High Activity Earthquake Swarm Event Monitoring and Impact Analysis on Underground High Energy Physics Research Facilities

Abstract

A seismic swarm is a series of earthquakes that occur in a small area over a short period of time. A sequence of earthquakes of this magnitude is unusual in Switzerland, and it is impossible to anticipate how it may unfold in the future.The seismic activity of such an event usually fades after a few days or weeks. Significantly greater earthquakes are likely to occur during the next several days, with up to a chance of 5 to 10%. For these reasons, the underground research facilities need tools to provide data on the impact of these events on their experiments. The paper presents the techniques implemented at The European Organization for Nuclear Research (CERN) to allow the tracking and monitoring of these unusual events. Additionally, the real effect of such an unusual event is presented together with the statistical approach to monitoring and effect evaluation. Considering the collision energy of the beams at 14 TeV, the energy stored in the magnets at 10 GJ (2400 kg of TNT), and the energy carried by the two beams at 724 MJ (173 kg of TNT), prolonged exposure to vibration close to or above the set alarm levels may result in serious safety issues. The presented evaluation of earthquake swarm impact on underground facilities together with the approach for data evaluation can be used for the design of future detectors and accelerators. Additionally, it provides tools for facilities users to present the data in an easy to understand way. This includes the Future Circular Collider, whose purpose is to significantly expand the energy and intensity frontiers of planned particle colliders, with the goal of reaching collision energies of 100 TeV in the quest for novel physics. As a result, even greater standards for beam size and stability will be required.

1. Introduction

To prove theories and look for new questions in the field of experimental physics, underground accelerators have been built around the world. The largest and most well-known, the Large Hadron Collider (LHC), is located near Geneva, on the French–Swiss border. This location has been chosen for many factors; however, one of the crucial aspects of the experiments is the seismic stability of the area [1]. Due to current and future advancements, the mechanical stability of high energy accelerator components will become increasingly important: high luminosity physics implies smaller beam sizes, improved structure vibration safety, and a better understanding of environmental mechanical noise to identify and resolve any harmful impacts on accelerator performance [2,3,4].

It is important to note that the vibration stability of structures is dependent on both the structure’s dynamic behaviour and the ambient circumstances in which it is installed. Dynamic behaviour of the LHC was extensively researched in the past [5,6], however, there is still much to be done in the case of describing the effects of specific environmental conditions and unique events. The effects of a seismic waves are often devastating for the experiments at CERN, which results in time and money lost. When a seismic wave hits the magnets in the LHC tunnels, it causes changes in beam position (orbit) all around the perimeter. Fast and large amplitude changes lead to a beam abort and a need to restart the acceleration process again, which lasts many hours [7,8]. The seismic effect of small-magnitude earthquakes is reduced by the LHC safety system, which may safely abort beams without causing damage to the accelerator components [9,10,11,12].

The integrated luminosity will be affected by the recurrence of beam aborts over time [11]. This is linked to the accelerator’s availability for the creation of physics data, and thus results in higher operating expenses. This is crucial for future projects and improvements such as the High-Luminosity LHC (HL-LHC), which, due to higher energies and lover beam sizes, will be affected not only by some natural events but also by civil engineering work (both surface and underground) and even the geothermal exploitation in the Geneva canton [13].

To monitor the dynamic condition regardless of its character, CERN has installed three seismic stations that are connected to the world seismic network with data availability when accessing the specific station. The network’s objective is to collect data on seismic activity. This data already allowed for monitoring the conditions during civil engineering work on new access shafts and chambers during LHC operation [4] and for developing new software and analyzing techniques for future monitoring and implementation [14].

Figure 1 shows an overview of a sensor chamber for one of the subterranean seismic stations (Figure 1a), as well as an example of LHC beam orbit oscillations driven by the moon’s gravity and a sequence of earthquake waves propagating through the Earth’s crust from the Australasia region (New Zealand) [2].

Although the primary aim of the CERN seismic station is to monitor dynamic conditions in the tunnel to characterize ambient noise, to gauge the influence of the tunnel work on amplitudes measured, and to get statistical knowledge on natural events like earthquakes, it is also important to analyze unexpected events. One such event was a high intensity earthquake swarm in which the epicentre was of a local character. Earthquake swarms are clusters of earthquakes that happen in rapid succession (hours or days) without a single mainshock [16,17]. During the examined case, in just one week, more than 300 earthquakes were recorded with magnitudes up to 3.3. This paper presents the effects of this event recorded in two crucial locations in the LHC tunnel. The data are especially useful to include the possibility of such a situation in case of future colliders.

2. Materials and Methods

2.1. CERN Seismic Network

Three seismic sites form the CERN seismic network (which is part of the Swiss Seismic Network). There are two subterranean units, one near the ATLAS detector (named Point 1), the other near the CMS detector (called Point 5), and one surface unit at the centre of the LHC accelerator loop and distant from any infrastructure or human impacts. During normal operating conditions and while the LHC is in operation periods and no other activity is done, the underground station detects even small amplitude vibration. In contrast, during technical operations like technical stops, so-called long LHC shut-downs (maintenance periods), or during civil engineering activities, these measurement systems experience a lot of background noise. This is the purpose of the surface station: to be a reference in case of any data analysis performed on the data from the tunnels. Finally, the CERN seismic station can be used to look into unique seismic events such as an earthquake swarm, which is highly unexpected in this region. Some possible data analyses for the earthquake swarm will be presented in the following sections.

The location of three CERN seismic stations can be seen in Figure 2a, together with the location of all Swiss Seismic Network stations. It must be pointed out that CERN applications require monitoring the dynamic conditions in a wider range than typical for seismic stations. As a result, one geophone is insufficient to cover the whole range of vibrations required by the CERN seismic station.

Strong motion sensors have considerable noise, which prevents them from monitoring low amplitude motions, but they might detect large ranges, such as the 2 g accelerations required for this network. For low amplitude tremors, a wideband seismometer is utilized. Seismometers that measure low frequency signals (30 s) are not often built to monitor up to 100 Hz [2].

Each of the three sensing systems consists of two sensors from the initiation of the CERN seismic station operation in 2017: a broadband seismometer and a strong motion sensor. These include GuralpTM 6T broadband seismometers for subterranean stations and 40 T broadband seismometers for surface stations, with a bandwidth extension up to 100 Hz. The Episensor ES-T from KinemetricsTM is a powerful motion sensor. A 12 V switching mode power source with fewer than 50 mVpp ripples provides power to them.

2.2. Alarm Level Threshold for CERN Underground Experiments

Considering the past observations of vibration effects on the LHC operation, especially the proton loss and luminosity, it was decided to calculate warning and alarm levels for velocity ground measurements. These have proven to be valuable to the team in charge of the projects as a rapid indicator of the level of vibration induced by the construction equipment [4]. Crossing the alarm level is associated with significant parameters for physic evaluation decrease that in many cases leads to experiments aboard and reaction from LHC protection system.

According to the previous observations, the limits were set as follows:

• Warning level 0.78 µm/s;
• Alarm level 1.56 µm/s;
• Additionally, the first two and most significant natural frequencies of the magnets positioned in the tunnel, 8 and 22 Hz, reduce dropout within the frequency spectrum of 7–28 Hz. The magnets and the beam travelling through them will be more affected by ground motions at these frequencies. The warning level was set at 0.39 m/s, and the alert level was set at 0.78 m/s.

Figure 3 presents the example of measurement where the alarm level was breached. The yellow and red horizontal lines represent the warning and alert levels, respectively, determined for the LHC vibration activity.

The consequence of going above the alarm level (red line) is the luminosity loss of the beam. Moreover, keeping in mind the beams’ collision energy of 14 TeV, the entire energy contained in the magnets is 10 GJ (2400 kilos of TNT), whereas the total amount of energy transported by the two beams is 724 MJ (173 kilos of TNT). Prolonged exposure to vibration close or above the alarm level may result in serious safety issues.

The evaluation of ground motion impact and placement of warning and alert levels has proven valuable to the team monitoring not only seismic occurrences but also work on the surface and in the tunnel as a rapid indicator of the degree of vibration produced by those events.

3. Results

According to previous analyses performed for natural events like earthquakes [19] or human activities like civil engineering work both in the underground chambers and on the surface [15], they have a significant effect on accelerators’ complex operations. The case of breaching safety limits (assigned alarm levels) often leads to experiment termination. The protection system reacts and the beam absorbs the energy of photons or other particles in an energetic beam. The so-called “beam dump” process is executed. The restart of the experiment takes several hours, which is associated with significant time and money lost.

It was also evaluated that even low magnitude but local seismic events will have a similar effect on CERN accelerators’ complex operation. Thus, the earthquake swarm will have an even more profound effect due to a large number of events over just several days. This kind of seismic occurrence has not been researched in case of impact on CERN facilities. In this section, the effect of such an event will be presented together with possible data analysis. The data can be used for designers of future colliders as one of the cases to include in their analysis.

3.1. Earthquake Swarm

An earthquake swarm can be described as a cluster of tremors that occur over a period of hours up to days with no identifiable mainshock. Such an accumulation of seismic events is atypical for Switzerland. However, during the night between 4 and 5 November 2019, an unusually active earthquake swarm began north of Sion near Saviese town in Switzerland. More than 100 incidents were logged in only one day by the Seismological Service at Zurich (SED) seismic network. Twelve of them had magnitudes of 2.5 or more and were felt by residents from Sion to Sierre. These tremors were also detected by CERN seismic network and in the case of the LHC operation period would have influenced the experiments. The four most powerful quakes had magnitudes ranging from 3.0 to 3.3. Figure 4 presents the origin of the earthquake swarm in comparison to the CERN underground facilities.

Forecasting the evolution of this series is impossible. Seismic activity often fades after a few days or weeks. Nevertheless, there is a 5 to 10% chance that substantially greater earthquakes will occur during the next few days [20].

Figure 5 presents the evolution of the analyzed earthquake swarm. The graph presents detected seismic events during one week together with the magnitude. The red line presents the accumulated number of events. The rapidity of event accumulation in the first 24 h, with more than 100 events recorded, is visible.

3.2. Analysis of the Swarm Event Impact on LHC Tunnel

According to SED, in Switzerland, the last strongly damaging quake, a magnitude of 5.8, occurred on 25 January 1946. Usually, the earthquakes in this region are not stronger than magnitude 3. For a typical year, the SED counts around 1200 tremors (in the whole of Switzerland). Approximately 10 to 20 earthquakes are felt per year with a magnitude of 2.5. In the case of the earthquake swarm, the number of felt earthquakes is considerably higher than for a whole typical year. Additionally, the impact of this kind of event can be observed in underground research facilities like CERN (France/Switzerland) and the European Synchrotron Radiation Facility (ESRF) located in Grenoble, France. In normal conditions, statistically, approximately only seven earthquakes per month trigger the LHC seismic stations alarm level. Statistically, only one of these can be considered a local seismic event.

Table 1. Major events of the swarm above LHC alarm level threshold.

No	Event Time [UTC]	Mag. [-]	V max P1/P5 [µm/s]	V RMS P1/P5 [µm/s]
1	5 November 2019 00.54	3.3	12.0/14.1	5.0/5.8
2	6 November 2019 03:36	2.7	14.2/16.8	5.9/7.2
3	5 November 2019 05:55	2.9	8.0/22.0	2.3/3.5
4	5 November 2019 06:00	2.9	8.0/15.0	2.6/4.8
5	5 November 2019 07:18	3.0	9.3/7.8	4.1/4.8
6	5 November 2019 07:47	2.7	6.7/10.8	3.5/4.9
7	5 November 2019 08:54	2.7	8.2/14.3	3.4/4.7
8	5 November 2019 19:21	2.8	7.4/12.5	3.7/5.3
9	5 November 2019 19:51	3.2	6.9/11.5	3.6/5.1
10	5 November 2019 19:52	2.8	6.9/11.5	3.6/5.1
11	5 November 2019 20:06	2.7	7.1/13.5	3.2/2.5
12	5 November 2019 20:10	2.5	4.0/10.1	2.0/2.9

displays the list of alarm lever thresholds surpassed on the first day of the earthquake swarm (5 November 2019) at two underground seismic station locations P1 (near the chamber of the ATLAS detector) and P5 (near the chamber of the CMS detector).

3.2.1. Impact Observation on the Underground LHC Tunnel

Although providing the large amount of data that are stored and easily accessible is useful, it is equally important to find a way to process these data to acquire useful insight into the vibration behaviour of the station. When evaluating vibration, a common strategy is to convert time-domain quantifiable information to frequency-domain qualitative information, such as power spectral density.

Earthquakes are rare events. How they arrive at the LHC tunnel depends on the magnitude and distance of the epicentre [21]. The beam response and potential resonant amplification of the induced ground movement by (parts of) the LHC elements depend on the propagation direction and wavelength. Many similar occurrences were seen on the LHC beams at various stages of the operating cycle, as illustrated in Figure 6. Effects were measured over time scales from a few seconds up to one hour and with varying impact. The main observables are individual orbit and loss signatures on the primary collimators, as well as variation of the luminosity if the event happened during the collision period. Each dataset clearly shows many ground excitation episodes.

During the earthquake swarm event, it was not possible to see the effect on the beam due to the service shutdown of the experiments (so-called Long Showdown 2). However, due to previous studies, observation and placement of the alarm threshold (whose breach is connected to visual effects on the beam data) it is known which events, in normal conditions, would have impacted the experiments (Table 1).

Additionally, some tools were developed to analyze the data in detail. Normally, the data from CERN seismic stations are transferred to SED servers continuously. However, with additional Python-based software, it is possible to plot velocity-time data over long periods of time.

Figure 7 shows the impact registered in the tunnel at Point 1 (ATLAS experiment) for one of the events above the alarm threshold for LHC (on 6 November 2019) of initial magnitude 2.7.

Observations during a long period of time allow studying aftershocks of lower amplitudes connected to other events from this earthquake swarm. The event observed on 6 November 2019, 03:36 AM (UTC) was detected by all three CERN seismic stations and can be plotted in detail as a velocity plot (Figure 8).

For the given event, maximal velocity over 1s and velocity RMS (Figure 9) can also be shown.

3.2.2. Probabilistic Power Spectral Density for the Period of Highest Activity of the Earthquake Swarm Event

To observe the impact of the vibrations over a longer time, it is possible to use probabilistic techniques. Calculating and visualizing probabilistic power spectral density (PPSD) is one option [14,22].

For a typical approach where only power spectral density is calculated (PSD), there is a limitation while using this procedure. The data obtained and used for creating the plot have to be stationary. Thus, commonly this kind of analysis is performed for short time blocks (few minutes) when usually no unexpected events are detected by the sensors. However, for extensive analysis when data are to be analyzed for hours, days, months, etc., this method is no longer valid. A possible alternative is the utilization of a probabilistic approach, especially the PPSD method.

The procedure is as follows:

• Long-time measurement data are partitioned into several minute blocks (segments);
• For each of the data blocks, a PSD function is calculated using a fast Fourier transform (FFT). These PSD functions are being stored for further processing. The amplitude of each PSD function is converted from typical units of power (as calculated from velocity) to dB range (under the assumption that the reference velocity);
• The next step is the assembly of the PPSD graph from the PSD data. This process is performed by discretizing the amplitude graphs by value (usually with a discretization step of 1 dB) and counting the number of times the value of amplitude falls within a specified range (e.g., between −99 and −100 dB);
• The same procedure is performed in the loop for every amplitude of every frequency octave range of every obtained PSD graph;
• Finally, based on the tallied data vs. the total amount, the percentages are calculated. These percentages are plotted to form the final PPSD graph at the specified time slot.

The PPSD method was used to obtain the data for the week when the highest number of events and amplitudes were observed during the analyzed earthquake swam. Figure 10 presents the vertical direction PPSD graphs obtained next to the ATLAS detector (CERN Point 1) and CMS detector (CERN Point 5) on a week where no specific seismic events were detected (Figure 10a,b) and a week of the highest activity of the earthquake swarm in 2019 (Figure 10c,d).

From the PPSD graphs, the characteristic of ground motion is noticeable as well as the statistical impact on the tunnel where the most precise machines and systems are installed. When looking at Figure 10, three major zones are crucial for analysis.

• Zone 1—has a frequency range of 0.1 to 1 Hz. This corresponds to the so-called microseismic motion induced by Earth’s ocean movement;
• Zone 2—has a frequency range of 1 Hz to 10 Hz. the range consists of “cultural noise”, which is associated with human activity, such as operating machines, manufacturing, public transport, etc., and in case of a statistically large number of events also seismic events;
• Zone 3—has a frequency range of 10 Hz to 100 Hz and is usually a stable line as seismic stations are located approximately 100 m below the surface and thus are not usually affected by surface noise or this effect is statistically negligible. As a result, this section of the graph depicts the tunnel’s true ambient noise. Due to earthquake swarm effects, however, the ambient noise in the tunnels also increased.

The grey lines in Figure 10 refer to the so-called updated high noise model (higher curve) and updated low noise model (bottom curve). These lines indicate the greatest and lowest observed amounts of natural Earth noise sources in the ambient environment.

4. Conclusions

Considering the plans for improvement of the current HL-LHC project and the construction of new accelerators and detectors (Future Circular Collider), the identification and proper analysis of seismic vibration impact are important. It is also essential for the accuracy of elementary particle behavior in particle accelerators. In the case of CERN infrastructure, the current tunnels will be also extensively used for future experiments that will require an even higher level of dynamic stability. Thus, the constant monitoring of the infrastructure and the consideration of extremally rare events are important. The ground motion impact is now one of the most critical elements restricting the operation of newer accelerators.

In normal conditions, approximately only seven earthquakes a month are of sufficient energy to breach the alarm threshold set for the CERN seismic stations. Approximately only one of these is considered a local quake (a distance less than 150 km from the facility). In the case of the examined situation, just in the first 24 h twelve such events were recorded, and considering that the whole swarm lasted over one week, in normal conditions this would have a significant influence on the experiments. Breaching the LHC alarm threshold is connected with visible parameter drops for LHC physics (proton loss, luminosity, etc.), and in many cases is a reason for triggering the accelerator protection system and beam extraction process. From the analysis presented in this paper, during the earthquake swarm, alarm levels were breached multiple times and most of the stronger events had additional aftershocks. The protection system, during normal LHC operations, would, in most cases, react and the beam would absorb the energy of photons or other particles in an energetic beam. The so-called “beam dump” process would be executed. The restart of the experiment takes several hours. During this time, further evens are to be expected, making it impossible to proceed with experiments. This is then an important situation to take into account in case of future systems designs.

Additionally, as noted in Section 3.2.2, the probabilistic approach to data analysis presents the data over long periods. The statistical data of the PPSD, normally, are not influenced by unusual events like earthquakes that normally only occur from time to time and, in the case of the CERN location area, are of low magnitude. However, an unusual event like a local earthquake swarm consisting of over 300 low magnitude tremors will affect the statistical data. It is visible, especially in the area of 1 Hz to 10 Hz. This is further proof that the accumulation of events, which normally would be mitigated, will have a profound effect and impact on the LHC tunnel.

Finally, the selected measurements presented in the paper will be useful for determining the possible vibration levels when an unusual local seismic event occurs. Due to the local character, fast wave propagation, and limited time for the reaction, an event like an earthquake swarm is challenging for current protection systems. The presented analysis will be also interesting for other research facilities where seismic protection systems are being designed.