Search

We present a phase retrieval algorithm for dispersion scan (d-scan), inspired by ptychography, which is capable of characterizing multiple mutually-incoherent ultrafast pulses (or modes) in a pulse train simultaneously from a single d-scan trace. In addition, a form of Newton's method is employed as a solution to the square root problem commonly encountered in second harmonic pulse measurement techniques. Simulated and experimental phase retrievals of both single-mode and multi-mode d-scan traces are shown to demonstrate the accuracy and robustness of the algorithm., Comment: 17 pages, 8 figures

Multi-scale, multi-fidelity numerical simulations form the pillar of scientific applications related to numerically modeling fluids. However, simulating the fluid behavior characterized by the non-linear Navier Stokes equations are often times computational expensive. Physics informed machine learning methods is a viable alternative and as such has seen great interest in the community [refer to Kutz (2017); Brunton et al. (2020); Duraisamy et al. (2019) for a detailed review on this topic]. For full physics emulators, the cost of network inference is often trivial. However, in the current paradigm of data-driven fluid mechanics models are built as surrogates for complex sub-processes. These models are then used in conjunction to the Navier Stokes solvers, which makes ML model inference an important factor in the terms of algorithmic latency. With the ever growing size of networks, and often times overparameterization, exploring effective network compression techniques becomes not only relevant but critical for engineering systems design. In this study, we explore the applicability of pruning and quantization (FP32 to int8) methods for one such application relevant to modeling fluid turbulence. Post-compression, we demonstrate the improvement in the accuracy of network predictions and build intuition in the process by comparing the compressed to the original network state.

The characteristics of an extensive air shower derive from both the mass of the primary ultra-high-energy cosmic ray that seeds its development and the properties of the hadronic interactions that feed it. With its hybrid detector design, the Pierre Auger Observatory measures both the longitudinal development of showers in the atmosphere and the lateral distribution of particles arriving at the ground, from which a number of parameters are calculated and compared with predictions from current hadronic interaction models tuned to LHC data. At present, a tension exists concerning the production of muons, in that the measured abundance exceeds all predictions. This discrepancy, measured up to center-of-mass energies of $\sim$ 140 TeV, is irresolvable through mass composition arguments, constrained by measurements of the depth of the electromagnetic-shower maximum. Here, we discuss a compilation of hadronically-sensitive shower observables and their comparisons with model predictions and conclude with a brief discussion of what measurements with the new detectors of the AugerPrime upgrade will bring to the table., Comment: 6 pages, 4 figures, presented at the ICHEP 2020 (Prague, Virtual Conference, August 2020)

Air showers, produced by the interaction of energetic cosmic rays with the atmosphere, are an excellent alternative to study particle physics at energies beyond any human-made particle accelerator. For that, it is necessary to identify first the mass composition of the primary cosmic ray (and its energy). None of the existing high energy interaction models have been able to reproduce coherently all air shower observables over the entire energy and zenith angle phase space. This is despite having tried all possible combinations for the cosmic ray mass composition. This proposal outlines a self-consistent strategy to study high energy particle interactions and identify the energy spectra and mass composition of cosmic rays. This strategy involves the participation of different particle accelerators and astrophysics experiments. This is important to cover the entire cosmic ray energy range and a larger phase-space of shower observables to probe the high energy interaction models., Comment: Snowmass2021 - Letter of Interest

Fluid turbulence is characterized by strong coupling across a broad range of scales. Furthermore, besides the usual local cascades, such coupling may extend to interactions that are non-local in scale-space. As such the computational demands associated with explicitly resolving the full set of scales and their interactions, as in the Direct Numerical Simulation (DNS) of the Navier-Stokes equations, in most problems of practical interest are so high that reduced modeling of scales and interactions is required before further progress can be made. While popular reduced models are typically based on phenomenological modeling of relevant turbulent processes, recent advances in machine learning techniques have energized efforts to further improve the accuracy of such reduced models. In contrast to such efforts that seek to improve an existing turbulence model, we propose a machine learning(ML) methodology that captures, de novo, underlying turbulence phenomenology without a pre-specified model form. To illustrate the approach, we consider transient modeling of the dissipation of turbulent kinetic energy, a fundamental turbulent process that is central to a wide range of turbulence models using a Neural ODE approach. After presenting details of the methodology, we show that this approach outperforms state-of-the-art approaches.

Knowledge of the mass composition of ultra-high-energy cosmic rays is understood to be a salient component in answering the open questions in the field. The AugerPrime upgrade of the Pierre Auger Observatory aims to enhance its surface detector with the hardware necessary to reconstruct primary mass for individual events. This involves placing a scintillation-based detector with an active area of $3.8 \,\mathrm{m}^2$ on top of each existing water-Cherenkov detector in its surface detector array. Here, we present the methods for simulating this Scintillator Surface Detector. These simulations have and will continue to aid in the interpretation of measurements with AugerPrime as well as the development and improvement of event reconstruction algorithms including primary mass., Comment: 5 pages, 6 figures, presented at the UHECR 2018 (Paris, October 2018)

Intrusion detection systems (IDSs) play a crucial role in the identification and mitigation for attacks on host systems. Of these systems, vehicular ad hoc networks (VANETs) are particularly difficult to protect due to the dynamic nature of their clients and their necessity for constant interaction with their respective cyber-physical systems. Currently, there is a need for a VANET-specific IDS that can satisfy these requirements. Spline function-based IDSs have shown to be effective in traditional network settings. By examining the various construction of splines and testing their robustness, the viability for a spline-based IDS can be determined.

We compute nodal centrality measures on the collaboration networks of students enrolled in three upper-division physics courses, usually taken sequentially, at the Colorado School of Mines. These are complex networks in which links between students indicate assistance with homework. The courses included in the study are intermediate Classical Mechanics, introductory Quantum Mechanics, and intermediate Electromagnetism. By correlating these nodal centrality measures with students' scores on homework and exams, we find four centrality measures that correlate significantly with students' homework scores in all three courses: in-strength, out-strength, closeness centrality, and harmonic centrality. These correlations suggest that students who not only collaborate often, but also collaborate significantly with many different people tend to achieve higher grades. Centrality measures between simultaneous collaboration networks (analytical vs. numerical homework collaboration) composed of the same students also correlate with each other, suggesting that students' collaboration strategies remain relatively stable when presented with homework assignments targeting different skills. Additionally, we correlate centrality measures between collaboration networks from different courses and find that the four centrality measures with the strongest relationship to students' homework scores are also the most stable measures across networks involving different courses. Correlations of centrality measures with exam scores were generally smaller than the correlations with homework scores, though this finding varied across courses., Comment: 10 pages, 4 figures, submitted to Phys. Rev. PER

We present a novel approach to search for origins of ultra-high energy cosmic rays. These particles are likely nuclei that initiate extensive air showers in the Earth's atmosphere. In large-area observatories, the particle arrival directions are measured together with their energies and the atmospheric depth at which their showers maximize. The depths provide rough measures of the nuclear charges. In a simultaneous fit to all observed cosmic rays we use the galactic magnetic field as a mass spectrometer and adapt the nuclear charges such that their extragalactic arrival directions are concentrated in as few directions as possible. Using different simulated examples we show that, with the measurements on Earth, reconstruction of extragalactic source directions is possible. In particular, we show in an astrophysical scenario that source directions can be reconstructed even within a substantial isotropic background., Comment: 14 pages, 15 figures

We use adversarial network architectures together with the Wasserstein distance to generate or refine simulated detector data. The data reflect two-dimensional projections of spatially distributed signal patterns with a broad spectrum of applications. As an example, we use an observatory to detect cosmic ray-induced air showers with a ground-based array of particle detectors. First we investigate a method of generating detector patterns with variable signal strengths while constraining the primary particle energy. We then present a technique to refine simulated time traces of detectors to match corresponding data distributions. With this method we demonstrate that training a deep network with refined data-like signal traces leads to a more precise energy reconstruction of data events compared to training with the originally simulated traces., Comment: 10 pages, 7 figures

This work derives how the convergence of stochastic Lagrangian/Eulerian simulations depends on the number of computational parcels, particularly for the case of spray modeling. A new, simple, formula is derived that can be used for managing the numerical error in two or three dimensional computational studies. For example, keeping the number of parcels per cell constant as the mesh is refined yields an order one-half convergence rate in transient spray simulations. First order convergence would require a doubling of the number of parcels per cell when the cell size is halved. Second order convergence would require increasing the number of parcels per cell by a factor of eight. The results show that controlling statistical error requires dramatically larger numbers of parcels than have typically been used, which explains why convergence has been so elusive.