arXiv subjects

Sayeef Salahuddin

Publications and source records attributed to Sayeef Salahuddin.

At least 37 records · Page 2Linked to original sources

arXiv2020

Logically Synthesized, Hardware-Accelerated, Restricted Boltzmann Machines for Combinatorial Optimization and Integer Factorization

The Restricted Boltzmann Machine (RBM) is a stochastic neural network capable of solving a variety of difficult tasks such as NP-Hard combinatorial optimization problems and integer factorization. The RBM architecture is also very compact; requiring very few weights and biases. This, along with its simple, parallelizable sampling algorithm for finding the ground state of such problems, makes the RBM amenable to hardware acceleration. However, training of the RBM on these problems can pose a significant challenge, as the training algorithm tends to fail for large problem sizes and efficient mappings can be hard to find. Here, we propose a method of combining RBMs together that avoids the need to train large problems in their full form. We also propose methods for making the RBM more hardware amenable, allowing the algorithm to be efficiently mapped to an FPGA-based accelerator. Using this accelerator, we are able to show hardware accelerated factorization of 16 bit numbers with high accuracy with a speed improvement of 10000x and a power improvement of 32x.

cs.LG↗

arXiv2020

Ising Model Optimization Problems on a FPGA Accelerated Restricted Boltzmann Machine

Optimization problems, particularly NP-Hard Combinatorial Optimization problems, are some of the hardest computing problems with no known polynomial time algorithm existing. Recently there has been interest in using dedicated hardware to accelerate the solution to these problems, with physical annealers and quantum adiabatic computers being some of the state of the art. In this work we demonstrate usage of the Restricted Boltzmann Machine (RBM) as a stochastic neural network capable of solving these problems efficiently. We show that by mapping the RBM onto a reconfigurable Field Programmable Gate Array (FPGA), we can effectively hardware accelerate the RBM's stochastic sampling algorithm. We benchmark the RBM against the DWave 2000Q Quantum Adiabatic Computer and the Optical Coherent Ising Machine on two such optimization problems: the MAX-CUT problem and finding the ground state of a Sherrington-Kirkpatrick (SK) spin glass. On these problems, the hardware accelerated RBM shows best in class performance compared to these other accelerators, with an empirical scaling performance of $\mathcal{O}(e^{-N})$ for probability of reaching the ground state compared to a similar empirical $\mathcal{O}(e^{-N})$ for the CIM (with the RBM showing a constant factor of improvement over the CIM) and empirical $\mathcal{O}(e^{-N^2})$ for the DWave Annealer. The results show up to $10^7$x and $10^5$x time to solution improvement compared to the DWave 2000Q on the MAX-CUT and SK problems respectively, along with a $150$x and $1000$x performance increase compared to the Coherent Ising Machine annealer on those problems. By using commodity hardware running at room temperature for acceleration, the RBM also has greater potential for immediate and scalable use.

cs.AR↗

arXiv2020

Negative Capacitance Enables FinFET Scaling Beyond 3nm Node

A comprehensive study of the scaling of negative capacitance FinFET (NC-FinFET) is conducted with TCAD. We show that the NC-FinFET can be scaled to "2.1nm node" and almost "1.5nm node" that comes two nodes after the industry "3nm node," which has 16nm Lg and is the last FinFET node according to the International Roadmap for Devices and Systems (IRDS). In addition, for the intervening nodes, NC-FinFET can meet IRDS Ion and Ioff target at target-beating VDD. The benefits of negative capacitance (NC) include improved subthreshold slope (SS), drain-induced barrier lowering (DIBL), Vt roll-off, transconductance over Id (Gm/Id), output conductance over Id (Gd/Id), and lower VDD. Further scaling may be achieved by improving capacitance matching between ferroelectric (FE) and dielectric (DE).

physics.app-ph↗

arXiv2020

One nanometer HfO$_2$-based ferroelectric tunnel junctions on silicon

In ferroelectric materials, spontaneous symmetry breaking leads to a switchable electric polarization, which offers significant promise for nonvolatile memories. In particular, ferroelectric tunnel junctions (FTJs) have emerged as a new resistive switching memory which exploit polarization-dependent tunnel current across a thin ferroelectric barrier. Here we demonstrate FTJs with CMOS-compatible Zr-doped HfO$_2$ (Zr:HfO$_2$) ferroelectric barriers of just 1 nm thickness, grown by atomic layer deposition on silicon. These 1 nm Zr:HfO$_2$ tunnel junctions exhibit large polarization-driven electroresistance (19000$\%$), the largest value reported for HfO$_2$-based FTJs. In addition, due to just a 1 nm ferroelectric barrier, these junctions provide large tunnel current (> 1 A/cm$^2$) at low read voltage, orders of magnitude larger than reported thicker HfO$_2$-based FTJs. Therefore, our proof-of-principle demonstration provides an approach to simultaneously overcome three major drawbacks of prototypical FTJs: a Si-compatible ultrathin ferroelectric, large electroresistance, and large read current for high-speed operation.

cond-mat.mtrl-sci↗

arXiv2020

Spin-orbit torque generated by amorphous Fe$_{x}$Si$_{1-x}$

While tremendous work has gone into spin-orbit torque and spin current generation, charge-to-spin conversion efficiency remains weak in silicon to date, generally stemming from the low spin-orbit coupling (low atomic number, Z) and lack of bulk lattice inversion symmetry breaking. Here we report the observation of spin-orbit torque in an amorphous, non-ferromagnetic Fe$_{x}$Si$_{1-x}$ / cobalt bilayer at room temperature, using spin torque ferromagnetic resonance and harmonic Hall measurements. Both techniques provide a minimum spin torque efficiency of about 3 %, comparable to prototypical heavy metals such as Pt or Ta. According to the conventional theory of the spin Hall effect, a spin current in an amorphous material is not expected to have any substantial contribution from the electronic bandstructure. This, combined with the fact that Fe$_{x}$Si$_{1-x}$ does not contain any high-Z element, paves a new avenue for understanding the underlying physics of spin-orbit interaction and opens up a new class of material systems - silicides - that is directly compatible with complementary metal-oxide-semiconductor (CMOS) processes for integrated spintronics applications.

cond-mat.mes-hall↗

arXiv2020

Anomalous Subthreshold Behaviors in Negative Capacitance Transistors

Recent measurements on ultra-thin body Negative Capacitance Field Effect Transistors have shown subthreshold behaviors that are not expected in a classical MOSFET. Specifically, subthreshold swing was found to decrease with increased gate bias in the subthreshold region for devices measured over multiple gate lengths down to 30 nm. In addition, improvement in the subthreshold swing relative to control devices showed a non-monotonic dependence on the gate length. In this paper, using a Landau-Khanatnikov ferroelectric gate stack model calibrated with measured Capacitance-Voltage, we show that both these anomalous behaviors can be quantitatively reproduced with TCAD simulations.

physics.app-ph↗

arXiv2020

Unifying femtosecond and picosecond single-pulse magnetic switching in GdFeCo

Many questions are still open regarding the physical mechanisms behind the magnetic switching in GdFeCo alloys by single optical pulses. Phenomenological models suggest a femtosecond scale exchange relaxation between sublattice magnetization as the driving mechanism for switching. The recent observation of thermally induced switching in GdFeCo by using both several picosecond optical laser pulse as well as electric current pulses has questioned this previous understanding. This has raised the question of whether or not the same switching mechanics are acting at the femo- and picosecond scales. In this work, we aim at filling this gap in the understanding of the switching mechanisms behind thermal single-pulse switching. To that end, we have studied experimentally thermal single-pulse switching in GdFeCo alloys, for a wide range of system parameters, such as composition, laser power and pulse duration. We provide a quantitative description of the switching dynamics using atomistic spin dynamics methods with excellent agreement between the model and our experiments across a wide range of parameters and timescales, ranging from femtoseconds to picoseconds. Furthermore, we find distinct element-specific damping parameters as a key ingredient for switching with long picosecond pulses and argue, that switching with pulse durations as long as 15 picoseconds is possible due to a low damping constant of Gd. Our findings can be easily extended to speed up dynamics in other contexts where ferrimagnetic GdFeCo alloys have been already demonstrated to show fast and energy-efficient processes, e.g. domain-wall motion in a track and spin-orbit torque switching in spintronics devices.

cond-mat.mtrl-sci↗

arXiv2019

Electric-Field Control of the Interlayer Exchange Coupling for Magnetization Switching

We propose an electric-field-controlled mechanism for magnetization switching assisted solely by the interlayer-exchange coupling (IEC) between the fixed and the free magnets, which are separated by two oxide barriers sandwiching a spacer material known for exhibiting large IEC. The basic idea relies on the formation of a quantum-well (QW) within the spacer material and controlling the transmission coefficient across the structure with an electric-field via the resonant tunneling phenomena. Using non-equilibrium Green's function (NEGF) method, we show that the structure can exhibit a bias-dependent oscillatory IEC that can switch the free magnet to have either a parallel or an antiparallel configuration with respect to the fixed magnet, depending on the sign of the IEC. Such bi-directional switching can be achieved with the same voltage polarity but different magnitudes. With proper choice of the spacer material, the current in the structure can be significantly reduced. Due to the conservative nature of the exerted torque by the IEC, the switching threshold of the proposed mechanism is decoupled from the switching speed, while the conventional spin-torque devices exhibit a trade-off due to the non-conservative nature of the exerted torque.

cond-mat.mes-hall↗

arXiv2019

Combining Learned Representations for Combinatorial Optimization

We propose a new approach to combine Restricted Boltzmann Machines (RBMs) that can be used to solve combinatorial optimization problems. This allows synthesis of larger models from smaller RBMs that have been pretrained, thus effectively bypassing the problem of learning in large RBMs, and creating a system able to model a large, complex multi-modal space. We validate this approach by using learned representations to create ``invertible boolean logic'', where we can use Markov chain Monte Carlo (MCMC) approaches to find the solution to large scale boolean satisfiability problems and show viability towards other combinatorial optimization problems. Using this method, we are able to solve 64 bit addition based problems, as well as factorize 16 bit numbers. We find that these combined representations can provide a more accurate result for the same sample size as compared to a fully trained model.

cs.LG↗

arXiv2018

Spin-Orbit Torque and Nernst Effect in Bi-Sb/Co Heterostructures

Harmonic measurements of the longitudinal and transverse voltages in Bi-Sb/Co bilayers are presented. A large second harmonic voltage signal due to the ordinary Nernst effect is observed. In experiments where a magnetic field is rotated in the film plane, the ordinary Nernst effect shows the same angular dependence in the transverse voltage as the damping-like spin-orbit torque and in the longitudinal voltage as the unidirectional spin-Hall magneto-resistance respectively. Therefore, the ordinary Nernst effect can be a spurious signal in spin-orbit torque measurements, leading to an overestimation of the spin-Hall angle in topological insulators or semimetals.

cond-mat.mes-hall↗

arXiv2018

Breaking the fundamental energy dissipation limit in ferroelectric-dielectric capacitors

Half of the energy is always lost when charging a capacitor. Even in the limit of vanishing resistance, half of the charging energy is still lost--to radiation instead of heat. While this fraction can technically be reduced by charging adiabatically, it otherwise places a fundamental limit on the charging efficiency of a capacitor. Here we show that this 1/2 limit can be broken by coupling a ferroelectric to the capacitor dielectric. Maxwell's equations are solved for the coupled system to analyze energy flow from the perspective of Poynting's theorem and show that (1) total energy dissipation is reduced below the fundamental limit during charging and discharging; (2) energy is saved by "recycling" the energy already stored in the ferroelectric phase transition; and (3) this phase transition energy is directly transferred between the ferroelectric and dielectric during charging and discharging. These results demystify recent works on low energy negative capacitance devices as well as lay the foundation for improving fundamental energy efficiency in all devices that rely on energy storage in electric fields.

physics.app-ph↗

arXiv2018

Voltage driven, local, and efficient excitation of nitrogen-vacancy centers in diamond

Magnetic sensing technology has found widespread application in industries as diverse as transportation, medicine, and resource exploration. Such use cases often require highly sensitive instruments to measure the extremely small magnetic fields involved, relying on difficult to integrate Superconducting Quantum Interference Device (SQUID) and Spin-Exchange Relaxation Free (SERF) magnetometers. A potential alternative, nitrogen vacancy (NV) centers in diamond, has shown great potential as a high sensitivity and high resolution magnetic sensor capable of operating in an unshielded, room-temperature environment. Transitioning NV center based sensors into practical devices, however, is impeded by the need for high power RF excitation to manipulate them. Here we report an advance that combines two different physical phenomena to enable a highly efficient excitation of the NV centers: magnetoelastic drive of ferromagnetic resonance (FMR) and NV-magnon coupling. Our work demonstrates a new pathway to combine acoustics and magnonics that enables highly energy efficient and local excitation of NV centers without the need for any external RF excitation, and thus could lead to completely integrated, on-chip, atomic sensors.

physics.app-ph↗

arXiv2017

Ferroelectric Negative Capacitance Domain Dynamics

Transient negative capacitance effects in epitaxial ferroelectric Pb(Zr$_{0.2}$Ti$_{0.8}$)O$_3$ capacitors are investigated with a focus on the dynamical switching behavior governed by domain nucleation and growth. Voltage pulses are applied to a series connection of the ferroelectric capacitor and a resistor to directly measure the ferroelectric negative capacitance during switching. A time-dependent Ginzburg-Landau approach is used to investigate the underlying domain dynamics. The transient negative capacitance is shown to originate from reverse domain nucleation and unrestricted domain growth. However, with the onset of domain coalescence, the capacitance becomes positive again. The persistence of the negative capacitance state is therefore limited by the speed of domain wall motion. By changing the applied electric field, capacitor area or external resistance, this domain wall velocity can be varied predictably over several orders of magnitude. Additionally, detailed insights into the intrinsic material properties of the ferroelectric are obtainable through these measurements. A new method for reliable extraction of the average negative capacitance of the ferroelectric is presented. Furthermore, a simple analytical model is developed, which accurately describes the negative capacitance transient time as a function of the material properties and the experimental boundary conditions.

cond-mat.mes-hall↗

arXiv2017

Differential voltage amplification from ferroelectric negative capacitance

It is well known that one needs an external source of energy to provide voltage amplification. Because of this, conventional circuit elements such as resistors, inductors or capacitors cannot provide amplification all by themselves. Here, we demonstrate that a ferroelectric can cause a differential amplification without needing such an external energy source. As the ferroelectric switches from one polarization state to the other, a transfer of energy takes place from the ferroelectric to the dielectric, determined by the ratio of their capacitances, which, in turn, leads to the differential amplification. {This amplification is very different in nature from conventional inductor-capacitor based circuits where an oscillatory amplification can be observed. The demonstration of differential voltage amplification from completely passive capacitor elements only, has fundamental ramifications for next generation electronics.

cond-mat.mes-hall↗

arXiv2017

Single shot ultrafast all optical magnetization switching of ferromagnetic Co/Pt multilayers

In a number of recent experiments, it has been shown that femtosecond laser pulses can control magnetization on picosecond timescales, which is at least an order of magnitude faster compared to conventional magnetization dynamics. Among these demonstrations, one material system (GdFeCo ferromagnetic films) is particularly interesting, as deterministic toggle-switching of the magnetic order has been achieved without the need of any symmetry breaking magnetic field. This phenomenon is often referred to as all optical switching (AOS). However, so far, GdFeCo remains the only material system where such deterministic switching has been observed. When extended to ferromagnetic systems, which are of greater interest in many technological applications, only a partial effect can be achieved, which in turn requires repeated laser pulses for full switching. However, such repeated pulsing is not only energy hungry, it also negates the speed advantage of AOS. Motivated by this problem, we have developed a general method for single-shot, picosecond timescale, complete all optical switching of ferromagnetic materials. We demonstrate that in exchange-coupled layers of Co/Pt and GdFeCo, single shot, switching of the ferromagnetic Co/Pt layer is achieved within 7 picoseconds after irradiation by a femtosecond laser pulse. We believe that this approach will greatly expand the range of materials and applications for ultrafast magnetic switching.

cond-mat.mes-hall↗

arXiv2017

Non Volatile MoS$_{2}$ Field Effect Transistors Directly Gated By Single Crystalline Epitaxial Ferroelectric

We demonstrate non-volatile, n-type, back-gated, MoS$_{2}$ transistors, placed directly on an epitaxial grown, single crystalline, PbZr$_{0.2}$Ti$_{0.8}$O$_{3}$ (PZT) ferroelectric. The transistors show decent ON current (19 $μA/μ$m), high on-off ratio (10$^{7}$), and a subthreshold swing of (SS ~ 92 mV/dec) with a 100 nm thick PZT layer as the back gate oxide. Importantly, the ferroelectric polarization can directly control the channel charge, showing a clear anti-clockwise hysteresis. We have selfconsistently confirmed the switching of the ferroelectric and corresponding change in channel current from a direct time-dependent measurement. Our results demonstrate that it is possible to obtain transistor operation directly on polar surfaces and therefore it should be possible to integrate 2D electronics with single crystalline functional oxides.

physics.app-ph↗

arXiv2017

Electrically Induced, Non-Volatile, Metal Insulator Transition in a Ferroelectric Gated MoS$_2$ Transistor

We demonstrate an electrically induced, non-volatile, metal-insulator phase transition in a MoS$_2$ transistor. A single crystalline, epitaxially grown, PbZr$_{0.2}$Ti$_{0.8}$O$_3$ (PZT) was placed in the gate of a field effect transistor made of thin film MoS$_2$. When a gate voltage is applied to this ferroelectric gated transistor, a clear transition from insulator to metal and vice versa is observed. Importantly, when the gate voltage is turned off, the remnant polarization in the ferroelectric can keep the MoS$_2$ in its original phase, thereby providing a non-volatile state. Thus a metallic or insulating phase can be written, erased or retained simply by applying a gate voltage to the transistor.

physics.app-ph↗

arXiv2017

Ultrafast Magnetic Switching of GdFeCo with Electronic Heat Currents

We report the magnetic response of Au/GdFeCo bilayers to optical irradiation of the Au surface. For bilayers with Au thickness greater than 50 nm, the great majority of energy is absorbed by the Au electrons, creating an initial temperature differential of thousands of Kelvin between the Au and GdFeCo layers. The resulting electronic heat currents between the Au and GdFeCo layers last for several picoseconds with energy flux in excess of 2 TW m-2, and provide sufficient heating to the GdFeCo electrons to induce deterministic reversal of the magnetic moment.

cond-mat.mtrl-sci↗