A Seismic Analysis of the XENON1T Detector
Falco Bijloo
10543449
December 6, 2016
Abstract
The largest earthquakes that occurred in central Italy as part of the 2016 sequence are discussed, with their sources explained. These earthquakes generated a gyrating move-ment of the XENON1T detector of the xenon collaboration housed in the underground laboratory LNGS in central Italy. The tilt of the XENON1T detector during some of the earthquakes is examined, and a natural frequency of 0.07 ± 0.02 Hz of the gyrating move-ment is found using Fourier analysis, which is lower than expected. Also one Innoseis’ Tremornet node, which was placed in Amsterdam, The Netherlands, has measured the ground movement of the 26 October and 30 October earthquakes, of which the frequency spectra are presented.
Name Supervisor: Auke-Pieter Colijn
Name Second Assessor: Marcel Vreeswijk
University, Faculty: UvA, FNWI
Institute: Nikhef
Group: Dark Matter
Study Program: B`eta-Gamma
Major: Physics
Report Bachelor Project Physics and Astronomy, size 15 EC, conducted between 01-09-2016 and 12-12-2016
Contents
1 Introduction 3 2 Theory 5 2.1 Earthquake source . . . 5 2.2 Seismic waves . . . 6 2.3 XENON1T detector . . . 11 3 Methods 12 3.1 Earthquake as accelerator . . . 12 3.2 Seismic noise . . . 15 3.3 Tremornet Node . . . 153.4 Assumptions and simplifications . . . 15
4 Results 16 4.1 Detector motion . . . 16 4.2 Noise . . . 17 4.3 Innoseis . . . 19 5 Discussion 23 6 Conclusion 26 7 Acknowledgements 27 8 Appendix: Summary 29 8.1 Summary . . . 29 8.2 Samenvatting . . . 29
F. Bijloo A Seismic Analysis of the XENON1T Detector
1
Introduction
Seismic waves are waves of energy that travel through the Earth, and are created mostly as a result of earthquakes, huge landslides, volcanic eruptions, magma movement or large man-made explosions [Kennett, 2009]. Smaller seismic waves, such as created by human traffic, wind hitting the trees or ocean waves crashing on to the beach, are called ambi-ent vibrations or seismic noise. The earth’s surface is not an inert surface, it moves and oscillates. This means in turn that every object that is on the earth feels these seismic waves and noises. If seismic waves are strong enough, objects shake or resonate. Earth-quakes are an example of the most common form of natural generators of large seismic waves [Kanamori and Brodsky, 2004]. With their frequency ranging from 0.001 Hz to 4 Hz, large earthquakes can be detected at enormous distances from their sources. It is ob-served that the seismic waves of very large earthquakes (>M6.0) can circulate the earth multiple times [Shearer, 2009].
In central Italy, approximately 50 km Southeast of Norcia, between l’Aquila and Ter-amo, is the mountain massif of Gran Sasso, which is part of the Apennines mountain range. The Apennines are a tectonically and geologically complex region, which is seis-mically very active [Galli et al., 2002]. Through the mountains of Gran Sasso a tunnel was drilled in 1984 to connect the cities Rome and Teramo, whereby an underground laboratory was built, named Laboratori Nazionali del Gran Sasso (LNGS). The first large experiments started in 1989. Since the mountain acts as a cosmic radiation shield, with a reduction by one million times, the laboratory is extremely suitable for low-background experiments [Votano, 2012]. In this laboratory the XENON1T experiment is housed. A 3500 kg dual-phase (liquid/gas) xenon detector is used to search for the elusive Dark Matter. As noticed before, Gran Sasso is located in the central Apennines, a seismi-cally very active region, with around 100.000 events per 20 years [Chiarabba et al., 2005]. At 24 August 2016, a large earthquake (>M6.0) occurred approximately 10 km South of Norcia, Central Italy [USGS, 2016b]. On 26 October 2016, another large earthquake (>M6.0) occurred approximately 10 km North of Norcia [USGS, 2016a]. The largest (at the time of writing) earthquake occurred 6 km North of Norcia on 30 October 2016, with a moment magnitude of M6.6 [USGS, 2016c]. The US Geological Survey (USGS) made a figure in which this earthquake sequence with aftershocks and foreshocks is visible. See figure 2 for this figure and a more detailed figure showing the largest earthquakes in the 2016 sequence.
The discussed earthquakes created seismic waves, which were felt by the XENON1T detector. The tiltmeters attached to the detector showed an oscillating movement. In this paper the effects of seismic waves on the XENON1T detector are described. Through the analysis of the effects of the recent earthquakes on the tilt of the detector a better understanding of the gyrating movement is gained and a natural frequency can be calcu-lated using Fourier analysis. This movement is compared to the gyrating movement of the detector caused by regular seismic noise. Also a proposition is made to research the
Figure 2: Top, a map made by USGS [USGS, 2016c] showing every noticeable (>M3.5)
after-shock and foreafter-shock of the 2016 sequence, with main after-shock being the M6.6 30 October, 2016
earthquake. Bottom, a map of central Italy showing the locations of the largest earthquakes
in the 2016 sequence (at the time of writing).
F. Bijloo A Seismic Analysis of the XENON1T Detector
inner structure of the mountain range and the seismic activity of the Gran Sasso region further with seismic sensors. Special highly sensitive seismic sensors, named Tremornet, are produced by Innoseis in the Netherlands, which can be used to study the upper earth’s inner structure. These sensors function with a geophone inside and are time-accurate us-ing constant GPS-checkus-ing. To provide insight in the functionus-ing of the sensors, one is tested at Nikhef, Amsterdam, and the results are presented in this paper.
2
Theory
First the source of the earthquakes in the central Italian region is briefly described. Af-terwards, to get an understanding of the effects on the detector, the creation, propagation and the measurement of seismic waves are concisely explained. Last, the setting of the detector in LNGS is described.
2.1
Earthquake source
At the 24th of August, 2016, a moment magnitude M6.2 earthquake occurred in Cen-tral Italy. The epicentre was concentrated 10.0 ± 4.4 kilometres South-East of Norcia [USGS, 2016b], above the hypocentre, with a depth of 4.4 ± 2.8 km. The largest shocks in the Apenninic region reveal normal faulting with extension perpendicular to the Apen-ninic chain, consistently with the tectonics of the internal sector of the northern Apennine belt and with previous earthquakes in adjacent regions [Chiarabba et al., 2009]. Normal faulting is an event where the earth’s crust is extended. Because of friction and rigidity, rocks cannot flow or glide past each other with ease, and therefore occasionally all move-ment ceases. This results in stress building up in the rocks. The momove-ment the level of stress exceeds the strain threshold, the accumulated potential energy will be dissipated by the release of the strain, which is focused into a plane along which relative motion is accom-modated [Sibson, 1977]. Faulting is the consequence, the crust reacts by fracture. The energy released by the instantaneous strain-release causes the creation of seismic waves. Most faults are non-vertical faults. The two sides of a non-vertical fault are known as the footwall and hanging wall, for historical reason [Jackson and McKenzie, 1983]. The dubbed footwall is located below the fault plane and the hanging wall above the fault plane. At a normal fault the hanging wall moves downward compared to the footwall, as is schematically depicted in figure 3. The 2016 earthquake sequence was the result of normal faulting on the North Northwest South Southeast oriented Apenninic chain [USGS, 2016b].
The Apennines are a geologically and tectonically complex area [Galadini and Galli, 2000], with subduction of the Adriatic micro-plate beneath the Eurasiatic plate and the Apen-nines from East to West, the opening of the Tyrrhenian basin to the West and the conti-nental collision of the Eurasian and African plate [Devoti et al., 2008]. This is
schemati-Figure 3: A graphical explanation of normal faulting. The crusts extends as the hanging wall
moves downward relatively to the foot wall.
cally depicted in figure 4.
The October 26 earthquake, M6.1, 3.0 ± 4.9 km West of Visso, hypocentre 10.0 ± 1.7 km deep, occurred as a result of normal faulting in the same structure. Only a few days later the largest, at the time of writing, of the sequence of earthquakes occurred 6.0 ± 5.1 km North of Norcia. It is measured at a moment magnitude of M6.6 and a hypocentre of 10.0 ± 1.7 km deep. Paulssen [Paulssen, 2016] pointed out that Meijer [Meijer, 2016] suggested that these earthquakes are not a direct expression of Africa-Eurasia tectonic convergence. This sequence of earthquakes is rather likely to reflect gravitational spreading. This happens when the force pushing inward on a piece of crust is inferior to the gravitational force of the mass, and therefore has the chance to spread out, collapsing slowly under its own weight, see figure 5. It is seen in the results of Faure-Walker et al. [Faure-Walker et al., 2012] that more elevated regions in the Apennines have a greater extension rate, which supports the claim of gravitational spreading.
2.2
Seismic waves
Earthquakes radiate seismic waves that travel throughout the earth [Shearer, 2009]. Seis-mic waves can be observed and detected by a variety of instruments. A seismograph is used to create a seismogram, which shows the displacement over time of the earth’s crust at a certain location on the earth. The patterns these seismograms show, largely consist of small movements from various ambient sources. From time to time the irregular pattern of the records is interrupted by a disturbance which rises above the background noise with a well defined waveform. This feature arises from the excitation of seismic waves by some natural or artificial source, in other words an earthquake. With their low frequency, 0.001 Hz to 4 Hz [Kennett, 2009]), seismic waves are measurable considerable distances
F. Bijloo A Seismic Analysis of the XENON1T Detector
Figure 4: A simplified schematic view of the tectonics of the Apenninic region. The Apennines
are a geologically and tectonically complex area, with subduction of the Adriatic
micro-plate beneath the Eurasiatic micro-plate and the Apennines from East to West, the opening of the
Tyrrhenian basin to the West and the continental collision of the Eurasian and African plate
in the South.
Figure 5: Gravitational spreading. When the force pushing the crust inward is omitted (or
inferior to the gravitational force), the gravitational force of the mass of the mountain results
in spreading of the crust [Meijer, 2016].
Figure 6: Different seismic waves. On the left side the body waves are depicted, which travel
through the earth and on the surface a) the compressional P wave and b) the transverse S
wave. On the right side the surface waves are depicted, which travel only in the upper part
of the crust c) the Love wave and d) the Rayleigh wave.
from the source. Seismic waves are divided in two main groups, body waves and surface waves. The different seismic waves are depicted in figure 6. Body waves are further dis-tinguished in compressional or pressure waves (P waves) and, slower moving, transverse or shear waves (S waves). Because the interior of the Earth is almost incompressible, P waves transmit their energy quite easily through the medium, and thus travel quickly [Shearer, 2009]. On the other hand, the transverse S waves, which direction of motion in the medium is perpendicular to the direction of propagation of the wave, transmit energy less easily through the medium, consequently S waves travel slower. Furthermore, surface waves are distinguished in Love waves and Rayleigh waves, named after their discoverers. Love waves move transverse horizontally to the direction of propagation. Rayleigh waves move both up and down as in the direction of propagation, and therefore are polarized, like ocean waves. Surface waves differ from body waves in some aspects. First, surface waves, and namely the Rayleigh waves, are felt more intense on the surface of the earth than P waves. That is because of the rolling movement generated by the Rayleigh waves. Second, surface waves move more slowly than body waves. Lastly, since surface waves are confined to the surface of a sphere, geometrical spreading effects are reduced compared to body waves and therefore attenuation is less. Consequently surface waves of large earth-quakes (>M6.0) are measurable for many hours and can circulate the earth many times. Different methods are used to visualize earthquakes. Today, practically all seismograms are recorded digitally to make analysis by computer easier. It is seen on a seismogram that the P wave arrives earlier than the S wave, which in turn arrives earlier than the
F. Bijloo A Seismic Analysis of the XENON1T Detector
Figure 7: A seismogram as recorded by NARS station, Fort Hoofddijk, Utrecht at 1700 km
from the source. The source was an earthquake of moment magnitude M6.8 at 21 May 2003
in Northern Algeria. The vertical displacement component is shown. The different arrival
times of the P, S and surface waves are visible.
surface waves. An example of a seismogram is seen in figure 7. In this figure the different arrival times of the different seismic waves are clearly visible. After the arrival of the first surface waves, the amplitude will oscillate heavily due to the many different surface waves and reflections and refractions of earlier waves. The 24 August earthquake was registered around the earth. Iris [IRIS, 2016] made a figure with seismograms of different seismological institutions. The different arrival times of the waves are clearly visible in this graph, as well as the spreading of the waves.
A more precise instrument to measure seismic waves is the Innoseis’ Tremornet. This instrument measures small variations in movement of the earth by use of a geophone. It is designed to autonomously record data with position and time stamping through GPS functionality, to enable accurate off-line data collection. Tremornet is an ultra-light weight wireless seismic sensor network. Its low power technology makes each sensor node significantly smaller and easier to handle, without compromising on sensing performance. Inside the sensor node is a geophone, GPS receiver, a small battery and some electronics. A geophone is a device that converts velocity, in this case ground movement, into voltage, by using a electromagnetic mass-spring system. The GPS receiver corrects the timing inside the sensor node. Furthermore the node records and stores data autonomously
Figure 8: Global displacement wavefield of the 24 August M6.2 earthquake near central Italy.
This figure shows the vertical displacement component of different seismological institutions
around the earth, with peak displacement 0.12 mm. The different arrival times of the seismic
waves are clearly visible in this graph. On the y-axis is the distance to the source (1 degree ∼
111 kilometres), starting from lat 42.714
◦N, 13.172
◦E. On the x-axis is the time in minutes
F. Bijloo A Seismic Analysis of the XENON1T Detector
Figure 9: Picture of an Innoseis’ Tremornet node, as depicted on the website of Innoseis.
inside. See figure 9 for a picture of an Innoseis node.
The node only reads data in the vertical component (not in the horizontal plane). For most purposes, like on-shore oil and gas exploration, this vertical component is enough to provide insight in the upper earth’s inner structure. For more complex studies, such as deep earth or inner core experiments, this vertical component alone could not be sufficient for a solidified experiment.
2.3
XENON1T detector
In this subsection the setting of the detector in LNGS is described. The xenon collabora-tion is searching for WIMP particles via direct deteccollabora-tion in the underground laboratory LNGS. It is located under the Gran Sasso mountain massif. The XENON1T detector is a dual-phase time projection chamber (TPC) of pure liquid xenon (LXe) with photomul-tiplier tubes inside on both sides for the simultaneous detection of scintillation light and ionization charge via corresponding gas scintillation, assembled inside a water Cherenkov muon veto [Votano, 2012]. More information about the LNGS and other experiments can be found in the article written by Votano [Votano, 2012]. Information about the XENON experiments and results from the previous XENON100 experiment can be found in the article written by Aprile and the xenon collaboration [Aprile et al., 2012]. The detector itself is a cylindrical vessel hanging from three steel wires suspended from a steel support structure. The outer vessel has a radius of 0.65 m and a height of 1.73 m. The steel wires are 4 m long each. The support structure rests on the ground inside a big water basin (figure 10). The mass of the vessel equals approximately 1185 kg, the mass of the xenon
inside the vessel equals approximately 3000 kg. These numbers are used to calculate any natural frequency later. Above the detector, between the steel wires, a tiltmeter is at-tached horizontally, which reads out data of the angle of tilt in mrad in two orthogonal directions (figure 11).
3
Methods
To understand the gyrating movement of the detector, the tilt is analyzed after and during an earthquake. Out of this gyrating movement, a natural frequency is extracted using Fourier analysis. The same method is used on standard irregular noise to check for a regular pattern in the tilt. The peaks found using Fourier analysis, should be in the region of the natural frequency [Almeida, 1994]. The value of the natural frequency is estimated in the next subsection, by applying buoyancy to a pendulum.
3.1
Earthquake as accelerator
When the seismic wave of an earthquake passes, it works as an accelerator for a brief moment. Every object that feels this wave, will have a bit of energy transmitted to it, and therefore will be accelerated. An acceleration happened to the XENON1T detector during the discussed earthquakes. The acceleration generated a gyrating movement of the detector. If the the acceleration is of a short duration, the absence of driving force will leave the detector free to swing and the system shall tend to oscillate in its natural frequency. To make an estimate of the natural frequency, the equation of a simple pen-dulum is attempted to extrapolate to a three-string-underwater penpen-dulum, which is the case of the detector. It is assumed the detector gyrates as a pendulum (figure 12 (a)) and not just tilts around its axis (figure 12 (b)), because of the rigidity of the steel wires.
An estimate of the natural frequency is calculated. By applying Newton’s second law for rotational systems [Giancoli, 2008], the equation of motion for a pendulum is obtained by
~
τ = I ~α (1)
with the torque ~τ , moment of inertia I and angular acceleration ~α. If the pendulum is treated like a mass on a string, the moment of inertia will be given by
I = mL2 (2)
with the mass m and length of string L. The torque ~τ for a simple pendulum is given by ~
τ = −mg sin(θ)L ˆθ (3)
Because the detector is underwater, the torque is reduced because of buoyancy of the vessel and therefore is given by
~
F. Bijloo A Seismic Analysis of the XENON1T Detector
Figure 10: A picture from within the water basin. Visible is the inner vessel which is hanging
from the three wires, which are suspended from the support structure.
Figure 11: A picture from the top, down into the watertank. Visible is the outer vessel which
is hanging from the three wires, labelled a
1, a
2and a
3, and the tiltmeter which is attached
Figure 12: Schematic view of the possible options of tilt of the detector. Because of the
rigidity of the steel wires, only pure pendulum motions of the system is assumed (a), and the
internal rotations are ignored(b).
with the density of water ρ and the volume of the displaced water V . Equation 1 can be rewritten that the angle must satisfy the equation
−(mg − ρgV ) sin(θ)L = mL2d 2θ dt2 (5) d2θ dt2 + (mg − ρgV ) mL θ = 0 (6)
where sin(θ) ' θ because of the small angle approximation. This differential equation holds the solution
θ(t) = θ0cos(ωt + φ) (7)
with some phase φ and ω = q
g(m−ρV )
mL being the natural frequency. For a first estimate
of the natural frequency the mass will be m = mvessel+ mxenon= 4185kg. The volume
is taken as V = 2.3 m3for a cylindrical outer vessel, and the length to the centre of mass
Lcom= Lstring+Lvessel2 = 4.7 m, as discussed in subsection 2.3.
ωest=
r
9.81(4185 − 1000 ∗ 2.3)
4.7 ∗ 4185 ' 1Hz (8)
This calculated natural frequency is tested with a Fourier analysis of the tilt of the detector after an acceleration of an earthquake. The tilt of the detector is analyzed from 15 seconds before the moment the seismic wave passes the detector until 135 seconds after the wave first arrived (150 seconds in total). This same timescale is Fourier transformed to find the frequency spectrum in which the tilt tends to oscillate. The average natural frequency is calculated and the broadness of the peaks is accounted for in the error. The underwater problem is partially solved by adding buoyancy, but drag force or viscous damping are neglected, which is discussed in subsection 3.4.
F. Bijloo A Seismic Analysis of the XENON1T Detector
3.2
Seismic noise
The earth is not an inert surface. Even when no earthquakes occur, earth’s surface vibrate continuously. Small ambient vibrations such as wind hitting trees, heavy traffic or machinery, ocean waves crashing on the beach or even doors slamming, are all noticeable seismic noise. This seismic noise shows irregular patterns. Because of these small irregular vibrations the detector still moves and the tilt oscillates. Perhaps it is possible to observe a regular pattern in the oscillation of tilt of the detector. One hour of data of the tiltmeters without seismic waves of an earthquake passing is used to test for a regular pattern.
3.3
Tremornet Node
One node of the Innoseis’ Tremornet is tested. It was put in the ground in Amsterdam (52.356◦ N, 4.951◦ E), The Netherlands, and measured the ground movement. To see earthquakes with this data, the velocity of ground movement will be filtered from 0.01 Hz to 3.0 Hz, because of the low-frequency of seismic waves. The geophone used in the Innoseis node is less sensitive at lower frequencies. This is not taken into account in the results presented in section 4.3. For a more solid result, the data should be weighted with the sensitivity response curve provided by the producer of the geophone. The results are presented in the form of figures which show the frequency spectrum over time. Two figures per event are shown, both filtered to 3.0 Hz as well as non-filtered spectra. The node has taken a datapoint of ground movement every 2 ms, and 50 datapoints are used to create the frequency spectrum of one block (100 ms).
3.4
Assumptions and simplifications
Already some assumptions and simplifications have already been noted. In this section all assumptions and simplifications are explained and discussed. First, the XENON1T detector is not a simple pendulum in the sense of its suspension. Equation 6 requires the pendulum to be a one-string pendulum. As discussed, this is definitely not the case for the XENON1T detector, as it is suspended from three rigid steel wires. This constraint will probably result in a restriction in direction of movement and a restriction in freedom of distance from its equilibrium position. Not much literature has been written about multiple-string pendulums and multiple-string pendulums with their strings not attached in the middle of the object. It was both because of limited time and limited knowledge not possible to take this problem into account and therefore the natural frequency is calculated as a simplified problem; a one-string underwater pendulum. The real data collected by the tiltmeters can therefore vary from the estimated frequency in section 3.1. In addition, the underwater problem is only partially solved by applying buoyancy. Drag force, or for small oscillations the viscous damping, is not implemented in equation 6. Martins, Silveira-neto & Steffen Jr. [Martins et al., 2007] have studied this problem of
an underwater pendulum. For small angle viscous damping equation 6 changes to d2θ dt2 + Cd mL dθ dt + (mg − ρgV ) mL θ = 0 (9)
with the drag co¨efficient Cd. For larger angle drag force correction it should change to
d2θ dt2 ± CdAp 2V ( dθ dt) 2+(mg − ρgV ) mL θ = 0 (10)
where Ap is the area projected to a plane normal to the direction of movement, V the
volume of the vessel, and the ± sign depending on which direction it is moving. The opposing drag force shall result in an attenuation. The potential effect of these damping terms is shifting the natural frequency to lower values. These equations are more accurate for the underwater pendulum problem, but again, through limited time and knowledge, too complex to solve and therefore simplified to equation 6.
Another simplification is the assumption the rotations of the detector commute. Nor-mally for considerable angles this is definitely not true. However the angles in which the detector is free to move is in order of magnitude 10−6 rad and therefore so small it is assumed the rotations do commute.
The last simplification discussed, is the assumption of an instant accelerator. An important fact is that seismic waves of an earthquake are present for a longer time than the instant accelerator discussed in section 3.1. This longer acceleration could interfere with the first seconds of the tilt of the detector. Therefore it will not be swinging in its natural frequency instantaneous, but first will be powered for a limited time by the earthquake. This can be accounted for by cutting off the first seconds of the data of gyrating movement recorded by the tiltmeters. This approach is not done in this paper, but could be done in further research.
4
Results
The results of the gyrating movement of the detector and Fourier analyses is presented in this section as well as the data recorded by the Innoseis node.
4.1
Detector motion
The tilt during the discussed earthquakes is shown and next to it, the Fourier analysis with the same timescale in figures 13 and 14.
The gyrating movement during the 24 August earthquake is depicted in figure 13. Tiltmeter A is represented with a blue line, whereas tiltmeter B is represented with a red line. In the left figure the damped waveform of the detector swing is clearly visible. The Fourier analysis, depicted in the right figure, shows a peak at 0.06 ± 0.01 Hz for tiltmeter A, and 0.05 ± 0.01 Hz for tiltmeter B. These results differ from the estimation and are discussed in the discussion, section 5.
F. Bijloo A Seismic Analysis of the XENON1T Detector
0 20 40 60 80 100 120 140 160
time in seconds since 2016-08-24 03:36:31UTC +2
0.001 0.002 0.003 0.004 0.005 0.006 0.007tilt (mrad)
Tilt A
Tilt B
0.0 0.1 0.2 0.3 0.4 0.5frequency (Hz)
0.000 0.005 0.010 0.015 0.020 0.025 0.030amplitude (a.u.)
Tilt A
Tilt B
Figure 13: The tilt of the detector during the 24 August, 2016 M6.2 earthquake. On the
y-axis the tilt of the detector in mrad and x-axis the time in seconds. The blue line represents
tiltmeter A, whereas the red line represents tiltmeter B. (Right) The Fourier analysis of the
tilt of the detector during the same event. Tiltmeter A peaks at 0.06 ± 0.01 Hz and tiltmeter
B peaks at 0.05 ± 0.01 Hz.
Furthermore two small aftershocks of the 24 August event, which have been named foreshocks of the largest 30 October M6.6 event, are depicted in figure 14. These shocks have not been discussed before, but these quakes have the same origin as the 24 August M6.2 event. The Fourier analysis of aftershock 1 tiltmeter A peaks at 0.09 ± 0.02 Hz and tiltmeter B peaks at 0.07 ± 0.02 Hz. The Fourier analysis of aftershock 2 tiltmeter A peaks at 0.09 ± 0.06 Hz and tiltmeter B peaks at 0.07 ± 0.01 Hz. The exact times of occurrence are given in the figure. The average of natural frequency is calculated with the three earthquakes and equals 0.07 ± 0.02 Hz.
4.2
Noise
One hour of data of the tiltmeters is presented and next to it the Fourier analysis in figures 15, 16, 17 and 18. This data was taken on 10 October, 2016. Due to different patterns in oscillation 2700 seconds after the start of the data (9:50 UTC +2), propably because of human activity around the basin, the data is split in two parts. The results of the first 2700 seconds of then noise is depicted in figures 15 and 16. The results of the last 900 seconds of noise is depicted in figures 17 and 18. In the first figures 15 and 16 the Fourier analysis shows the highest peak around respectively 0.002 ± 0.001 Hz and 0.0007 ± 0.0001 Hz. As this is really low, the secondary peaks are calculated which are located at 0.02±0.07 Hz for tiltmeter A and 0.05±0.05 Hz for tiltmeter B. These relatively large errors are a consequence of the not well distinguished, broad, peaks. The Fourier analysis of last part of the noise, depicted in figures 17 and 18, peaks at 0.08 ± 0.02 Hz for tiltmeter A and 0.06 ± 0.02 Hz for tiltmeter B, which have better distinguished peaks.
0 20 40 60 80 100 120 140 160
time in seconds since 2016-08-24 03:55:56UTC +2
0.0034 0.0036 0.0038 0.0040 0.0042 0.0044 0.0046tilt (mrad)
Tilt A
Tilt B
0.0 0.1 0.2 0.3 0.4 0.5frequency (Hz)
0.000 0.001 0.002 0.003 0.004 0.005amplitude (a.u.)
Tilt A
Tilt B
0 20 40 60 80 100 120 140 160time in seconds since 2016-08-24 04:33:31UTC +2
0.0032 0.0034 0.0036 0.0038 0.0040 0.0042 0.0044 0.0046 0.0048tilt (mrad)
Tilt A
Tilt B
0.0 0.1 0.2 0.3 0.4 0.5frequency (Hz)
0.000 0.002 0.004 0.006 0.008 0.010amplitude (a.u.)
Tilt A
Tilt B
Figure 14: Tilt and Fourier analysis of the two aftershocks of the 24 August, 2016 M6.2
earthquake. The blue line represents tiltmeter A, whereas the red line represents tiltmeter
B. (Top left) The tilt of the detector during the first aftershock. On the y-axis the tilt of the
detector in mrad and x-axis the time in seconds. (Top right) The Fourier analysis of the tilt
of the detector during the same event. Tiltmeter A peaks at 0.09 ± 0.02 Hz and tiltmeter B
peaks at 0.07 ± 0.02 Hz. (Bottom left) Tilt of the detector during the second aftershock, and
(bottom right) Fourier analysis, where tiltmeter A peaks at 0.09 ± 0.06 Hz and tiltmeter B
peaks at 0.07 ± 0.01 Hz.
F. Bijloo A Seismic Analysis of the XENON1T Detector
0 500 1000 1500 2000 2500
time in seconds since 10/10/16 09:50UTC +2
0.004410 0.004415 0.004420 0.004425 0.004430 0.004435 0.004440tilt (mrad)
TILT A1
0.0 0.1 0.2 0.3 0.4 0.5frequency (Hz)
0.0000 0.0005 0.0010 0.0015 0.0020amplitude (a.u.)
Figure 15: Tilt and Fourier analysis of the first 2700 seconds of the noise, recorded by tiltmeter
A. The Fourier analysis peaks at 0.002 ± 0.001 Hz and secondary at 0.02 ± 0.07 Hz.
0 500 1000 1500 2000 2500
time in seconds since 10/10/16 09:51UTC +2
0.003795 0.003800 0.003805 0.003810 0.003815 0.003820 0.003825 0.003830 0.003835tilt (mrad)
TILT B1
0.0 0.1 0.2 0.3 0.4 0.5frequency (Hz)
0.0000 0.0005 0.0010 0.0015 0.0020amplitude (a.u.)
Figure 16: Tilt and Fourier analysis of the first 2700 seconds of the noise, recorded by tiltmeter
B. The Fourier analysis peaks at 0.0007 ± 0.0001 Hz and secondary at 0.05 ± 0.05 Hz.
The average of the secondary peaks of the first part and the normal peaks of the second part of the noise is calculated to be 0.05 ± 0.03 Hz.
4.3
Innoseis
The Innoseis’ Tremornet node was placed next to the Nikhef building in Amsterdam, The Netherlands. The exact location was 52.356◦ N, 4.951◦ E. The node measured the ground velocity from 22 October through the 30 October 2016. In this time two major earthquakes occurred in central Italy as discussed in the previous sections. In figures 19 and 20, with timescales respectively 26 October, 19:16:00 - 19:35:59 UTC and 30 October, 06:30:00 - 06:59:59 UTC, these events are shown. In both figures two subfigures are shown. In the right subfigures small variations are seen in the low-frequency region (0.0 - 3.0 Hz),
0 100 200 300 400 500 600 700 800 900
time in seconds since 10/10/16 10:34UTC +2
0.004410 0.004415 0.004420 0.004425 0.004430 0.004435 0.004440 0.004445 0.004450tilt (mrad)
TILT A2
0.0 0.1 0.2 0.3 0.4 0.5frequency (Hz)
0.0000 0.0005 0.0010 0.0015 0.0020amplitude (a.u.)
Figure 17: Tilt and Fourier analysis of the last 900 seconds of the noise, with noise probably
created due to human activity around the water basin of the detector, recorded by tiltmeter
A. The Fourier analysis peaks at 0.08 ± 0.02 Hz.
0 100 200 300 400 500 600 700 800 900
time in seconds since 10/10/16 10:35UTC +2
0.00372 0.00374 0.00376 0.00378 0.00380 0.00382 0.00384 0.00386tilt (mrad)
TILT B2
0.0 0.1 0.2 0.3 0.4 0.5frequency (Hz)
0.0000 0.0005 0.0010 0.0015 0.0020amplitude (a.u.)
Figure 18: Tilt and Fourier analysis of the last 900 seconds of the noise, with noise probably
created due to human activity around the water basin of the detector, recorded by tiltmeter
B. The Fourier analysis peaks at 0.06 ± 0.02 Hz.
F. Bijloo A Seismic Analysis of the XENON1T Detector
Figure 19: This figure shows the frequency spectrum measured by the Innoseis node. On the
vertical axis is given the time in seconds after 19:16:00 UTC, 26 October. On the horizontal
axis is the frequency in Hz. Red shows higher amplitudes at a given frequency, while blue
shows lower amplitudes. Because of the lesser sensitivity of the node to lower frequencies, the
same timescale is shown twice (filtered to 2.5 Hz and non-filtered). The left filtered (filtered
from 0.01 - 2.5 Hz) figure ranges from 0.0 to 3.0 Hz. The right unfiltered figure ranges from
0.0 to 30.0 Hz.
this frequency region is enlarged and shown in the left subfigures.
The 26 October earthquake occurred on 19:18:08. The first wave arriving to the Innoseis node is at 300 seconds after 19:16:00, so 19:21:00, which is seen in figure 19. The time difference between occurrence of the earthquake and the wave passing the node is ∆t ' 172 ± 5 seconds, while the distance from point of occurrence of the earthquake to the node is ∆x ' 1210 ± 20 kilometres, therefore
v1.1=
∆x
∆t ' 7.1 ± 0.2 km/s (11) is the speed of the first arriving wave. The curvature of the earth has not been accounted for. It has frequencies of 0.5 Hz to 1.0 Hz. After 480 seconds (312 ± 5 seconds after the occurrence of the earthquake) the most amplitude oscillation is seen, at lower frequencies
Figure 20: This figures shows the frequency spectrum measured by the Innoseis node. On the
vertical axis is given the time in seconds after 06:30:00 UTC, 30 October. On the horizontal
axis is the frequency in Hz. Red shows higher amplitudes at a given frequency, while blue
shows lower amplitudes. Because of the lesser sensitivity of the node to lower frequencies, the
same timescale is shown twice (filtered to 2.5 Hz and non-filtered). The left filtered (filtered
from 0.01 - 2.5 Hz) figure ranges from 0.0 to 3.0 Hz. The right unfiltered figure ranges from
0.0 to 30.0 Hz.
F. Bijloo A Seismic Analysis of the XENON1T Detector
(0.1 Hz - 0.6 Hz). This arrival speed is
v1.2 ' 3.9 ± 0.1 km/s (12)
The 30 October earthquake occurred on 06:40:19. The first wave arriving to the Innoseis node is at 780 seconds after 06:30:00, so 06:43:00, which is seen in figure 20. The time difference between occurrence of the earthquake and the wave passing the node is ∆t ' 161 ± 5 seconds, while the distance from point of occurrence of the earthquake to the node is ∆x ' 1220 ± 20 kilometres, therefore
v2.1 ' 7.6 ± 0.3 km/s (13)
was the speed of the first arriving wave. After 960 seconds (341 ± 5 seconds after the earthquake) the most amplitude oscillation is seen at lower frequencies. The arrival speed of these later arriving waves is
v2.2 ' 3.6 ± 0.1 km/s (14)
Both of the first waves arriving in figures 19 and 20 range between 0.5 Hz and 2.0 Hz. The second waves arriving range between 0.1 Hz and 0.5 Hz.
Also more data is presented to compare the results. The timescale for this data is 26 October 00:00:00 - 00:19:59 UTC. In this timescale no earthquake occurred, thus no large low-frequency seismic wave is observed, also around this time, less human activity was present. This is presented in figure 21.
5
Discussion
First the results are discussed and interpreted, afterwards some assumptions and simpli-fications are examined that could have lead to a misunderstanding of the results. Then mistakes and inaccuracies are discussed. At last some suggestions will be done for follow-up studies.
The measured natural frequency of 0.07 ± 0.02 Hz is a lot lower than the estimated frequency of 1 Hz, in equation 8. The restriction of one datapoint per second recorded by the tiltmeters, does not allow the Fourier analysis to be reliable above 0.5 Hz and not possible above 1.0 Hz. Therefore the estimated frequency of 1 Hz can not properly be tested with the current sampling frequency. Furthermore, though one datapoint is recorded every second, notice the value changes only every three seconds, therefore the reliability of the frequency spectrum is highly debatable and not accurate above 1
6 Hz. For
a proper analysis more datapoints per second are needed, which of course change value accordingly per datapoint.
Another possibility of misunderstanding the results could be a moving support struc-ture. In section 3.1, the analysis of the swing and calculation of the natural frequency
Figure 21: This figures shows the frequency spectrum measured by the Innoseis node, without
the occurence of an earthquake. On the vertical axis is given the time in seconds after 00:00:00
UTC, 26 October. On the horizontal axis is the frequency in Hz. Red shows higher amplitudes
at a given frequency, while blue shows lower amplitudes. Because of the lesser sensitivity of
the node to lower frequencies, the same timescale is shown twice (filtered to 2.5 Hz and
non-filtered). The left filtered (filtered from 0.01 - 2.5 Hz) figure ranges from 0.0 to 3.0 Hz. The
right unfiltered figure ranges from 0.0 to 30.0 Hz.
F. Bijloo A Seismic Analysis of the XENON1T Detector
require the assumption the detector is hanging from a stationary support structure. In reality, the support structure is free to oscillate and move. This will have an effect on the natural frequency and has to be further researched.
The Fourier analysis of the second part of the noise recorded by the tiltmeters show peaks (0.08 ± 0.02 Hz and 0.06 ± 0.02 Hz) that are similar to the peaks found in the earth-quake Fourier analysis. The Fourier analysis of tiltmeter B shows a well distinguished, but broad, peak, whereas the Fourier analysis of tiltmeter A shows a less distinguished peak. Furthermore the measured peaks of the first part of the noise show peaks at very low frequencies. The peak measured by tiltmeter B, 0.0007 ± 0.0001 Hz, correspond to a period of 1
0.0007 ' 1400 seconds. This period is actually clearly visible in figure 16. The
frequency peak of tiltmeter A of the first part of the noise, 0.002 ± 0.001, corresponds to a period of 0.0021 = 500 seconds. Also this waveform is visible in figure 15. The sec-ondary peaks of the first part were calculated, but the peaks are too low to be properly distinguished.
A misinterpretation of the results of the noise could be related to the accuracy of the tiltmeter. The tilt of noise in figures 15 and 16 is so small, it could be smaller than the accuracy of the tiltmeter itself. Therefore the possibility exists the measured tilt is not actual real tilt of the detector but noise created in the cables or digitalizers etcetera. This has not been accounted for and needs further research.
The Innoseis Tremornet node sensed and recorded the two earthquakes which are visible in the frequency spectrum in figures 19 and 20. It is clearly visible the node sensed lower frequencies of ground movement when an earthquake occurred, than without an earth-quake, figure 21. It is visible that the waves arriving first are of higher frequency than the later arriving waves, ranging from 0.5 Hz to 2.0 Hz, which is typical for large earthquakes [Shearer, 2009, Kennett, 2009]. This is as expected because of higher wave velocity and frequencies of P and S waves, compared to surface waves. At both earthquakes, the later arriving wave velocities vx.2 are lower than 60% of vx.1, with their frequency range from
0.1 Hz to 0.5 Hz, which is typical for surface waves of large earthquakes. Therefore it may be concluded that the first part of arriving waves represent the body waves, and the second part represent the surface waves. No distinction can be made between P and S waves in figures 19 and 20. To solidify these results, a sensitivity response curve for the geophone need to be taken into account. Because of time limitation, this has not been done. A response curve could be obtained by the producer of the geophones. It would mean that lower frequencies will be weighted more, therefore it will be easier to recognize an earthquake.
Obviously more activity is visible in the right subfigures of the Innoseis frequency spectra. For example a lot of movement is seen around 12 Hz and a concentrated line is seen at 23.1 Hz in all figures. Comparing figure 21 to figures 19 and 20, it is seen that less activity is observed in the frequency region of 3.5 Hz to 12.0 Hz in figure 21. This
figure shows the ground movement at night, at which human activity is less than at day. This has not been further researched. To know exactly which activity shows which peak could be a point of interest for further research.
6
Conclusion
In the second half of the year 2016, a sequence large earthquakes occurred in central Italy. At the 24thof August a moment magnitude M6.2 earthquake occurred, at a depth
of 4.4 ± 2.8 km, 10.0 ± 4.4 km Southeast of Norcia. At the 26th of October, a moment
magnitude M6.1 earthquake occurred, at a depth of 10.0 ± 1.7 km, 3.0 ± 4.9 km West of Visso. These were the largest foreshocks of the moment magnitude M6.6 main earthquake (at the time of writing) at the 30th of October, which occurred at a depth of 10.0 ± 1.7
km, 6.0 ± 5.1 km North of Norcia. These earthquake are the result of normal faulting in the Northwest-Southeast oriented mountain range of the Apennines, with extension perpendicular to the mountain chain. The Apennines are a geologically and tectonically complex region and in what way these earthquakes and the rest of the sequence is related to each other needs further research. A likely cause of the seismicity of the Apennines is gravitational spreading. An event where the inwards pushing force is removed, due to changing tectonics. Therefore the weight of the mountain causes the mountain to spread out, which is associated with normal faulting inside the rock, which is the most probable origin of the intraplate earthquakes discussed.
These earthquakes were sensed by the tiltmeters attached to the XENON1T detector in the Laboratory Nazionali del Gran Sasso (LNGS), a laboratory which is located inside the largest mountain of the Apennines. The seismic waves of the 24 August M6.2 earthquake and two of its aftershocks were treated as an instant accelerator to the XENON1T de-tector. The gyrating movement recorded by the tiltmeters is examined and using Fourier analysis an average natural frequency of 0.07 ± 0.02 Hz was found, which is lower than expected. The tiltmeters only recorded one datapoint every second, and changed value every three seconds. This does not allow the Fourier analysis to show frequencies higher than 0.5 Hz due to the recording of one datapoint per second, and frequencies above 16 Hz are unreliable, due to the non-changing value problem. The estimated natural frequency can therefore not be tested. To get a better understanding of the gyrating movement of the detector, a measurement should be repeated at a higher sampling frequency.
The discussed earthquakes were also sensed and measured by a Innoseis Tremornet node that was located in Amsterdam, lat: 52.3557 deg N, long: 4.9507 deg E. Variation in am-plitude at low frequencies (0.1 Hz - 2.0 Hz) was found at the expected times. The average velocity of the first arriving waves, P waves, were calculated to have been 7.1 ± 0.2 km/s and 7.6 ± 0.3 km/s, for the 26 October and 30 October events respectively. The frequency of these first arriving waves ranged between 0.5 and 2.0 Hz, which is as expected for P and
F. Bijloo A Seismic Analysis of the XENON1T Detector
S waves of earthquakes this size (M6.0 - M7.0). The lower frequency later arriving waves ranged between 0.1 and 0.5 Hz, which is as expected for surface waves of large earthquakes (>M6.0), and the velocity was calculated to have been 3.9 ± 0.1 km/s and 3.6 ± 0.1 km/s, for the 26 October and 30 October events respectively.. A proper distinction between the P and S waves has not been found. A network of these nodes could be installed in the area of the central Apennines to get a proper understanding of the internal structure of the mountain range, ruptures, fault lines and the earth’s crust. The advantages compared to installing seismographs is the size, weight, straightforwardness, robustness and expense of the nodes. A disadvantage is the reduced sensitivity to low frequencies, which makes the nodes less advisable for the use of earthquake research.
7
Acknowledgements
I would like to give credit to Mark Beker and the rest of Innoseis for lending us the Tremornet node and helping us interpret the results. Also the help of Hanneke Paulssen of the Geosciences group at Utrecht University was indispensable for the creation of this thesis, as she helped us understanding the 2016 earthquake sequence and seismology as a whole. My supervisor, Auke-Pieter Colijn, made my time at Nikhef highly enjoyable and helped me greatly with his enthusiastic, though critical, view. At last I would like to thank the Dark Matter group at Nikhef for taking me into the group and helping me where needed.
References
[Almeida, 1994] Almeida, L. B. (1994). The fractional fourier transform and time-frequency representations. IEEE Transactions on signal processing, 42(11):3084–3091. [Aprile et al., 2012] Aprile, E., Alfonsi, M., Arisaka, K., Arneodo, F., Balan, C., Baudis, L., Bauermeister, B., Behrens, A., Beltrame, P., Bokeloh, K., et al. (2012). Dark matter results from 225 live days of xenon100 data. Physical review letters, 109(18):181301. [Chiarabba et al., 2009] Chiarabba, C., Amato, A., Anselmi, M., Baccheschi, P., Bianchi,
I., Cattaneo, M., Cecere, G., Chiaraluce, L., Ciaccio, M., De Gori, P., et al. (2009). The 2009 l’aquila (central italy) mw6. 3 earthquake: Main shock and aftershocks. Geophysical Research Letters, 36(18).
[Chiarabba et al., 2005] Chiarabba, C., Jovane, L., and DiStefano, R. (2005). A new view of italian seismicity using 20 years of instrumental recordings. Tectonophysics, 395(3):251–268.
[Devoti et al., 2008] Devoti, R., Riguzzi, F., Cuffaro, M., and Doglioni, C. (2008). New gps constraints on the kinematics of the apennines subduction. Earth and Planetary Science Letters, 273(1):163–174.
[Galadini and Galli, 2000] Galadini, F. and Galli, P. (2000). Active tectonics in the central apennines (italy)–input data for seismic hazard assessment. Natural Hazards, 22(3):225–268.
[Galli et al., 2002] Galli, P., Galadini, F., Moro, M., and Giraudi, C. (2002). New pale-oseismological data from the gran sasso d’italia area (central apennines). Geophysical Research Letters, 29(7).
[Giancoli, 2008] Giancoli, D. C. (2008). Physics for scientists and engineers with modern physics, volume 2. Pearson Education.
[IRIS, 2016] IRIS (2016). Mwc 6.2 central italy. Retrieved from http://ds.iris.edu/spud/event/5191298.
[Jackson and McKenzie, 1983] Jackson, J. and McKenzie, D. (1983). The geometrical evolution of normal fault systems. Journal of Structural Geology, 5(5):471–482. [Kanamori and Brodsky, 2004] Kanamori, H. and Brodsky, E. E. (2004). The physics of
earthquakes. Reports on Progress in Physics, 67(8):1429.
[Kennett, 2009] Kennett, B. (2009). Seismic wave propagation in stratified media. ANU Press.
[Martins et al., 2007] Martins, Silveira-Deto, and Jr, S. (2007). A pendulum-based model for fluid structureinteraction analyses. Engenharia Trmica (Thermal Engineering), 6:76–83.
[Meijer, 2016] Meijer, P. (2016). Tectonic context of the recent earthquakes of central italy. Presentation.
[Paulssen, 2016] Paulssen, H. (2016). Personal communication.
[Shearer, 2009] Shearer, P. M. (2009). Introduction to seismology. Cambridge University Press.
[Sibson, 1977] Sibson, R. (1977). Fault rocks and fault mechanisms. Journal of the Geological Society, 133(3):191–213.
[USGS, 2016a] USGS (2016a). M 6.1 - 3km nw of visso, italy. http://earthquake.usgs.gov/earthquakes/eventpage/us1000725y.
[USGS, 2016b] USGS (2016b). M 6.2 - 10km se of norcia, italy. Retrieved from http://earthquake.usgs.gov/earthquakes/eventpage/us10006g7d.
[USGS, 2016c] USGS (2016c). M6.6 - 6km n of norcia, italy. http://earthquake.usgs.gov/earthquakes/eventpage/us1000731j.
[Votano, 2012] Votano, L. (2012). The gran sasso laboratory. The European Physical Journal Plus, 127(9):1–12.
[Walker et al., 2012] Walker, J. F., Roberts, G. P., Cowie, P., Papanikolaou, I., Michetti, A., Sammonds, P., Wilkinson, M., McCaffrey, K., and Phillips, R. (2012). Relationship between topography, rates of extension and mantle dynamics in the actively-extending italian apennines. Earth and Planetary Science Letters, 325:76–84.
F. Bijloo A Seismic Analysis of the XENON1T Detector
8
Appendix: Summary
8.1
Summary
The largest earthquakes that occurred in central Italy as part of the 2016 sequence are discussed, with their sources explained. During these earthquakes the XENON1T detector of the xenon collaboration housed in the underground laboratory LNGS in central Italy oscillated. The tilt of the XENON1T detector during some of the earthquakes is examined and a natural frequency of 0.07 ± 0.02 Hz of the generated movement was found through Fourier analysis, which is lower than expected. Also one Innoseis’ Tremornet node, which was placed in Amsterdam, The Netherlands, has measured the 26 October and 30 October earthquakes, of which the frequency spectra are presented.
8.2
Samenvatting
Itali¨e is een seismisch zeer actief gebied. Er komen veel zware aardbevingen voor met een magnitude groter dan M6.0. In de tweede helft van 2016 vond een serie zware aardbevingen plaats in de Apennijnen, centraal Itali¨e. De oorsprong van deze serie aardbeving is in deze paper beschreven. Tijdens deze aardbevingen schommelde de XENON1T detector in het ondergrondse laboratorium in Gran Sasso, centraal Itali¨e. Een eigenfrequentie, de frequentie waarin het systeem schommelt zonder aandrijving, is gevonden door middel van een Fourier analyse van de tilt van de detector tijdens de 24 augustus M6.2 aardbeving en twee naschokken. De eigenfrequentie van het schommelen van de XENON1T detector is vastgesteld op 0.07 ± 0.02 Hz. Daarnaast is een seismische sensor van Innoseis’ Tremornet getest in Amsterdam en de resultaten, waarin de 26 oktober 2016 en 30 oktober 2016 aardbevingen te zien zijn, worden in deze paper gepresenteerd.
(a) Frequentie spectrum van de grondbeweging voor het Nikhef gebouw, Amsterdam, op 30 okto-ber van 06:30:00 tot 06:59:00. Op de linkerfiguur is duidelijk de beweging door de M6.6 30 oktober 2016 aardbeving te zien. 0.0 0.1 0.2 0.3 0.4 0.5 frequency (Hz) 0.000 0.001 0.002 0.003 0.004 0.005 amplitude (a.u.) Tilt A Tilt B
(b) Een Fourier analyse van de tilt van de XENON1T detector in Gran Sasso, tijdens een naschok van de M6.2 24 augustus 2016 aardbev-ing in Itali¨e.