Bioluminescence at the KM3NeT neutrino telescope

Bioluminescence at the KM3NeT

neutrino telescope

THESIS

submitted in partial fulfillment of the requirements for the degree of

BACHELOR OF SCIENCE

in

PHYSICS

Author : Karlijn Kruiswijk

Student ID : 1854380

Bioluminescence at the KM3NeT

neutrino telescope

Karlijn Kruiswijk

Huygens-Kamerlingh Onnes Laboratory, Leiden University P.O. Box 9500, 2300 RA Leiden, The Netherlands

June 28, 2019

Abstract

The KM3NeT Neutrino telescope is a telescope in the Mediterranean sea researching neutrino events to study both neutrino properties and neutrino sources in the universe. While searching for neutrinos, the detector observes noise from bioluminescence and40K decay as well. The

40_{K noise is constant and homogeneous, and the bioluminescence}

consists of high peaks in the signal (bursts) and is periodic in nature. In this research the periodicity of the bioluminescence was researched for the different detectors, together with the general noise characteristics of

the detectors and their effect on the muon track reconstruction. The periodicity was found to be consistent with the current around the detectors and the inertial waves caused by the Coriolis force, though

Chapter

1

Introduction

Nowadays the Mediterranean Sea is house not only to many different fish and other sea creatures, but to neutrino telescopes as well. The neutrino telescopes of KM3NeT are looking at neutrinos to research cosmic events and to investigate the neutrino mass ordering. To do this research the noise present around the detectors needs to be carefully considered, which will be done in this research. Part of the noise is caused by bioluminescent creatures. These creatures are said to give a signal periodic in nature, but this had never been found on the neutrino detectors themselves. This re-search aims to to characterise the noise on the detector signal. This also includes investigating the periodic signal of the bioluminescent creatures and their effect on neutrino event reconstruction.

1.1

KM3NeT

KM3NeT is a neutrino telescope, or rather a set of neutrino telescopes, consisting of detectors at different locations deep in the Mediterranean sea.The Goal of KM3NeT is to observe cosmic neutrinos and discover the mass ordering of the neutrinos. This is done by building a detector of a few cubic kilometres capable of detecting and processing neutrino events through their resulting Cherenkov radiation. Right now there are two lo-cations of detector sites; of the coast of Sicily in Italy and of the coast of Toulon in France. At the Italian site the ARCA (Astroparticle Research with Cosmics in the Abyss) telescope is placed, and at the France site the ORCA (Oscillation Research with Cosmics in the Abyss) telescope is lo-cated close the KM3NeT’s predecessor ANTARES.

6 Introduction

therefore focuses on the atmospheric neutrinos. These neutrinos can hold the key to the ordering of the neutrino masses[1], which is still unknown.

ARCA instead focuses on the cosmic neutrinos to try and find their origin, energy spectrum, and flavours. The ARCA detector is able to map 87% of the sky including the Galactic centre, where it might help discover new phenomena and give an new way to look at astronomical sources, with not just photons, but neutrinos as well[1].

Lastly it is also important to mention ANTARES, the predecessor of ARCA and ORCA designed as a first step towards a km3 scale detector. This detector searches for cosmic neutrinos as well[2].

1.1.1

Detection principle

Neutrinos are famously hard to detect on their own, as they have a very small cross-section and therefore almost never react with other particles. This makes them very interesting to study, as neutrinos allow for observ-ing far away sources that would otherwise be obstructed, but also very challenging. The small cross-section is usually accounted for by making a very large detector area, so the chances of neutrinos reacting is larger. In the case of KM3NeT the large detector area is a part of the Mediter-ranean sea, where the detectors are planned to cover several kilometres of seawater[1].

The reactions that allow us to actually detect the neutrinos are scat-tering interactions through the weak force. These interactions can be cate-gorised as either charged current events or neutral current events. If a neu-trino collides with a nucleus, this is often a neutral current event, where the neutrino is able to continue on while the nucleus breaks apart and causes a hadronic shower. Charged current events contain a collision with a nucleus as well, but have the neutrino change into a different lepton. For an electron neutrino this means that apart from the hadronic shower, the new electron can also cause an electromagnetic shower. For muon neu-trinos the new muon causes a muon track,which is much longer than the hadronic shower and can emit Cherenkov radiation. The tau-muon cre-ates a tauon which can travel a short time before decaying. This decay then causes a second shower to occur behind the first one and can create a muon as well [2].

Of these reactions, the electromagnetic showers produce photons on their own while hadronic showers and muon tracks can do this through a specific kind of radiation. If a charged particle resulting from a neutrino interaction has an energy above it Cherenkov limit in the sea water, some-6

1.1 KM3NeT 7

thing special can occur. The velocity of the charged particle is then higher than that of light in the sea water, resulting in the production of Cherenkov radiation. This Cherenkov radiation has a cone shape with a specific an-gle depending on the environments refractive index (n) and the particles velocity compared to the speed of light in vacuum (β), namely [2]

θC =arccos  1 nβ  . (1.1)

These cones of light can then be observed by the detectors, after which the track if the particles can be reconstructed. The detectors best observe these events at wavelengths between 300 to 600nm, where the water is most transparent [2].

To make the neutrino event reconstruction more manageable, the de-tector splits these different events up into showers and tracks. Showers contain neutral current and charged current electron and tauon neutrino events,while tracks contain the muon events from charged current muon and tauon neutrino events [3].

The track events are particularly interesting as the muons give a clear track of the muon path, which in turn gives an indication of the neutrino path as well. Reconstructing these events to their tracks is possible by comparing the data of several different detectors in the telescope, and are what makes KM3NeT a neutrino telescope instead of a neutrino detector.

1.1.2

Detector Setup

At both the ARCA and ORCA detectors, the general setup of the detectors is the same. The actual detectors that observe the Cherenkov radiation are 3 inch diameter photo-multiplier tubes (PMTs), which are capable of detecting single photons. To properly capture the Cherenkov radiation, these PMTs have the highest quantum efficiency between 300 and 550nm [4]. 31 of these PMTs are placed inside of a digital optical module (DOM) in rings of different elevations to maximise the visual area of the detectors, as can be seen in figure 1.1.1. To compare, the older neutrino detector named ANTARES is made up of PMTs as well, though these are a bit larger and not placed inside DOMs. These PMT have a 10 inch diameter instead of ARCA’s and ORCA’s 3 inch. The PMTs each have their own housing and are hung in sets of 3 over 25 floors.

The DOMs are placed in the water on a string, with on the underside an anchor and on top a buoy to keep the 18 DOMs in place. The string also holds the cabling that transports the data of the detector units to an

8 Introduction

Figure 1.1.1: A DOM on the string with the PMTs placed inside it. Courtesy KM3NeT.

onshore location to be processed. One such string is called a detector unit (DU), and the plan is for several of such DUs to form the detector infras-tructure, as can be seen in figure 1.1.2. The intent is to make three detector blocks, with detector areas larger than a square kilometer and consisting of 115 DUs. Two of these blocks will be placed are ARCA and one AT ORCA [3]. Presently both ARCA and ORCA are working with single strings.

Figure 1.1.2:The Setup of the different DUs. Copyright Edward Berbee/Nikhef

To be able to detect neutrinos of very low energies (GeV), the ORCA detectors are spaced closer together than the ARCA detectors. The dis-tance between ORCA DOMs is 9m vertically and horizontally 20m, while the vertical distance between ARCA DOMs is 36m and the horizontal distance is 90m[3]. Another difference is that the ARCA detectors are placed deeper than the ORCA detectors, with ARCA at 3500m deep versus ORCA’s 2450m [1].

8

1.2 Noise 9 0 50 100 150 200 250 300 350 400 3 10 × Time (s) 0 20 40 60 80 100 120 140 3 10 × Hit rate (Hz) L1 Signal DOM 1 Full signal Base rate L1 Signal DOM 1

Figure 1.2.1:The L1 time stream of DOM 1 during the 2017 run. Both the hit rate and the base rate are visible.

1.2

Noise

As the detectors are placed in the sea, they will always pick up other noise signals among the neutrino events. In the hit rate over time of the detec-tors the noise seems to be almost the sole contributor, as seen in figure 1.2.1. The two most important noise sources are in this case the40K decay in seawater and the bioluminescent signals of different deep sea creatures.

40_{K decay creates a constant amount of photons, with a Gaussian}

distri-bution. Therefore is average hit rate induced by 40K constant in time as well and can be defined as the base rate. This baserate is useful for com-paring different DOMs as the decay is homogeneous. Everything higher than this Gaussian noise can be called a burst and is caused by biolumi-nescence. These bursts are not constant and can have different hit rates, making them more difficult to quantify. However there different methods possible for doing so.

The neutrino events can be distinguished from the noise of biolumi-nescence and40K by comparing the signal of all PMTs. One40K or biolu-minescence event happens in a relatively small detector volume, and few PMTs will catch the same event. However, neutrino tracks are made of fast charged particles, and will therefore create hits on many PMTs in a short time. If two PMTs get a hit within a certain time interval, this can be defined as a two-fold coincidence of an event. These two-fold coincide rates can then be called L1 data, and while40K is able to get even six-fold coincidences, higher coincidence rates happen with muons only [5].

10 Introduction

1.2.1

K Decay

Potassium is only one of the many minerals present in seawater, but is nevertheless very important for the detectors. Especially the isotope40K is noteworthy, which makes up 0.0117% of the potassium in seawater. This isotope with a half-life of 1.25·109 years has a 89.52% possibility of beta decaying into40Ca and a 10.48% of decaying 40Ar through electron capture[6]. The beta decay into40Ca creates high energy electrons, which are able to produce Cherenkov radiation. This radiation is picked up by the PMTs of the detector, and causes the steady base rate as seen in figure 1.2.1.

As said before, the signal from the40K decays are not just noise. As the sea is very homogeneous, the concentration of the potassium is the same everywhere, which gives a clear baseline for the detector efficiencies by comparing this signal. Furthermore, as the DOM is able to detect mul-tiple photons from a single 40K decay, it is an excellent process for time calibrations between PMTs in a single DOM[5].

1.2.2

Bioluminescence

In the deep sea, bioluminescence is very common, with estimations that 90% of all deep sea creatures are able to use it for varying purposes [7]. Bioluminescence is both used by bacteria and plankton.

The bioluminescent bacteria can glow continuously for up to days un-der specific conditions, and cause bioluminescence blooms [8] where whole parts of the sea light up for a long period. The bioluminescence glow was found to be stronger with increasing current speed and with increasing salinity and temperature (dense water). The increase of salinity and tem-perature are caused by sinking cold surface water with warmer deeper water. The renewal of this deep and dense water fuels the local ecosys-tem and can cause bioluminescence blooms [8]. These blooms are a steady glow that can affect the mean rates of the detectors.

When the bioluminescent (mega)plankton hit (the turbulent water around) the detectors, they can trigger bioluminescent behaviour, causing a sud-den burst of light[7]. This signal is not a steady glow like the bacteria or

40_{K radiation, but does seem to have a certain rhythm in its occurrences.}

These bursts can be seen in figure 1.2.1 as well, as the high peaks of the signal that stand out above the base rate.

The amount of Bioluminescence differs from location to location, with in the Mediterranean sea more activity towards western waters and shal-low seabeds[7]. This means ORCA is more sensitive to bioluminescent 10

1.3 Goal 11

noise than ARCA, as ARCA is placed lower in the sea and in the east as well.

The bioluminescent signals are mostly dependent on the current speed and direction [9], so it would seem logical the signal is similar to the cur-rent which is periodic in nature. In the Mediterranean this periodicity is not caused by the tides, but is mostly due to the inertial waves inside of the sea itself. These waves are most affected by the Coriolis force. As the wa-ter is moving and the earth rotates, the wawa-ter can feel a Coriolis force. This force can cause currents to change and rotate, causing changes in the over-all current. The strength and therefore the period of this force depends on the latitude, with the following relation:

ω =2Ω sin λ. (1.2)

With ω the angular frequency at the latitude, Ω the angular frequency of the earth, and λ the latitude. Rewriting this formula for the period gives [2]

P= 12h

sin λ. (1.3)

From this the periodicities of the detectors can be calculated. ARCA, with a latitude of 36◦16’is expected to have a periodicity of 20.3 hours. ORCA and ANTARES are both on the same latitude of 42◦48’, and therefore share the expected period of 17.7 hours.

While a certain periodicity of the Bioluminescence has in some places been found to be that of inertial fluctuations [10] (with possibly a tidal fluctuation added [2]), these have not been studied with the help of the neutrino detectors yet.

1.3

Goal

The goal of this research is to characterise the effects of40K and Biolumi-nescent noise sources in the data. This means that the sub-goals consist of:

• Defining the differences in noise spectra across the different KM3NeT telescopes

• Finding the periodicity of the bioluminescent noise

• Finding the effect of bioluminescence on the muon event reconstruc-tion

Chapter

2

Data and general method

For this research long periods of data from the neutrino detectors was used to characterise the noise and its possible effects.

Data from ORCA, ARCA and from ANTARES was handled, together with current MII measurements taken close to ORCA. Here the calibration runs and broken runs were filtered out and filled with a fake constant sig-nal. The data of ORCA and ARCA consisted of the hit rate of the DOMs over time.

Of the detector data, 2 periods of measurements were used; from the 16th of October till the 13th of December in 2017, and from the 21st of February till the 10th of March in 2019. The first period contained data of all three detectors, while the second period contained only data of ORCA. ORCA had different DU’s used over these two periods as well(DU2 for 2017 and DU1 for 2019). The ORCA and ARCA data during the 2017 runs consisted of different runs of L0 and L1 hit rates over time of the 18 DOMs while the 2019 runs consist of only L1 hit rates (though ARCA L0 runs only cover a small part of the L1 runs used).

Here an L1 hit rate means that there must be a second hit of a photon on a different PMTs in a certain time frame (10ns for the 1st part of 2017, after which it was halfway switched to 25ns) for the hit to be registered. This also means there is some form of ”double counting” for the hit rates of the DOMs, as there many PMTs are contributing to the same L1 rate. The hit of a PMT can then be used to register a hit of a second PMT, while the second PMT is used to register the hit of the first. The DOM L1 hit rate then gets these hits from both PMTs, while they should only contribute one hit in total. In comparison the L0 has no coincidence rates, as it simply consists of the hit rates of the PMTs and can therefore be considered as raw data. This means that part of the data is not completely raw data, but the

14 Data and general method

40_{K and bioluminescent noise should still be visible in the L1 data.}

The ANTARES data consists of the base rate and burstfraction of one of three PMTs on a certain floor, with ANTARES containing 25 floors. This data is from the 1st of October 2017 till the 30th of December 2017, and thus overlaps with part of the ORCA data.

To see if the periodicity of the bioluminescence is in fact caused by the current, current measurements near ORCA were taken for the first period as well. This data consisted of information of several different variables, of which the horizontal current velocity and the pressure were used in this research.

To properly evaluate the effect of the bioluminescent noise on the track reconstruction, both track reconstruction files of the same runs of the first period of ORCA measurements were used. These track reconstruction files contained information about the starting position, direction, and quality of the muon events. Furthermore were Monte Carlo simulations of the re-construction of these runs utilised to account for different settings, like the run time of a single run or the bioluminescent noise present. The Monte Carlo simulations have the same hits as the data, and then simulate the background and signal, which should match the data. These simulations are based on a parametric model of muon distributions in the sea [11].

The main variables that were researched were the zenith angle Theta and the quality factor Q. The quality factor gives an indication of good-ness of fit with a higher quality factor indicating a better fit. The quality factor is defined as the likelihood for the track hypothesis. This means that it is the sum of the contribution of all hits, and is not independent of the number of hits in the fit. The zenith angle Theta is important as it allows for distinguishing upwards and downwards muons and therefore for up-wards and downup-wards neutrinos, whose differences give insight to the hierarchy problem. Therefore Theta needs good resolution and no bias.

2.1

Periodogram

To find the periodicity of a signal in the data different methods could be applied. For example fitting the signal to a sine function is possible if only one frequency is expected. Furthermore is Fourier transforming the signal to find its frequency components a possibility as well, but requires a reasonable smooth and continuous signal (without large gaps). Lastly a periodogram can be made, where the variances of a frequency show whether it is significant or not [12]. For the last method a longer period of data is needed for averaging compared to the Fourier transform, but 14

2.1 Periodogram 15

the data can be less smooth and gaps are no large problem. Therefore the periodogram was utilised in this research.

This Enright periodogram is often used by biologists to find rhythms of animals, especially rhythms with periods of about a day. The method is very useful in that it does not need extra smoothing to function and can deal with partially missing data[12], though the data taking does need to be at a constant frequency.

This periodogram should not be confused with the astronomical defini-tion of a periodogram, which is simply the Fourier transform of the signal squared. The Enright periodogram does not use the Fourier transform, and is useful for testing single or multiple periods for if they are signif-icantly present in the signal. This is done by making an average period of the data points, and comparing the variance of this period to the over-all variance of over-all the data points. If the periodic component is present in the data, the period is clearly found in the average period, with variance higher than that of the whole signal. If the period is not present in the data, the data points will average to an average of the total signal with a variance much lower than that of the total signal.

Making the average period can be seen as filling a matrix with as many columns as the period has data points. The average period is then found by averaging over the columns. For a signal with N data points the data points of the average period (Xc,p) look like

Xc,p= 1 K K

∑

Where X is the discrete signal, p is the number of data points in the ex-pected period, c is the column number,and K = N/p the number of rows in the signal matrix.

The data points of the average period can then be compared to all the data points of the signal to see if the period is significantly present in the data. This is done through calculating the quality factor (Qp) of the period;

Qp =

K∗N∑_cp₌₁(Xc,p−X)2

∑N

i=1.(Xi−X)2

(2.2) Here X is the average of the total signal.

The quality factor Qp conforms to a χ2distribution of p−1 degrees of

freedom if the signal consists of random noise [12]. However, if there is a periodic component to the signal, the quality factor of this period will

16 Data and general method

Figure 2.1.1: A periodogram of an 18h sine function. On the x-axis is the period in hours, and on the y-axis the variance of the period.

be much higher than expected for a χ2 distribution, through which the periodicity in the signal can be detected.

In this research the Enright periodogram comprises of the quality fac-tors of periods between 10 and 28 hours, with half-hour increments. To see if any of these periods were present in the signal, the 97.5% and 99% intervals of a χ2 distribution were used. These intervals were chosen as they represent the 2 and 3 sigma measurements respectively. The peri-odogram can be used on preferably 10 days worth of data or more, but is also applicable up to even 5 days [12]. Therefore in this research, both the whole time stream (>400 hours) and 200 hour periods are checked for periodicity.

An example of the Enright periodogram can be seen in figure 2.1.1. Here a periodogram of of an 18 hour sine function is shown. Seen is that the 18 hour peak is far above the rest of the signal and far above the 99.9%

χ2 interval, indicating a very high quality factor which does agree with

the sinusoidal function. However there is a side lobe at 18.5 hours which also seems significant, but is in fact some leftover from the 18h periodicity. This can be made clear by looking at a smaller scale (15 min or shorter increments) or can just be identified by eye.

With this data and the periodogram the following characteristics were observed. First the noise spectrum was researched in chapter 3. The base 16

2.1 Periodogram 17

rates were calculated in section 3.2, which in turn was used to calculate the burstfraction and burst rates in section 3.3 and 3.4. The noise on single PMTs is discussed in section 3.5, while the periodicity of the burst rate is observed in chapter 4. Lastly the effect of the bioluminescent bursts are discussed in section 5.

Chapter

3

Noise Spectrum

Of all three detectors the data was observed to find the base rate, burst rate and burstfraction and their evolution over time. These variables give insight to the two most prominent noise sources, the 40K decay and the bioluminescent creatures.

3.1

Methods

As seen in figure 3.1.1, the hit rate of a DOM contains the Gaussian noise of

40_{K decay and some non Gaussian noise from the bioluminescence. This}

bioluminescent noise always has a higher hit rate than the40K noise and is less frequent. However, to properly analyse both components, they need to be separated from each other. This is done by calculating the base rate and the burst rate. As mentioned before the base rate is the mean of the Gaussian noise of 40K . This means the base rate of a signal can be eas-ily calculated by filling a histogram with the signal and fitting a Gaus-sian function [5]. However, as our signal is several days long and con-tains many different observation runs the base rate can slightly vary over time depending on changes in run setup or by bioluminescent interfer-ence. Therefore the base rate is calculated over 30 min intervals. These base rates are able to make statements about the detectors and their com-ponents when compared, though it should be noted that the baserate of ANTARES should be different as it has a larger PMT diameter.

To observe the bursts of bioluminescence is a somewhat more exten-sive process, as the bursts are everything but constant in timing and can overlap. Therefore there are different methods of calculating said bursts.

20 Noise Spectrum 100 150 200 250 300 350 400 450 500 550 600 hit rate (kHz) 0 1000 2000 3000 4000 5000 3 10 × Histogram Signal

Figure 3.1.1:A histogram of the hit rate of one of the DOMs L0 data.

the signal is higher than 120% of the base rate, this can be categorised as a burst. The burstfraction was then defined to be the fraction of a time frame that had a hit rate higher than 120% of the burst rate.

However, this method does not differentiate between many small bursts or one large burst and is sometimes unable to find smaller bursts. This is why a different method is also used in this research, which should be able to count the number of bursts in a time frame as well.

In this research the burst rate is defined with the derivative of the sig-nal. This is because a sharp rise in hits is very characteristic in the signal of a burst. For a burst the derivative of the signal passes a certain thresh-old, by which it can be identified. Here two burst that happen at the exact same time can of course still be missed, but partially overlapping bursts can be differentiated [2].

For this definition, the threshold was defined differently for the L0 and L1 data. For the L1 data the threshold was T1 = 250·

√

base rate , and for the L0 data the threshold was T0 = 80·

√

base rate. These threshold differentiated the bursts from the40K noise, as seen in fig 3.1.2. The high bursts are found high in their derivative as well, while large variances that are not bursts has less high peaks in their derivative, which do not make it past the threshold.

In this research the base rate was used for the ORCA and ARCA data to look for the periodicity of bioluminescence. The burstfraction was used 20

3.2 Base rates 21 5000 10000 15000 20000 25000 30000 time (s) 2000 − 1000 − 0 1000 2000 3000

derivative hit rate(kHz/s)

Signal

baserate

derivative signal

Figure 3.1.2:A graph of the derivative of the signal, and the threshold for L0 data.

in the ANTARES data and to compare burstfractions of ORCA and ARCA to ANTARES, where the time frame over which they calculate the base rate and burstfraction is 15 minutes. This is because the ANTARES data was already processed and just the base rate and burstfraction is given. For proper comparison the 15 minute interval were placed upon the burst rates and burstfractions of the other data as well.

3.2

Base rates

Fitting the data histogram with a Gaussian gives the base rate of the data. The base rate is necessary to derive the burst rate, but it has some interest-ing qualities on its own as well.

3.2.1

ORCA 2017 runs

For the 2017 runs the base rates had some special behaviour. While figures 3.2.1 and 3.2.2 show an average baserate of 6kHz and 230kHz respectively, the first six DOMs have significantly lower mean base rates than the others in both L1 and L0 data. This is caused by the lower six DOMs having a lower efficiency. Luckily, all base rates do exhibit Gaussian behaviour, with no large skewness towards a particular side, as can be seen in figure

22 Noise Spectrum

3.2.3. This would be consistent with the idea that the base rate is mainly consistent of40K noise, which should be Gaussian. The L0 data in figure 3.2.2 shows some intervals of higher base rates, which is most likely a residue from the bioluminescence, which has not completely filtered out. This could happen if a multitude of bursts occur at the same time and obscures the40K noise.

2 4 6 8 10 12 14 Base rate 0 200 400 600 800 1000 1200 time (h) 2 4 6 8 10 12 14 16 18 DOM Base rate

Figure 3.2.1: The time evolution of the base rate of the 2017 ORCA L1 data, with on the x-axis the time in hours and on the y-axis the DOM number. The colour indicates the base rate in kHz.

50 100 150 200 250 300 350 Base rate 0 100 200 300 400 500 600 700 800 900 time (h) 2 4 6 8 10 12 14 16 18 DOM Base rate

Figure 3.2.2: The time evolution of the base rate of the 2017 ORCA L0 data. The colour indicates the base rate in kHz. The time in hours is on the x-axis and the DOM number on the y-axis.

50 100 150 200 250 300 350 400 base rates (kHz) 2 4 6 8 10 12 14 16 18 DOM 0 200 400 600 800 1000

Histogram base rates

Figure 3.2.3: The histogram of the base rates of all ORCA DOM’s L0 data. The lower 6 DOMs have lower means. The higher DOMs have approximately equal baserates

In figure 3.2.1 a clear shift in values is visible from 1.5KHz to 6KHz, which cannot be seen in the L0 data of figure 3.2.2. This was because of a 22

3.2 Base rates 23 10 12 14 16 18 20 22 24 26_{Period (h)}28 0 10 20 30 40 50 60 70 80 DOM1 DOM2 DOM3 DOM4 DOM5 DOM6 DOM7 DOM8 DOM9 95% interval periodogram 10 12 14 16 18 20 22 24 26 28 Period (h) 10 20 30 40 50 60 70 80 DOM10DOM11 DOM12 DOM13 DOM14 DOM15 DOM16 DOM17 DOM18 95% interval periodogram

Figure 3.2.4: The periodogram of the base rate of the 2017 ORCA L1 runs. All DOMs and the 95% χ2interval are visible

change in routine for gathering the L1 data. As mentioned in section 2, the L1 data is made up of double hits. For an L1 hit to be registered, there need to be two hits on two PMTs within a certain time frame. The shift in base rates for L1 was caused by a change in that time frame. Namely, the first half of the data was taken with a time window of 10ns to get a hit, while the second half was taken with a time window of 25ns. This changed the rate of the L1 data, while the L0 data was unaffected.

The base rate is supposed to consist only of the noise from 40K and is therefore not supposed to be periodic in nature. This can be checked with a periodogram as explained in section2.1. Figure 3.2.4 shows that there is indeed no periodicity, as no peak surpasses the χ2interval.

3.2.2

ORCA 2019 runs

For L1 data of the 2019 runs figure 3.2.5 shows the base rate had been around 6.5kHz. However, here the base rates were not equal everywhere. figure 3.2.6 shows DOMs 12,16, and 17 appear to be somewhat lower than the average, while DOM 2, 4 and 5 seem to be higher than average. The

24 Noise Spectrum 0 50 100 150 200 250 300 350 400 time (h) 2 4 6 8 10 12 14 16 18 DOM 5 6 7 8 9 10 Base rate

Figure 3.2.5: A time evolution of all the base rates from the 2019 ORCA runs. On the x-axis is the time in hours, on the y-axis the DOM number and the colour indicates the base rate in kHz.

0 50 100 150 200 250 300

Histogram base rate

0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 Base rate (Hz) 2 4 6 8 10 12 14 16 18 DOM

Histogram base rate

Figure 3.2.6: A histogram of all the base rates from the 2019 ORCA runs. The x-axis indicates the base rate in Hz, while the y-axis shows the DOMs.

base rates here also follow a Gaussian noise just like the 2017 data, as is expected for40K noise. However the Gaussian distribution does appear to show a certain skewness towards higher values which can come from a bi-oluminescence contribution. This is also visible in the small changes in the base rate over time in figure 3.2.5, where the bioluminescent noise must be contributing as the40K noise is supposed to be reasonably constant.

Even if the baserate seemed to be somewhat sensitive to the biolumi-nescent noise, the base rate has no specific periodicity for the 2019 runs either, as can be seen from figure 3.2.7. No signal is higher than the 95%

χ2interval.

3.2.3

ARCA

Some data was used from the ARCA detector as well, to confirm previous studies and to compare to the ORCA detectors from before. Both L1 and L0 data was studied, though the L0 data covered only part of the L1 data. Figure 3.2.8 shows that the base rate of the different DOMs in the ARCA detector are not all similar to each other. DOM 13 and 17 are lower than the average at 4kHz, while DOM 4,5,7, and 16 are higher than average at 5.5kHz. In figure 3.2.9 this is visible as well, though to a lesser extend. Here the difference appears to be between 180 kHz and 205kHz. The av-erages are 5kHz for the L1 data and 200 kHz for the L0 data. Fortunately the base rates are stable over time for each DOM, which means there is probably a difference in efficiency between the DOMs.

24

3.2 Base rates 25 10 12 14 16 18 20 22 24 26period(h) 28 0 10 20 30 40 50 60 70 80 dom1dom2 dom3 dom4 dom5 dom6 dom7 dom8 dom9

95% interval Chi squared

periodogram

95% interval Chi squared

periodogram

Figure 3.2.7: The periodogram of the different base rates of the ORCA 2019 L1 data. 2500 3000 3500 4000 4500 5000 5500 Base rate 0 50 100 150 200 250 time (h) 2 4 6 8 10 12 14 16 18 DOM Base rate

Figure 3.2.8: The base rates of the different DOMs for the ARCA L1 data. On the x-axis is the time in hours displayed and on the y-axis the DOM number. 0 50 100 150 200 250 300 350 Base rate 0 20 40 60 80 100 time (h) 2 4 6 8 10 12 14 16 18 DOM Base rate

Figure 3.2.9: The base rates of the different DOMs for the ARCA L0 data. The x-axis dis-plays the time in hours and the y-axis the DOM number

26 Noise Spectrum 0 50 100 150 200 250 300 350 400 Base rate 0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 time (h) 5 10 15 20 25 floor Base rate

Figure 3.2.10:The base rates of a PMT on differ-ent floors of ANTARES in kHz. On the x-axis is the time in hours displayed and on the y-axis the floor number. The colour indicates the base rate in kHz. If multiple PMTs are available on a floor, the PMT highest in number is shown. The black lines signify the start and end of the ORCA data.

0 1000 2000 3000 4000 5000 6000 7000 8000

Histogram base rate

0 50 100 150 200 250 300 base rate (kHz) 5 10 15 20 25 floor

Histogram base rate

Figure 3.2.11: The 2D histogram of the ANTARES base rates, with on the y-axis the floor number and on the x-axis the base rate in kHz.

3.2.4

ANTARES

For the ANTARES data, the base rate was already calculated, and was relatively constant around 50 kHz, with some small peaks (figure 3.2.10). The small peaks are outside of the normal Gaussian shape of the40K noise, as is visible in figure 3.2.11. This might indicate that bioluminescence has some component in this base rate. Noticeable as well is that there are some white periods, where the data is missing.

The base rate of the ANTARES data is not completely constant, and the periodogram reflects this (figure 3.2.12). While the signal from most PMTs shows no periodic component, some do have significant peaks. These peaks can just be caused as random noise, as section 2.1 explained that there is a 5% chance random noise added up is higher than the 95% in-terval. This is a lot of data as well, so there is a large chance on outliers. However it is possible that some peaks are in fact caused by biolumines-cent periodicity, especially the high peaks of floor 19 at 21 hours and 23 at 25 hours, which even make the 99.9% χ2interval.

3.2.5

Comparison

While all these different runs used different detectors (even the 2017 and 2019 ORCA runs used a different DU), they can still be compared as they are based on the same principle. Noticeable is that the L0 data has much 26

3.2 Base rates 27 10 12 14 16 18 20 22 24 26 period (h) 28 0 20 40 60 80 100 120 97.5% Chi2 interval 99.9% Chi2 interval floor 1 floor 3 floor 4 floor 5 floor 6 floor 7 floor 8

Periodogram base rate

10 12 14 16 18 20 22 24 26 period (h) 28 20 40 60 80 100 120 97.5% Chi2 interval 99.9% Chi2 interval floor 9 floor 10 floor 11 floor 12 floor 13 floor 14 floor 15 floor 16 floor 17

Periodogram base rate

10 12 14 16 18 20 22 24 26 period (h) 28 0 20 40 60 80 100 120 97.5% Chi2 interval 99.9% Chi2 interval floor 18 floor 19 floor 20 floor 21 floor 22 floor 23 floor 24 floor 25

Periodogram base rate

Figure 3.2.12: Periodograms of the base rates from the ANTARES PMTs, sorted on floor. The χ2_{intervals are visible as well.}

28 Noise Spectrum

higher base rates than the L1 data, for both ORCA and ARCA. This is because the L1 data is not raw data, and needs two photons in a short time frame to register a hit, while the L0 data registers all photons.

Furthermore the L0 base rates differ between detectors. For the L0 base rates the ANTARES base rate is smallest at 50 kHz, though this is mostly because it is the noise of a single PMT, and not 31 PMTs. Second up is the ARCA data at 200 kHz, with ORCA having the highest L0 baserate at 230 kHz. The difference between ARCA and ORCA is small, as they both use the same kind of DOMs. If we calculate these average base rates of ARCA and ORCA to the base rates per PMT we get 6.5kHz and 7.4kHz respec-tively. This is a great deal lower than that of ANTARES, but seems logical if the size of PMTs is taken into account. Whereas ARCA and ORCA have a PMT diameter of 3 inches, ANTARES has a PMT diameter of 10 inches.

The L1 base rates are not equal between the detectors either. The first part of the ORCA L1 data in 2017 shows the lowest base rate at around 1.5kHz. The reason for this is the sharper time window in which an L1 hit was registered. ARCA has a lower base rate at 5kHz, followed by the base rate of the ORCA runs of the second part of 2017 and 2019 with 6kHz and 7kHz respectively. These differences could be caused by detector efficien-cies differing between the detectors, as40K decay and its noise is supposed to be homogeneous over the sea.

For further analysis the average base rates of all detectors can be found in appendix A.

3.3

Burstfractions

From the base rates the burstfraction can be calculated. This gives a notion of how much bioluminescence is present and can be used to compare the bioluminescent noise between the detectors. The burstfraction was de-fined as the time fraction (for 15 minutes intervals) that the hit rate was above 120% of the base rate.

3.3.1

ORCA

The calculated burstfraction of the ORCA L1 data in 2017 seen in figure 3.3.1 shows some remarkable characteristics. The burstfraction, which seems to vary mostly from 0 to 0.7 shows some large waves at the 400 and 600 hour mark and smaller waves all around it. These smaller waves could coincide with the periodic nature of the bioluminescence due to the Corio-lis force, which should create waves with a period of 17.7 hours. Another 28

3.3 Burstfractions 29 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Burstfraction 0 200 400 600 800 1000 1200 time (h) 2 4 6 8 10 12 14 16 18 DOM Burstfraction

Figure 3.3.1: The burstfractions of the L1 data 2017 ORCA runs, with the burstfraction shown by the colour, the DOM on the y-axis and the time in hours on the x-axis.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Burstfraction 0 100 200 300 400 500 600 700 800 900 2 4 6 8 10 12 14 16 18 Burstfraction

Figure 3.3.2: The burstfraction of the L0 data of the 2017 ORCA runs. The x-axis contains the time in hours, while the y-axis contains the DOM number.

interesting property is that the graph shows the burstfraction of ORCA is almost constantly at around 0 after the 900 hour mark. This is most likely caused by the down-scaling in the L1 data, where there now is a factor of 20 less bursts possible. In further research this could be further investigated if mediating this difference can cause the bioluminescence to reappear for the down-scaled part.

The L0 data of ORCA shares the long bursts of the L1 data, shown in figure 3.3.2. The L0 data also seems to be much more sensitive to biolu-minescent noise in the burst fraction than the L1 data, with many more small waves of bursts visible. This data also shows burstfractions mostly between 0 and 0.7, though it does average more towards higher values than the L1 data.

ORCA data of 2019 the burstfraction was calculated as well, resulting in interestingly low values compared to the 2017 ORCA runs (see figure 3.3.3). The burstfraction here shows only peaks as high as 0.2 instead of the high burstfractions of 0.7 for ORCA. DOM 11 shows this lower signal even more, with lower values than the rest of the DOMs. There are two larger waves of 100 hours with some small waves in between that do show up, but for the second half there is almost no bioluminescence present.

A reason for the difference in bioluminescence in the same place could be the season, as the 2019 runs were recorded in winter, and the 2017 runs in autumn. The different seasons could change the activity and amount of the bioluminescent creatures present. Another explanation could be a different run setup being used which could partly filter out the biolumi-nescence.

30 Noise Spectrum

Figure 3.3.3: The burstfractions of the 2019 ORCA runs, with the burstfraction shown by the colour, the DOM on the y-axis and the time in hours on the x-axis.

3.3.2

ARCA

The same routine was done for ARCA data. However, as ARCA is less deep than ORCA, it detects less bioluminescent noise. This of course im-plicates lower burstfraction and a larger difficulty for finding periodicity.

In figures 3.3.4 and 3.3.5 these low values are indeed visible, with the highest burstfraction being 0.15 for L1 and 0.3 for L0. These peaks are almost insignificant when compared to the peaks of 0.7 for the 2017 ORCA runs, but somewhat compare to the burstfractions of the ORCA runs of 2019. It should be noted that the ARCA data was taken at approximately the same time as the ORCA 2017 runs, so there is no seasonal difference between the ARCA data and the ANTARES data. The difference is most likely caused by the difference in depth, which in turn causes a decreased amount of bioluminescent creatures.

Interesting as well is that there seems to be more bursts for the higher DOMs at ARCA, especially for the L1 data in figure 3.3.4. This also could indicate a depth-dependence of the bioluminescent creatures. However it is noticeable hat this is less visible in the L0 data of ARCA in figure 3.3.5, where the first part seems to show similar burstfractions for all DOMs.

Lastly DOMs 17 and 13 in figure 3.3.4 seem to have higher burstfrac-tions in general. This is most likely caused by their base rates being much lower than the other DOMs as well, as seen in figure 3.2.8.

30

3.3 Burstfractions 31 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 Burst fraction 0 50 100 150 200 250 time (h) 2 4 6 8 10 12 14 16 18 DOM Burst fraction

Figure 3.3.4: The burstfraction of the L1 ARCA data. The x-axis contains the time in hours, while the y-axis contains the DOM number.

0 0.05 0.1 0.15 0.2 0.25 0.3 All Burstfractions 0 20 40 60 80 100 2 4 6 8 10 12 14 16 18 All Burstfractions

Figure 3.3.5: The burstfraction of the L0 ARCA data. The x-axis contains the time in hours, while the y-axis contains the DOM number.

3.3.3

ANTARES

The burstfraction of ANTARES was pre-calculated, and the parts where there was no base rate present are cut out or set to zero. As a result, visible

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Burstfraction 0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 time (h) 5 10 15 20 25 floor Burstfraction

Figure 3.3.6: The burstfractions of the ANTARES runs, with the burstfraction shown by the colour, the floor on the y-axis and the time in hours on the x-axis. The black lines indicate the start and end of the ORCA run.

in figure 3.3.6, the burstfraction of ANTARES is changing constantly, over both large (hundreds of hours) and small (tens of hours) time periods. These changes do seem to be happening equally fast on all floors, with no floor seeming out of phase. Especially interesting are the large bursts at 300, 800, 1000, and 2000 hours. Of these bursts the 800 hour and 1000 hour

32 Noise Spectrum

burst are visible in the ORCA 2017 (L0) data as well, which could indicate that the ORCA and ANTARES detectors share the same current. The large bursts are not visible in the base rates of ANTARES either (figure 3.2.10), showing the baserate can filter out most of these bursts.

Furthermore shows the burstfraction of ANTARES fractions from 0 to 0.9, meaning that on 15 minute intervals, the data can consist of nearly none to nearly all bioluminescent bursts. This shows that overall ANTARES seems to have the most bioluminescent noise compared to the ORCA and ARCA detectors. This difference could be explained by the different de-tector structures.

3.4

Burst rates

As the burstfraction sometimes has difficulty picking up on biolumines-cent noise, especially in the second halves of the ORCA L1 data of 2017 and 2019, another method can be useful to express the bioluminescent noise; the burst rate. The burst rate, which indicates the number of bursts in a 15 minute interval, shows some better contrast, making it easier to identify high bioluminescent noise times against low bioluminescent noise times.

3.4.1

ORCA

The burst rate of the ORCA 2017 runs can be seen in figures 3.4.1 and 3.4.2. The second part of the 2017 L1 data in figure 3.4.1 especially shows improvement compared to the burst fraction. Here both burst rates show a large contrast with the smaller waves of the expected periodicity of ap-proximately 18 hours. These waves are most apparent in the L0 data in figure 3.4.2, but are clear in the middle of the L1 data as well (figure 3.4.1). Interesting to note is that the high bursts of bioluminescence seen in the burstfraction are here much less visible. Instead a high burst rate is mostly noted around the 600 hour mark.

For the 2019 ORCA data the burst rate in figure 3.4.3 shows more bio-luminescent noise and variance therein than the burst fraction, especially the second half of the data. The burst rate does seem to be lower than the burst rate of the 2017 runs, with a maximum of 250 bursts per interval compared to a maximum of 600 bursts per interval. Therefore it seems that during these runs the detector definitely did find less bioluminescent noise than in the 2017 runs.

32

3.4 Burst rates 33 0 100 200 300 400 500 600 Burst rate 0 200 400 600 800 1000 1200 time (h) 2 4 6 8 10 12 14 16 18 DOM Burst rate

Figure 3.4.1:The burst rate of the L1 2017 ORCA data, with on the x-axis the time in hours, on the y-axis the DOM number and the burst rate indi-cated by colour. 0 20 40 60 80 100 120 140 160 180 200 220 Burst rate 0 100 200 300 400 500 600 700 800 900 time (h) 2 4 6 8 10 12 14 16 18 DOM Burst rate

Figure 3.4.2: The burst rate of the L0 data of the 2017 ORCA runs. The x-axis contains the time in hours, while the y-axis contains the DOM num-ber. 0 50 100 150 200 250 Burst rate 0 50 100 150 200 250 300 350 400 time (h) 2 4 6 8 10 12 14 16 18 DOM Burst rate

Figure 3.4.3:The burst rate of the L1 2019 ORCA data. The x-axis shows the time in hours, the y-axis the DOM number and the colour the burst rate.

34 Noise Spectrum 0 5 10 15 20 25 Burst rate 0 50 100 150 200 250 time (h) 2 4 6 8 10 12 14 16 18 DOM Burst rate

Figure 3.4.4:The burst rate of the L0 ARCA data. The x-axis contains the time in hours, while the y-axis contains the DOM number.

0 10 20 30 40 50 60 Burst rate 0 20 40 60 80 100 time (h) 2 4 6 8 10 12 14 16 18 DOM Burst rate

Figure 3.4.5:The burst rate of the L0 ARCA data. The x-axis contains the time in hours, while the y-axis contains the DOM number.

3.4.2

ARCA

As ARCA is placed deeper underwater than ORCA, the burstrate is much less than that of ORCA as well. The burst rates in figures 3.4.4 and 3.4.5 do show some variance, but especially the lower levels show no clear periodic bioluminescence. The top DOMs do however show some semblance of a wave forming and seem to have higher burst rates in general as well, but the lower DOMS almost pick up nothing, especially in the L0 data.

The burst rate is much lower in general than that of ORCA as well, with the L1 data having a maximum of 25 bursts per interval and the L0 data having a maximum of 50 bursts per interval. This burst rate is significantly smaller compared to the 200 bursts per 15 min for the lowest parts of the ORCA data.

3.5

Noise on individual PMTs

While the base rate and burst rate can be relatively stable over time, they seem to be less stable when compared over DOMs. This can be taken one step further by comparing the detectors inside the DOMs; the PMTs. Ac-cording to their circular design the individual PMTs should not be favoured more for the40K noise or bioluminous noise.

However in practice there are some differences in base and burst rates, as seen in figure 3.5.1 and 3.5.2. Here the average base and burst rate of the individual PMTs was calculated to compare the different PMTs. The base rates in figure 3.5.1 show some outliers, but it seems clear that the lower PMTs of higher DOMs (specifically those of DOM 8 and higher) 34

3.5 Noise on individual PMTs 35 20 40 60 80 100 120 Average baserates 2 4 6 8 10 12 14 16 18 DOM nr 5 10 15 20 25 30 PMT nr Average baserates

Figure 3.5.1:The base rates of the different PMTs of the different DOMs.

50 100 150 200 250 300 350 400 burstrates 2 4 6 8 10 12 14 16 18 DOM nr 5 10 15 20 25 30 PMT nr burstrates

Figure 3.5.2: The burst rates of the different PMTs of the different DOMs.

have higher base rates. The most probable reason for this is that the lower PMTs of all DOMs are supposed to have higher base rates, but the lower 6 DOMs have lower efficiencies which make the difference invisible. How-ever this still leads to the question of why the lower PMTs observe higher base rates, as the40K nois is supposed to be the same everywhere.

The average burst rates shown in figure 3.5.2 show a similar higher burst rate for lower PMT numbers and higher DOMs compared to normal. However, here it seems that higher PMT numbers do show even higher burst rates at places, though this is with much more variance. It is clear that the middle PMTs from 14 to 20 have the lowest burst rates over all DOMs.

Chapter

4

The periodicity of bioluminescence

With the burst rate clearly showing the bioluminescence and its wavelike pattern over time, the periodicity of this signal can be observed. The pe-riodic components of all 3 detectors was researched,each with different results.

4.1

Methods

To find the periodic component of the bioluminescence, periodograms were used on the burst rates, as explained in section 2.1. However, just using the whole time stream is not the only way to find the periodicity of the burst rate.

As the periodogram is able to work with 200 hour periods of data, it can be used to observe the change in periodicity over time as well. For this the periodicity can be checked for 200 hour parts of the data, each with a later starting time or offset compared to the beginning of the data taking. If some periodicity gives a quality factor higher than the 97.5% and 99% intervals of a χ2distribution respectively in this period, the periodicity is noted down in a graph. By moving this 200 hour window over the time stream a time evolution of the periodicity is then created, as can be seen in figure 4.1.1. When the data of a window with a certain offset has a significant periodicity, either above the 97.5% interval or the 99% interval of the χ2 distribution, this is visible in the graph with a black or red dot respectively. For the data in figure 4.1.1, it is clear that at the beginning the period is around 16.5 hours, after which it lowers to even 15 hours. At the end of the data it does come back to a periodicity around 18 hours, but there are some outliers at around 400h of 26.5 hours. From this it is

38 The periodicity of bioluminescence 0 100 200 300 400 500 600_offset(h) 700 10 12 14 16 18 20 22 24 26 28 periodicity(h)

Time evolution periodogram DOM 3 Time evolution periodogram DOM 3

Figure 4.1.1: The time evolution of the periodogram. The x-axis shows the offset compared to the beginning of the data taking and the y-axis shows the period found at that offset. black dots signify the 97.5% interval and red ones the 99% interval.

clear the period can in fact change over time and the time evolution of the periodogram is able to see this.

4.2

Results

With these methods the periodicity of the different detectors were ob-served. Selections of different DOMs or floors of these detectors will be discussed here. The graphs showing the periodicity of all the DOMs and floors can be found in appendix B.

4.2.1

ORCA 2019

The analysis of the ORCA runs of 2019 revealed a small periodicity around 18h, which does match with the hypothesis of the inertial tide being the main component. This periodicity is somewhat visible in the periodogram, as seen in figure 4.2.1. However, in the periodogram the peak is to small to be significant against random noise, even though it does stand out against the other periods.

38

4.2 Results 39 10 12 14 16 18 20 22 24 26period(h) 28 0 10 20 30 40 50 60 70 80 dom1_dom2 dom3 dom4 dom5 dom6 dom7 dom8 dom9

95% interval Chi squared

periodogram

95% interval Chi squared

periodogram

Figure 4.2.1: The periodograms of all DOMs during the 2019 ORCA runs, with the lowest nine in the top graph and the highest nine in the bottom graph.

Furthermore is there a slight periodicity around the 16h mark visible in some of the DOMs as well. However, this is also below the significance mark. Noticeable is that while the peaks do not make the significance interval, they do stand out against the rest of the noise. This could be caused by the noise destructively interfering with each other at the other periods.

To see if this was always the case for the whole run a time evolution of the periodogram was made as well, as seen in figure 4.2.2. However this time evolution used the 95% and 97.5% intervals instead of the nor-mal 97.5% and 99% intervals, as both the periodicity peaks and the noise seemed very low in figures 4.2.1. This shows some periodic components for the different DOMs, especially for the first parts of DOM 8 and 16. However some DOMs barely show any signal like DOM 4 or show both the 18 hour signal but with some other noise as well as seen for DOM18. Over all it is clear that the beginning of the data shows the most

periodic-40 The periodicity of bioluminescence 20 30 40 50 60 70 80 offset (h) 90 10 12 14 16 18 20 22 24 26 28 periodicity (h) DOM4 DOM4 0 20 40 60 80 100 offset (h) 120 10 12 14 16 18 20 22 24 26 28 periodicity (h) DOM8 DOM8 0 10 20 30 40 50 60 70 80 offset (h) 90 10 12 14 16 18 20 22 24 26 28 periodicity (h) DOM16 DOM16 0 20 40 60 80 100 120 140 160 180offset (h) 200 10 12 14 16 18 20 22 24 26 28 periodicity (h) DOM18 DOM18

Figure 4.2.2:Time evolution of the periodogram of four DOMs of the 2019 ORCA L1 Data. The x-axis shows the time offset and the y-axis the period found, with black marks indicating the 95% interval and red marks the 97.5% interval.

ity, specifically the first 270 hours (as the window is 200 hours long and the periodicities are clear until the 70 hour offset mark). For the 97.5% interval the periods found gave an average of 17.7±0.1 hours. This is the same as the expected value from the Coriolis force of 17.7 for ORCA.

4.2.2

ORCA 2017

The analysis of the ORCA runs of 2017 revealed a large periodicity around 15.5h, which does not match the inertial tide hypothesis. This periodicity does not match any other water movement phenomena as well. However the Periodicity is clearly significant for all DOMs (figure 4.2.3), with the exception of DOM 16 not making the 99,9% χ2 Mark. This does not mean that the periodicity is absent there, it is just more difficult to differentiate it from random noise. Furthermore have the lower nine DOMs a larger periodicity at the 15.5h mark, with DOM 6 having the highest. This can be because of the bioluminescence being more abundant on this level, or the non-periodic noise was less high at this spot.

40

4.2 Results 41

Figure 4.2.3:The periodograms of all the DOMs of the ORCA 2017 runs, with the lowest nine in the top graph and the highest nine in the bottom graph. The x-axis shows the periodicity in hours, and the y-axis the quality factor

42 The periodicity of bioluminescence 0 200 400 600 800 1000 1200 _{offset (h)} 10 12 14 16 18 20 22 24 26 28 periodicity (h) DOM1 DOM1 0 200 400 600 800 1000 1200 _{offset (h)} 10 12 14 16 18 20 22 24 26 28 periodicity (h) DOM7 DOM7 0 200 400 600 800 1000 1200 _{offset (h)} 10 12 14 16 18 20 22 24 26 28 periodicity (h) DOM12 DOM12 0 200 400 600 800 1000 1200 _{offset (h)} 10 12 14 16 18 20 22 24 26 28 periodicity (h) DOM17 DOM17

Figure 4.2.4: The time evolution of the periodogram of 4 DOMs of the 2017 L1 ORCA data, with black marks indicating the 97.5% interval and red marks the 99% interval.

In figure 4.2.4 the time evolution of the periodicity is visible, where it is clear that the 16 hour periodicity is mostly present at the middle and end of the data. There are some parts where a 18 hour periodicity is found as well, as seen in the different DOMs at the 800 hour mark. This gives an average periodicity of 16.2±0.2 hours for the whole detector. While this is close to the expected periodicity of 17.8h, it is still no match. Furthermore do the graphs show that the periodicity can change over time, and is not constant. Interesting is the gap present in nearly all DOMs at the 400 hour mark, caused by a very low and constant burst rate as can be found back in figure 3.3.1.

For the L0 data the time evolution of the period was researched as well, and gave somewhat similar results. This can be seen in figure 4.2.5. The first part here shows a time evolution similar in shape to the one of L1 (fig-ure 4.2.4), with a 16 hour periodicity slowly changing into an 18 hour pe-riodicity. This should be the case as they are of the same time period. The average periodicity of the periods found with a significance level higher than the 99% interval is calculated to be 16.3±0.3. This is similar to the 42

4.2 Results 43 0 100 200 300 400 500 600 700 offset(h) 10 12 14 16 18 20 22 24 26 28 periodicity(h) DOM2 DOM2 0 100 200 300 400 500 600 700 offset(h) 10 12 14 16 18 20 22 24 26 28 periodicity(h) DOM4 DOM4 0 100 200 300 400 500 600 700 offset(h) 10 12 14 16 18 20 22 24 26 28 periodicity(h) DOM10 DOM10 0 100 200 300 400 500 600 700 offset(h) 10 12 14 16 18 20 22 24 26 28 periodicity(h) DOM17 DOM17

Figure 4.2.5:The time evolution of the periodogram for the L0 data of the ORCA 2017 period. The time evolution of the different DOMs are shown with the x-axis describing the offset and the y-axis the period found. The black marks indicate the 97.5% interval and the red marks indicate the 99% interval.

average period from the L1 data, and while it is slightly higher it does not match the 17.8 hour periodicity that should be caused by the Coriolis force.

4.2.3

ARCA

ARCA had a significantly lower burst rate than ORCA or ANTARES. How-ever, while the burst rate is low it can still be analysed for any periodic signals with the periodogram. As a result of this, visible in figure 4.2.6, some periodicity could be found. The low burst rate does cause for lesser quality of periodograms, especially on the lower level DOMs, like DOM 6, which do not show many coherent periods. However the higher DOMs (12 and up) do show a ongoing periodicity around 20.5 hours, as expected by ARCA’s latitude. The average periodicity of the periods found at 99% significance was found to be 20.8±0.3 hours for the ARCA L1 data, which is close to the expected 20.3 hours, but still a little off.

44 The periodicity of bioluminescence 0 20 40 60 80 100 120 140 160 180 200 offset(h) 220 10 12 14 16 18 20 22 24 26 28 periodicity (h) DOM6 DOM6 0 20 40 60 80 100 120 140 160 180 200 offset(h) 220 10 12 14 16 18 20 22 24 26 28 periodicity (h) DOM12 DOM12 0 20 40 60 80 100 120 140 160 180 200 offset(h) 220 10 12 14 16 18 20 22 24 26 28 periodicity (h) DOM13 DOM13 0 20 40 60 80 100 120 140 160 180 200 offset(h) 220 10 12 14 16 18 20 22 24 26 28 periodicity (h) DOM17 DOM17

Figure 4.2.6: The time evolution of the periodogram of four different DOMs of ARCA, with period on the y-axis and time on the x-axis, with black marks indi-cating the 97.5% interval and red marks the 99% interval.

44

4.2 Results 45 0 2 4 6 8 10 12 14 16 18 DOM number 10 12 14 16 18 20 22 24 26 28

period found (hours)

Periodic compontents of the DOMs

Periodic compontents of the DOMs

Figure 4.2.7: The results of the periodogram for all DOMs of ARCA. The x-axis shows the DOM number and the y-axis the period found.

The L0 data of ARCA had similar results to the L1 data, with no dis-cernible signal in the lower levels but a clear signal for a 20.5 period for the higher DOMs, as can be seen in figure 4.2.7. There is a large spread in peri-odicity present in the data, which is most likely caused by the low amount of data used. A reason for this absence in periodic bioluminescence on the lower levels could be because of lower levels of bioluminescence there, as could be seen in figures 3.3.5 and 3.4.5. The difference is small but appar-ently significant. Taking all periods higher than the 99% χ2 interval, the average period around the ARCA detector was found to be 20±1 hours for the L0 data. This value is within the significance of the expected peri-odicity from the Coriolis force of 20.3 hours for ARCA.

4.2.4

ANTARES

The periodic nature of the bioluminescence around ANTARES should be similar to that of ORCA, as they are on the same latitude. This similarity is in fact shown as well in the time evolution periodogram of ANTARES (fig-ure 4.2.8). Visible in these graphs is the same pattern as for ORCA. At first

46 The periodicity of bioluminescence 400 600 800 1000 1200 1400 1600 1800 2000 offset(h) 10 12 14 16 18 20 22 24 26 28 periodicity(h) floor7 floor7 0 200 400 600 800 1000 1200 1400 1600 1800offset(h) 10 12 14 16 18 20 22 24 26 28 periodicity(h) floor18 floor18 0 200 400 600 800 1000 1200 1400 1600 1800offset(h) 10 12 14 16 18 20 22 24 26 28 periodicity(h) floor21 floor21 0 200 400 600 800 1000 1200 1400 1600 1800 2000 offset(h) 10 12 14 16 18 20 22 24 26 28 periodicity(h) floor23 floor23

Figure 4.2.8:The time evolution of the periodogram of all the PMTs on four dif-ferent floors of ANTARES, with black marks indicating the 97.5% interval and red marks the 99% interval.

the periodicity seems to be at 18 hours which changes into a periodicity of 16 or 15.5 hours, but it changes later on towards 17 hours.

Visible is that for ANTARES the periodicity of the bioluminescent sig-nal is a bit weaker than that of ORCA, with large parts of data not showing any periodicity at all, especially in the range from 50 to 600 hours. How-ever, from 600 hours onwards there is a clear signal of around 16 hours. This range is during the ORCA period, which does explain some similari-ties, but does not explain the absence of periodicity outside of this range. The average period on the whole detector taking the 99% χ2 interval was found to be 16.15±0.17 hours. This is comparable to the average period found for ORCA in the same time period, which could indicate they were in the same current flow. However the value is much lower than the ex-pected 17.7 hour periodicity.

46

4.3 Currents 47 0 200 400 600 800 1000 1200 time (h) 0.02 0.04 0.06 0.08 0.1 0.12 velocity (m/s)

Horizontal Current velocity Horizontal Current velocity

0 200 400 600 800 _offset(h) 1000 10 12 14 16 18 20 22 24 26 28 periodicity(h) Current Current

Figure 4.3.1: Two graphs of the horizontal current velocity. In the first graph the horizontal current during the 2017 run is shown. The lower graph contains the time evolution of the periodogram of the data, with black marks indicating the 97.5% interval and red marks the 99% interval.

4.3

Currents

While the ORCA data does not initially seem to match the periodicity ex-pected from literature, it does have some similarities with the current mea-surements at that time, especially the horizontal current velocity, seen in figure 4.3.1. The graphs show a large burst in the data around the 1100h mark, which also causes a lack of periodicity in the time evolution past an offset of 1000h. The beginning and end of the time evolution of the peri-odogram do not show clear periodicities, but in the middle there is first a periodicity of 16h present, which then turn into an 18h periodicity.

Compared to the horizontal current velocity, the pressure in the sea wa-ter shows a very clear periodicity, but this periodicity is different from that of the bioluminescence. The periodicity of the pressure instead follows the tidal frequency, as seen in figure 4.3.2. In this figure the periodograms show a periodicity at both 12.5 and 25 hours. The 25 hour peak can be ex-plained by the fact that it 25 hours consists of two 12.5 hour periods. The

48 The periodicity of bioluminescence 0 200 400 600 800 1000 1200 time (h) 2509.7 2509.8 2509.9 2510 2510.1 2510.2 velocity (m/s) Pressure Pressure 0 200 400 600 800 1000_offset(h) 10 12 14 16 18 20 22 24 26 28 periodicity(h) Current Current

Figure 4.3.2: Two graphs showing the pressure characteristics. The first graph contains the raw pressure data during the 2017 runs, while the second contains the time evolution of the periodogram of the data, with black marks indicating the 97.5% interval and red marks the 99% interval.

periodogram can thus not differ between multiplicities of periodicities. This shows that the bioluminescence is most dependent on the current, or at least the horizontal component of it. However the pressure of the water does not correlate to the horizontal current or the bioluminescent signal. The fact that the bioluminescence correlates to the (horizontal) cur-rent and not the pressure does support the theory that bioluminescence bursts are caused by the bioluminescent creatures hitting the detector, as more creatures would hit the detector if the current was stronger.

However the actual correlation to the burst rate or burstfraction is not as strong. The correlation between the burstfraction and the current ve-locity was found to be 0.26 for most DOMs, indicating a weak correlation between the two.

48

4.3 Currents 49 0 100 200 300 400 500 600 700 800 900 time (h) 0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014 0.0016 0.0018 0.002 0.0022 0.0024

Normalized Burstrate and Current Velocity

Burstrate Horizontal current velocity

Normalized Burstrate and Current Velocity

Figure 4.3.3:The horizontal current velocity and the burst rate of the first DOM of ORCA.

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

current and burstrate

0 20 40 60 80 100 120 140 160 180 200 220 burstrate (Hz) 0 0.02 0.04 0.06 0.08 0.1 0.12 current velocity (m/s)

current and burstrate

Figure 4.3.4: The correlation between the hori-zontal current velocity and the burst rate of the first DOM of ORCA.

Chapter

5

Effect on track reconstruction

When reconstructing tracks, it is important to note any biases that may appear in the reconstruction, as they can be misinterpreted as track prop-erties later on. Bioluminescence has a possibility to create such biases, as the burst can create hits that can be misconstrued for muon hits, or can hide hits through the high rate veto implemented on most data. This is why different track variable distributions were studied to see if there was a bias present in runs with high or low amounts of bursts. The zenith angle Theta and the quality factor are discussed here.

5.1

Methods

Bioluminescent bursts could have an effect on the track reconstructions as a burst cannot be differentiated from a neutrino event on its own. This means that high or low amounts of bursts can cause problems in recon-struction by adding outliers to the data. To see if that is the case and to see if the Monte Carlo simulations properly combat this difference 600 runs were split on high burst rate (>300 burst per run) or low burst rate (<300 burst per run) for the L1 time stream. The out of sync runs were left out, resulting in distributions for the track variables seen in the results.

Focus was mostly on the following two track variables: the zenith an-gle theta, and the quality factor Q. Here the quality factor indicates how many of the hits that indicated the event fit in the track, and gives a mea-sure of how well the track is fitted. Theta gives the zenith angle of the direction of the muon, and therefore an idea of where the muon came from. The angular resolution of these events is important, as it gives an indication of where the neutrino sources are.