Einstein Telescope

The Einstein Telescope (ET) is a planned European third-generation gravitational-wave observatory. Building on the discoveries made by LIGO and Virgo, it aims to achieve approximately ten times better sensitivity than current gravitational-wave detectors and to extend observations to frequencies as low as a few Hertz. The underground observatory will use highly sensitive laser interferometers to measure the minute distortions of spacetime produced by gravitational waves.

The Einstein Telescope will open new possibilities for studying the Universe, including observations of merging black holes and neutron stars across cosmological distances, precision tests of general relativity, and investigations of the early Universe. Its unprecedented sensitivity will also enable detailed studies of the properties of compact objects and improve the prospects for multi-messenger astronomy, in which gravitational-wave observations are combined with electromagnetic and other astronomical signals.

The challenge of Newtonian noise

One of the major technological challenges for the Einstein Telescope is achieving the required sensitivity at low frequencies. Seismic waves propagating through the ground produce small variations in the local mass distribution. These variations exert fluctuating gravitational forces on the interferometer mirrors, giving rise to so-called Newtonian noise. Unlike mechanical vibrations, gravitational forces cannot be shielded, making this noise particularly difficult to suppress.

A promising solution is active noise mitigation (ANM). Arrays of seismometers installed around the interferometer mirrors measure the surrounding seismic wave field. Advanced algorithms, including artificial neural networks, can use these measurements to predict the resulting gravitational disturbance. The predicted noise can then be subtracted from the recorded data or compensated in real time through the interferometer control system. Real-time mitigation is particularly attractive because it can improve the sensitivity of low-latency gravitational-wave searches and enable more rapid follow-up observations.

Our research activities at Siegen

Our group has recently joined the Einstein Telescope research effort, contributing expertise in detector electronics, real-time data processing, and machine learning developed in our work on the ATLAS experiment at CERN and the Pierre Auger Observatory. Our activities focus on developing hardware systems that allow AI-based noise-mitigation algorithms to operate continuously with very low and predictable processing latency.

Within the BMFTR-funded project FPGA-based Building Blocks for Active Noise Mitigation, we are developing real-time processing architectures based on field-programmable gate arrays (FPGAs). These programmable electronic devices can execute many operations in parallel and are particularly well suited to demanding real-time applications.

Our research addresses three closely connected areas:

• Hardware-aware optimisation of neural networks: We adapt machine-learning models to the limited numerical precision and computing resources available on FPGAs. Techniques such as quantisation-aware training and high-granularity quantisation allow us to reduce the complexity of neural networks while preserving their prediction accuracy.
• Real-time FPGA inference: We develop streaming processing architectures capable of handling multiple seismic sensor channels simultaneously. Particular attention is given to deterministic low-latency execution, efficient use of FPGA resources, and scalable integration into the interferometer control infrastructure.
• Hardware validation and data emulation: We develop FPGA-based systems that reproduce realistic seismic sensor data streams. By replaying simulated and recorded signals, we can test the complete processing chain and evaluate its latency, throughput, resource requirements, and scalability under controlled conditions.

Collaboration and experimental validation

Our activities are part of a coordinated research programme on Newtonian-noise mitigation with partners at RWTH Aachen University, FH Aachen, Forschungszentrum Jülich, and TU Dortmund University. While our partners develop noise-prediction algorithms, sensor-level preprocessing, and distributed processing concepts, the Siegen group concentrates on the central real-time FPGA implementation.

An important part of the programme is the experimental validation of the developed methods. This includes laboratory-scale interferometers and the ETpathfinder facility in Maastricht, which provides a test environment for technologies required by future gravitational-wave observatories. Although Newtonian noise itself cannot be measured directly at ETpathfinder, analogous noise-mitigation techniques can be investigated using acoustic disturbances and environmental sensors.

Through this research, we aim to help translate promising machine-learning algorithms into practical, energy-efficient, real-time systems capable of meeting the demanding requirements of the future Einstein Telescope.