Characterizing Newtonian Noise for ET

Karlsruhe Institute of Technology (KIT) is contributing to the Einstein Telescope (ET) preparatory program through a two-pronged project focused on solving the challenge of Newtonian Noise (NN) cancellation — a key technical barrier for third-generation gravitational-wave observatories. The program combines experimental testbeds at the world-renowned Black Forest Observatory (BFO) with cutting-edge numerical simulation and artificial intelligence approaches.

This work is part of the Preparatory Program for the Einstein Telescope (ET-VorPro), within the joint project Newtonian Noise Cancelling Headphones — Propagation, Reconstruction, Evaluation, and Digital Twin Integration for Characterizing Newtonian Noise for ET.

The Challenge

Newtonian Noise is caused by changes in the spatial distribution of masses — such as bulk compression and displacement of mass in heterogeneous geological structures near the telescope's test masses. Cancelling this noise requires a precise, and potentially real-time, representation of the deformation field of the Earth's crust in the vicinity of the ET test masses, measured with an optimal array of seismometers, strainmeters, and similar sensors.

Project 1: BFO as a Testbed for Wavefield Interpolation

The Black Forest Observatory — a former silver mine in granite with exceptionally low background noise (down to −150 dB) and up to 170 m of overburden — serves as an ideal testbed. The project will:

  • Expand the existing sensor infrastructure to build a comprehensive local array with a three-dimensional sensor layout
  • Establish best-practice guidelines for sensor installation and coupling to achieve the accuracy needed for NN cancellation
  • Test newly developed sensors (including volume strainmeters) from internal and external partners
  • Investigate the optimal spatial arrangement and minimum number of sensors required

Key research questions include how to best combine seismometers and strainmeters, how to ensure accurate sensor coupling and calibration, and how to measure cavern deformation with a finite sensor count.

Project 2: Seismic Wavefield Simulation and Newtonian Noise

This project combines physics-based wave-propagation modeling with deep-learning AI for deterministic, quantitative estimation of Newtonian noise in 3D heterogeneous structures. Its two main objectives are:

  1. 3D Seismic Wavefield Simulation — Using state-of-the-art spectral element software (SPEFEM3D and Salvus) to model multi-source wavefields and explore how complex geological features (topography, sedimentary cover, tunnels) influence the seismic wavefield on which NN computation is based.
  2. Semi-Real-Time NN Mitigation via Foundation Models — Developing a novel masked diffusion model approach for seismic waveform interpolation and extrapolation. This is a first-of-its-kind foundation model targeting multi-station and large-N seismic datasets, aiming to generate NN estimates and predict seismic noise levels at the mirror position.

Synthetic scenarios will be validated against observed seismic wavefields at BFO and the ET candidate sites (Lausitz, Sardinia, EMR).

Collaboration

The project draws on extensive collaboration with partner institutions across Germany and international partners, including Gran Sasso, INFN, NIKEF, and the University of Montpellier.