Search

The evolution of a deformed subcritical fast magnetosonic shock front is compared between two two-dimensional PIC simulations with different orientations of the magnetic field relative to the simulation box. All other initial and simulation conditions are kept identical. Shock boundary oscillations are observed in the simulation where the magnetic field direction is resolved. This oscillation is caused by the reformation of the shock front. One part of the front acts as a shock, while the other functions as a magnetic piston, with both halves changing their states in antiphase. The oscillation period corresponds to the time required for one shock wave to grow as the other collapses. In contrast, the corrugated fast magnetosonic shock does not oscillate in the second simulation, where the magnetic field is oriented out of the simulation plane. This dependence on magnetic field orientation suggests that the shock oscillation is induced by magnetic tension, which is only effective in the first simulation. In both simulations, the shock perturbation does not grow over time, indicating that the shocks are stable. The potential relevance of these findings for the Alfv\'enic oscillations of the supercritical Earth's bow shock, detected by the MMS multi-spacecraft mission, is also discussed., Comment: 16 pages, 7 figures, accepted for publication in Plasma Physics and Controlled Fusion

Ablating a target with an ultraintense laser pulse can create a cloud of collisionless plasma. A density ramp forms, in which the plasma density decreases and the ion's mean speed increases with distance from the plasma source. Its width increases with time. Electrons lose energy in the ion's expansion direction, which gives them a temperature anisotropy. We study with one-dimensional particle-in-cell simulations the expansion of a dense plasma into a dilute one, yielding a density ramp similar to that in laser plasma experiments and a thermal-anisotropy-driven instability. Non-propagating Weibel-type modes grow in the simulation with no initial magnetic field. Their magnetic field diffuses across the the shock and expands upstream. Circularly polarized propagating Whistler waves grow in a second simulation, in which a magnetic field is aligned with the ion expansion direction. Both wave modes are driven by non-resonant instabilities, they have a similar exponential growth rates, and they can leave the density ramp and expand into the dilute plasma. Their large magnetic amplitude should make them detectable in experimental settings., Comment: 18 pages, 9 figures, accepted for publication in Physica Scripta

We report the creation of ultracold ground state $^{6}\textrm{Li}^{40}\textrm{K}$ polar molecules with high efficiency. Starting from weakly-bound molecules state, stimulated Raman adiabatic passage (STIRAP) is adopted to coherently transfer the molecules to their singlet ro-vibrational ground state $|\textrm{X}^{1}\Sigma^{+},v=0,J=0>$. By employing a singlet STIRAP pathway and low-phase-noise narrow-linewidth lasers, we observed a one-way transfer efficiency of 96(4)\,\%. Held in an optical dipole trap, the lifetime of the ground-state molecules is measured to be 5.0(3)\,ms. The large permanent dipole moment of LiK is confirmed by applying a DC electric field on the molecules and performing Stark shift spectroscopy of the ground state. With recent advances in the quantum control of collisions, our work paves the way for exploring quantum many-body physics with strongly-interacting $^{6}\textrm{Li}^{40}\textrm{K}$ molecules., Comment: 6 pages, 6 figures

Plant-frugivore interactions play a central role for plant persistence and spatial distribution by promoting the long-range dispersal of seeds by frugivores. However, plant-frugivore interactions are increasingly being threatened by anthropogenic activities. An important anthropogenic threat that could expose plant-frugivore systems to extinction risk is fruit harvesting. Here, we develop an individual-based and a pair-approximation model of plant-frugivore-human interactions to elucidate the effects of human harvesting of fruits on plant establishment, persistence, and spatial distribution. Our results show that frugivores strongly affect global density of plants and gradually shift their spatial distribution from aggregated to random, depending on the attack rate and dispersal efficiency of frugivores. We find that, in the absence of frugivores, plants experiencing intense fruit harvesting cannot persist even if their fecundity is high. In the presence of frugivores, fruit harvesting profoundly affects the global dispersal of seeds and thus changes the spatial distributions of plants from random to aggregated, potentially causing plant extinction. Our results demonstrate that sufficiently efficient frugivores mitigate the negative impact of fruit harvesting on plant populations and enable plant establishment precluded by harvesting. Taken together, these results draw attention to previously underappreciated impacts of fruit harvesting in plant-frugivore-human interactions.

We study with particle-in-cell (PIC) simulations the stability of fast magnetosonic shocks. They expand across a collisionless plasma and an orthogonal magnetic field that is aligned with one of the directions resolved by the 2D simulations. The shock speed is 1.6 times the fast magnetosonic speed when it enters a layer with a reduced density of mobile ions, which decreases the shock speed by up to 15\% in 1D simulations. In the 2D simulations, the density of mobile ions in the layer varies sinusoidally perpendicularly to the shock normal. We resolve one sine period. This variation only leads to small changes in the shock speed evidencing a restoring force that opposes a shock deformation. As the shock propagates through the layer, the ion density becomes increasingly spatially modulated along the shock front and the magnetic field bulges out where the mobile ion density is lowest. The perturbed shock eventually reaches a steady state. Once it leaves the layer, the perturbations of the ion density and magnetic field oscillate along its front at a frequency close to the lower-hybrid frequency; the shock is mediated by a standing wave composed of obliquely propagating lower-hybrid waves. We perform three 2D simulations with different box lengths along the shock front. The shock front oscillations are aperiodically damped in the smallest box with the fastest variation of the ion density, strongly damped in the intermediate one, and weakly damped in the largest box. The shock front oscillations perturb the magnetic field in a spatial interval that extends by several electron skin depths upstream and downstream of the shock front and could give rise to Whistler waves that propagate along the shock's magnetic field overshoot. Similar waves were observed in hybrid and PIC simulations and by the MMS satellite mission., Comment: 25 pages, 12 figures, accepted for publication in Physica Scripta

We study with a 3D PIC simulation discontinuities between an electron-positron pair plasma and magnetized electrons and protons. A pair plasma is injected at one simulation boundary with a speed 0.6$c$ along its normal. It expands into an electron-proton plasma and a magnetic field that points orthogonally to the injection direction. Diamagnetic currents expel the magnetic field from within the pair plasma and pile it up in front of it. It pushes electrons, which induces an electric field pulse ahead of the magnetic one. This initial electromagnetic pulse (EMP) confines the pair plasma magnetically and accelerates protons electrically. The fast flow of the injected pair plasma across the protons behind the initial EMP triggers the filamentation instability. Some electrons and positrons cross the injection boundary and build up a second EMP. Electron-cyclotron drift instabilities perturb the plasma ahead of both EMPs seeding a Rayleigh-Taylor-type instability. Despite equally strong perturbations ahead of both EMPs, the second EMP is much more stable than the initial one. We attribute the rapid collapse of the initial EMP to the filamentation instability, which perturbed the plasma behind it. The Rayleigh-Taylor-type instability transforms the planar EMPs into transition layers, in which magnetic flux ropes and electrostatic forces due to uneven numbers of electrons and positrons slow down and compress the pair plasma and accelerate protons. In our simulation, the expansion speed of the pair cloud decreased by about an order of magnitude and its density increased by the same factor. Its small thickness implies that it is capable of separating a relativistic pair outflow from an electron-proton plasma, which is essential for collimating relativistic jets of pair plasma in collisionless astrophysical plasma., Comment: 25 pages, 12 figures, provisionally accepted for publication by the New Journal of Physics

To apply federated learning to drug discovery we developed a novel platform in the context of European Innovative Medicines Initiative (IMI) project MELLODDY (grant n{\deg}831472), which was comprised of 10 pharmaceutical companies, academic research labs, large industrial companies and startups. The MELLODDY platform was the first industry-scale platform to enable the creation of a global federated model for drug discovery without sharing the confidential data sets of the individual partners. The federated model was trained on the platform by aggregating the gradients of all contributing partners in a cryptographic, secure way following each training iteration. The platform was deployed on an Amazon Web Services (AWS) multi-account architecture running Kubernetes clusters in private subnets. Organisationally, the roles of the different partners were codified as different rights and permissions on the platform and administrated in a decentralized way. The MELLODDY platform generated new scientific discoveries which are described in a companion paper., Comment: 9 pages, 4 figures, to appear in AAAI-23 ([IAAI-23 track] Deployed Highly Innovative Applications of AI)

By means of a particle-in-cell (PIC) simulation, we study the interaction between a uniform magnetized ambient electron-proton plasma at rest and an unmagnetized pair plasma, which we inject at one simulation boundary with a mildly relativistic mean speed and temperature. The magnetic field points out of the simulation plane. The injected pair plasma expels the magnetic field and piles it up at its front. It traps ambient electrons and drags them across the protons. An electric field grows, which accelerates protons into the pair cloud's expansion direction. This electromagnetic pulse separates the pair cloud from the ambient plasma. Electrons and positrons, which drift in the pulse's nonuniform field, trigger an instability that disrupts the current sheet ahead of the pulse. The wave vector of the growing perturbation is orthogonal to the magnetic field direction and magnetic tension cannot stabilize it. The electromagnetic pulse becomes permeable for pair plasma, which forms new electromagnetic pulses ahead of the initial one. A transition layer develops with a thickness of a few proton skin depths, in which protons and positrons are accelerated by strong electromagnetic fields. Protons form dense clumps surrounded by a strong magnetic field. The thickness of the transition layer grows less rapidly than we would expect from the typical speeds of the pair plasma particles and the latter transfer momentum to protons; hence, the transition layer acts as a discontinuity, separating the pair plasma from the ambient plasma. Such a discontinuity is an important building block for astrophysical pair plasma jets., Comment: Accepted for publication in Physics of Plasmas, 13 pages, 15 figures

Current text-to-speech algorithms produce realistic fakes of human voices, making deepfake detection a much-needed area of research. While researchers have presented various techniques for detecting audio spoofs, it is often unclear exactly why these architectures are successful: Preprocessing steps, hyperparameter settings, and the degree of fine-tuning are not consistent across related work. Which factors contribute to success, and which are accidental? In this work, we address this problem: We systematize audio spoofing detection by re-implementing and uniformly evaluating architectures from related work. We identify overarching features for successful audio deepfake detection, such as using cqtspec or logspec features instead of melspec features, which improves performance by 37% EER on average, all other factors constant. Additionally, we evaluate generalization capabilities: We collect and publish a new dataset consisting of 37.9 hours of found audio recordings of celebrities and politicians, of which 17.2 hours are deepfakes. We find that related work performs poorly on such real-world data (performance degradation of up to one thousand percent). This may suggest that the community has tailored its solutions too closely to the prevailing ASVSpoof benchmark and that deepfakes are much harder to detect outside the lab than previously thought., Comment: Interspeech 2022

Deepfakes are synthetically generated media often devised with malicious intent. They have become increasingly more convincing with large training datasets advanced neural networks. These fakes are readily being misused for slander, misinformation and fraud. For this reason, intensive research for developing countermeasures is also expanding. However, recent work is almost exclusively limited to deepfake detection - predicting if audio is real or fake. This is despite the fact that attribution (who created which fake?) is an essential building block of a larger defense strategy, as practiced in the field of cybersecurity for a long time. This paper considers the problem of deepfake attacker attribution in the domain of audio. We present several methods for creating attacker signatures using low-level acoustic descriptors and machine learning embeddings. We show that speech signal features are inadequate for characterizing attacker signatures. However, we also demonstrate that embeddings from a recurrent neural network can successfully characterize attacks from both known and unknown attackers. Our attack signature embeddings result in distinct clusters, both for seen and unseen audio deepfakes. We show that these embeddings can be used in downstream-tasks to high-effect, scoring 97.10% accuracy in attacker-id classification., Comment: Submitted to Insterspeech 2022

Context. Relativistic jets are ubiquitous in astrophysics. High Mass Microquasars (HMMQs) are useful labs to study such jets, as they are relatively close and evolve over observable time scales. The ambient medium into which the jet propagates is, however, far from homogeneous. Corresponding simulation studies to date consider various forms of a wind-shaped ambient medium but typically neglect radiative cooling and relativistic effects. Aims. We investigate the dynamical and structural effects of radiative losses and system parameters on relativistic jets in HMMQs, from the jet launch to its propagation over several tens of orbital separations. Methods. We use 3D relativistic hydrodynamical simulations including parameterized radiative cooling derived from relativistic thermal plasma distribution to carry out parameter studies around two fiducial cases inspired by Cygnus X-1 and Cygnus X-3. Results. Radiative losses are found to be more relevant in Cygnus X-3 than Cygnus X-1. Varying jet power, jet temperature, or the wind of the donor star tends to have a larger impact at early times, when the jet forms and instabilities initially develop, than at later times when the jet has reached a turbulent state. Conclusions. Radiative losses may be dynamically and structurally relevant at least in Cygnus X-3 case, and thus should be examined in more detail., Comment: 28 pages, 25 figures, 8 tables. Accepted for publication in A&A

We report on a high-resolution spectroscopic survey of ${}^{6}\textrm{Li}{}^{40}\textrm{K}$ molecules near the $2\textrm{S}+4\textrm{P}$ dissociation threshold and produce a fully empirical representation for the $\textrm{B}^{1}\Pi$ potential by connecting available short- and long-range data. The purpose is to identify a suitable intermediate state for a coherent Raman transfer to the absolute ground state, and the creation of a molecular gas with dipolar interactions. Starting from weakly bound ultracold Feshbach molecules, the transition frequencies to twenty-six vibrational states are determined. Our data are combined with long-range measurements [Ridinger et al., EPL, 2011, 96, 33001], and near-dissociation expansions for the spin-orbit coupled potentials are fitted to extract the $C_6$ dispersion coefficients. A suitable vibrational level is identified by resolving its Zeeman structure and by comparing the experimentally attained g-factor to our theoretical prediction. Using mass-scaling of the short-range data for the $\textrm{B}^{1}\Pi$ [Pashov et al., Chem. Phys. Lett., 1998, 292, 615-620] and an updated value for its depth, we model the short- and the long-range data simultaneously and produce a Rydberg-Klein-Rees curve covering the entire range.

We present our analysis of a significant data artifact in the official 2019/2021 ASVspoof Challenge Dataset. We identify an uneven distribution of silence duration in the training and test splits, which tends to correlate with the target prediction label. Bonafide instances tend to have significantly longer leading and trailing silences than spoofed instances. In this paper, we explore this phenomenon and its impact in depth. We compare several types of models trained on a) only the duration of the leading silence and b) only on the duration of leading and trailing silence. Results show that models trained on only the duration of the leading silence perform particularly well, and achieve up to 85% percent accuracy and an equal error rate (EER) of 15.1%. At the same time, we observe that trimming silence during pre-processing and then training established antispoofing models using signal-based features leads to comparatively worse performance. In that case, EER increases from 3.6% (with silence) to 15.5% (trimmed silence). Our findings suggest that previous work may, in part, have inadvertently learned thespoof/bonafide distinction by relying on the duration of silence as it appears in the official challenge dataset. We discuss the potential consequences that this has for interpreting system scores in the challenge and discuss how the ASV community may further consider this issue.