arXiv subjects

J. Lehmann

Publications and source records attributed to J. Lehmann.

5 recordsLinked to original sources

arXiv2022

A vertical inertial sensor with interferometric readout

High precision interferometers such as gravitational-wave detectors require complex seismic isolation systems in order to decouple the experiment from unwanted ground motion. Improved inertial sensors for active isolation potentially enhance the sensitivity of existing and future gravitational-wave detectors, especially below 30 Hz, and thereby increase the range of detectable astrophysical signals. This paper presents a vertical inertial sensor which senses the relative motion between an inertial test mass suspended by a blade spring and a seismically isolated platform. An interferometric readout was used which introduces low sensing noise, and preserves a large dynamic range due to fringe-counting. The expected sensitivity is comparable to other state-of-the-art interferometric inertial sensors and reaches values of $10^{-10}\,\text{m}/\sqrt{\text{Hz}}$ at 100 mHz and $10^{-12}\,\text{m}/\sqrt{\text{Hz}}$ at 1 Hz. The potential sensitivity improvement compared to commercial L-4C geophones is shown to be about two orders of magnitude at 10 mHz and 100 mHz and one order of magnitude at 1 Hz. The noise performance is expected to be limited by thermal noise of the inertial test mass suspension below 10 Hz. Further performance limitations of the sensor, such as tilt-to-vertical coupling from a non-perfect levelling of the test mass and nonlinearities in the interferometric readout, are also quantified and discussed.

physics.ins-det↗

arXiv2020

Pretrained Transformers for Simple Question Answering over Knowledge Graphs

Answering simple questions over knowledge graphs is a well-studied problem in question answering. Previous approaches for this task built on recurrent and convolutional neural network based architectures that use pretrained word embeddings. It was recently shown that finetuning pretrained transformer networks (e.g. BERT) can outperform previous approaches on various natural language processing tasks. In this work, we investigate how well BERT performs on SimpleQuestions and provide an evaluation of both BERT and BiLSTM-based models in datasparse scenarios.

cs.CL↗

arXiv2017

Huddle test measurement of a near Johnson noise limited geophone

In this paper the sensor noise of two geophone configurations (L-22D and L-4C geophones from Sercel with custom built amplifiers) was measured by performing two huddle tests. It is shown that the accuracy of the results can be significantly improved by performing the huddle test in a seismically quiet environment and by using a large number of reference sensors to remove the seismic foreground signal from the data. Using these two techniques, the measured sensor noise of the two geophone configurations matched calculated predictions remarkably well in the bandwidth of interest (0.01 Hz to 100 Hz). Low noise operational amplifiers OPA188 were utilized to amplify the L-4C geophone to give a sensor that was characterized to be near Johnson noise limited in the bandwidth of interest with a noise value of $10^{-11} \text{m}/\sqrt{\text{Hz}}$ at 1 Hz.

physics.ins-det↗

arXiv2010

Effect of pressure on the kinetics of bulge bases in small RNAs

Using molecular dynamics simulations, we study the effect of pressure on the binding propensity of small RNAs by calculating the free energy barrier corresponding to the looped out conformations of unmatched base, which presumably acts as the binding sites for ligands. We find that the free energy associated with base looping out increases monotonically as the pressure is increased. Furthermore, we calculate the mean first passage time of conformational looping out of the base bulge using the diffusion of reaction coordinate associated with the base flipping on the underlying free energy surface. We find that the mean first passage time associated with bulge looping out increases slowly upon increasing pressures $P$ upto $2$~kbar but changes dramatically for $P>2$~kbar. Finally, we discuss our results in the light of the role of hydration shell of water around RNA.

cond-mat.soft↗

arXiv2006

Development of an Indium Bump Bond Process for Silicon Pixel Detectors at PSI

The hybrid pixel detectors used in the high energy physics experiments currently under construction use a three dimensional connection technique, the so-called bump bonding. As the pitch below 100um, required in these applications, cannot be fullfilled with standard industrial processes (e.g. the IBM C4 process), an in-house bump bond process using reflown indium bumps was developed at PSI as part of the R&D for the CMS-pixel detector. The bump deposition on the sensor is performed in two subsequent lift-off steps. As the first photolithographic step a thin under bump metalization (UBM) is sputtered onto bump pads. It is wettable by indium and defines the diameter of the bump. The indium is evaporated via a second photolithographic step with larger openings and is reflown afterwards. The height of the balls is defined by the volume of the indium. On the readout chip only one photolithographic step is carried out to deposit the UBM and a thin indium layer for better adhesion. After mating both parts a second reflow is performed for self alignment and obtaining a high mechanical strength. For the placement of the chips a manual and an automatic machine was constructed. The former is very flexible in handling different chip and module geometries but has a limited throughput while the latter features a much higher grade of automatisation and is therefore much more suited for producing hundreds of modules with a well defined geometry. The reliability of this process was proven by the successful construction of the PILATUS detector. The construction of PILATUS 6M (60 modules) and the CMS pixel barrel (roughly 800 modules) will start in 2005.

physics.ins-det↗