Dense analog synaptic crossbar arrays are a promising candidate for neuromorphic hardware accelerators due to the ability to mitigate data movement by performing in-situ vector-matrix products and weight updates within the storage array itself. However, many analog weight storage cells suffer from long latencies or low dynamic ranges, limiting the achievable performance. In this work, we demonstrate that the voltage-controlled partial polarization switching dynamics in ferroelectric-field-effect transistors (FeFET) can be harnessed to enable a 32 state non-volatile analog synaptic weight cell with large dynamic range (67×) and low latency weight updates (50 ns) for an amplitude modulated pulse scheme.

We report the production of ultracold heteronuclear 7 Li 85 Rb molecules in excited electronic states by photoassociation (PA) of ultracold 7 Li and 85 Rb atoms. PA is performed in a dual-species 7 Li-85 Rb magnetooptical trap (MOT) and the PA resonances are detected using trap loss spectroscopy. We identify several strong PA resonances below the Li (2s 2 S 1/2 ) + Rb (5p 2 P 3/2 ) asymptote and experimentally determine the long range C 6 dispersion coefficients. We find a molecule formation rate (P LiRb ) of 3.5×10 7 s -1 and a PA rate coefficient (K PA ) of 1.3×10 -10 cm 3 /s, the highest among heteronuclear bi-alkali molecules. At large PA laser intensity, we observe the saturation of the PA rate coefficient (K PA ) close to the theoretical value at the unitarity limit. DOI:PACS number (s): 34.50.-s, 37.10.Mn, 33.20.-t Heteronuclear polar molecules have recently attracted enormous attention [1-17] owing to their ground state having a large electric dipole moment [16]. The long range anisotropic dipole-dipole interaction in such systems is the basis for a variety of applications including quantum computing [13], precision measurements [14], ultracold chemistry [2] and quantum simulations [15]. Heteronuclear bi-alkali molecules (XY, where X and Y are two different alkali atom species), only a small subset of polar molecules, have received special attention mainly because the constituent alkali atoms are easy to laser cool and can be easily associated to form molecules at ultracold temperatures. The two primary methods for production of heteronuclear bialkali molecules have been magneto-association (MA), as in the case of KRb, NaK and NaLi [1][2][3][4], and photoassociation (PA), as in the case of LiCs, RbCs, NaCs, KRb and LiK [5][6][7][8][9][10][11][12]. Such molecules can be transferred to their absolute ground state where they have significant dipole moment, for example, by Stimulated Raman Adiabatic Passage (STIRAP) [1,12]. There is considerable interest in other heteronuclear combinations either due to their higher dipole moments, different quantum statistics or the possibility of finding simpler or more efficient methods for the production of ultracold molecules.In this Rapid Communication, we report a highly efficient production of ultracold 7 Li 85 Rb molecules by PA. Prior to our work (also see [18]), LiRb was one of the few bi-alkali molecules yet to be produced at ultracold temperatures. There is considerable interest in LiRb because the rovibronic ground state LiRb molecule is predicted to have a relatively high dipole moment of 4.1 Debye (exceeded only by LiCs and NaCs) [16], which makes it a strong candidate for many of the applications mentioned above. It is also interesting to note that bosonic 85 Rb, 87 Rb and 7 Li, and the fermionic 6 Li are among the more commonly trapped alkali atomic species. This can make LiRb molecules readily available in both fermionic and bosonic forms (depending on the Li isotope chosen), broadening the range of physics that can be studied. We provide the first step tow...

Computationally hard problems, including combinatorial optimization, can be mapped into the problem of finding the ground-state of an Ising Hamiltonian. Building physical systems with collective computational ability and distributed parallel processing capability can accelerate the ground-state search. Here, we present a continuous-time dynamical system (CTDS) approach where the ground-state solution appears as stable points or attractor states of the CTDS. We harness the emergent dynamics of a network of phase-transition nanooscillators (PTNO) to build an Ising Hamiltonian solver. The hardware fabric comprises of electrically coupled injection-locked stochastic PTNOs with bi-stable phases emulating artificial Ising spins. We demonstrate the ability of the stochastic PTNO-CTDS to progressively find more optimal solution by increasing the strength of the injection-locking signal -akin to performing classical annealing. We demonstrate in silico that the PTNO-CTDS prototype solves a benchmark non-deterministic polynomial time (NP)-hard Max-Cut problem with high probability of success. Using experimentally calibrated numerical simulations, we investigate the performance of the hardware with increasing problem size. We show the best-in-class energy-efficiency of 3.26x10 7 solutions/sec/Watt which translates to over five orders of magnitude improvement when compared with digital CMOS, superconducting qubit and photonic Ising solver approaches. We also demonstrate an order of magnitude improvement over a discrete-time memristor-based Hopfield network approach. Such an energy efficient CTDS hardware exhibiting high solutionthroughput/Watt can find application in industrial planning and manufacturing, defense and cyber-security, bioinformatics and drug discovery.Combinatorial optimization is ubiquitous in various fields such as artificial intelligence, bioinformatics, drug discovery, cryptography, quantitative finance, operations research, resource allocation, trajectory and route planning. Such problems belong to the NP-hard or NP-complete complexity class, requiring computational resources (time and/or energy) that scale exponentially with the problem size. Interestingly, many combinatorial optimization problems can be translated into another fundamental physics problem of finding the ground state of an Ising model (1) (or its equivalent Quadratic Unconstrained Binary Optimization (QUBO) problem ( 2)). The Ising model, describing the property of spin glass, was put forward as a tool of statistical physics to explain the phenomenon of ferromagnetism. The Ising Hamiltonian with 𝑁𝑁 discrete spins 𝜎𝜎 1≤𝑖𝑖≤𝑁𝑁 ∈ {−1, +1} 𝑁𝑁

The two possible pathways toward artificial intelligence (AI)-(i) neuroscience-oriented neuromorphic computing [like spiking neural network (SNN)] and (ii) computer science driven machine learning (like deep learning) differ widely in their fundamental formalism and coding schemes (Pei et al., 2019). Deviating from traditional deep learning approach of relying on neuronal models with static nonlinearities, SNNs attempt to capture brainlike features like computation using spikes. This holds the promise of improving the energy efficiency of the computing platforms. In order to achieve a much higher areal and energy efficiency compared to today's hardware implementation of SNN, we need to go beyond the traditional route of relying on CMOS-based digital or mixed-signal neuronal circuits and segregation of computation and memory under the von Neumann architecture. Recently, ferroelectric field-effect transistors (FeFETs) are being explored as a promising alternative for building neuromorphic hardware by utilizing their nonvolatile nature and rich polarization switching dynamics. In this work, we propose an all FeFET-based SNN hardware that allows low-power spike-based information processing and co-localized memory and computing (a.k.a. in-memory computing). We experimentally demonstrate the essential neuronal and synaptic dynamics in a 28 nm high-K metal gate FeFET technology. Furthermore, drawing inspiration from the traditional machine learning approach of optimizing a cost function to adjust the synaptic weights, we implement a surrogate gradient (SG) learning algorithm on our SNN platform that allows us to perform supervised learning on MNIST dataset. As such, we provide a pathway toward building energy-efficient neuromorphic hardware that can support traditional machine learning algorithms. Finally, we undertake synergistic device-algorithm co-design by accounting for the impacts of device-level variation (stochasticity) and limited bit precision of on-chip synaptic weights (available analog states) on the classification accuracy.

The possibility of using spin waves for information transmission and processing has been an area of active research due to the unique ability to manipulate the amplitude and phase of the spin waves for building complex logic circuits with less physical resources and low power consumption. Previous proposals on spin wave logic circuits have suggested the idea of utilizing the magneto-electric effect for spin wave amplification and amplitude- or phase-dependent switching of magneto-electric cells. Here, we propose a comprehensive scheme for building a clocked non-volatile spin wave device by introducing a charge-to-spin converter that translates information from electrical domain to spin domain, magneto-electric spin wave repeaters that operate in three different regimes - spin wave transmitter, non-volatile memory and spin wave detector, and a novel clocking scheme that ensures sequential transmission of information and non-reciprocity. The proposed device satisfies the five essential requirements for logic application: nonlinearity, amplification, concatenability, feedback prevention, and complete set of Boolean operations.