LIGO-India EMR Projects

One year of uninterrupted seismic data was collected at the planned LIGO-India observatory site using an array of five seismometers.

The objective of this study is to use that data to understand the seismic environment at the LIGO-India observatory site, specifically:

• (i) to understand the underground geological structures at and around the site
• (ii) to identify and characterize any low-threshold local seismic events
• (iii) to identify and localize any continuous and transient seismic noise sources arising from natural and human activities around the site
• (iv) to identify and characterize any low-frequency seismic noise produced by the nearby water reservoir

The LIGO-India beam tube, which is 4 km long with a 1.24-meter inner diameter, is maintained in a UHV (Ultra High Vacuum) regime during detector operation. Past experiences from the LIGO observatories in the USA have shown that locating tiny leaks can be a time-consuming and challenging task. Therefore, the design of the LIGO-India beam tube must consider this issue to make leak detection more manageable. The size of the leaks we need to localise is as small as 1 x 10-9 mbar L/s, which is critical for maintaining the integrity of the UHV environment necessary for the detector's operation.

To achieve this, we need a model that simulates both the steady-state and transient pressures for this geometry, incorporating all sources and sinks for several gas species. The primary goal is to determine optimal locations for Residual Gas Analysers (RGAs) based on this analysis, enabling us to localise leaks to within 100 meters through static analysis. Such models will also facilitate the localisation of leaks using Markov Chain Monte Carlo (MCMC) methods to find the maximum likelihood posteriors based on the collected data. This includes modelling the beam tube layout to evaluate the placement of RGAs and other accessories.

Proposals are invited in this domain to prepare models of the LIGO-India Vacuum envelope and propose dynamic and static leak testing methods to localise leaks. The techniques and methods proposed may be tested at the LI-VISTA facility at IPR.

The LI-VISTA facility features an 80K Cryopump as a key component of its assembly. The internal bath of this pump has a reservoir capacity of approximately 800 liters, which is cooled by liquid nitrogen (LN2) to effectively absorb water molecules. The 80K Cryopump plays a crucial role in removing heat transferred through radiation, conduction, and the accumulation of condensate on its cold surface. However, during the evaporation of LN2, bubbles can form, potentially inducing vibrations that may adversely affect the performance of the LIGO Detector.

This project will be divided into two main parts:

• i) Simulation of Bubble Formation: We invite proposals to simulate the process of bubble formation based on previous studies in the scientific literature. This will involve analyzing how condensate accumulation and changes in the emissivity of the LN2 reservoir surface contribute to bubble dynamics.
• ii) Vibration Measurements: We invite proposals to conduct vibration measurements of the cryopump using accelerometers to quantify the vibrations induced by bubble formation. This data will help us understand the impact of these vibrations on the performance of the LIGO Detector.

By addressing these two critical aspects, we aim to enhance the operational stability of the LIGO-India Detector and gain valuable insights into the dynamics of cryogenic systems.

Proposals are invited to develop accurate simulation models for the LIGO-India detector, recognising its unique characteristics compared to LIGO-US The objective is to integrate existing simulation models for the mechanical, electronic, optical, and control systems to create comprehensive frequency and time domain models. A key component of this initiative is the implementation of a **Digital Twin** for the LIGO-India detector. This Digital Twin will function as a time-domain, nonlinear model that combines the integrated frequency domain models with real-time data, providing a dynamic representation of the detector's behavior to enhance monitoring and optimization capabilities.

Additionally, projects which aim to train neural networks to emulate the existing physics-based models may also be proposed within this domain, By incorporating these neural networks into the Digital Twin framework, the goal is to facilitate faster simulations and generate valuable insights for the ongoing operation and optimisation of the LIGO-India detector. Proposals should focus on innovative approaches to achieve these objectives and contribute to the successful implementation and performance of LIGO-India in its unique operational environment.

One of the persistent problems in suspended interferometers of the Advanced LIGO type is the prestabilisation of the arms of the Michelson Interferometer. Each of these arms, comprised of two suspended mirrors (test masses), separated by 4 km, forms a Fabry Perot optical cavity and enhances the effective length of the arms to about 400 km each. Prestabilising the length of these arms is necessary before the main length control loops of the Dual-recycled-Fabry-Perot-Michelson interferometer can be locked. The Arm Length Stabilisation System (ALS), which presently serves this function, utilises four km-long optical fibres to lock the auxiliary lasers at the ends of the arms to the main laser in the corner station. One of the performance-limiting factors of this ALS is phase noise picked up by these optical fibres due to their exposure to ambient sources. Proposals are invited in this domain aimed at reducing phase noise between two lasers, which are locked by an optical fibre of several kilometres in length, which is exposed to realistic ambient sources of noise. A technique such as this may also be used as a distributed seismic noise sensor in the future.

The LIGO India detector will utilise extensive seismic isolation systems with active feedback in order to mitigate the ground motion from reaching the suspended test-masses.

These acive feedback systems require careful modelling and servo fiter design, both of which depend on an accurate representation of the "plant" which needs to be controlled. While analytical models of these electro-mechanical systems exist and are used extensively in the design and development of these systems, in practical day-to-day work, the dynamical response of these sub-systems in LIGO are measured and a mathematical model is extracted from these measurements heuristically.

However, since the frequency domain of interest often spans 0.001 Hz to 100 Hz, the measurement of this frequency response takes a considerable amount of interferometer time. These MIMO response functions often have sharp resonances and cross couplings and, furthermore, the cross-couplings may arise from unmodelled structural flexure and deformities. In addition to the loss of time, there is also a cost in performance as the accuracy to which we can determine these transfer functions suffers, when they are measured over long durations of time. Often the plants we are characterising are also slowly evolving in time as the day proceeds. Given all these difficulties, and the crucial role that plant identification plays in servo control, it is imperative that we develop efficient methods of plant identification by optimal sampling of transfer functions. We may also employ realtime data as the interferometer is running in order to develop a plant model for any given subsystem such that interferometer duty cycle is not adversely affected. Proposals are invited in this domain to improve accuracy of plant identification with optimal sampling and minimise the time needed.

All the present-day gravitational wave interferometers make extensive use of feedback systems to hold the detector at its working point and linearise the plant. Recent work has shown that technical noise sources arising from non-optimal servo filter design are a major cause of noise in the low-frequency domain. Due to the highly coupled nature of the interferometer control loops, both due to cross couplings within the optical plant, as well as in seismic isolators and suspensions, it is hard to analytically model these systems or even measure them using traditional stimulus-response architectures. One of the promising ways forward is to utilise modern computational techniques in the Artificial Intelligence and Machine Learning domains to speed up the plant identification processes and improve their accuracy as well. Each interferometer continuously records about 250,000 channels of data, which contain a wealth of information regarding the causal pathways and plants' response in various sub-systems as well as the integrated system as a whole. While building analytical models for such a plants would be fobiddingly difficult, an AI/ML method which uses the recorded data for "learning" the behaviour of the plant may be able to effiiciently converge on a well-approximated plant models. Projects in this domain would attempt to develop measurement and modelling techniques for optimal plant identification utilising AI/ML techniques and prove their efficacy predicting interferometer behavior with a high degree of success.