Showing papers by "Igor I. Soloviev published in 2018"

Adiabatic superconducting artificial neural network: Basic cells

Abstract: We consider adiabatic superconducting cells operating as an artificial neuron and synapse of a multilayer perceptron (MLP). Their compact circuits contain just one and two Josephson junctions, respectively. While the signal is represented as magnetic flux, the proposed cells are inherently nonlinear and close-to-linear magnetic flux transformers. The neuron is capable of providing the one-shot calculation of sigmoid and hyperbolic tangent activation functions most commonly used in MLP. The synapse features both positive and negative signal transfer coefficients in the range ∼ ( − 0.5 , 0.5 ). We briefly discuss implementation issues and further steps toward the multilayer adiabatic superconducting artificial neural network, which promises to be a compact and the most energy-efficient implementation of MLP.

Adiabatic Superconducting Artificial Neural Network: Basic Cells.

Abstract: We consider adiabatic superconducting cells operating as an artificial neuron and synapse of a multilayer perceptron (MLP). Their compact circuits contain just one and two Josephson junctions, respectively. While the signal is represented as magnetic flux, the proposed cells are inherently nonlinear and close-to-linear magnetic flux transformers. The neuron is capable of providing a one-shot calculation of sigmoid and hyperbolic tangent activation functions most commonly used in MLP. The synapse features by both positive and negative signal transfer coefficients in the range ~ (-0.5,0.5). We briefly discuss implementation issues and further steps toward multilayer adiabatic superconducting artificial neural network which promises to be a compact and the most energy-efficient implementation of MLP.

Protected 0-π states in SIsFS junctions for Josephson memory and logic

Abstract: We study the peculiarities in current-phase relations (CPR) of the SIsFS junction in the region of 0 to π transition. These CPR consist of two independent branches corresponding to 0- and π-states of the contact. We have found that depending on the transparency of the SIs tunnel barrier, the decrease in the s-layer thickness leads to transformation of the CPR shape going in the two possible ways: either one of the branches exists only in discrete intervals of the phase difference φ or both branches are sinusoidal but differ in the magnitude of their critical currents. We demonstrate that the difference can be as large as 10% under maintaining superconductivity in the s layer. An applicability of these phenomena for memory and logic application is discussed.

Protected 0-pi states in SIsFS junctions for Josephson memory and logic

Abstract: We study the peculiarities in current-phase relations (CPR) of the SIsFS junction in the region of $0$ to $\pi $ transition. These CPR consist of two independent branches corresponding to $0-$ and $\pi-$ states of the contact. We have found that depending on the transparency of the SIs tunnel barrier the decrease of the s-layer thickness leads to transformation of the CPR shape going in the two possible ways: either one of the branches exists only in discrete intervals of the phase difference $\varphi$ or both branches are sinusoidal but differ in the magnitude of their critical currents. We demonstrate that the difference can be as large as $10\%$ under maintaining superconductivity in the s layer. An applicability of these phenomena for memory and logic application is discussed.

Energy Efficient Superconducting Neural Networks for High-Speed Intellectual Data Processing Systems

Abstract: We present the results of circuit simulations for the adiabatic flux-operating neuron. The proposed cell with one-shot calculation of activation function is based on a modified single-junction superconducting quantum interferometer. In comparison, functionally equivalent elements of the artificial neural network (ANN) in the semiconductor-based implementations consist of approximately 20 transistors. Also in the article, we present the connecting synapse based on the adiabatic quantum flux parametron. These neurons and synapses allow constructing ANNs with a magnetic representation of information in the form of direction and/or magnitude of the magnetic flux in the superconducting circuit. We discuss the dissipation of energy during operations in the frame of the proposed concept. This value in superconducting neurons and synapses with sub-nanosecond timescale can be reduced down to 10 and 0.1 aJ, respectively. The use of the adiabatic superconducting logic circuits in our approach promises compatibility with superconducting quantum information processing systems.

Determination of the Current-Phase Relation in Josephson Junctions by Means of an Asymmetric Two-Junction SQUID

Abstract: An analytical approach that makes it possible to reconstruct the current–phase relation (CPR) in Josephson structures included in one of the arms of a two-junction superconducting quantum interference device (SQUID), where the second junction has a significantly higher critical current and a known (sinusoidal) CPR, has been developed. The developed methods of analytical and numerical studies of current–flow transformations in two-junction SQUIDs make it possible to reconstruct the CPR of a junction with a low critical current taking into account both the existence of the self-inductance of the interferometer contour and a possible asymmetry in the supply current system. The efficiency of this approach has been confirmed by the experimental study of niobium–aluminum/aluminum oxide–niobium test structures with the known CPR.

Compact Josephson φ-Junctions

Abstract: This chapter is devoted to the study of controllable proximity effects in superconductors (S), in terms of both fundamental aspects and applications. As a part of the work, theoretical description was suggested for a number of structures with superconducting electrodes and multiple interlayers with new physics related to the proximity effect and nanoscale φ-junctions. They are Josephson structures with the phase of the ground state φ g , 0 < φ g < π φ-junctions can be created on the basis of longitudinally oriented normal metal (N) and ferromagnetics (F) layers between superconducting electrodes. Under certain conditions, the amplitude of the first harmonic in the current-phase relation (CPR) is relatively small due to F layer. The coupling across N layer provides negative sign of the second harmonic. To derive quantitative criteria for realization of a φ-junction, we have solved two-dimensional boundary-value problem in the frame of Usadel equations for overlap and ramp geometries of different structures with NF bilayer. This chapter is focused on different geometries of nanoscale φ-structures of the size much less than Josephson penetration depth λ J . At the same time, φ-state cannot be realized in conventional SNS and SFS sandwiches. Proximity effect between N and F layers limits minimal possible size of φ-junction. In the case of smaller junctions, NF bilayer becomes almost homogeneous, φ-state is prohibited, and junction exists in 0- or π-state. The conditions for realization of φ-junctions in ramp-type S–NF–S, overlap-type SFN–FN–NFS, and RTO-type SN–FN–NS geometries are discussed in the chapter. It is shown that RTO-type SN–FN–NS geometry is most suitable for practical realization. It is also shown in this chapter that the parameter range of φ-state existence can be sufficiently broadened. It allows to realize Josephson φ-junctions using up-to-date technology. By varying the temperature, we can slightly shift the region of 0-π transition and, consequently, we can control the mentioned phase of the ground state. Furthermore, sensitivity of the ground state to an electron distribution function permits applications of φ-junctions as small-scale self-biasing single-photon detectors. Moreover, these junctions are controllable and have degenerate ground states +φ and −φ, providing necessary condition for the so-called silent quantum bits.