arXiv subjects

Thomas Bsaibes

Publications and source records attributed to Thomas Bsaibes.

4 recordsLinked to original sources

arXiv2026

Dual mass milligram-scale torsion oscillator for vibration-free optomechanical sensing

Chip-scale optomechanical devices are driving the miniaturization of inertial sensors and next generation fundamental physics experiments. However, precision at the theoretical limit is often unattainable due to extraneous vibrations. One solution is tailoring the device to isolate a degree of freedom from the environment while maintaining coupling to the signal of interest. To this end, we introduce a dual milligram-mass torsion oscillator, formed by mass loading a strained silicon nitride nanoribbon. The antisymmetric torsion mode suppresses vibrations by over an order of magnitude to achieve a thermally limited torque sensitivity of $10^{-18}$ Nm/$\sqrt{\rm Hz}$ while maintaining ultralow loss. We demonstrate the sensing ability by detecting an optical radiation pressure torque of $10^{-16}$ Nm over a 30 Hz bandwidth. We also characterize the device for frequency-based gravimetry, demonstrating $10^{-6}g_0$ ($g_0=9.8$ $\rm m s^{-2}$) precision in 30 seconds with an oscillation amplitude of only 100 $\mu$rad. This device demonstrates a technique for overcoming vibration noise, with broad implications for optomechanical sensing from commercial applications to fundamental physics experiments.

physics.app-ph↗

arXiv2026

Nanofabricated torsion pendulums for tabletop gravity experiments

Measurement of mutual gravitation on laboratory scales is an outstanding challenge and a prerequisite to probing theories of quantum gravity. A leading technology in tabletop gravity experiments is the torsion balance, with limitations due to thermal decoherence. Recent demonstrations of lithographically defined suspensions in thin-film silicon nitride with macroscale test masses suggest a path forward, as torsion pendulums dominated by gravitational stiffness may achieve higher mechanical quality factors through dilution of material losses. Here we demonstrate a 250 micron by 5 mm by 1.8 micron torsion fiber supporting 87 grams and forming a Cavendish-style torsion pendulum with tungsten test masses that -- to our knowledge -- is the largest thin-film silicon-nitride-based oscillator to date. Torsion pendulums with thin-film, nanofabricated suspensions provide a test bed for near-term tabletop experiments probing classical and quantum gravitational interaction between oscillators.

physics.ins-det↗

arXiv2025

Lithographically Defined Si$_3$N$_4$ Torsional Pendulum

Torsion pendulums provide an opportunity to trap large masses in a potential weak enough to explore two-body gravitation. Cooled to, and then released from a ground state, weak quantum effects, including those from gravity, might reveal themselves in the evolving decoherence of a torsion pendulum, if its baseline dissipation were sufficiently dilute for quantum coherent oscillation. Monolithic ribbon-like, or multi-filar suspension geometries provide a key to such dilution in torsion, but are challenging to make. As a solution, we introduce a lithographically defined silicon nitride (Si$_3$N$_4$) ribbon suspension in a wafer-scale approach to pendulum fabrication that is conducive to such 2-D geometries, making extreme aspect ratios, and even multi-filar designs, a possibility. A monofilar, monolithic, centimeter scale torsion pendulum is fabricated and released in a first proof of concept. Mounted in vacuum, it is optically excited and cooled using measurement based feedback. Though only 37 mg, the device displays a fundamental frequency of 162 mHz and an undiluted Q of 12000, demonstrating a foundational step towards ultra-coherent, ultra-low frequency torsion pendulums.

physics.app-ph↗

arXiv2024

Towards Predicting the Success of Transfer-based Attacks by Quantifying Shared Feature Representations

Much effort has been made to explain and improve the success of transfer-based attacks (TBA) on black-box computer vision models. This work provides the first attempt at a priori prediction of attack success by identifying the presence of vulnerable features within target models. Recent work by Chen and Liu (2024) proposed the manifold attack model, a unifying framework proposing that successful TBA exist in a common manifold space. Our work experimentally tests the common manifold space hypothesis by a new methodology: first, projecting feature vectors from surrogate and target feature extractors trained on ImageNet onto the same low-dimensional manifold; second, quantifying any observed structure similarities on the manifold; and finally, by relating these observed similarities to the success of the TBA. We find that shared feature representation moderately correlates with increased success of TBA (\r{ho}= 0.56). This method may be used to predict whether an attack will transfer without information of the model weights, training, architecture or details of the attack. The results confirm the presence of shared feature representations between two feature extractors of different sizes and complexities, and demonstrate the utility of datasets from different target domains as test signals for interpreting black-box feature representations.

cs.CV↗