Prototype Data Analysis of the KM3NeT Neutrino Telescope

(1)

University of Amsterdam

Bachelor Thesis

Prototype Data Analysis of the

KM3NeT Neutrino Telescope

Author:

Rick Hutten

10189939

Supervisor:

prof. dr. P.M. Kooijman

Second supervisor:

dr. M. Vreeswijk

Bachelor Thesis Physics and Astronomy, 15 EC

30-03-2015 - 29-06-2015

submitted on 08-10-2015

NIKHEF University of Amsterdam FNWI

(2)

Abstract

Detection of neutrinos from outer space can reveal new information that is un-obtainable with modern day telescopes. The KM3NeT neutrino telescope aims to pinpoint sources of neutrino production by detecting Cherenkov radiation in-duced by muons which are formed in neutrino collisions. In this research, data is analysed coming from two different prototypes and shows the improvements made in the newest prototype, the PPM-DU. Results show that the newest pro-totype, the PPM-DU, suffers less from bioluminescence than the PPM-DOM and is ready for final deployment.

(3)

Populair wetenschappelijke samenvatting

Artist impression van de KM3NeT detector in de Mid-dellandse Zee.

De aarde wordt continu bestookt met deeltjes vanuit de ruimte met enorme energie. Deze deeltjes, kosmische straling genoemd, bestaan voornamelijk uit protonen en atoom kernen. Het is in de astronomie nog onduidelijk waar deze kosmische straling vandaan komt en door welke processen deze deeltjes hun hoge energie te danken hebben. Het KM3NeT project pro-beert deze vraag te beantwoorden door mid-del van een grote onderzeese telescoop.

Omdat kosmische straling bestaat uit ge-laden deeltjes, worden deze afgebogen door de magnetische velden in het heelal en het magnetische veld van de aarde. De richting waarmee deze deeltjes de aardbodem bereiken is dus niet hetzelfde als de richting van de oor-sprong van deze kosmische deeltjes. Hierdoor is het bijna onmogelijk om de precieze oor-sprong van de kosmische straling te achter-halen. De verwachting is dat er bij de

pro-cessen die leiden tot kosmische straling ook andere soorten deeltjes worden gemaakt, namelijk neutrino’s. Neutrino’s zijn elementaire deeltjes en reageren zo weinig met materie dat ze zonder moeite door de hele aarde kunnen reizen zonder een enkel spoor achter te laten. Hierdoor zal het overgrote deel onge-hinderd van de plek van creatie naar de aarde kunnen reizen. Ook zijn de deeltjes ongeladen waardoor ze niet door magneetvelden kunnen worden afge-bogen. Deze eigenschappen maken het deeltje ideaal om bronnen van kosmische straling te achterhalen.

Om een neutrino te kunnen detecteren, is het echter wel vereist dat een neutrino botst met andere deeltjes. Omdat dit zeer zelden gebeurt is dit ook een groot obstakel in het observeren van neutrino’s. Gelukkig zijn er zo veel neutrino’s om ons heen dat er wel af en toe botsingen plaats vinden. Het aantal neutrino’s dat waargenomen kan worden schaalt met de grootte van de detector. De KM3NeT neutrino telescoop wordt een reusachtig telescoop op de bodem van de Middellandse Zee met een volume van enkele kubieke kilometers. Neutrino’s botsen met atomen in de aarde of zee en vormen deeltjes genaamd muonen. Muonen zijn deeltjes equivalent aan elektronen maar zijn een stuk zwaarder. De muonen hebben een hele hoge snelheid en zullen sneller gaan dan het licht in zeewater. De muonen zullen hierbij licht gaan uitstralen wat bekend staat als het Tsjerenkov-effect. De KM3NeT neutrino telescoop probeert dit licht waar te nemen en hieruit de richting van het originele neutrino te reconstrueren.

(4)

De neutrino telescoop is nog in aanbouw maar er zijn al wel twee werkende prototypes. In dit verslag is er een data analyse gemaakt op de data afkomstig van deze twee prototypes. Resultaten laten onder andere zien dat het eerste prototype erg last heeft van bioluminisentie, het uitzenden van licht door diertjes die onder water bij de telescoop leven. Het tweede prototype heeft aanzienlijk minder last van bioluminisentie en het experiment is klaar voor de plaatsing van de eerste onderdelen van de neutrino telescoop op de bodem van de Middelandse Zee.

(5)

Introduction

In 1930 the existence of a new particle was postulated by Wolfgang Pauli in order to explain the continuous energy spectrum of particles produced in beta decay. The new particle was meant to conserve the laws of energy and momentum. Wolfgang Pauli first named the mysterious particle the neutron, which was later renamed by Enrico Fermi to neutrino to avoid confusion with the more massive baryon. The neutrino has no charge and only interacts through the weak nuclear force. Since the only force by which the neutrino can interact with other particles is the weak nuclear force, it is able to easily pass through matter without interactions with the surrounding particles. This lack of interaction with other particles makes the neutrino very hard to detect, since it could pass through a detector without leaving a signal. Although it’s existence is confirmed, many properties of the neutrino still remain unclear. Experiments have shown that the neutrino has a small but nonzero mass. Theory has provided only lower and upper bounds of the mass. Research has to be done to get a better understanding of the standard model.

The KM3NeT (short for Cubic Kilometre Neutrino Telescope) collabora-tion’s objective is to locate sources of high energy neutrino production in the universe. By trying to pinpoint the direction of incoming neutrino’s it can be traced back from where the neutrino originated. Only two confirmed objects are known to be sources of neutrinos, the sun and supernova 1987A.1,2 _High

energy neutrinos are believed to be created in the same processes which create cosmic rays. Sources of cosmic ray production are unknown because the rays are deflected by the magnetic fields of the Earth and the galactic magnetic field. Studying sources of neutrino production may also lead to the discovery of new cosmic ray sources.

As of writing this thesis, two prototypes have been deployed. The first stage of deployment of the final production model is expected to be within sev-eral weeks. This research is about providing an analysis of data coming from the two deployed prototypes, the PPM-DU and PPM-DOM. First, a theoret-ical background and an introduction to the KM3NeT project will be provided. Secondly, the results of the data analysis of the prototypes will be given and

(7)

discussed. Lastly, there will be a conclusion and recommendations for future work.

1.1 Neutrino Astronomy

When humankind first started using telescopes to study the universe in the early 17th century, the only information that was studied was visible light. It was not until the 1930s that scientists became aware of radio waves emitted by celestial bodies. Modern astronomy makes use of a wider range of the electromagnetic wave spectrum. There are several branches in astronomy that research electro-magnetic waves from a specific range in wavelengths. These include gamma-ray, x-ray, ultraviolet, optical, infrared and radio astronomy. Since some wavelengths are absorbed or disturbed in the Earth’s atmosphere most wavelengths are best observed from space. Not only the atmosphere limits the electromagnetic waves from being observed. Photons can be absorbed by interstellar particles causing them to never even reach the Earth. Most photons that are observed from Earth are emitted by the surface of celestial bodies.

In contrast to photons, neutrinos are only affected by the weak nuclear force and therefore can travel through matter undisturbed. The neutrinos coming from celestial bodies can travel through all obstacles photons may encounter. This property of the neutrino makes it possible to gather information from processes that can not be obtained by electromagnetic waves. However, this property is also a disadvantage for detecting the particle since the neutrino needs some kind of interaction to be detected. Neutrino telescopes are therefore very large in size to increase the chance of a collision.

KM3NeT is a successor of the ANTARES3_{, NEMO}4 _{and NESTOR}5

pro-jects. All telescopes are located in deep water and utilize the water as collision material for neutrino interactions. This way it is possible to create a detector that is several cubical kilometers in volume as is the case for KM3NeT. Together with neutrino telescope IceCube6 _{which is located under the ice in Antarctica}

searching for neutrinos from the northern hemisphere, the KM3NeT neutrino telescope will look for neutrinos coming from the southern hemisphere and will complete full coverage of the sky.

(8)

Chapter 2

Detection Method

The idea behind detecting neutrino’s in the KM3NeT experiment relies on an upward going neutrino passing through the Earth and colliding with an atom in the outer layers of the Earths crust or in the water. Neutrino’s can only interact through the weak nuclear force by a Z or a W boson. Interaction mediated by exchanging a Z boson is called neutral current and only transfers energy to the particle it collided with. The neutrino bounces off the particle but it will not change flavor. Interactions mediated by a Z boson are therefore not of interest in this experiment. Interactions mediated by a W boson, called charged current, are different because the W boson carries charge. A W boson will change the flavor of a neutrino to the charged lepton of the same generation. The outcome of a collision may thus produce an electron, a muon or a tau lepton, depending on the flavor of the neutrino. These charged leptons are possible to detect by the light produced by Cherenkov radiation.

Electrons produced by an electron neutrino will not propagate very far in matter. High energised electrons produce electromagnetic showers caused by bremsstrahlung and pair creation. The electron is therefore not able to travel long distances in the water. This is why an electron neutrino is hard to detect, the neutrino has to collide inside the detector for the electron to be detected.

Tau leptons have a mass that is about 3.5 · 103 _{times higher than the mass}

of the electron. The energy loss by bremsstrahlung is therefore lower than that of the electron as the total radiated power P ∝_m14 when the acceleration ~a ⊥ ~v

and P ∝ _m16 when ~a k ~v.7,8 However, tau leptons have a very small mean

lifetime of 2.9 · 10−13 seconds. If a tau is created, it will almost immediately decay into a tau neutrino. Only tau leptons with an energy in the PeV range or above can travel a considerable distance to be able to collide outside of the detector, effectively increasing the volume in which a collision can occur. The short mean lifetime of the tau lepton makes detecting tau neutrinos hard.

Muons are about 207 times heavier than electrons making them lose less energy by bremsstrahlung than electrons do. Muons have a mean lifetime of 2.2 · 10−6 seconds which is significantly longer than the mean lifetime of the tau lepton. Therefore they can travel a much longer distance before decaying.

(9)

Because the muon can travel a long distance before it decays, the collision of the muon neutrino does not necessarily have to be inside of the detector, as long as the created muon is directed towards the detector. Because of this increased volume in which the neutrino can collide, muon neutrinos have by far the highest chance of being detected.

Figure 2.1: Feynman diagram showing a muon neutrino interacting with a proton or neutron to create a muon and a hadron shower. This type of weak interactions is called the charged current because of the charged W boson.

Figure 2.1 shows a typical collision where a muon neutrino collides with an atom in which it creates an muon and a hadron shower. The muon is a charged particle with a great penetration depth. The muon will travel in roughly the same direction as the neutrino and travels through the water in the Mediterranean Sea, as shown in figure 2.2. This particle’s velocity depends on the energy of the neutrino. If the velocity of the muon is large enough, the particle will induce Cherenkov radiation. The photons of the Cherenkov radiation will eventually be detected by the detector.

2.1 Cherenkov radiation

A particle is able to induce photons when it is in a dielectric medium, with a refractive index n, and is moving with a velocity that is higher than the phase velocity of photons through the same medium, c

n. This process is known as

Cherenkov radiation. As the particle travels through the dielectric medium, the electric field of the particle excites surrounding electrons. These electrons will radiate off photons when they fall back to their original state. If the particle moves slower than _nc, the photons are are annihilated by destructive interfer-ence. When the particle travels faster than _nc, the photons will cause form a coherent wavefront caused by constructive interference. The wavefront is pro-duced under a specific angle that depends on the speed of the charged particle and the refractive index of the medium. A schematic view of Cherenkov radi-ation is given in figure 2.3. The angle at which the light is emitted is called the Cherenkov angle given by equation 2.1.

(10)

Figure 2.2: At the bottom, a muon neutrino enters the Earth’s crust and collides with another particle to create a muon. This muon flies off into the water and induces Cherenkov radiation which is detected by the KM3 Neutrino Telescope. cos(θc) = 1 βn = c vn (2.1) Here v denotes the velocity of the charged particle and n the refractive index of the medium.

Figure 2.3: Cherenkov radiation caused by a charged particle (red line) traveling faster than _nc. The blue arrows point in the direction (angle θc)

(11)

Chapter 3

The KM3NeT detector

The detector used in the KM3NeT experiment consists of multiple smaller build-ing blocks. In total, the neutrino telescope will have about 8100 Digital Optical Modules that will collect photons from the bottom of the Mediterranean Sea. A Digital Optical Module (DOM) is a glass sphere containing 31 photomultiplier tubes. Photomultiplier tubes (PMTs) are very sensitive photon detectors and are further discussed in section 3.3. Multiple DOMs are connected by a line and are spaced several meters apart, the actual spacing will vary. This line is called a detection unit (DU) and is attached to the seabed. The final version of the KM3NeT detector will house 18 DOMs on a single detection unit. The de-tection unit is partially kept upright by the DOMs which are vacuum pumped. Some extra buoys are used to create extra buoyancy so the string will stand up straight. In total, the neutrino telescope will have 450 detection units in two separate locations and will cover a volume of several cubic kilometers. Cur-rently there are two prototypes active for testing purposes, described in sections 3.1 and 3.2. The locations of the final detector are the same as where the two prototypes are deployed.

3.1 The PPM-DU prototype

The Pre-Production Model Detection Unit is a prototype consisting of a detec-tion unit with just 3 DOMs. It was deployed on May 6 2014, 80 kilometers off the coast of Portopalo di Capo Passero in Italy at a depth of 3457 meters.9_The

DOMs are separated 36 meters apart with the bottom DOM located 72 meters above the seabed. Two empty glass spheres are attached to the top to provide buoyancy to keep the structure in a vertical position.

The prototype uses two different types of PMTs. The PMTs placed in DOM 0 and DOM 1 (the bottom and middle DOMs) are made by ETEL, the PMTs in DOM 2 (top DOM) are made by Hamamatsu. The PMTs are similar performance wise but have different reflector rings. On the outside on the front of the PMTs where the light enters the PMT, a reflective ring is placed at an

(12)

angle of 45° to increase the amount of light hitting the PMT.10 This reflective ring has a width of 16mm on the PMTs produced by ETEL and a width 17mm of the PMTs produced by Hamamatsu. Because of this, it is expected that the DOM 2 will have a higher average signal than the two other DOMs.

3.2 The PPM-DOM prototype

Figure 3.1: PPM-DOM at-tached to an ANTARES Optical Module Frame.

The second prototype is called the Pre-Production Model Digital Optical Module or PPM-DOM for short. The prototype consists of only one single DOM attached to an ANT-ARES frame called an Optical Module Frame. It is located 40 kilometers off the coast near Toulon in france at a depth of 2375 meters near the ANTARES neutrino telescope.11 _A

picture of the PPM-DOM just before it was deployed is shown in figure 3.1.

3.3 Photomultiplier tubes

A photomultiplier tube (or PMT for short) is a very sensitive photon detector capable of detecting single photons. It combines the principles of the photoelec-tric effect and secondary emission to multiply the signal coming from a single photon for over a million times. Photons that are headed to the PMT will strike the photocathode, this is a negatively charged cathode that absorbs incoming light and produces electrons. The energy of the outgoing electron is the same as the photon energy minus the binding energy of the electron in the photocath-ode. The outgoing electron is then directed to the first dynphotocath-ode. This dynode is positively charged so that the potential between the cathode and the dynode will cause the electron to gain a lot of energy proportional to the potential dif-ference. When an electron hits an electrode in a vacuum tube it is able to emit additional electrons. The number of electrons emitted is larger than the number of electrons hitting the electrode. This effect is called secondary emission.12_The

newly created electrons are being pulled to the next dynode which has a higher voltage than the first. The electrons will then crash into the second dynode creating even more electrons. This process is repeated several times until the electrons reach the anode. Here the number of electrons is greatly increased and the signal from the anode is large enough to be read out. A schematic view of this system is shown in figure 3.2.

The voltage on the anode is constantly being read out. When a photon causes emission of electrons in the PMT, the voltage peaks on the anode. The size of this peak is proportional to the energy of the photon and is recorded by its Time over Threshold (ToT). This is the time the voltage is above a pre-configured threshold and is a characteristic of the strength of a signal; a strong

(13)

Figure 3.2: Schematic view of a PMT.

signal will have a larger ToT than a weak signal. The threshold of is set at 1.2 × the baseline of the PMT.13

3.4 PMT numbering

The PMTs in the DOM are numbered in a specific way. One of the 31 PMTs is pointed straight downwards, this PMT is numbered 0. Then there are five rows of six PMTs, three rows on the bottom half and two rows on the top half of the DOM. Figure 3.3 shows how the PMTs are numbered. The numbering of the PMTs in every row start at the same direction. This way, PMTs that are above each other differ in number by six. The PMTs are equally distributed in every row resulting in an angle of 60° between two PMTs. Alternate rows are shifted 30° to make more room to fit the PMTs. Because of this, a PMT that is ’directly above’ another PMT may differ even 5, 6 or 7 in number.

Figure 3.3: Digital Optical Module overlapped with the numbering of the PMTs.

(14)

3.5 Data acquisition

The signal coming from the PMTs is analog. Therefore the data has to be converted to a digital signal. Normally, when the PMT is not hit and the voltage of the anode is not above the pre-configured threshold, the PMT will send a low signal to the Central Logic Board (CLB). When electrons in the PMT cause the signal of the anode to be higher than the threshold, the PMT will send a high signal to the CLB until the signal of the anode falls below the threshold. The CLB then processes the signal from the PMT and calculates the Time over Threshold (ToT). The time of the hit and the ToT are saved to a buffer. When a buffer is full, the buffer is send through optical fibers to the shore station and a new buffer is created. After every ∼134 ms which is called a timeslice, the remaining buffer is send to shore. On the shore station, the buffers of the same timeslice are filtered by special software and stored on disk. When there are two hits on a DOM with a maximum time difference of 25 ns this is called a L1 hit. Hits that are caused outside this time-frame of 25 ns are not likely to be caused by an event. All L1 hits within a time-frame of 330 ns trigger a physics event which is consistent with charged bypassing particles inducing Cherenkov radiation. The data flux through the optical fibers to the shore station is highly dependent on the amount of ToTs that have been recorded by the CLB. Therefore the more a PMT is hit with light, the more data flows through the cables. This stream of data can get too much for the optical fibers, which have a maximum capacity of 10 Gb/s14_{, if there is too}

much bioluminescence. This is a problem since the CLB does not have enough memory to store all the data that can’t be transferred in time, this will result in loss of data. To prevent this, the data stream gets cut when the signal gets too high. The data that is thrown away is not of importance because such high signals can only be caused by the bioluminescence of larger animals. In this period, muon tracks cannot be recorded because the background signal is too high.

(15)

Chapter 4

Background signal

Light produced by the Cherenkov effect is not the only signal the neutrino telescope detects. There are several other sources that produce photons that contribute to a background signal. Sunlight does not extend to whole depth of the Mediterranean Sea. At the depths at which the experiments take place, sunlight is negligible. Radioactive potassium and bioluminescence are the most abundant sources of the background signal. It is important to study these light sources in order to classify and process the data.

4.1 Dark count

Photomultiplier tubes register hits when the voltage of the anode gets above a certain threshold as described in section 3.3. This can happen even if the PMT is not hit by a photon. This leads to an intrinsic dark count which can be caused by straggler electrons. The dark count is tested in the laboratory and has a rate of 600-1500 Hz9_.

4.2 Radioactive Potassium-40

Radioactive40_{K is by far the most abundant form of radioactive noise. There}

are other radioactive elements in the water, but these signals are so small that they are neglected. Potassium-40 emits β±radiation as well as electron capture which will result in the emission of a gamma ray. Most of the time (89.28%)

40_{K decays to} 40_{Ca with the emission of a β}− _{particle through the following}

reaction:

40 19K −→

40

20Ca + e− + ¯νe

The emitted electron will usually have a high enough energy to induce Cher-enkov radiation, which can be detected by the PMTs. Potassium-40 decay by electron capture happens for about 10.72% of the time.

(16)

40

19K + e− −→ 4018Ar + νe

Very rarely, about 0.001% of the time, potassium-40 decays by β+ emission through the following reaction:

40 19K −→ 40 18Ar + e + _{+ ν} e

4.3 Bioluminescence

The ability for an animal to emit light can be very helpful. Bioluminescence is a form of chemiluminescence, a chemical reaction that causes emission of light. Although bioluminescence does not occur frequently in organisms that live on land, deep in the water where light from the sun is reduced it is much more common. These animals also live at the bottom of the sea where the experiments take place and can cause interference with the detector. The light is produced by a chemical compound called luciferin. Luciferin is a generic term for compounds that make use of the enzyme luciferase that causes a redox reaction to form oxyluciferin and light.15 _{The exact structure of the molecule varies per animal}

so the color of the light may differ. There are multiple reasons how animals can benefit from bioluminescence, which include defence against predators, warning signals and attracting prey. Bursts of light that are recorded with the KM3NeT detector are caused by vertebrates and invertebrates and can cause the system to be overflowed by background signal making it unable to record distinct muon tracks. Bioluminescent bacteria occur both free and in symbiosis with other animals which allow them to make use of bioluminescence even if they do not have the specific organs for it. Free bacteria can group together in a cloud that creates an area that is slightly brighter than the background. If such a cloud covers a DOM, the PMTs in this DOM will have an increased base photon rate. The effects of these bioluminescent bacteria clouds will become visible when looking at the change in base rates over time.

(17)

Chapter 5

Data analysis

The goal of this research is to make an analysis of the background signal of the PPM-DU and PPM-DOM prototypes. The two prototypes differ in both structure and location, the data analysis performed on them is roughly equal. The data shown up until section 5.4 is only from the PPM-DU as this prototype best represents the final version of the neutrino telescope. Section 5.5 shows the differences of the two prototypes. The idea was to also analyse the data coming from the first string being deployed for the final version of KM3NeT. However, this was not possible due to the fact that the deployment was postponed several times. The results shown will be discussed in this section.

5.1 Signal fluctuation and standard deviation

The data in this section is taken from the PPM-DU on on October 9 2014. Figure 5.1 shows an example of the data when the rates are plotted as a function of time.

Figure 5.1: Frequency/counting rate per second plotted as a function of time. This example is from PMT 19 on DOM 0.

(18)

As seen in figure 5.1, the signal fluctuates around 5 kHz. These random fluctuations are due to background signal caused by radioactive40K and dark current. In order to differentiate between large and small peaks that will have a small and large chance to be caused by random fluctuations respectively, a histogram is made that correlates the frequencies and their occurrences in the run. A histogram of the same data as in figure 5.1 is shown in figure 5.2.

Figure 5.2: Histogram of the same data as in figure 5.1. Here the fre-quency is plotted on the x axis and their occurrence on the y axis. The blue lines belong to the histogram itself. The frequencies are normally distributed and can be fitted using a Gaussian function (red line).

This distribution of frequencies has a Gaussian shape. Fitting this histogram with a Gaussian function will yield important information about the signal such as the mean frequency and the amount of spread in the fluctuations. The function used for fitting the histogram is shown in equation 5.1,

f (x) = a · exp −(x − x0) 2 2σ2 (5.1) where a is the height of the peak, x0the position of the peak’s center and σ

the standard deviation (root mean square, RMS). The standard deviation of the fit gives information about the likelihood of a point to be a certain distance away from the center due to random fluctuations. Points far away from the center have a smaller chance to be due to random fluctuations than points close to the center. The surface under the Gaussian between +1σ and -1σ away from the center is equal to about 68.27% of the total surface of the Gaussian (from −∞

(19)

to +∞). This means that 100% - 68.27% = 31.73% of the points are outside of the 1σ border.

In this paper, the standard deviation is set at 4σ. If a data point is above this value, it is marked as an significant peak. Data files from the PPM-DU are approximately 30 minutes long. One data point contains the amount of photons that hit the PMT in 227_{ns (about 134 ms). The number of data points for one}

PMT in one data file is

1800

0.134 ≈ 1.34 · 10

4_. _(5.2)

This is in accordance with the data of figure 5.2 where there are 13542 data points. The fraction of a Gaussian surface between -4σ and +4σ is 1−6.334·10−5

of the total. The amount of peaks expected to be above or below -4σ and +4σ in a data set of 1.34·104_{data points is equal to}

6.334 · 10−5· 1.34 · 104_{≈ 0.849.} _(5.3)

Because the Gaussian function is symmetrical, this implies that in every data file ∼0.42 peaks are expected to be above 4σ that are due to random fluctuation. With this information, the graph in figure 5.1 can be analysed. The results of this analysis is shown in figure 5.3. The vertical blue lines shows where the signal exceeds 4σ.

Figure 5.3: Same data as in figure 5.1 analysed for peaks above 4σ. The horizontal green line corresponds to the peak in the fitted histogram shown in figure 5.2. The two horizontal red lines are the -4σ and +4σ values. Vertical blue lines mark points where the signal exceeds +4σ.

(20)

5.2 Peak classification

There are several reasons of how a peak can occur. This can be due to random fluctuations of the background signal as discussed above, a high-energy charged particle or bioluminescence. All these different causes have their own unique characteristics. The following section describes the different shapes of peaks and how they occur.

5.2.1 Peaks from bioluminescence

From all peaks in the data, the peaks that are caused by bioluminescence are the ones that are most easy to recognise. The light that is emitted by an an-imal takes some time to reach its maximum brightness and when this maximum brightness is reached, it takes an even longer time for the light source to disap-pear. Because this increase and decrease in brightness typically takes a couple of seconds, there are multiple data points where the event is recorded. A wide peak spanning more than one data point of 134 ms is the clearest sign of biolu-minescence, peaks caused by either a high-energy charged particle or by random fluctuations are only one data point wide. Figure 5.4 shows four different signs of bioluminescence.

Figure 5.4a shows a relatively small peak caused by bioluminescence. These peaks can either be caused by a creature that emits relatively little light or by one that emits a lot but is not directly in front of the PMT. This can be checked by looking at the signal in the other PMTs. Figure 5.4b shows a peak with about the same height as in picture 5.4a but it has a cut in the middle of the peak. This is caused by the fact that the detector will not transfer data if the signal gets too big, as described in section 3.5. The count rate at this point is too high for the system to process. Instead of a very large peak, the data now shows a cut in the signal. This is a clear sign for bioluminescence: such a high signal can not be caused by a charged lepton inducing Cherenkov radiation. Figure 5.4c is to illustrate that signals really do can reach a high frequency before being cut off by the system. The signal get cut off at around 1496 seconds but is turned on shortly after. The animal that caused the burst of light is still active and the light is recorded by the system. The gap in the signal of figure 5.4d is despite the lack of any high signal still due to bioluminescence. Note that the events in figure 5.4c and 5.4d happened at the same time but the data is taken from a different PMT. The event that caused the peak in figure 5.4c is the same that caused the gap in figure 5.4d. These are caused by larger animals that light up a big part of the DOM. The brighter an object becomes, the more likely it is for another PMT to catch photons emitted by it. Some PMTs will capture more light that others depending on the placement of the PMT relative to the light source. The PMTs that are directly aligned towards the source capture the most photons and cause the largest peaks, or if the peak gets too large, cause a big gap in the signal. By looking at which PMTs are hit, it is possible to estimate the light source.

(21)

(a) Small bioluminescence peak. (b) Peak with a dip in the middle.

(c) Very large peak with a dip. (d) Big gap with little or no peak.

Figure 5.4: Four different symptoms of bioluminescence on the signal.

5.2.2 Peaks from random fluctuation or Cherenkov

radi-ation

Peaks from bioluminescence are easy to recognise due to the length of the signal and the intensity. Peaks that are caused by random fluctuation or Cherenkov radiation are not easily distinguishable: both peaks are only one data point wide and can have a very low intensity, just above 4σ. Figure 5.5 shows a peak where the signal just exceeds the threshold of 4σ.

Whether the peak was caused by random fluctuations or by Cherenkov ra-diation is hard to determine. Random fluctuations will occur across every PMT equally so the distribution will be homogeneous from every direction. Muon tracks will not be homogeneous distributed, and therefore it is still possible to locate neutrino sources in the universe. Cherenkov radiation induced by high-energy muons are likely to hit more than just one PMT. So one other way to

(22)

Figure 5.5: Close up of figure 5.3. One single data point exceeding the value of 4σ.

make a distinction between random fluctuation and real muon tracks is to look only at peaks that occur at the same time on multiple PMTs. By counting only these hits, most of the fluctuations are filtered out.

5.3 Simultaneous hits

Light that is caused by physical events like bioluminescence or Cherenkov radi-ation will likely be recorded by more than one PMT. By looking at PMTs that are hit simultaneously it is possible to reconstruct the origin of the light source. A plot of simultaneous PMT hits is given in figure 5.6. The data is summed up of 25 data files ranging from October 9 2014 till December 4 2014. No more than one data file per day was taken.

Figure 5.6a shows two areas that are significantly larger than the rest. To interpret the results of this, it is necessary to know how the PMTs are numbered as described in section 3.4. The lower area of the graph is from all the PMTs in the bottom half of the DOM which contains PMT 0-19. The upper area corresponds to all simultaneous PMT hits in the upper half of the DOM which contains PMT 20-30. Of these areas, the number of hits is lower in the center than in the corner because the angle between the PMTs is larger near the equator of the DOM. The angle is the largest because every row contains 6 PMTs and the radius is the largest on the equator of the DOM. The reason for including figure 5.6b is to show that there are distinct lines that cause a pattern in the two areas that seem empty in figure 5.6a. There are four different lines parallel to the diagonal. These lines occur twice in the figure, on both sides of the diagonal, because the graph is symmetrical. The closest line to the diagonal

(23)

(a) Simultaneous PMT hits.

(b) Simultaneous PMT hits zoomed in on the z-axis.

Figure 5.6: Matrix of simultaneous PMT hits. Shows the number of times two PMTs are hit in the same timeslice (∼134 ms). Figure 5.6b has the same data as figure 5.6a but is zoomed in on the z-axis labeled ’Hits’ to show more structure.

is caused PMTs that have a difference in PMT numbering of 5, 6 or 7. This corresponds to PMTs that are above each other in a neighbouring row. The other three lines are caused by the same phenomenon but represent two, three or four rows in difference.

5.4 Change in mean signal

The mean signal of a PMT in one run is different for all PMTs. The PMTs are all slightly different which is also why the dark rates are different for every PMT. The values of the mean signal can also change over time. Figure 5.7 shows the change of the mean signal of every PMT per DOM over time. At first sight, it is clear that the mean signal is fairly constant over the course of two months. The change in mean signal on specific dates tends to be same across all PMTs. One clear example of this is in figure 5.7a where there is an increase in

(24)

mean signal on October 12 and 15 of almost every PMT compared to the mean signal in the days surrounding these dates. These rises in signal are also visible on DOM 1, but are less significant. On DOM 2, these rises are barely visible. Rises and drops in general seem to be strongest in DOM 0, less in DOM 1, and even less in DOM 2. Note that the sudden rise (as seen in DOM 0 and DOM 1) on November 17 is not as sudden as it seems. There is a gap of 20 days where there was no available data to be used. The mean values of DOM 2 are slightly higher than the other two DOMs, which was expected as described in section 3.1.

In figure 5.7 more is visible than just simple rises and drops in mean signal. Figure 5.7b has one PMT that seems to be out of the ordinary, a big rise in mean signal on October 16 and on October 21. The PMT in question (PMT 18) seems to have some irregular technical issues. The data from this PMT cannot be used for analysis in some data files as the noise is too high. On November 20, five PMTs on DOM 2 turned off and turned back on a day later, which is visible in figure 5.7c.

5.5 Differences between PPM-DU and PPM-DOM

The two prototypes share a lot of similarities. The biggest difference between the two is caused by the different locations of the prototypes and the different suspensions. The DOM suffers from more bioluminescence than the PPM-DU which results in whole data files that can not be analysed properly. Figure 5.8 shows a typical counting rate over time of one PMT of the PPM-DOM analogous to figure 5.3.

Figure 5.8 does not only show a higher amount of bioluminescence, the base rate of the signal is also higher than the base rate the PPM-DU. This will most likely be caused by the difference in location. The amount of bioluminescence dominates the signal and therefore it is hard/impossible to detect small signals from muon tracks. Figure 5.9 shows a histogram of the counting rates in figure 5.8. The shape of the histogram can not longer be approximated as a normal distribution. The Gaussian function to fit the histogram returns an inaccurate base rate and standard deviation. Figure 5.10 shows the change in mean signal of all the PMTs of the PPM-DOM. It clearly shows that some runs are more affected by bioluminescence than other runs. The runs on 5-21 and 5-22 are relatively clean runs with low bioluminescence. The mean values of the PMTs are close together. The runs in the period between 5-23 and 5-27 suffer from bioluminescence, the mean values of the PMTs are more separated and are higher in general. From the 10 data files used in analysing the PPM-DOM, 4 of the data files where not dominated by bioluminescence and could be accurately analysed. This is different from the PPM-DU where there were no runs that could not be analysed by the amount of bioluminescence. Keep in mind that the data used in this research is not representative of an average sample. On average, only a quarter of the data files are dominated by bioluminescence.16

(25)

(a) DOM 0

(b) DOM 1

(c) DOM 2

Figure 5.7: Mean signal change over time per DOM. Every line represents the mean signal of one single PMT. Date is noted mm-dd of the year 2014, e.g. 10-11 is October 11 2014.

(26)

Figure 5.8: Frequencies/counting rate per second plotted as a function of time. Multiple overlapping hits caused by bioluminescence dominate the signal. Data was taken on May 25 2013.

Figure 5.9: Frequencies of figure 5.8 plotted in a histogram. The shape of the histogram makes it harder for a normal distribution to fit this data than the data of figure 5.2 because of the amount of high frequencies. The result of the fit returns an inaccurate base rate and standard deviation.

Figure 5.11 shows the total number of hits per PMT of the PPM-DOM. There are five peaks evenly spaced by an average of 6 PMT numbers. This is a clear sign that the signal is not equally distributed around the DOM. One direction generates more hits than the others. Most likely this is caused by the suspension of the PPM-DOM (see figure 3.1). The exact cause of the signal is not clear. The increase of hits could be caused by animals interacting with the suspension causing more bioluminescent bursts from that specific direction or the increase of hits could be caused by the frame itself. However, the suspension of the PPM-DOM seems to be a limiting factor of the prototype.

(27)

Figure 5.10: Mean signal change over time of the PPM-DOM. Every line represents the mean signal of one single PMT. Date is noted mm-dd of the year 2013, e.g. 05-22 is May 22 2013.

Figure 5.11: Total number of hits per PMT. The number of hits is summed up of 10 data files ranging from May 21 2013 till May 31 2013. No more than one data file per day was taken.

(28)

Chapter 6

Conclusion

Detection of neutrinos from outer space can reveal new information that is un-obtainable with modern day telescopes. The KM3NeT neutrino telescope aims to pinpoint sources of neutrino production by detecting Cherenkov radiation induced by muons which are formed when neutrinos collide with atoms in the water or the Earths crust. The project is still in development although two prototypes are already taking data.

Results of the data analysis performed on these prototypes showed typical runs of the PPM-DU and PPM-DOM. It showed the difference of the effects of bioluminescence and muon tracks on the signal. The relative positioning of PMTs was identified by looking at different PMTs that where simultaneously hit. The PPM-DOM suffered more from bioluminescence than the PPM-DU which can partially be explained by the different suspension used. The base rate of the PMTs in the PPM-DOM is significantly higher than the base rate of the PPM-DU which is likely to be caused by the difference in location of the two prototypes.

The deployment of the first Detection Unit of the final KM3NeT neutrino telescope was planned during this thesis. Unfortunately, the deployment has been postponed and no data of the final Detection Unit could have been taken. The data of the two prototypes look promising and within a few months, the KM3NeT neutrino telescope will have started a new period in neutrino astro-nomy.

(29)

Chapter 7

Discussion

The software developed in this thesis made assumptions that may have affected some of the results. When a signal crossed the 4σ threshold is was registered as a peak. Afterwards, the peak had to be a certain amount of time in between the -4σ and +4σ levels before a new peak could be registered. This was done because the signal could fluctuate around the +4σ value and would register multiple hits while instead the signal was actually caused by one event of bioluminescence. This amount of time was set to 10 timeslices (≈1.34 seconds) while is seemed to deliver the best results. A disadvantage of this method is that after every peak, there is a period of at least 1.34 seconds where no other peaks can be registered on that specific PMT.

Histograms like in figure 5.9 are still fitted using a Gaussian function even though the data is not normally distributed. This could cause the standard deviation and base rate to be very inaccurate though these values are still used for finding peaks in the signal.

Recommended areas of future research include a better peak recognition algorithm, focusing primarily on runs that are dominated by bioluminescence. Also, more data files need to be analysed to improve the accuracy of figures like 5.6 and 5.11 and would also increase the accuracy of the fraction of runs that can’t be processed by the amount of bioluminescence.

(30)

Acknowledgements

First, I would really like to thank my supervisor Paul Kooijman for the enthu-siasm that he put into this project and for all the pleasant talks we had about it. I could see how much he cared for KM3NeT which really motivated me.

Special thanks goes to Robert Bormuth for helping me with Linux and all the other basic questions I bothered him with. I really appreciate the help that he has given me. I also would like to thank Karel Melis for showing me around Nikhef on the first day, letting me eat a pie with the chewiest crust I’ve ever seen, but most for helping me getting set up at Nikhef.

I really enjoyed working with the other bachelor students Bas, Maarten, Olmo and Matthijs. The pies we ate every week were delicious and I will really miss working with you guys.

(31)

Bibliography

[1] Raymond Davis Jr, Don S Harmer, and Kenneth C Hoffman. Search for neutrinos from the sun. Physical Review Letters, 20(21):1205, 1968. [2] RM Bionta, G Blewitt, et al. Observation of a neutrino burst in

coincid-ence with supernova 1987a in the large magellanic cloud. Physical Review Letters, 58(14):1494, 1987.

[3] M Ageron, JA Aguilar, et al. Antares: the first undersea neutrino telescope. Nuclear Instruments and Methods in Physics Research Section A: Acceler-ators, Spectrometers, Detectors and Associated Equipment, 656(1):11–38, 2011.

[4] Antonio Capone, Sebastiano Aiello, et al. Recent results and perspect-ives of the nemo project. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 602(1):47–53, 2009.

[5] Petros A Rapidis, NESTOR collaboration, et al. The nestor underwater neutrino telescope project. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 602(1):54–57, 2009.

[6] Albrecht Karle, J Ahrens, et al. Icecube—the next generation neutrino telescope at the south pole. Nuclear Physics B-Proceedings Supplements, 118:388–395, 2003.

[7] JJ Shea. A plasma formulary for physics, technology, and astrophysics [book review]. IEEE Electrical Insulation Magazine, 1(19):46,48, 2003. [8] David Jeffrey Griffiths and Reed College. Introduction to electrodynamics,

volume 3. prentice Hall Upper Saddle River, NJ, 1999.

[9] D. F. E. Samtleben S. Biagi, A. Creusot. The prototype detection unit of the km3net neutrino telescope. 2015.

[10] KM3NeT Collaboration et al. Expansion cone for the 3-inch pmts of the km3net optical modules. Journal of Instrumentation, 8(03):T03006, 2013.

(32)

[11] Km3net sees first light. http://www.km3net.org/images/ press-releases/2013-06-03.pdf, June 2013. Press release.

[12] R Kollath. Sekund¨arelektronen-emission fester k¨orper bei bestrahlung mit elektronen. In Electron-Emission Gas Discharges I/Elektronen-Emission Gasentladungen I, pages 232–303. Springer, 1956.

[13] A. Creusot R. Bormuth. Km3net data conventions, December 2014. Un-finished at time of consultation.

[14] Km3net: Technical design report, 2010.

[15] J Woodland Hastings. Chemistries and colors of bioluminescent reactions: a review. Gene, 173(1):5–11, 1996.

[16] No official records exist. Estimates where obtained in a conversation with prof. dr. P.M. Kooijman.

Prototype Data Analysis of the KM3NeT Neutrino Telescope

University of Amsterdam

Bachelor Thesis