skip to main content

Title: Deep multi-task mining Calabi–Yau four-folds
Abstract

We continue earlier efforts in computing the dimensions of tangent space cohomologies of Calabi–Yau manifolds using deep learning. In this paper, we consider the dataset of all Calabi–Yau four-folds constructed as complete intersections in products of projective spaces. Employing neural networks inspired by state-of-the-art computer vision architectures, we improve earlier benchmarks and demonstrate that all four non-trivial Hodge numbers can be learned at the same time using a multi-task architecture. With 30% (80%) training ratio, we reach an accuracy of 100% forh(1,1)and 97% forh(2,1)(100% for both), 81% (96%) forh(3,1), and 49% (83%) forh(2,2). Assuming that the Euler number is known, as it is easy to compute, and taking into account the linear constraint arising from index computations, we get 100% total accuracy.

Authors:
; ; ;
Award ID(s):
2019786
Publication Date:
NSF-PAR ID:
10303335
Journal Name:
Machine Learning: Science and Technology
Volume:
3
Issue:
1
Page Range or eLocation-ID:
Article No. 015006
ISSN:
2632-2153
Publisher:
IOP Publishing
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    We use ALMA observations of CO(2–1) in 13 massive (M*≳ 1011M) poststarburst galaxies atz∼ 0.6 to constrain the molecular gas content in galaxies shortly after they quench their major star-forming episode. The poststarburst galaxies in this study are selected from the Sloan Digital Sky Survey spectroscopic samples (Data Release 14) based on their spectral shapes, as part of the Studying QUenching at Intermediate-z Galaxies: Gas, anguLarmomentum, and Evolution (SQuIGGLE) program. Early results showed that two poststarburst galaxies host large H2reservoirs despite their low inferred star formation rates (SFRs). Here we expand this analysis to a larger statistical sample of 13 galaxies. Six of the primary targets (45%) are detected, withMH2109M. Given their high stellar masses, this mass limit corresponds to an average gas fraction offH2MH2/M*7%or ∼14% using lower stellar masses estimates derived from analytic, exponentially declining star formation histories. The gas fraction correlates with theDn4000 spectral index, suggesting that the cold gas reservoirs decrease with time since burst, as found in local K+A galaxies. Star formation histories derived from flexible stellar population synthesis modeling support thismore »empirical finding: galaxies that quenched ≲150 Myr prior to observation host detectable CO(2–1) emission, while older poststarburst galaxies are undetected. The large H2reservoirs and low SFRs in the sample imply that the quenching of star formation precedes the disappearance of the cold gas reservoirs. However, within the following 100–200 Myr, theSQuIGGLEgalaxies require the additional and efficient heating or removal of cold gas to bring their low SFRs in line with standard H2scaling relations.

    « less
  2. Abstract

    Cosmic reionization was the last major phase transition of hydrogen from neutral to highly ionized in the intergalactic medium (IGM). Current observations show that the IGM is significantly neutral atz> 7 and largely ionized byz∼ 5.5. However, most methods to measure the IGM neutral fraction are highly model dependent and are limited to when the volume-averaged neutral fraction of the IGM is either relatively low (x¯HI103) or close to unity (x¯HI1). In particular, the neutral fraction evolution of the IGM at the critical redshift range ofz= 6–7 is poorly constrained. We present new constraints onx¯HIatz∼ 5.1–6.8 by analyzing deep optical spectra of 53 quasars at 5.73 <z< 7.09. We derive model-independent upper limits on the neutral hydrogen fraction based on the fraction of “dark” pixels identified in the Lyαand Lyβforests, without any assumptions on the IGM model or the intrinsic shape of the quasar continuum. They are the first model-independent constraints on the IGM neutral hydrogen fraction atz∼ 6.2–6.8 using quasar absorption measurements. Our results give upper limits ofx¯more »mathvariant='normal'>HI(z=6.3)<0.79±0.04(1σ),x¯HI(z=6.5)<0.87±0.03(1σ), andx¯HI(z=6.7)<0.940.09+0.06(1σ). The dark pixel fractions atz> 6.1 are consistent with the redshift evolution of the neutral fraction of the IGM derived from Planck 2018.

    « less
  3. Abstract

    The best upper limit for the electron electric dipole moment was recently set by the ACME collaboration. This experiment measures an electron spin-precession in a cold beam of ThO molecules in their metastableH(3Δ1)state. Improvement in the statistical and systematic uncertainties is possible with more efficient use of molecules from the source and better magnetometry in the experiment, respectively. Here, we report measurements of several relevant properties of the long-livedQ(3Δ2)state of ThO, and show that this state is a very useful resource for both these purposes. TheQstate lifetime is long enough that its decay during the time of flight in the ACME beam experiment is negligible. The large electric dipole moment measured for theQstate, giving rise to a large linear Stark shift, is ideal for an electrostatic lens that increases the fraction of molecules detected downstream. The measured magnetic moment of theQstate is also large enough to be used as a sensitive co-magnetometer in ACME. Finally, we show that theQstate has a large transition dipole moment to theC(1Π1)state, which allows for efficient population transfer between the ground stateXmore »width='0.50em'/>(1Σ+)and theQstate viaXCQStimulated Raman Adiabatic Passage (STIRAP). We demonstrate 90 % STIRAP transfer efficiency. In the course of these measurements, we also determine the magnetic moment ofCstate, theXCtransition dipole moment, and branching ratios of decays from theCstate.

    « less
  4. Abstract

    The genericity of Arnold diffusion in the analytic category is an open problem. In this paper, we study this problem in the followinga prioriunstable Hamiltonian system with a time-periodic perturbationHε(p,q,I,φ,t)=h(I)+i=1n±12pi2+Vi(qi)+εH1(p,q,I,φ,t),where(p,q)Rn×Tn,(I,φ)Rd×Tdwithn,d⩾ 1,Viare Morse potentials, andɛis a small non-zero parameter. The unperturbed Hamiltonian is not necessarily convex, and the induced inner dynamics does not need to satisfy a twist condition. Using geometric methods we prove that Arnold diffusion occurs for generic analytic perturbationsH1. Indeed, the set of admissibleH1isCωdense andC3open (a fortiori,Cωopen). Our perturbative technique for the genericity is valid in theCktopology for allk∈ [3, ∞) ∪ {∞,ω}.

  5. Abstract

    We present a chemodynamical study of the Grus I ultra-faint dwarf galaxy (UFD) from medium-resolution (R∼ 11,000) Magellan/IMACS spectra of its individual member stars. We identify eight confirmed members of Grus I, based on their low metallicities and coherent radial velocities, and four candidate members for which only velocities are derived. In contrast to previous work, we find that Grus I has a very low mean metallicity of 〈[Fe/H]〉 = −2.62 ± 0.11 dex, making it one of the most metal-poor UFDs. Grus I has a systemic radial velocity of −143.5 ± 1.2 km s−1and a velocity dispersion ofσrv=2.50.8+1.3km s−1, which results in a dynamical mass ofM1/2(rh)=84+12×105Mand a mass-to-light ratio ofM/LV=440250+650M/L. Under the assumption of dynamical equilibrium, our analysis confirms that Grus I is a dark-matter-dominated UFD (M/L> 80M/L). However, we do not resolve a metallicity dispersion (σ[Fe/H]< 0.44 dex). Our results indicate that Grus I is a fairly typical UFD with parameters that agree with mass–metallicity and metallicity-luminosity trends for faint galaxies. This agreement suggests that Grus I has not lost an especially significant amount of mass from tidal encounters with the Milky Way, in linemore »with its orbital parameters. Intriguingly, Grus I has among the lowest central densities (ρ1/23.52.1+5.7×107Mkpc−3) of the UFDs that are not known to be tidally disrupting. Models of the formation and evolution of UFDs will need to explain the diversity of these central densities, in addition to any diversity in the outer regions of these relic galaxies.

    « less