Predictions of nuclear properties far from measured data are inherently imprecise because of uncertainties in our knowledge of nuclear forces and in our treatment of quantum many-body effects in strongly-interacting systems. While the model bias can be directly calculated when experimental data is available, only an estimate can be made in the absence of such measurements. Current approaches to compute the estimated bias quickly lose predictive power when input variables such as proton or neutron number are extrapolated, resulting in uncontrolled uncertainties in applications such as nucleosynthesis simulations. In this letter, we present a novel technique to identify the input variables of machine learning algorithms that can provide robust estimates of model bias. Our process is based on selecting input variables, or features, based on their probability distribution functions across the entire nuclear chart. We illustrate our approach on the problem of quantifying the model bias in nuclear binding energies calculated with Density Functional Theory (DFT). We show that feature selection can systematically improve theoretical predictions without increasing uncertainties.
We discuss a variational calculation for nuclear shell-model calculations and propose a new procedure for the energy-variance extrapolation (EVE) method using a sequence of the approximated wave functions obtained by the variational calculation. The wave functions are described as linear combinations of the parity, angular-momentum projected Slater determinants, the energy of which is minimized by the conjugate gradient method obeying the variational principle. The EVE generally works well using the wave functions, but we found some difficult cases where the EVE gives a poor estimation. We dis…
Calculations of nuclei are often carried out in finite model spaces. Thus, finite-size corrections enter, and it is necessary to extrapolate the computed observables to infinite model spaces. In this work, we employ extrapolation methods based on artificial neural networks for observables such as the ground-state energy and the point-proton radius. We extrapolate results from no-core shell model and coupled-cluster calculations to very large model spaces and estimate uncertainties. Training the network on different data typically yields extrapolation results that cluster around distinct values…
A new two-dimensional procedure for extrapolation of the values of matter, neutron, and proton radii obtained in no-core shell model (NCSM) calculations to infinite size of its basis is proposed. A relationship between the radii is used as an additional test. Together with the JISP16 potential, which is frequently used in NCSM calculations of the radii, the Daejeon16 potential is applied for these purposes for the first time. Halo nucleus 6He is the object of studies. The small spread of radii values, successful testing using relationship between the values of these three radii, and reasonable…
We investigate two distinct sources of uncertainty in low-energy nuclear physics calculations and develop ways to account for them. Harmonic oscillator basis expansions are widely used in ab-initio nuclear structure calculations. Finite computational resources usually require that the basis be truncated before observables are fully converged, necessitating reliable extrapolation schemes. We show that a finite oscillator basis effectively imposes a hard-wall boundary condition. We accurately determine the position of the hard-wall as a function of oscillator space parameters, derive extrapolati…
Iowa State University（爱荷华州立大学）；Horia Hulubei National Institute for Physics and Nuclear Engineering（霍里亚·胡卢贝伊国家物理与核工程研究所）；Moscow State University（莫斯科国立大学）；Pacific National University（太平洋国立大学）；Institute for Basic Science（基础科学研究所）；Lawrence Berkeley National Laboratory（劳伦斯伯克利国家实验室）
Ab initio approaches in nuclear theory, such as the no-core shell model (NCSM), have been developed for approximately solving finite nuclei with realistic strong interactions. The NCSM and other approaches require an extrapolation of the results obtained in a finite basis space to the infinite basis space limit and assessment of the uncertainty of those extrapolations. Each observable requires a separate extrapolation and most observables have no proven extrapolation method. We propose a feed-forward artificial neural network (ANN) method as an extrapolation tool to obtain the ground state ene…
We present a method to extrapolate nuclear binding energies from known values for neighbouring nuclei. We select four specific mass relations constructed to eliminate smooth variation of the binding energy as function nucleon numbers. The fast odd-even variations are avoided by comparing nuclei with same parity. The mass relations are first tested and shown to either be rather accurately obeyed or revealing signatures of quickly varying structures. Extrapolations are initially made for a nucleus by applying each of these relations. Very reliable estimates are then produced either by an average…
Projection Monte Carlo calculations of lattice Chiral Effective Field Theory suffer from sign oscillations to a varying degree dependent on the number of protons and neutrons. Hence, such studies have hitherto been concentrated on nuclei with equal numbers of protons and neutrons, and especially on the alpha nuclei where the sign oscillations are smallest. Here, we introduce the "symmetry-sign extrapolation" method, which allows us to use the approximate Wigner SU(4) symmetry of the nuclear interaction to systematically extend the Projection Monte Carlo calculations to nuclear systems where th…
The extrapolation of self-consistent nuclear mean-field models to the region of superheavy elements is discussed with emphasis on the extrapolating power of the models. The predictions of modern mean-field models are confronted with recent experimental data. It is shown that a final conclusion about the location of the expected island of spherical doubly-magic superheavy nuclei cannot be drawn on the basis of the available data.
Nuclear many-body calculations are computationally demanding. An estimate of their accuracy is often hampered by the limited amount of computational resources even on present-day supercomputers. We provide an extrapolation method based on perturbation theory, so that the binding energy of a large basis-space calculation can be estimated without diagonalizing the Hamiltonian in this space. The extrapolation method is tested for 3H and 6Li nuclei. It will extend our computational abilities significantly and allow for reliable error estimates.
The prediction of nuclear observables beyond the finite model spaces that are accessible through modern ab initio methods, such as the no-core shell model, pose a challenging task in nuclear structure theory. It requires reliable tools for the extrapolation of observables to infinite many-body Hilbert spaces along with reliable uncertainty estimates. In this work we present a universal machine learning tool capable of capturing observable-specific convergence patterns independent of nucleus and interaction. We show that, once trained on few-body systems, artificial neural networks can produce …
The mass, or binding energy, is the basis property of the atomic nucleus. It determines its stability, and reaction and decay rates. Quantifying the nuclear binding is important for understanding the origin of elements in the universe. The astrophysical processes responsible for the nucleosynthesis in stars often take place far from the valley of stability, where experimental masses are not known. In such cases, missing nuclear information must be provided by theoretical predictions using extreme extrapolations. Bayesian machine learning techniques can be applied to improve predictions by taki…
Sequences of experimental ground-state energies for both odd and even $A$ are mapped onto concave patterns cured from convexities due to pairing and/or shell effects. The same patterns, completed by a list of excitation energies, give numerical estimates of the grand potential $Ω(β,μ)$ for a mixture of nuclei at low or moderate temperatures $T=β^{-1}$ and at many chemical potentials $μ.$ The average nucleon number $<{\bf A} >(β,μ)$ then becomes a continuous variable, allowing extrapolations towards nuclear masses closer to drip lines. We study the possible concavity of several thermodynamical …
The kernel ridge regression (KRR) approach is extended to include the odd-even effects in nuclear mass predictions by remodulating the kernel function without introducing new weight parameters and inputs in the training network. By taking the WS4 mass model as an example, the mass for each nucleus in the nuclear chart is predicted with the extended KRR network, which is trained with the mass model residuals, i.e., deviations between experimental and calculated masses, of other nuclei with known masses. The resultant root-mean-square mass deviation from the available experimental data for the 2…
Recent work has shown that a finite harmonic oscillator basis in nuclear many-body calculations effectively imposes a hard-wall boundary condition in coordinate space, motivating infrared extrapolation formulas for the energy and other observables. Here we further refine these formulas by studying two-body models and the deuteron. We accurately determine the box size as a function of the model space parameters, and compute scattering phase shifts in the harmonic oscillator basis. We show that the energy shift can be well approximated in terms of the asymptotic normalization coefficient and the…
The kernel ridge regression (KRR) method with Gaussian kernel is used to improve the description of the nuclear charge radius by several phenomenological formulae. The widely used $A^{1/3}$, $N^{1/3}$ and $Z^{1/3}$ formulae, and their improved versions by considering the isospin dependence are adopted as examples. The parameters in these six formulae are refitted using the Levenberg-Marquardt method, which give better results than the previous ones. The radius for each nucleus is predicted with the KRR network, which is trained with the deviations between experimental and calculated nuclear ch…
Extrapolations in Euclidean time form a central part of Nuclear Lattice Effective Field Theory (NLEFT) calculations using the Projection Monte Carlo method, as the sign problem in many cases prevents simulations at large Euclidean time. We review the next-to-next-to-leading order NLEFT results for the alpha nuclei up to $^{28}$Si, with emphasis on the Euclidean time extrapolations, their expected accuracy and potential pitfalls. We also discuss possible avenues for improving the reliability of Euclidean time extrapolations in NLEFT.
The properties of symmetric nuclear matter are investigated within the Green's functions approach. We have implemented an iterative procedure allowing for a self-consistent evaluation of the single-particle and two-particle propagators. The in-medium scattering equation is solved for a realistic (non-separable) nucleon-nucleon interaction including both particle-particle and hole-hole propagation. The corresponding two-particle propagator is constructed explicitely from the single-particle spectral functions. Results are obtained for finite temperatures and an extrapolation to T=0 is presented…
We provide an overview of recent work exploring the quark-mass dependence of hadronic observables and the associated role of chiral nonanalytic behavior due to the meson-cloud of hadrons. In particular, we address an issue of great current interest, namely the degree of model independence of results obtained through a controlled extrapolation of lattice QCD simulation results. Physical insights gained from this research are highlighted. We emphasize how chiral effective field theory formulated with a finite-range regulator provides a reliable and model-independent extrapolation to the physical…
Directional correlations in nuclear charge-radius model residuals enable extrapolation with calibrated uncertainties
Indian Institute of Technology (Indian School of Mines)（印度理工学院（印度矿业学院））；GSI Helmholtzzentrum für Schwerionenforschung GmbH（吉森亥姆霍特重离子研究中心有限公司）；Justus-Liebig-Universität Gießen（吉森尤斯图斯-李比希大学）；Saint Mary’s University（圣玛丽大学）；TRIUMF（特里厄姆夫实验室）
I review the application of self-consistent Green's functions methods to study the properties of infinite nuclear systems. Improvements over the last decade, including the consistent treatment of three-nucleon forces and the development of extrapolation methods from finite to zero temperature, have allowed for realistic predictions of the equation of state of infinite symmetric, asymmetric and neutron matter based on chiral interactions. Microscopic properties, like momentum distributions or spectral functions, are also accessible. Using an indicative set of results based on a subset of chiral…
We present three different methods to estimate error bars on the predictions made using a neural network. All of them represent lower bounds for the extrapolation errors. For example, we did not include an analysis on robustness against small perturbations of the input data. At first, we illustrate the methods through a simple toy model, then, we apply them to some realistic cases related to nuclear masses. By using theoretical data simulated either with a liquid-drop model or a Skyrme energy density functional, we benchmark the extrapolation performance of the neural network in regions of the…
The Garvey-Kelson relations are used in an iterative process to predict nuclear masses in the neighborhood of nuclei with measured masses. Average errors in the predicted masses for the first three iteration shells are smaller than those obtained with the best nuclear mass models. Their quality is comparable with the Audi-Wapstra extrapolations, offering a simple and reproducible procedure for short range mass predictions. A systematic study of the way the error grows as a function of the iteration and the distance to the known masses region, shows that a correlation exists between the error a…
The chart of the nuclides is limited by particle drip lines beyond which nuclear stability to proton or neutron emission is lost. Predicting the range of particle-bound isotopes poses an appreciable challenge for nuclear theory as it involves extreme extrapolations of nuclear masses beyond the regions where experimental information is available. Still, quantified extrapolations are crucial for a variety of applications, including the modeling of stellar nucleosynthesis. We use microscopic nuclear mass models and Bayesian methodology to provide quantified predictions of proton and neutron separ…
The problem of analytic continuation of the scattering data to the negative-energy region to obtain information on asymptotic normalization coefficients (ANCs) of bound states is discussed. It is shown that a recently suggested $Δ$ method [O.L.Ram\'ırez Suárez and J.-M. Sparenberg, Phys. Rev. C {\bf 96}, 034601 (2017)] is not strictly correct in the mathematical sense since it is not an analytic continuation of a partial-wave scattering amplitude to the region of negative energies. However, it can be used for practical purposes for sufficiently large charges and masses of colliding particles. …
A multi-task learning (MTL) framework, called gradient kernel ridge regression, for nuclear masses and separation energies is developed by introducing gradient kernel functions to the kernel ridge regression (KRR) approach. By taking the WS4 mass model as an example, the gradient KRR network is trained with the mass model residuals, i.e., deviations between experimental and theoretical values of masses and one-nucleon separation energies, to improve the accuracy of theoretical predictions. Significant improvements are achieved by the gradient KRR approach in both the interpolation and the extr…
We review recent progress in studies of nuclear final-state interactions in deep inelastic scattering (DIS) off the lightest nuclei tagged by a recoil nucleon. These processes hold a lot of potential for resolving the outstanding issues related to the dynamics of hadronization in QCD. Within the minimal Fock component framework, valid at large Bjorken $x$, the main features of the theoretical approach based on the virtual nucleon approximation are elaborated. In this approach, the strong final-state interaction of the DIS products with the nuclear fragments is described by an effective eikonal…
We investigate the energy dependence of the astrophysical $S$ factor for the reaction $^7$Be$(p,γ)^8$B, the primary source of high-energy solar neutrinos in the solar $pp$ chain. Using simple models we explore the model dependence in the extrapolation of the experimental data to the region of astrophysical interest near 20 keV. We find that below approximately 400 keV the energy dependence is very well understood and constrained by the data for the elastic scattering of low energy neutrons from $^7$Li. Above 400 keV nuclear distortion of the wave function of the incident proton introduces a si…
The effects of the variation of the dimensionless strong interaction parameter Xq=mq/Lambda{QCD} (mq is the quark mass, Lambda{QCD} is the QCD scale) are enhanced about 1.5 x 10**5 times in the 7.6 eV "nuclear clock" transition between the ground and first excited states in the 229Th nucleus and about 1 x 10**8 times in the relative shift of the 0.1 eV compound resonance in 150Sm.The best terrestrial limit on the temporal variation of the fundamental constants, |delta(Xq)/Xq| < 4 x 10**-9 at 1.8 billion years ago (|d(Xq/Xq)/dt| < 2.2 x 10**-18 y**-1), is obtained from the shift of this Sm reso…
The mass of $^{101}$Sn and Bayesian extrapolations to the proton drip line
Facility for Rare Isotope Beams, Michigan State University（稀有同位素束流设施，密歇根州立大学）；Department of Physics and Astronomy, Michigan State University（密歇根州立大学物理与天文学系）；Department of Physics and Astronomy, University of Notre Dame（圣母大学物理与天文学系）
In the process of understanding nuclear collisions, reliable extrapolations from pp collisions, based on Glauber models, are highly desirable, though seldomly accurate. We review the inclusion of diffractive excitations and argue that they provide an important contribution to centrality observables in pA collisions. We present a method for distinguishing between diffractively and non-diffractively wounded nucleons, and a proof-of-principle for an extrapolation of multiparton interaction models built on this.
This contribution reviews the present status of the Skyrme-Hartree-Fock (SHF) approach as one of the leading self-consistent mean-field models in the physics of atomic nuclei. It starts with a brief summary of the formalism and strategy for proper calibration of the SHF functional. The main emphasis lies on an exploration of the reliability of predictions, particularly in the regime of extrapolations. Various strategies are discussed to explore the statistical and systematic errors of SHF. The strategies are illustrated on examples from actual applications. Variations of model and fit data are…
The neutron and proton drip lines represent the limits of the nuclear landscape. While the proton drip line is measured experimentally up to rather high $Z$-values, the location of the neutron drip line for absolute majority of elements is based on theoretical predictions which involve extreme extrapolations. The first ever systematic investigation of the location of the proton and neutron drip lines in the covariant density functional theory has been performed by employing a set of the state-of-the-art parametrizations. Calculated theoretical uncertainties in the position of two-neutron drip …
The fusion probability for the production of superheavy nuclei in cold fusion reactions was investigated and compared with recent experimental results for $^{48}$Ca, $^{50}$Ti, and $^{54}$Cr incident on a $^{208}$Pb target. Calculations were performed within the fusion-by-diffusion model (FbD) using new nuclear data tables by Jachimowicz et al. It is shown that the experimental data could be well explained within the framework of the FbD model. The saturation of the fusion probability at bombarding energies above the interaction barrier is reproduced. It emerges naturally from the physical eff…
Product units in gated recurrent units improve nuclear-mass prediction
University of Applied Sciences Koblenz（应用科学大学科伦兹大学）；Technical University of Munich（慕尼黑技术大学）；University of Madeira（马德拉大学）；Washington University in St. Louis（圣路易斯华盛顿大学）
With the help of radial basis function (RBF) and the Garvey-Kelson relation, the accuracy and predictive power of some global nuclear mass models are significantly improved. The rms deviation between predictions from four models and 2149 known masses falls to about 200 keV. The AME95-03 and AME03-Border tests show that the RBF approach is a very useful tool for further improving the reliability of mass models. Simultaneously, the differences from different model predictions for unknown masses are remarkably reduced and the isospin symmetry is better represented when the RBF extrapolation is co…
The limits of the nuclear landscape are determined by nuclear binding energies. Beyond the proton drip lines, where the separation energy becomes negative, there is not enough binding energy to prevent protons from escaping the nucleus. Predicting properties of unstable nuclear states in the vast territory of proton emitters poses an appreciable challenge for nuclear theory as it often involves far extrapolations. In addition, significant discrepancies between nuclear models in the proton-rich territory call for quantified predictions. With the help of Bayesian methodology, we mix a family of …
The DNA of nuclear models: How AI predicts nuclear masses
NSF AI Institute for Artificial Intelligence and Fundamental Interactions（NSF人工智能与基本相互作用研究所）；Laboratory for Nuclear Science, MIT, Cambridge, MA 02139, USA（MIT核科学实验室）；Theoretical Physics Department, CERN, Geneva, Switzerland（欧洲核子研究中心理论物理部门）
The radial basis function (RBF) approach is applied in predicting nuclear masses for 8 widely used nuclear mass models, ranging from macroscopic-microscopic to microscopic types. A significantly improved accuracy in computing nuclear masses is obtained, and the corresponding rms deviations with respect to the known masses is reduced by up to 78%. Moreover, strong correlations are found between a target nucleus and the reference nuclei within about three unit in distance, which play critical roles in improving nuclear mass predictions. Based on the latest Weizsäcker-Skyrme mass model, the RBF a…
Machine learning methods and uncertainty quantification have been gaining interest throughout the last several years in low-energy nuclear physics. In particular, Gaussian processes and Bayesian Neural Networks have increasingly been applied to improve mass model predictions while providing well-quantified uncertainties. In this work, we use the probabilistic Mixture Density Network (MDN) to directly predict the mass excess of the 2016 Atomic Mass Evaluation within the range of measured data, and we extrapolate the inferred models beyond available experimental data. The MDN not only provides m…
Ruling out nucleonic direct Urca cooling in low-mass neutron stars using nuclear data
Université de Caen Normandie, ENSICAEN, CNRS/IN2P3, LPC Caen UMR6534（卡昂诺曼底大学，ENSICAEN，CNRS/IN2P3，LPC Caen UMR6534）；Dipartimento di Fisica “Aldo Pontremoli”, Università degli Studi di Milano（米兰大学物理系“Aldo Pontremoli”）；INFN, Sezione di Milano（意大利国家核物理研究所，米兰分部）；Department of Physics, University of Houston（休斯顿大学物理系）；Grand Accélérateur National d’Ions Lourds (GANIL), CEA/DRF-CNRS/IN2P3（重离子国家大型加速器（GANIL），CEA/DRF-CNRS/IN2P3）
Theoretical predictions of physical observables often involve extrapolations to regions that are poorly constrained by laboratory experiments and astrophysical observations. Without properly quantified theoretical errors, such model predictions are of very limited utility. In this contribution we use maximum-likelihood estimation to compute theoretical errors and assess correlations between physical observables. We illustrate the power and elegance of these methods using examples of both pedagogical and realistic interest. In particular, we implement a gaussian approximation to the likelihood …
Mimar Sinan Fine Arts University（米马尔·希南美术大学）；University of Leuven (KU Leuven)（鲁汶大学（KU Leuven））；Yildiz Technical University（耶尔德兹技术大学）
Neural networks have become popular in many fields of science since they serve as promising, reliable and powerful tools. In this work, we study the effect of data augmentation on the predictive power of neural network models for nuclear physics data. We present two different data augmentation techniques, and we conduct a detailed analysis in terms of different depths, optimizers, activation functions and random seed values to show the success and robustness of the model. Using the experimental uncertainties for data augmentation for the first time, the size of the training data set is artific…
