Resource
2019 EN
D. Adamová · G. Aglieri Rinella · M. Agnello
+395 more
The present document discusses plans for a compact, next-generationmulti-purpose detector at the LHC as a follow-up to the present ALICEexperiment. The aim is to build a nearly massless barrel detector consisting oftruly cylindrical layers based on curved wafer-scale ultra-thin silicon sensorswith MAPS technology, featuring an unprecedented low material budget of 0.05%X$_0$ per layer, with the innermost layers possibly positioned inside the beampipe. In addition to superior tracking and vertexing capabilities over a widemomentum range down to a few tens of MeV/$c$, the detector will provideparticle identification via time-of-flight determination with about 20~psresolution. In addition, electron and photon identification will be performedin a separate shower detector. The proposed detector is conceived for studiesof pp, pA and AA collisions at luminosities a factor of 20 to 50 times higherthan possible with the upgraded ALICE detector, enabling a rich physics programranging from measurements with electromagnetic probes at ultra-low transversemomenta to precision physics in the charm and beauty sector.
Resource
2019 EN
Minyoung Kim · Yuting Wang · Pritish Sahu
+1 more
We propose a novel VAE-based deep auto-encoder model that can learndisentangled latent representations in a fully unsupervised manner, endowedwith the ability to identify all meaningful sources of variation and theircardinality. Our model, dubbed Relevance-Factor-VAE, leverages the totalcorrelation (TC) in the latent space to achieve the disentanglement goal, butalso addresses the key issue of existing approaches which cannot distinguishbetween meaningful and nuisance factors of latent variation, often the sourceof considerable degradation in disentanglement performance. We tackle thisissue by introducing the so-called relevance indicator variables that can beautomatically learned from data, together with the VAE parameters. Our modeleffectively focuses the TC loss onto the relevant factors only by toleratinglarge prior KL divergences, a desideratum justified by our semi-parametrictheoretical analysis. Using a suite of disentanglement metrics, including anewly proposed one, as well as qualitative evidence, we demonstrate that ourmodel outperforms existing methods across several challenging benchmarkdatasets.
Resource
2019 EN
Rachel Akeson · Lee Armus · Etienne Bachelet
+103 more
The Wide Field Infrared Survey Telescope (WFIRST) is a 2.4m space telescopewith a 0.281 deg^2 field of view for near-IR imaging and slitless spectroscopyand a coronagraph designed for > 10^8 starlight suppresion. As backgroundinformation for Astro2020 white papers, this article summarizes the currentdesign and anticipated performance of WFIRST. While WFIRST does not have the UVimaging/spectroscopic capabilities of the Hubble Space Telescope, for widefield near-IR surveys WFIRST is hundreds of times more efficient. Some of themost ambitious multi-cycle HST Treasury programs could be executed as routineGeneral Observer (GO) programs on WFIRST. The large area and time-domainsurveys planned for the cosmology and exoplanet microlensing programs willproduce extraordinarily rich data sets that enable an enormous range ofArchival Research (AR) investigations. Requirements for the coronagraph aredefined based on its status as a technology demonstration, but its expectedperformance will enable unprecedented observations of nearby giant exoplanetsand circumstellar disks. WFIRST is currently in the Preliminary Design andTechnology Completion phase (Phase B), on schedule for launch in 2025, withseveral of its critical components already in production.
Resource
2019 EN
Saurabh Kumar Srivastav · Manas Ranjan Sahu · K. Watanabe
+3 more
The universal quantization of thermal conductance provides information on thetopological order of a state beyond electrical conductance. Such measurementshave become possible only recently, and have discovered, in particular, thatthe value of the observed thermal conductance of the 5/2 state is notconsistent with either the Pfaffian or the anti-Pfaffian model, motivatingseveral theoretical articles. The analysis of the experiments has been madecomplicated by the presence of counter-propagating edge channels arising fromedge reconstruction, an inevitable consequence of separating the dopant layerfrom the GaAs quantum well. In particular, it has been found that the universalquantization requires thermalization of downstream and upstream edge channels.Here we measure the thermal conductance in hexagonal boron nitride encapsulatedgraphene devices of sizes much smaller than the thermal relaxation length ofthe edge states. We find the quantization of thermal conductance within 5%accuracy for {\nu} = 1, 4/3, 2 and 6 plateaus and our results strongly suggestthe absence of edge reconstruction for fractional quantum Hall in graphene,making it uniquely suitable for interference phenomena exploiting paths ofexotic quasiparticles along the edge.
Resource
2019 EN
Minyoung Kim · Pritish Sahu · Behnam Gholami
+1 more
In unsupervised domain adaptation, it is widely known that the target domainerror can be provably reduced by having a shared input representation thatmakes the source and target domains indistinguishable from each other. Veryrecently it has been studied that not just matching the marginal inputdistributions, but the alignment of output (class) distributions is alsocritical. The latter can be achieved by minimizing the maximum discrepancy ofpredictors (classifiers). In this paper, we adopt this principle, but propose amore systematic and effective way to achieve hypothesis consistency viaGaussian processes (GP). The GP allows us to define/induce a hypothesis spaceof the classifiers from the posterior distribution of the latent randomfunctions, turning the learning into a simple large-margin posterior separationproblem, far easier to solve than previous approaches based on adversarialminimax optimization. We formulate a learning objective that effectively pushesthe posterior to minimize the maximum discrepancy. This is further shown to beequivalent to maximizing margins and minimizing uncertainty of the classpredictions in the target domain, a well-established principle in classical(semi-)supervised learning. Empirical results demonstrate that our approach iscomparable or superior to the existing methods on several benchmark domainadaptation datasets.
Resource
2019 EN
V. K. B. Kota · R. Sahu · P. C. Srivastava
Rotational $SU(3)$ algebraic symmetry continues to generate new results inthe shell model (SM). Interestingly, it is possible to have multiple $SU(3)$algebras for nucleons occupying an oscillator shell $\eta$. Several differentaspects of the multiple $SU(3)$ algebras are investigated using shell model andalso deformed shell model based on Hartree-Fock single particle states withnucleons in $sdg$ orbits giving four $SU(3)$ algebras. Results show that one ofthe $SU(3)$ algebra generates prolate shapes, one oblate shape and the othertwo also generate prolate shape but one of them gives quiet small quadrupolemoments for low-lying levels. These are inferred by using the standard form forthe electric quadrupole transition operator and using quadrupole moments and$B(E2)$ values in the ground $K=0^+$ band in three different examples. Multiple$SU(3)$ algebras extend to interacting boson model and using $sdg$IBM, thestructure of the four $SU(3)$ algebras in this model are studied by coherentstate analysis and asymptotic formulas for $E2$ matrix elements. The resultsfrom $sdg$IBM further support the conclusions from the $sdg$ shell modelexamples.
Resource
2019 EN
Chien-Hsiu Lee · Rachel Street · Kailash Sahu
+1 more
Microlensing offers a unique opportunity to probe exoplanets that aretemperate and beyond the snow line, as small as Jovian satellites, atextragalactic distance, and even free floating exoplanets, regimes where thesensitivity of other methods drops dramatically. This is because microlensingdoes not depend on the brightness of the planetary host star. The microlensingmethod thus provides great leverage in studying the exoplanets beyond the snowline, posing tests to the core accretion mechanism, especially on the run-awayphase of gas accretion to form giant planets. Here we propose to robustly androutinely measure the masses of exoplanets beyond 1 AU from their host starswith the microlensing method; our experiment relies on directly imaging andresolving the host star (namely the lens) from the background source of themicrolensing events, which requires the high spatial resolution delivered bythe ELTs. A direct result from this project will be planet occurrence ratebeyond the snow line, which will enable us to discern different planetformation mechanisms.
Resource
2019 EN
Ran Xin · Anit Kumar Sahu · Usman A. Khan
+1 more
In this paper, we study distributed stochastic optimization to minimize a sumof smooth and strongly-convex local cost functions over a network of agents,communicating over a strongly-connected graph. Assuming that each agent hasaccess to a stochastic first-order oracle ($\mathcal{SFO}$), we propose a noveldistributed method, called $\mathcal{S}$-$\mathcal{AB}$, where each agent usesan auxiliary variable to asymptotically track the gradient of the global costin expectation. The $\mathcal{S}$-$\mathcal{AB}$ algorithm employs row- andcolumn-stochastic weights simultaneously to ensure both consensus andoptimality. Since doubly-stochastic weights are not used,$\mathcal{S}$-$\mathcal{AB}$ is applicable to arbitrary strongly-connectedgraphs. We show that under a sufficiently small constant step-size,$\mathcal{S}$-$\mathcal{AB}$ converges linearly (in expected mean-square sense)to a neighborhood of the global minimizer. We present numerical simulationsbased on real-world data sets to illustrate the theoretical results.
Resource
2019 EN
Matthew T. Penny · Etienne Bachelet · Samson Johnson
+17 more
Free-floating planets are the remnants of violent dynamical rearrangements ofplanetary systems. It is possible that even our own solar system ejected alarge planet early in its evolution. WFIRST will have the ability to detectfree-floating planets over a wide range of masses, but it will not be able todirectly measure their masses. Microlensing parallax observations can be usedto measure the masses of isolated objects, including free-floating planets, byobserving their microlensing events from two locations. The intra-L2 separationbetween WFIRST and Euclid is large enough to enable microlensing parallaxmeasurements, especially given the exquisite photometric precision that bothspacecraft are capable of over wide fields. In this white paper we describe howa modest investment of observing time could yield hundreds of parallaxmeasurements for WFIRST's bound and free-floating planets. We also describe howa short observing campaign of precursor observations by Euclid can improveWFIRST's bound planet and host star mass measurements.
Resource
2019 EN
Aparna Bhattacharya · Rachel Akeson · Jay Anderson
+33 more
Microlensing is the only method that can detect and measure mass of wideorbit, low mass, solar system analog exoplanets. Mass measurements of suchplanets would yield massive science on planet formation, exoplanetdemographics, free floating planets, planet frequencies towards the galaxy.High res follow-up observations of past microlens targets provide a massmeasurement of microlens planets and hosts at an uncertainty of <20%. This willbe primary method for mass measurement with WFIRST. We advocate for the factthat high resolution observations with AO, HST and JWST(in future) remainnecessary in coming decade to develop the methods, to determine the field andfilter selection, understand the systematics and to develop a robust pipelineto release high quality data products from WFIRST microlensing survey such thatthe astronomy community can promptly engage in the science. We also supportfuture high res obs with US ELTs with advanced Laser AO systems in context ofenhancing the science return of WFIRST microlensing survey. We endorse the 2018 Exoplanet Science Strategy report published by theNational Academy. This white paper extends and complements the materialpresented therein. In particular, this white paper supports the recommendationof the National Academy Exoplanet Science Strategy report that: NASA shouldlaunch WFIRST to conduct its microlensing survey of distant planets and todemonstrate the technique of coronagraphic spectroscopy on exoplanet targets.This white paper also supports to the finding from that report which states "Anumber of activities, including precursor and concurrent observations usingground- and space-based facilities, would optimize the scientific yield of theWFIRST microlensing survey."