin. It shows, that more inclined the muon/air shower have higher measured/estimated energies.

Figure 6.22 shows the zenith of observed air showers binned in cos(θ) on the x-axis and IceTop energy estimator (left) or in-ice muon energy (right) on the -axis. log10(S125) y It clearly shows, that up to θ ≈ 3◦ air showers and muons are observed for all energy regimes, but higher energetic air showers/particles can be observed even when being more

70 6.3 IceAct Telescope Performance inclined. This is expected since higher energetic showers have more particles with a high pT emitting Cherenkov light outside the main cone centered on the shower axis. This causes some additional smearing of the zenith distribution in gures 6.18 and 6.20.

Summary

The coincident events of IceAct, IceTop and IceCube prove that the IceAct telescope demonstrator has observed cosmic rays. The impact points are distributed as expected for an small air Cherenkov telescope. Also the direction of coincident air showers is compatible with the limited eld of view of the telescope. The Likelihood approach to determine the telescope tilt might be "overpowered" for the small 7-pixel demonstrator, but it proves the possibility to determine the telescope alignment after deployment using coincident data: important for future versions of the telescope (61-pixel camera) that will be used to improve IceTop air shower reconstructions.

71 6 Analysis of Coincident Events

6.3.2 Energy threshold of the IceAct Telescope

IceTop coincidences Due to dierent snow heights on all IceTop tanks that change from year to year, the conversion from S125 to an air shower energy needs to be calibrated using Monte-Carlo simulations of air showers. For the data of 2016 there is no such conversion obtained from simulations yet. Therefore, the energy estimator S125 is used as well as the IceTop station multiplicity, both directly proportional to air shower and thus primary energy.

Roughly, the conversion from S125 is:

log10(S125) ≈ 0 → E / 1 PeV

log10(S125) ≈ 1 → E ≈ 10 PeV

log10(S125) ≈ 2 → E ≈ 100 PeV .

N = 320 35 N = 320 IceTop_StandardFilter IceTop_StandardFilter 35 N = 287 N = 287 30 30 25 25

# 20

15 15 frequency [#]

10 10

5 5

0 0 3 2 1 0 1 2 3 0 10 20 30 40 50 60 70 80 log10(S125) station multiplicity [#] (a) Histogram of . (b) IceTop station hit multiplicity. log10(S125)

Figure 6.23: IceTop energy estimators of air shower energy. Shown is log10(S125) (left) and the station multiplicity (right) for all coincident events passing the quality cuts of sections 6.2.2 and 6.2.1 in blue and those that would pass the additional IceTop_StandardFilter in orange (dashed).

The distribution of is shown in gure 6.23 (left) along with the station log10(S125) multiplicity (right) for all IceTop coincident events. It peaks at a value of , log10(S125) ≈ 0 roughly 1 PeV. Shown are all coincident events that are selected as described in sections 6.2.2 and 6.2.1. Overlaid are events that would pass the IceTop_StandardFilter applied in the IceTop analysis. This lter removes events close to the energy threshold of IceTop resulting in larger uncertainties of energy and angular reconstruction. Since this lter

72 6.3 IceAct Telescope Performance would cut away events that passed the other quality cuts of IceTop and IceAct, it proves that the telescope's energy threshold is very close or even lower than IceTop's. The same conclusion can be drawn from the station hit multiplicity shown right in gure 6.23.

70 N = 320 25 N = 130 IceTop_StandardFilter N = 287 60 20

50 ]

# 15 [

s 40 n o # i t a t S N 30 10

20 5 10

0 0.5 0.0 0.5 1.0 1.5 2.0 102 103 104 105 106 107 108 log10(S125) energy [GeV] (a) IceTop station multiplicity vs. . (b) In-ice muon energy of coincident events. log10(S125) Coincident events of this analysis (IceTop The energy peaks at about 100 TeV due to station multiplicity vs. ) are strong cuts on direct DOM hits (especially log10(S125) shown in blue while orange events would ≥ 2 strings). pass the standard IceTop lters as well.

Figure 6.24: IceTop energy estimators and in-ice muon energy.

By plotting the station multiplicity of IceTop versus a nice correlation is log10(S125) visible as shown in gure 6.24a. This correlation demonstrates the possibility of con- verting both quantities to an air shower energy with a sucient amount of Monte-Carlo simulations. On the other hand it shows, that both observables are good energy estimators. It also shows that the energy measurement of IceTop can be improved by IceAct: IceAct can add additional information on the air shower, especially for low energy events that would not pass the standard IceTop analysis cuts.

IceCube coincidences The muon energy in the IceCube in-ice array was determined using the "DOMs-method" described in [78]. Obtained muon energies are shown in a histogram in gure 6.24b. They clearly peak at 105 GeV = 100 TeV. Compared to the energy threshold for muons in IceCube of about 100 GeV and about 10 GeV for the DeepCore inll, these are high energetic muons. However, one has to take into account that muons are produced in air cosmic rays air showers and must therefore traverse the ice up to a depth of 1.5 km to reach the in-ice array. Moreover, there are strong cuts applied (e.g. direct DOM hits on two dierent strings) that raise the in-ice energy threshold for this selection. On the

73 6 Analysis of Coincident Events other side these cuts are required to ensure a good, safe reconstruction as muons with only one string hit have large uncertainties in their angular reconstruction.

6.4 Correlation between IceAct and IceCube/IceTop

Although the IceAct telescope demonstrator has a 7-pixel camera only, there should be a (small) correlation of telescope data and IceTop and IceCube. Figure 6.25 shows the IceTop energy estimator (left) and the in-ice muon energy (right) versus the log10(S125) total charge of IceAct (sum of pulse integral of all channels for each event) deposited in the camera for each event. Figure 6.26 shows the same but with the sum of pulse amplitudes instead of pulse integrals.

For the in-ice muon energy no clear correlation is visible. There might be a weak correlation for the IceTop energy estimator and the IceAct charge, but the log10(S125) statistic is to low to quantify this qualitatively. Keeping in mind, that the SiPMs of the telescope demonstrator are not calibrated and corrected for their dierent gain and the IceTop measurement is not corrected for dierent snow heights above each station (yet) the correlation could be improved a lot in the future. This shows that it is possible for to improve IceTop's energy reconstruction by future versions of the telescope.

2.5 median median 6 15 2.0 106

1.5 4 10 ) 5

2 1.0 1

S 5 ( 10 0 1 g

o 0.5 l 2 5 frequency [#], N = 130 frequency [#], N = 317

0.0 InIce muon energy [GeV]

104 0.5 0 0 1.0 101 102 101 102 - IceAct integral [10 9 V s] - IceAct integral [10 9 V s]

(a) IceTop energy estimator . (b) IceCube in-ice muon energy. log10(S125)

Figure 6.25: Air shower energy estimator vs. IceAct pulse integral. The pulse integrals of all pixels are summed up for each event and shown in a log-log plot with the air shower energy estimator (left) and the in-ice muon energy (right). log10(S125)

74 6.4 Correlation between IceAct and IceCube/IceTop

2.5 median median 6

2.0 15 106

1.5 4 )

5 1.0 2

1 10

S 5 ( 10 0 1 g

o 0.5 l 2 frequency [#], N = 130 frequency [#], N = 317

0.0 5 InIce muon energy [GeV]

104 0.5 0 0 1.0 102 103 102 103 - IceAct amplitude [mV] - IceAct amplitude [mV] (a) IceTop energy estimator . (b) IceCube in-ice muon energy. log10(S125)

Figure 6.26: Air shower energy estimator vs. IceAct pulse amplitude. The pulse amplitudes of all pixels are summed up for each event and shown in a log-log plot with the air shower energy estimator (left) and the in-ice muon energy (right). log10(S125)

Figure 6.27 shows the total charge deposited in IceTop stations by each air shower compared to the charge measured in the IceAct telescope. The charge for IceTop is given in units of VEM: vertical muon equivalents. One VEM is the charge deposited in an IceTop tank of a vertical down-going, minimal ionizing muon. Again, the 7-pixel demonstrator can not improve the energy measurement of IceTop but this might be possible for a full 61-pixel IceAct telescope with a much higher resolution and calibrated gain for each pixel.

104 104 median 10 median

10 103 103

2 10 102 5 frequency [#], N = 317 frequency [#], N = 317 IceTop total charge [VEM] IceTop total charge [VEM]

1 10 101 0 0 1 2 10 10 102 103 9 - IceAct integral [10 V s] - IceAct amplitude [mV] (a) IceTop total charge vs. IceAct integral. (b) IceTop total charge vs. IceAct amplitude.

Figure 6.27: IceTop total charge vs. IceAct pulse integral/amplitude. The pulse integrals/amplitudes of all pixels are summed up for each event and shown in a log-log plot with the total charge deposited in IceTop stations for each event.

75 6 Analysis of Coincident Events

76 7 Summary and Outlook

The IceAct telescope demonstrator has been successful operated after its deployment (December 2015) and taken data in 2016. The telescope was connected to the IceCube Neutrino Observatory and its surface detector IceTop. After applying quality cuts there are 501 events remaining that were taken in coincidence with IceCube and IceTop. These events show that the IceAct telescope has observed cosmic rays and that air Cherenkov telescopes can be remotely operated even in the harsh environment at the South Pole. The rst part of this thesis introduced a method to synchronize the data stream of IceAct with the combined data stream of IceTop and IceCube. It is exible and not specic for the IceAct telescope. It can be used for future versions as well as for dierent applications where two data streams need to be aligned/synchronized. Afterwards a rst analysis of the coincident events was performed proving the successful operation and observation of cosmic rays by the telescope. The data, especially the direction and impact point of coincident air showers is in good agreement with the small eld of view of the IceAct demonstrator telescope and its position in the center of the IceTop surface detector array. It also shows that a successor telescope with 61-pixels can improve IceTop's air shower energy measurement that is not yet possible with the limited resolution of the 7-pixel demonstrator. The software developed during this thesis, for time synchronization as well as for IceAct, IceCube and IceTop data processing is applicable to run in an automated way. It is possible to search for coincidences and to process IceCube and IceTop raw data to a higher analysis level on demand. This is important to analyze future data of coincident events in a fast way after it is taken. A successor telescope with a full 61-pixel camera was deployed in the beginning of 2018 and will be replaced by a further improved (especially DAQ improvements) telescope version this season (2019).

77 7 Summary and Outlook

78 Acknowledgment

Nach einem Jahr am III. Physikalisches Institut B möchte ich einigen Personen einen besonderen Dank aussprechen: Zunächst möchte ich mich bei Prof. Wiebusch bedanken, der mir diese Masterarbeit in seiner Arbeitsgruppe angeboten hat. Ihm danke ich insbesondere für die vielen konstruk- tiven Gespräche und Anregungen in den zahlreichen Meetings und die Möglichkeit nach meiner Bachelorarbeit auch meine Masterarbeit in seiner Arbeitsgruppe anzufertigen. Auÿerdem danke ich Ihm für die Möglichkeit einen Teil meiner Arbeit auf dem IceCube Collaboration Meeting 2018 in Stockholm vortstellen zu können und dort wertvolle Er- fahrungen zu sammeln sowie Kontakte knüpfen zu können. Jan Auenberg als meinem Betreuer gilt auch ein besonderer Dank für Anregungen zu dieser Arbeit und die stetige Erreichbarkeit bei Fragen jeglicher Art. Weiter bedanke ich mich bei allen Mitgliedern der Aachener IceCube Gruppe für die vielen interessanten Meetings sowie den gegenseitigen Wissensaustausch. Insbesondere Merlin Schaufel soll hier erwähnt sein der mir als Bürokollege jederzeit weiterhelfen kon- nte. Martin Rongen und Christian Haack danke ich besonders für Ihre Hilfe bei dem Prozessieren der IceCube Daten. Auch Serap Tilav von der University of Delaware gilt mein besonderer Dank für ihre Mithilfe bei den IceTop Event-Rekonstruktionen und ihre Hilfe bei der Analyse der IceTop Daten. Zuletzt möchte ich mich bei meiner Familie, meinen Freunden und meiner Freundin bedanken die mir während der ganzen Zeit dieser Arbeit tatkräftig zur Seite gestanden haben. Ohne ihren Zuspruch, insbesondere während der letzten Wochen, wäre es mir nicht möglich gewesen diese Arbeit anzufertigen.

79 7 Summary and Outlook

80 81 A Appendix A Appendix

(a) 2d-histogram of channel 1 board 1. (b) 2d-histogram of channel 2 board 1.

(e) 2d-histogram of channel 2 board 0. (f) 2d-histogram of channel 3 board 0.

Figure A.1: 2d-histogram of IceAct waveforms.

82 Bibliography

[1] V. F. Hess. Über Beobachtungen der durchdringenden Strahlung bei sieben Freiballonfahrten. In: Phys. Z. 13 (1912), pages 10841091. url: https://www. mpi-hd.mpg.de/hfm/HESS/public/HessArticle.pdf (visited on 09/06/2018). [2] Johannes Blümer, Ralph Engel, and Jörg R. Hörandel. Cosmic rays from the knee to the highest energies. In: Progress in Particle and Nuclear Physics 63.2 (2009), pages 293338. issn: 0146-6410. doi: 10.1016/j.ppnp.2009.05.002. [3] M. Tanabashi et al. Review of Particle Physics. In: Phys. Rev. D 98 (3 Aug. 2018), page 030001. doi: 10.1103/PhysRevD.98.030001. [4] T. K. Gaisser, T. Stanev, and S. Tilav. Cosmic ray energy spectrum from measurements of air showers. In: Frontiers of Physics 8.6 (Dec. 2013), pages 748 758. issn: 2095-0470. doi: 10.1007/s11467-013-0319-7. [5] The Pierre Auger Cosmic Ray Observatory. In: Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associ- ated Equipment 798 (2015), pages 172213. issn: 0168-9002. doi: https://doi. org/10.1016/j.nima.2015.06.058. [6] Jörg R. Hörandel. Models of the knee in the energy spectrum of cosmic rays. In: Astroparticle Physics 21.3 (2004), pages 241265. issn: 0927-6505. doi: https: //doi.org/10.1016/j.astropartphys.2004.01.004. [7] Arno A. Penzias and Robert Woodrow Wilson. A Measurement of excess antenna temperature at 4080-Mc/s. In: Astrophys. J. 142 (1965), pages 419421. doi: 10.1086/148307. [8] Kenneth Greisen. End to the Cosmic-Ray Spectrum? In: Phys. Rev. Lett. 16 (17 Apr. 1966), pages 748750. doi: 10.1103/PhysRevLett.16.748. [9] G. T. Zatsepin and V. A. Kuzmin. Upper limit of the spectrum of cosmic rays. In: JETP Lett. 4 (1966). [Pisma Zh. Eksp. Teor. Fiz.4,114(1966)], pages 7880. [10] W. Heitler. The Quantum Theory of Radiation. 3rd edition. London: Oxford University Press, 1954. Chapter 38, page 386.

83 Bibliography

[11] J. Matthews. A Heitler model of extensive air showers. In: Astroparticle Physics 22.5 (2005), pages 387397. issn: 0927-6505. doi: 10.1016/j.astropartphys. 2004.09.003. [12] P. A. erenkov. Visible Radiation Produced by Electrons Moving in a Medium with Velocities Exceeding that of Light. In: Phys. Rev. 52 (4 Aug. 1937), pages 378 379. doi: 10.1103/PhysRev.52.378. [13] I. M. Frank and I. E. Tamm. Coherent visible radiation of fast electrons passing through matter. In: Compt. Rend. Acad. Sci. URSS 14.3 (1937), pages 109114. doi: 10.3367/UFNr.0093.196710o.0388. [14] The Nobel Prize in Physics 1958. Nobel Media AB 2018. url: https : / / www . nobelprize.org/prizes/physics/1958/summary/ (visited on 09/17/2018). [15] Merlin Schaufel. HAWC's Eye - Implementing Hybrid Detection by Combining a Compact Air-Cherenkov Telescope With the HAWC Gamma-Ray Observatory. Master's thesis. RWTH Aachen University, 2017. [16] Michael Doring et al. Measurement of the Cherenkov light spectrum and of the polarization with the HEGRA IACT system. In: (2001). arXiv: astro-ph/0107149 [astro-ph]. [17] Jamie Holder. Latest Results from VERITAS: Gamma 2016. In: AIP Conf. Proc. 1792.1 (2017), page 020013. doi: 10.1063/1.4968898. arXiv: 1609.02881 [astro-ph.HE]. [18] Ryan C. G. Chaves. Highlights from H.E.S.S. In: AIP Conf. Proc. 1792.1 (2017), page 020012. doi: 10.1063/1.4968897. [19] Javier Rico. Overview of MAGIC results. In: Nuclear and Particle Physics Proceedings 273-275 (2016). 37th International Conference on High Energy Physics (ICHEP), pages 328333. issn: 2405-6014. doi: 10.1016/j.nuclphysbps.2015. 09.046. [20] F. A. Aharonian et al. Measurement of the radial distribution of Cherenkov light generated by TeV gamma-ray air showers. In: Astropart. Phys. 10 (1999), pages 2129. doi: 10.1016/S0927-6505(98)00039-5. arXiv: astro-ph/9807119 [astro-ph].

84 Bibliography

[21] IceCube Collaboration: M.G. Aartsen et al. Evidence for High-Energy Extraterrestrial Neutrinos at the IceCube Detector. In: Science 342.6161 (2013). arXiv:1311.5238. issn: 0036-8075. doi: 10.1126/science.1242856. eprint: http: / / science . sciencemag . org / content / 342 / 6161 / 1242856 . full . pdf. url: http://science.sciencemag.org/content/342/6161/1242856. [22] IceCube Collaboration: M.G. Aartsen et al. Neutrino emission from the direction of the blazar TXS 0506+056 prior to the IceCube-170922A alert. In: Science (2018). issn: 0036-8075. doi: 10.1126/science.aat2890. eprint: http: //science.sciencemag.org/content/early/2018/07/11/science.aat2890. full.pdf. url: http://science.sciencemag.org/content/early/2018/07/11/ science.aat2890. [23] The IceCube Collaboration et al. Multimessenger observations of a ar- ing blazar coincident with high-energy neutrino IceCube-170922A. In: Science 361.6398 (2018). issn: 0036-8075. doi: 10.1126/science.aat1378. eprint: http: / / science . sciencemag . org / content / 361 / 6398 / eaat1378 . full . pdf. url: http://science.sciencemag.org/content/361/6398/eaat1378. [24] physicus. IceCube Collaboration. url: https://gallery.icecube.wisc.edu/ internal/v/GraphicRe/graphics/physicus.pdf.html (visited on 09/18/2018). [25] M.G. Aartsen et al. The IceCube Neutrino Observatory: instrumentation and online systems. In: Journal of Instrumentation 12.03 (2017), P03012. url: http: //stacks.iop.org/1748-0221/12/i=03/a=P03012. [26] M.G. Aartsen et al. Measurement of South Pole ice transparency with the Ice- Cube LED calibration system. In: Nuclear Instruments and Methods in Physics Research Section A 711 (2013), pages 7389. issn: 0168-9002. doi: https://doi. org/10.1016/j.nima.2013.01.054. url: http://www.sciencedirect.com/ science/article/pii/S0168900213001460. [27] P.B. Price and K. Woschnagg. Role of group and phase velocity in high-energy neutrino observatories. In: Astroparticle Physics 15.1 (2001), pages 97100. issn: 0927-6505. doi: https://doi.org/10.1016/S0927- 6505(00)00142- 0. url: http://www.sciencedirect.com/science/article/pii/S0927650500001420. [28] M. G. Aartsen et al. Observation of the cosmic-ray shadow of the Moon with Ice- Cube. In: Phys. Rev. D 89 (10 May 2014), page 102004. doi: 10.1103/PhysRevD. 89.102004. url: https://link.aps.org/doi/10.1103/PhysRevD.89.102004.

85 Bibliography

[29] F. Bos et al. Observation of the Cosmic-Ray Shadow of the Moon and Sun with IceCube. In: ASTRA Proceedings 2 (2015), pages 58. doi: 10.5194/ap-2-5-2015. url: https://www.astra-proc.net/2/5/2015/. [30] M. G. Aartsen et al. Energy reconstruction methods in the IceCube neutrino telescope. In: Journal of Instrumentation 9.03 (2014), P03009. arXiv: 1311.4767. url: http://stacks.iop.org/1748-0221/9/i=03/a=P03009. [31] University of Wisconsin-Madison Physical Sciences Lab. Digital Optical Modules. Board of Regents of the University of Wisconsin System. url: http://www.psl. wisc.edu/projects/large/icecube/more-icecube/dom (visited on 09/10/2018). [32] The IceCube Collaboration, R. Abbasi, et al. IceCube sensitivity for low- energy neutrinos from nearby supernovae. In: Astronomy & Astrophysics 535 (2011), A109. doi: 10.1051/0004-6361/201117810. Erratum ibid. In: A&A 563 (2014), page C1. doi: 10.1051/0004-6361/201117810e. [33] R. Abbasi et al. IceTop: The surface component of IceCube. In: Nuclear In- struments and Methods in Physics Research A 700 (2013), pages 188220. issn: 0168-9002. doi: https://doi.org/10.1016/j.nima.2012.10.067. [34] M. G. Aartsen et al. Measurement of the cosmic ray energy spectrum with IceTop-73. In: Phys. Rev. D 88 (4 Aug. 2013), page 042004. doi: 10 . 1103 / PhysRevD.88.042004. [35] Thomas K. Gaisser. Primary spectrum and composition with IceCube/IceTop. In: Nuclear and Particle Physics Proceedings 279-281 (2016). Proceedings of the 9th Cosmic Ray International Seminar, pages 4755. issn: 2405-6014. doi: https: //doi.org/10.1016/j.nuclphysbps.2016.10.008. [36] J. van Santen. IceCube-Gen2: the next-generation neutrino observatory for the South Pole. In: Proceedings, 35th International Cosmic Ray Conference (ICRC2017): Bexco, Busan, Korea, July 10-20, 2017. PoS(ICRC2017)991. IceCube Collabora- tion. 2017. [37] IceCube-Gen2 Collaboration, : M. G. Aartsen, et al. IceCube-Gen2: A Vi- sion for the Future of Neutrino Astronomy in Antarctica. 2014. eprint: arXiv : 1412.5106. [38] Jamie Yang. IceCube Gen2 Facility. url: https://gallery.icecube.wisc. edu / internal / v / IceCubeGen2 / IceCube _ Gen2 _ Facility / ?g2 _ GALLERYSID = 3c02b78e23dd8dffb0ce78b1332408d4 (visited on 09/18/2018).

86 Bibliography

[39] M. G. Aartsen et al. All-sky Search for Time-integrated Neutrino Emission from Astrophysical Sources with 7 yr of IceCube Data. In: The Astrophysical Journal 835.2 (2017), page 151. [40] Aya Ishihara. A search for extremely high energy neutrinos in 6 years (2008- 2014) of IceCube data. In: PoS ICRC2015 (2015), page 1064. doi: 10.22323/1. 236.1064. [41] Samridha Kunwar et al. The IceTop Scintillator Upgrade. In: PoS ICRC2017 (2017), page 401. doi: 10.22323/1.301.0401. [42] Jan Auffenberg et al. Motivations and Techniques for a Surface Detector to Veto Air Showers for Neutrino Astronomy with IceCube in the Southern Sky. In: The 34th International Cosmic Ray Conference. PoS(ICRC2015)1156. 2015. [43] A. Balagopal V. et al. Search for PeVatrons at the Galactic Center using a radio air-shower array at the South Pole. In: Eur. Phys. J. C78.2 (2018), page 111. doi: 10.1140/epjc/s10052-018-5537-2. arXiv: 1712.09042. [44] Jan Auffenberg. IceAct: Imaging Air Cherenkov Telescopes with SiPMs at the South Pole for IceCube-Gen2. In: PoS ICRC2017 (2018), page 1055. doi: 10.22323/1.301.1055. [45] W. Galbraith and J. V. Jelly. Light Pulses from the Night Sky associated with Cosmic Rays. In: Nature 171 (Feb. 1953), pages 349350. doi: 10.1038/171349a0. [46] Jan Auffenberg. Private communication. RWTH Aachen University, 2018. [47] Merlin Schaufel et al. Small size air-Cherenkov telescopes for ground detection arrays - a possible future extension? In: PoS ICRC2017 (2018), page 786. doi: 10.22323/1.301.0786. [48] Thomas Bretz et al. FAMOUS - A uorescence telescope using SiPMs. In: PoS ICRC2015 (2016), page 649. doi: 10.22323/1.236.0649. [49] A. U. Abeysekara et al. The 2HWC HAWC Observatory Gamma Ray Catalog. In: Astrophys. J. 843.1 (2017), page 40. doi: 10.3847/1538-4357/aa7556. arXiv: 1702.02992 [astro-ph.HE]. [50] Jan Audehm. Simulation und Test von Winston Cones für das abbildende Luft- Tscherenkow-Teleskop IceACT. Bachelor's Thesis. RWTH Aachen University, Aug. 2016.

87 Bibliography

[51] SCHOTT BOROFLOAT R 33. The versatile oated borosilicate glass - with an innite number of applications. SCHOTT Technical Glass Solutions GmbH. url: https://psec.uchicago.edu/glass/borofloat_33_e.pdf (visited on 09/06/2018). [52] Positive Fresnel Lenses. ORAFOL Fresnel Optics GmbH. 2013. url: https:// www.orafol.com/energy/europe/en/positive- fresnel- lenses (visited on 09/06/2018). [53] Hans Michael Eichler. Characterisation studies on the optics of the prototype uorescence telescope FAMOUS. Master's thesis. RWTH Aachen University, Mar. 2014. [54] Roland Winston, Juan C. Miñano, and Pablo Benítez. Nonimaging Optics. Burlington: Academic Press, 2005. isbn: 978-0-12-759751-5. doi: 10.1016/B978- 0-12-759751-5.X5000-3. [55] Jan Paul Koschinsky. Development of a 61-Pixel Camera for the IceAct Imag- ing Air Cherenkov Telescope. Master's thesis. RWTH Aachen University, Sept. 2017. [56] Steven H Simon. The Oxford Solid State Basics. Oxford, United Kingdom: Ox- ford Univ. Press, 2013. [57] J-Series High PDE and Timing Resolution, TSV Package. USER MANUAL. Rev. 1.0, December 2016. SensL Technologies Ltd. 2016. url: http : // sensl . com / downloads/ds/UM-MicroJ.pdf (visited on 09/22/2018). [58] Hamamatsu Optosemiconductor Handbook. Editorial Committee. Hamamatsu Pho- tonics K.K. 2014. url: https://www.hamamatsu-news.de/hamamatsu_optosemiconductor_ handbook/files/assets/basic-html/index.html (visited on 09/21/2018). [59] J-Series High PDE and Timing Resolution, TSV Package. DATASHEET. Rev. 3.1, Preliminary, December 2017. SensL Technologies Ltd. 2017. url: http://sensl. com/downloads/ds/DS-MicroJseries.pdf (visited on 09/22/2018). [60] A. Biland et al. Calibration and performance of the photon sensor response of FACT The First G-APD Cherenkov telescope. In: JINST 9.10 (2014), P10012. doi: 10.1088/1748-0221/9/10/P10012. arXiv: 1403.5747 [astro-ph.IM]. [61] M. Petasecca et al. Thermal and electrical characterization of silicon photomultiplier. In: Proceedings, 2007 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC 2007): Honolulu, Hawaii, October 28-November 3, 2007. Volume 2. Oct. 2007, pages 12751278. doi: 10.1109/NSSMIC.2007.4437235.

88 Bibliography

[62] C-Series Low Noise, Blue-Sensitive Silicon Photomultipliers. DATASHEET. Rev. 3.0, Preliminary, April 2018. SensL Technologies Ltd. 2014. url: http://sensl. com/downloads/ds/DS-MicroCseries.pdf (visited on 09/22/2018). [63] IceAct Commissioning at South Pole. RWTH Aachen University. url: http:// www . institut3b . physik . rwth - aachen . de / cms / ParticlePhysics3B / Das - Institut/Aktuelle-Meldungen/Aktuelle-Meldungen-intern/~jvrm/IceAct- Commission-at-South-Pole/ (visited on 08/30/2018). [64] Tim Niggemann et al. Status of the Silicon Photomultiplier Telescope FAMOUS for the Fluorescence Detection of UHECRs. In: Proceedings, 33rd International Cosmic Ray Conference (ICRC2013): Rio de Janeiro, Brazil, July 2-9, 2013. 2015, page 0014. arXiv: 1502.00792 [astro-ph.IM]. [65] Datenblatt UG 11. SCHOTT AG. 2014. url: https : / / www . schott . com / d / advanced_optics/2ac9d777-241a-4b09-940f-4dcf5f7aa898/1.2/schott-uv- bandpass-ug11-jun-2017-de.pdf (visited on 09/22/2018). [66] DRS4 Evaluation Board. Paul Scherrer Institut. url: https://www.psi.ch/drs/ evaluation-board (visited on 09/01/2018). [67] Stefan Ritt. DRS4 Evaluation Board User's Manual. Board Revision 5 as of January 2014. Last revised: April 27, 2016. Paul Scherrer Institut. url: https:// www.psi.ch/drs/DocumentationEN/manual_rev50.pdf (visited on 09/01/2018). [68] Merlin Schaufel. Private communication. RWTH Aachen University, 2018. [69] Tim Niggemann. The silicon photomultiplier telescope FAMOUS for the detection of uorescence light. PhD thesis. RWTH Aachen University, 2016. [70] Johannes Schumacher et al. Dedicated power supply system for silicon photomultipliers. In: PoS ICRC2015 (2016), page 605. doi: 10.22323/1.236.0605. [71] Maurice Guender. Analyse der Zeitsynchronisation des IceACT-Luft-Cerenkov- Teleskop Prototypen mit dem IceCube Neutrino Observatorium anhand erster Feld- daten. Bachelor's Thesis. RWTH Aachen University, Aug. 2016. [72] Lasse Halve. Analysis of the rst Data from the SiPM-Camera of the Air Cherenkov Telescope IceAct at the South Pole. Bachelor's Thesis. RWTH Aachen University, Apr. 2016. [73] M. G. Aartsen et al. Measurement of the cosmic ray energy spectrum with IceTop-73. In: Phys. Rev. D88.4 (2013), page 042004. doi: 10.1103/PhysRevD. 88.042004. arXiv: 1307.3795 [astro-ph.HE].

89 Bibliography

[74] Serap Tilav. Private communication. University of Delaware, 2018. [75] J. Ahrens et al. Muon track reconstruction and data selection techniques in AMANDA. In: Nucl. Instrum. Meth. A524 (2004), pages 169194. doi: 10.1016/ j.nima.2004.01.065. arXiv: astro-ph/0407044 [astro-ph]. [76] Kai Schatto. Stacked searches for high-energy neutrinos from blazars with Ice- Cube. PhD thesis. Johannes Gutenberg-Universität Mainz, June 2014. [77] Leif Rädel. Measurement of High-Energy Muon Neutrinos with the IceCube Neutrino Observatory. PhD thesis. RWTH Aachen University, 2017. [78] R. Abbasi et al. An improved method for measuring muon energy using the truncated mean of dE/dx. In: Nucl. Instrum. Meth. A703 (2013), pages 190198. doi: 10.1016/j.nima.2012.11.081. arXiv: 1208.3430 [physics.data-an].

90 List of Figures

2.1 Energy spectrum of cosmic rays...... 4 2.2 Schematic of the Heitler model for air shower description...... 6 2.3 Illustration of polarization yielding to Cherenkov light in a dielectric medium. Taken from [15] ...... 8 2.4 Cherenkov spectrum and illustration of emission angle...... 9 2.5 Light pool of a γ-ray induced air shower as measured on ground...... 10

3.1 Artistic view of cosmological particle propagation...... 11 3.2 Sketch of the IceCube Neutrino Observatory ...... 12 3.3 Examples of the two event topologies in the IceCube array: ...... 13 3.3a Shower event ...... 13 3.3b Track-like event ...... 13 3.4 Picture and a sketch of an IceCube DOM...... 14 3.5 Array layout of the IceTop surface detector and a detailed sketch of one of its stations. Taken from [33]...... 16 3.6 Top and side view of proposed Gen2 extensions of the IceCube Neutrino observatory...... 17

4.1 Photograph of the IceAct imaging air Cherenkov telescope installed at South Pole in 2017...... 19 4.2 Mechanical sketch of the IceAct imaging air Cherenkov telescope...... 21 4.3 Picture of IceAct's camera circuit board...... 22 4.5 Close-up photography of Winston cones used for IceAct telescopes. . . . . 24 4.6 Sketch of a p-n junction...... 25 4.7 Structure of a single GAPD cell used in SiPMs...... 26 4.8 Image and simplied schematic of an SiPM...... 27 4.9 Overlay of measured SiPM waveforms...... 27 4.10 Deployment of the IceAct telescope demonstrator at South Pole ...... 29 4.11 Picture of the 7-pixel camera showing the SiPMs and Winston cones. . . . 30

91 List of Figures

4.12 Sketch of the single pixel of the 7-pixel camera with Winston cone. . . . . 31 4.13 Picture of the 7-pixel camera of the IceAct telescope demonstrator . . . . 32 4.14 DRS4 Evaluation boards used in the IceAct telescope demonstrator . . . . 33 4.15 Overview of data runs taken with the IceAct telescope demonstrator in 2016...... 35

5.1 Schematic of the trigger connection between IceAct and IceCube . . . . . 38 5.2 Visualization of the time synchronization algorithm ...... 39 5.3 Time synchronization results using all coincident events ...... 42 5.4 Time synchronization results using only coincident physic events . . . . . 43 5.5 Event rate of IceAct - IceCube coincidences ...... 44

6.1 Example signal trace for software trigger algorithm...... 46 6.2 Example trace for signal extraction...... 47 6.3 Trigger threshold setting for the analyzed IceAct run...... 49 6.4 Pixel multiplicity of all coincident physics trigger...... 51 6.5 Time dierence for IceAct-IceCube event pairs after time synchronization. 52 6.6 Histogram of the baseline RMS for all channels and example of noise. . . . 53 6.8 Histogram of direct DOM and string hits...... 58 6.9 Event rate of IceAct-IceCube/IceTop coincident events...... 59 6.10 2d-histogram of waveforms of channel 0 and 3 of board 1...... 60 6.11 Trigger multiplicity and mean charge of each camera pixel...... 60 6.12 IceAct event display of two coincident events...... 61 6.13 IceCube event display of two coincident events...... 62 6.14 Histogram of IceAct pulse integral and minimum for all events...... 63 6.15 Shower core position of coincident events...... 64 6.16 Air shower surface positions with color coded energy...... 65 6.17 Air shower energy estimator versus distance...... 66 6.18 Histogram of zenith and azimuth of coincident air showers...... 67 6.19 Polar scatter plot of zenith and azimuth of coincident air showers...... 68 6.20 Histogram of tilted zenith and azimuth of coincident air showers...... 69 6.21 LLH t of telescope tilt using coincident air showers...... 70 6.22 Zenith of coincident air showers versus energy estimator...... 70 6.23 IceTop energy estimators of the air shower energy...... 72 6.24 IceTop energy estimators and in-ice muon energy...... 73 6.24a IceTop station multiplicity vs...... 73 log10(S125) 6.24b In-ice muon energy of coincident events...... 73

92 List of Figures

6.25 Air shower energy estimator vs. IceAct pulse integral...... 74 6.26 Air shower energy estimator vs. IceAct pulse amplitude...... 75 6.27 IceTop total charge vs. IceAct pulse integral/amplitude...... 75

A.1 2d-histogram of IceAct waveforms...... 82

93 List of Figures

94 List of Tables

6.1 Summary of applied IceAct quality cuts...... 53 6.2 Summary of applied IceTop quality cuts...... 55 6.3 Summary of applied InIce quality cuts...... 57

Zentrales Prüfungsamt/Central Examination Office

Eidesstattliche Versicherung Statutory Declaration in Lieu of an Oath

______Name, Vorname/Last Name, First Name Matrikelnummer (freiwillige Angabe) Matriculation No. (optional) Ich versichere hiermit an Eides Statt, dass ich die vorliegende Arbeit/Bachelorarbeit/ Masterarbeit* mit dem Titel I hereby declare in lieu of an oath that I have completed the present paper/Bachelor thesis/Master thesis* entitled

______selbstständig und ohne unzulässige fremde Hilfe (insbes. akademisches Ghostwriting) erbracht habe. Ich habe keine anderen als die angegebenen Quellen und Hilfsmittel benutzt. Für den Fall, dass die Arbeit zusätzlich auf einem Datenträger eingereicht wird, erkläre ich, dass die schriftliche und die elektronische Form vollständig übereinstimmen. Die Arbeit hat in gleicher oder ähnlicher Form noch keiner Prüfungsbehörde vorgelegen. independently and without illegitimate assistance from third parties (such as academic ghostwriters). I have used no other than the specified sources and aids. In case that the thesis is additionally submitted in an electronic format, I declare that the written and electronic versions are fully identical. The thesis has not been submitted to any examination body in this, or similar, form.

______Ort, Datum/City, Date Unterschrift/Signature *Nichtzutreffendes bitte streichen *Please delete as appropriate Belehrung: Official Notification:

§ 156 StGB: Falsche Versicherung an Eides Statt Wer vor einer zur Abnahme einer Versicherung an Eides Statt zuständigen Behörde eine solche Versicherung falsch abgibt oder unter Berufung auf eine solche Versicherung falsch aussagt, wird mit Freiheitsstrafe bis zu drei Jahren oder mit Geldstrafe bestraft. Para. 156 StGB (German Criminal Code): False Statutory Declarations Whoever before a public authority competent to administer statutory declarations falsely makes such a declaration or falsely testifies while referring to such a declaration shall be liable to imprisonment not exceeding three years or a fine. § 161 StGB: Fahrlässiger Falscheid; fahrlässige falsche Versicherung an Eides Statt (1) Wenn eine der in den §§ 154 bis 156 bezeichneten Handlungen aus Fahrlässigkeit begangen worden ist, so tritt Freiheitsstrafe bis zu einem Jahr oder Geldstrafe ein. (2) Straflosigkeit tritt ein, wenn der Täter die falsche Angabe rechtzeitig berichtigt. Die Vorschriften des § 158 Abs. 2 und 3 gelten entsprechend. Para. 161 StGB (German Criminal Code): False Statutory Declarations Due to Negligence (1) If a person commits one of the offences listed in sections 154 through 156 negligently the penalty shall be imprisonment not exceeding one year or a fine. (2) The offender shall be exempt from liability if he or she corrects their false testimony in time. The provisions of section 158 (2) and (3) shall apply accordingly.

Die vorstehende Belehrung habe ich zur Kenntnis genommen: I have read and understood the above official notification: ______Ort, Datum/City, Date Unterschrift/Signature