Observations of astrophysical transients and transits of exoplanets on small telescopes of INASAN and the system of their automatic planning
Observations of astrophysical transients and transits of exoplanets on small telescopes of INASAN and the system of their automatic planning
A. N. Tarasenkov$^{1,2}$
A network of small-diameter robotic telescopes is being created at the Institute of Astronomy of the Russian Academy of Sciences to study variable stars, dangerous asteroids, space debris, exoplanets, and astrophysical transients. This paper describes the architecture and operating principles of the operational planning system for photometric observations at INASAN telescopes. It describes the principles of constructing a database of objects, methods for prioritizing observations of various types of objects, and calculating an observation plan. It also describes the results of observations of astrophysical transients of various natures and exoplanetary transits at INASAN small telescopes, obtained using the observation planning system.
Show AbstractCorrelations of neutron star characteristics with properties of nucleon and hyperon interactions
Correlations of neutron star characteristics with properties of nucleon and hyperon interactions
A. I. Nasakin$^{1,2}$, D. E. Lanskoy$^1$, S. A. Mikheev$^{1,2}$, A. M. Potokin$^1$, T. Yu. Tretyakova$^{1,2}$
The hyperons appearance at densities several times greater than the nuclear one in the massive neutron stars interior has a significant impact on neutron star characteristics. Using various parameterizations of Skyrme forces, we calculated the equations of state of neutron star matter and analyzed possible correlations between the observed characteristics of neutron stars and the properties of nucleon and hyperon interactions. The $\Lambda N$ interaction contracting power, which characterizes the ability of the $\Lambda$-hyperon to modify the nucleon core of the hypernucleus, has the strongest influence on the density of matter at the point of hyperon appearance and, consequently, on the characteristics of neutron stars. The relationship between astrophysical quantities and the properties of nucleon interactions weakens with the appearance of hyperons, but in most cases remains. The incompressibility of nuclear matter, quantities reflecting the behavior of the symmetry energy at high densities, and the hyperon interaction contracting power determine the maximum mass of a neutron star.
Show AbstractApplication of machine learning methods for identification of spin centers in metal oxides
Application of machine learning methods for identification of spin centers in metal oxides
E. V. Kytina, A. V. Vasiliev, E. A. Konstantinova, D. R. Khokhlov
In this paper, we solve the problem of classifying electron paramagnetic resonance (EPR) spectra of nanocrystalline metal oxides (using the example of aluminum and titanium nanoxides) and identifying spin centers using machine learning methods. Based on the literature data and the performed study of metal nanoxides by the EPR method, spin centers (radicals) were selected, which are most often found in nanoscale oxides, in particular in aluminum and titanium oxides. Since the available experimental spectra turned out to be insufficient to form a representative dataset, a synthetic dataset was used to train the models, obtained by simulating experimental spectra and adding Gaussian noise in order to approximate the simulated EPR spectra to real ones. The test dataset consisted of real experimental EPR spectra. To solve the classification problem, classical machine learning models (SVM, Random Forest, LGBM, CatBoost, XGBoost) and convolutional neural networks (ResNet18, Resnet34, Efficientnetb0, Efficientnetb3, Mobilenetv2) were tested. A preliminary data preprocessing was carried out, which consisted in extracting additional features from spectral data for classical machine learning tasks. The best average accuracy of 98% was obtained using the CatBoost gradient boosting model and the ResNet18 neural network model. The results obtained will automate the processing of EPR spectra, which will greatly simplify and accelerate the work of scientists conducting scientific research in the fields of condensed matter physics, and will also contribute to the popularization of the EPR method in the scientific community.
Show AbstractStudy of optical properties of nanostructured germanium after cluster ion bombardment
Study of optical properties of nanostructured germanium after cluster ion bombardment
I. V. Nikolaev, I. A. Azarov, N. G. Korobeishchikov
The optical properties of single-crystal germanium nanostructured by an argon cluster ion beam were studied. The initial germanium surfaces were bombarded with argon cluster ions with a low specific energy (10 eV/atom). The ion fluence was 1.0×10^16, 1.4×10^16 and 4.2×10^16 cluster ion/cm^2. Using spectral ellipsometry, dispersion curves were obtained, and a comparison was made between crystalline, amorphous and nanostructured germanium after cluster ion bombardment. Anisotropy of the effective refractive and absorption indices of nanostructured germanium samples along and across the direction of the wave vector of nanostructure was demonstrated.
Show AbstractLocal chemical reactions involving fluorine atoms on the Pt(111) surface
Local chemical reactions involving fluorine atoms on the Pt(111) surface
D. A. Muzychenko$^1$, S. I. Oreshkin$^2$, M. N. Petukhov$^3$, M. A. Ardamin$^1$, М. Yu. Alexandrov$^1$, V. I. Panov$^1$, A. I. Oreshkin$^1$
The adsorption of fluorine atoms on the Pt(111) surface was studied using ultra-high-vacuum scanning tunneling microscopy (UHV STM) and X-ray photoelectron spectroscopy (XPS). Fluorofullerene molecules were chosen as fluorine sources. When deposited on Pt(111), fluorine atoms are cleaved from the carbon backbone, forming new surface structures. Varying the stoichiometric composition, as well as the number of deposited fluorofullerene molecules, allows for varying fluorine atom concentrations on the surface. Using C60F18 molecules, the formation of a fluorine-induced surface structure was observed. The observed structure was unstable and completely disappeared within 3-4 days of continuous STM monitoring. Deposition of C60F48 molecules on Pt(111) also resulted in the formation of unstable structures on the surface. XPS spectra clearly demonstrated the absence of chemical interaction between platinum and fluorine. The presence of hydrogen on the platinum surface and its interaction with fluorine to form volatile hydrogen fluoride explains the disappearance of the observed structures over time.
Show AbstractWave Function and S Factor of Nucleon into Quark-Diquark Model with Chromodynamical Interaction
Wave Function and S Factor of Nucleon into Quark-Diquark Model with Chromodynamical Interaction
Yu. D. Chernichenko
Finite-difference and integral forms of relativistic quasipotential equations in the configuration representation for the wave function of a nucleon in the framework of its quark-diquark model are obtained. In this model, the nucleon is considered as a two-particle composite system in which the quark has spin 1/2 and the diquark has spin 0. Approximate solutions of the scattering problem and the spectral problem for the radial wave function of the s-state of a nucleon in a quark-diquark model with a Coulomb (chromodynamical) potential are found. The condition for quantization of the energy levels of the nucleon corresponding to the Coulomb chromodynamical potential is determined. An expression for the relativistic threshold resummation S factor of a nucleon is obtained and its properties are investigated. The new regularities of behavior for the threshold S factor of the nucleon are established. The consideration is carried out within the framework of a relativistic quasipotential approach based on the covariant Hamiltonian formulation of quantum field theory, via a transition from the momentum formulation in Lobachevsky space to a three-dimensional relativistic configurational representation for the case of a composite system of two relativistic spin particles of arbitrary masses.
Show AbstractActivation functions for deep learning based on generalized entropies
Activation functions for deep learning based on generalized entropies
R. A. Rudamenko, A. M. Savchenko, K. M. Semenov
The Shannon (Boltzmann-Gibbs) entropy is the foundation of classical statistical mechanics and deep learning, but it has difficulties in describing the dynamics of non-extensive systems. This paper proposes the application of generalized entropies to construct new fundamental blocks in deep neural network architectures. The proposed approach generalizes the classical softmax layer using the parametric entropies of Renyi, Tsallis, and Sharma-Mittal. The parameters \(q\) and \(r\) control the shape of the distribution: for \(q \to 1\), the optimal distribution converges to softmax, and for \(q = 2\), it converges to sparsemax. In particular, a variant corresponding to \(q\)-entmax is considered, where adaptivity is achieved by varying the parameter \(q\) for a fixed \(r\). The study includes obtaining analytical expressions for the Jacobian in terms of the parameters \(q\) and \(r\) for the purpose of optimization via explicit differentiation methods. A comparative analysis with existing approaches -- softmax, sparsemax, and entmax (for \(q \in \{1.25,1.5,1.75\}\)) -- was conducted. The obtained results demonstrate an increase in quality metrics relative to the softmax, sparsemax, and q-entmax approaches for the classification problem with correlated class labels, which allows us to conclude that the Sharma-Mittal-based method is superior to the task at hand.
Show AbstractNeural Network Modeling of Single-Pion Electroproduction Observables on the Proton in the Resonance Region
Neural Network Modeling of Single-Pion Electroproduction Observables on the Proton in the Resonance Region
A. V. Golda$^1$, E. L. Isupov, A. A. Rusova$^1$, V. V. Chistyakova$^1$
This work presents a neural-network–based approach for predicting differential cross sections of the exclusive single-pion electroproduction reaction $\gamma^* p \rightarrow n \pi^+$. The problem is formulated as a multidimensional regression task in the kinematic variables of the reaction. A deep fully connected neural network is trained on experimental data from the CLAS detector without imposing any a priori theoretical assumptions on the reaction dynamics. Particular attention is paid to systematic validation of the model and to the assessment of its physical reliability. The predictive performance of the network is validated through a comparison with the phenomenological MAID2007 model. This comparison demonstrates that the neural network correctly reproduces the characteristic angular dependence and the overall behavior of the cross sections, including regions of phase space not represented in the training dataset. In addition, a replica method is employed, based on constructing an ensemble of neural-network models trained on statistically fluctuated datasets, which allows for a quantitative evaluation of the stability of the predictions and the associated uncertainties. The results show that the proposed model provides physically consistent predictions of differential cross sections, remains robust against statistical fluctuations in the input data, and accurately reproduces the multidimensional correlations in the kinematic distributions of single-pion electroproduction on the proton.
Show AbstractModelling of the SPD end-cap detector
Modelling of the SPD end-cap detector
V. A. Kuzmin
Using the example of the straw tube end-cap tracker of the SPD (Spin Physics Detector), which is currently being developed at JINR (Dubna), we simulate the operation of this type of detector to evaluate its effectiveness and accuracy of track reconstruction for various configurations of its layers. It is shown that a tracker with a two-layer construction of coordinate planes will not provide unambiguous information for track reconstruction due to the large range of angles of entry of tracks into the detector. A method for reconstructing tracks using layered detectors made of drift tubes, in which the layers do not combine to form coordinate planes, is proposed. Only signals from individual detector layers are used for reconstructing the tracks. To determine the optimal design geometry of the tracker, the efficiency and accuracy of the track reconstruction are calculated for 20 different geometrical configurations of the detector, with different levels of accuracy in its electronics. The sufficiency of the modeling methodology used is shown. Simulation results indicate the possibility of creating a single-layer tracker with coordinate planes with accuracy characteristics similar to those of a detector consisting of perfect two-layer planes.
Show AbstractOptical Simulator For Quantum Key Distribution
Optical Simulator For Quantum Key Distribution
L. V. Biguaa, S. P. Kulik
In this paper, we propose and test a scheme of a novel optical system for the numerical and experimental study of quantum key distribution (QKD) systems, as well as various attacks on the implementation of known protocols in a plug-and-play mode. Moreover, the system exploits the standard laboratory tools such as general purpose diode lasers and photodiode detectors. This allows to perform QKD demonstration without need for deploying an expensive and technically complex actual QKD system. Such a paradigm seems to be useful in the context of intensive developing of QKD systems for research purposes Firstly, it is linked to the emerging need in numerical simulation of QKD systems for research purposes. In turn, the experimental realization of such an approach is essential for practical demonstration of QKD systems as well as training the specialists in the field of quantum ciphering. Moreover, the currently available systems only allow researchers to experimentally reproduce the simplest eavesdropper (Eve) attacks on QKD protocols, such as the intercept-and-resend attacks. They can also only numerically simulate a wide range of attacks, but using enough sophisticated systems. In contrast, the suggested system allows for both numerical and experimental replication of attacks ranging from simple ones to technically challenging PNS-like ones in a simple plug-and-play manner.
Show Abstract