Memory Matrix Research Articles

(Invited) Resistive Memories (RRAM) Variability: Challenges and Solutions

IntroductionResistive random access memory (RRAM) technologies have experienced an increasing interest as next generation non-volatile memory devices. Thanks to their promising characteristics, RRAM are believed to become a good choice for Storage Class Memory (SCM) [1-3]. Indeed, nowadays, due to popularity of cloud storage and mobile devices, it is urgent to develop lower energy consuming devices with high speed, high density and low cost. On the other hand, as they can be integrated in the BEOL of CMOS logic, they are strongly envisaged for future embedded memory generation [4-5]. RRAM working principle is based on the reversible formation and dissolution of a conductive filament in a resistive layer, resulting from oxygen vacancies generation and diffusion (OxRAM) or dissolution of an active electrode (CBRAM). As filament formation and dissolution is governed by stochastic mechanisms, RRAM features have to be described by statistic laws, and variability has to be carefully considered for RRAM reliability evaluation before industrialization could be envisaged. In this work, we address Resistive RAM (RRAM) variability. To this aim, we investigate various classes of RRAM (Oxide RAM and Conductive Bridging RAM), integrated on kb 1T1R arrays. Impact of variability is evaluated and discussed for various RRAM features: window margin, switching speed, consumption, retention and endurance. Solutions are proposed in order to improve overall RRAM performances. Results and discussionFirst, we present a general overview of RRAM characteristics, and investigate how performances are correlated. In particular, we show that window margin, endurance and retention are linked. Using electrical characterization combined with atomistic calculations, we demonstrate that the tradeoff is associated to the ion migration energy barrier of the species constituting the conductive filament, and we propose guidelines to optimize RRAM material stack depending on the targeted application. Then we deeply investigate variability, and we propose solutions to improve RRAM performances. Window margin variability – We describe RRAM resistance distributions for high and low resistive states and quantify how resistance dispersion affects RRAM window margin, and impacts maximum memory array size. Optimized programming schemes are proposed to reduce resistance dispersion and enlarge window margin. Consumption and speed – Variability has an impact on RRAM consumption and speed. Indeed, as RRAM switching time is dispersed, long programming pulses may be required to program all the array of a large memory matrix. In order to improve RRAM speed and consumption, Ramp Voltage Stress (RVS) is proposed to program the memory instead of Constant Voltage Stress (CVS) method. Voltage increase in RVS gradually increases programming strength, reducing impact of intrinsic RRAM switching time dispersion. Retention variability – Filament dissolution in the resistive layer is follows a stochastic behavior, leading to dispersion of memory retention time. In order to improve RRAM retention, we propose to reduce the energy provided to the system during programming modes, reducing ion migration rate and improving RRAM stability during bake. Endurance variability – We experimentally demonstrate that RRAM maximum number of cycles follows a log normal distribution law. Maximum endurance can be described using physics of dielectric breakdown. We investigate how SET and RESET programming conditions can affect RRAM endurance statistics. Then we propose optimized programming schemes to reduce electrical stress applied to the dielectrics, improving endurance, but maintaining sufficient RRAM window margin.

Micromechanics of stress transfer in shape memory alloy reinforced three-phase composites

Due to the weak interface in shape memory alloy wire–reinforced composites, the influence of interphase on the mechanical properties and stress distribution of hybrid composites is of considerable importance. In this article, a three-cylinder axisymmetric model using a pull-out test is developed to predict stress transfer and interfacial behavior between shape memory alloy wire, interphase, and matrix. In this article, only superelasticity behavior of the shape memory alloy wire is considered. Based on the stress function method and the principle of minimum complementary energy, stress distribution is derived for three different cases in terms of loading and boundary conditions (thermal loading model, intact model, and partially debonded model). Inhomogeneous interphase and different radial and hoop stress components in each phase are considered to achieve deeper physical understanding. Finite element analysis also performed to simulate stress transfer from the wire to the matrix through the interphase. To evaluate the accuracy of this model, the results of the work are compared with the results of the two-cylinder model proposed by Wang et al. and finite element results.

3D memory matrix based on a composite memristor-diode crossbar for a neuromorphic processor

An electrical circuit, topology, and a fabrication technology have been developed for an ultra large multilayer memory matrix having a non-volatile memory, low energy consumption and large-scale integration of elements based on the cells incorporating complementary bipolar memristors and a Zener diode. Unlike memory matrices applicable in information technology, the proposed matrix can not only store information but also allow for voltage weighing of input signals that pass through memristors and their summing. What is more, low signal degradation when summing is achieved by selecting a special circuit for sending input impulses.Technologically, large-scale integration of elements is implemented by sequential vertical build-up of planar bilayer memory matrices on a crystal into a 3D structure consisting of identical horizontal mutually mirrored composite crossbars that include an active memristor layer and a semiconductor Zener diode layer. Both layers as well as the conductive paths can be fabricated by a vacuum technology, using a magnetron process module. A SPICE model, which allows for optimization of parameters and operating modes of elements of the cell and of the matrix in general, has been constructed for a composite memristor crossbar with a Zener diode.

Experimental Demonstration of Operational Z2-FET Memory Matrix

In this letter, a functional Z2-FET DRAM memory matrix is experimentally demonstrated for the first time. Word-level operation with simultaneous reading and programming accesses is successfully proved. Disturbance is also explored, and the results demonstrate bitline disturbance immunity. Furthermore, the fabricated memory matrix, which includes only one selector per word-line, facilitates the scalability of the memory for increasing array dimensions.

Computing the sparse matrix vector product using block-based kernels without zero padding on processors with AVX-512 instructions.

The sparse matrix-vector product (SpMV) is a fundamental operation in many scientific applications from various fields. The High Performance Computing (HPC) community has therefore continuously invested a lot of effort to provide an efficient SpMV kernel on modern CPU architectures. Although it has been shown that block-based kernels help to achieve high performance, they are difficult to use in practice because of the zero padding they require. In the current paper, we propose new kernels using the AVX-512 instruction set, which makes it possible to use a blocking scheme without any zero padding in the matrix memory storage. We describe mask-based sparse matrix formats and their corresponding SpMV kernels highly optimized in assembly language. Considering that the optimal blocking size depends on the matrix, we also provide a method to predict the best kernel to be used utilizing a simple interpolation of results from previous executions. We compare the performance of our approach to that of the Intel MKL CSR kernel and the CSR5 open-source package on a set of standard benchmark matrices. We show that we can achieve significant improvements in many cases, both for sequential and for parallel executions. Finally, we provide the corresponding code in an open source library, called SPC5.

EEG Reflection of Clinical Effects Dynamics during Botulinum Toxin Therapy of Movement Disorders under Long-Term Consciousness Disorders

When studying the dynamics of brain electric activity (BEA) in the course of botulinum toxin (Incobotulinumtoxin A) injections into all hypertonic muscles of 16 patients suffering from post-coma longterm consciousness disorders, BEA changes were observed to start within the first minutes after the first intramuscular injection of botulinum toxin, while not being associated with any pain stimulus response. Despite the diversity of brain lesions in these patients, the dynamics of BEA reorganisation during botulinum toxin therapy showed the presence of pathological sensorimotor integration involving the whole brain in case of long-term consciousness disorders. In our opinion, it is expressed in a pathological state developing in the brain, which affects the polyfunctional capabilities of neurons. This state is unbalanced due to destabilizing effect at the beginning of the reduction by pathological hyperafferentation from muscles and, in principle, can be disrupted at a complete reduction. It is, however, maintained by memory matrix, which manifests itself in the shape of oscillation process during its disruption when the newly build-up BEA is occasionally displaced by the previous one. During the reduction of sensory hyperafferentation, BEA reorganisation initially involved the motor cortex (in 15 out of 16 patients) and then spread to other brain areas, including those responsible for the higher mental functions, vision, and hearing. The first transiently decreased muscle tone (in all patients) and improvement of neurological signs of awareness (in some patients) were observed immediately after injecting all hypertonic muscles with widespread BEA changes. The appearance of high-frequency EEG activity in the frontal areas was considered as a universal marker of an improved functional state of the brain during the disruption of pathological sensorimotor integration even before clinical signs, suggesting neuronal readiness to maintain various activities. In view of data obtained, the first-stage clinical measures in patients with disorders of consciousness should focus on disrupting pathological sensorimotor integration, opening a way for restoring neuronal functional activity, provided that all “pathological afferents” are reduced.

Damage Detection on Steel Plane Trusses via Harmony Search Algorithm

A damage detection study is presented on steel plane truss structures via harmony search algorithm which is one of the basic metaheuristic methods. This algorithm method mimics better musical harmony procedures. Optimum solutions are carried out by a harmony memory matrix consisting of a predetermined number of solution vectors. Scenario damages are considered on truss models. The locations of the damages are tried to determine by using the dynamic parameters. To obtain damage locations, a program based on the finite element method is coded in MATLAB programming. The dynamic parameters of damaged models are obtained by SAP2000 software. By using MATLAB programming, the scenario damage locations considered in SAP2000 programs are tried to determine by approaching the dynamic values of the damaged model as a result of a series of iterations. The results are presented by tables and figures. The results show that scenario damages on truss models are successfully detected by using dynamic parameters in MATLAB programming.

Condition Assessment Approach of Hydraulic Brake for Large Crane Based on State Estimation Algorithm

Hydraulic brake was widely used for mechanical brake of port crane. The run-state of the brake affects the safety of the crane because of the sudden accident. Because of the complex structure and unsuited to use in real time of traditional assessment model, the condition assessment approach of hydraulic brake was constructed based on state estimation algorithm in this paper. The oil temperature, dynamic friction coefficient, spring stiffness coefficient, brake shoe clearance and contact area were chosen as the state components of the memory matrix based on the analysis of the structure and failure reasons of the brake. Considering the correlation between the state components, the Mahalanobis distance was chosen as the nonlinear operator of algorithm, and the uncertainty factors and random disturbances in state assessment were eliminated by the sliding window residual statistics method. The dynamics simulation of hydraulic brake was constructed for confirming the validity of the approach in this paper. The result shows that it can be accurately judged by the method if the brake is in abnormal state.

Parameter identification for structural dynamics based on interval analysis algorithm

A parameter identification method using interval analysis algorithm for structural dynamics is presented in this paper. The proposed uncertain identification method is investigated by using central difference method and ARMA system. With the help of the fixed memory least square method and matrix inverse lemma, a set-membership identification technology is applied to obtain the best estimation of the identified parameters in a tight and accurate region. To overcome the lack of insufficient statistical description of the uncertain parameters, this paper treats uncertainties as non-probabilistic intervals. As long as we know the bounds of uncertainties, this algorithm can obtain not only the center estimations of parameters, but also the bounds of errors. To improve the efficiency of the proposed method, a time-saving algorithm is presented by recursive formula. At last, to verify the accuracy of the proposed method, two numerical examples are applied and evaluated by three identification criteria respectively.

ISOLATION, SLEEP, COGNITION AND NEUROPHYSIOLOGICAL RESPONSES – AN INVESTIOGATION IN THE HUMAN EXPLORATION RESEARCH ANALOG (HERA)

ISOLATION, SLEEP, COGNITION AND NEUROPHYSIOLOGICAL RESPONSES – AN INVESTIOGATION IN THE HUMAN EXPLORATION RESEARCH ANALOG (HERA)

Singular value Decomposition applied to Associative Memory of Hopfield Neural Network

Content-addressable memories are useful for storage and data retrieval from arrays of sensors. The Hopfield neural network is a model of associative content-addressable memory with a simple flexible structure. Design of this artificial neural network is capable of memorizing large quantity of data and recalling the same from available information. In this work, we have shown that with corrupted input dataset, the correct set of data can be retrieved from approximated or compressed associative memory matrix. The idea has been explained through an experiment using binary datasets. The proposed methodology will increase the storage capacity of associative memory. In the experiment we have shown that it will minimize the number of arithmetic operations involved in recovery process of the Hopfield model but will also ensure correctness of information retrieval from the synaptic matrix from a corrupted dataset as input.

Impact of Line Resistance Combined with Device Variability on Resistive RAM Memories

In this paper, the performance and reliability of oxide-based Resistive RAM (ReRAM) memory is investigated in a 28nm FDSOI technology versus interconnects resistivity combined with device variability. Indeed, common problems with ReRAM are related to high variability in operating conditions and low yield. At a cell level ReRAMs suffer from variability. At an array level, ReRAMs suffer from different voltage drops seen across the cells due to line resistances. Although research has taken steps to resolve these issues, variability combined with resistive paths remain an important characteristic for ReRAMs. In this context, a deeper understanding of the impact of these characteristics on ReRAM performances is needed to propose variability tolerant designs to ensure the robustness of the technology. The presented study addresses the memory cell, the memory word up to the memory matrix.

Complementary memristive diode cells for the memory matrix of a neuromorphic processor

This research covers the sphere of neuromorphic electronics, with a new memory cell developed. Its circuit design and topology is based on a complementary memristor, which is switched through a Zener diode. The above-mentioned cells can be implemented on-chip in the form of a parallel matrix or on-bit access type to a storage device in a neuromorphic processor. The suggested topology of a memristor-diode cell enables a high degree of integration in the case these cells get combined in a very large matrix. In this regard, all the square matrices are filled with memristor cells of a nanometric size. The rest of large elements are moved to the matrix periphery to vacate the specific square area of memristor technological placement. The cell under discussion implicates the parallel connection of electrical loads in the matrix related to simple memristor circuit crossbars since the total resistance of each cell always remains high, and the current resistance through complementary memristors is always minimal.

Exploring the sources and mechanisms of cognitive errors in medical diagnosis with associative memory models.

One of the central challenges of third millennium medicine is the abatement of medical errors. Among the most frequent and hardiest causes of misdiagnosis are cognitive errors produced by faulty medical reasoning. These errors have been analyzed from the perspectives of cognitive psychology and empirical medical studies. We introduce a neurocognitive model of medical diagnosis to address this issue. We construct a connectionist model based on the associative nature of human memory to explore the non-analytical, pattern-recognition mode of diagnosis. A context-dependent matrix memory associates signs and symptoms with their corresponding diseases. The weights of these associations depend on the frequencies of occurrence of each disease and on the different combinations of signs and symptoms of each presentation of that disease. The system receives signs and symptoms and by a second input, the degree of diagnostic uncertainty. Its output is a probabilistic map on the set of possible diseases. The model reproduces different kinds of well-known cognitive errors in diagnosis. Errors in the model come from two sources. One, dependent on the knowledge stored in memory, varies with the accumulated experience of the physician and explains age-dependent errors and effects such as epidemiological masking. The other is independent of experience and explains contextual effects such as anchoring. Our results strongly suggest that cognitive biases are inevitable consequences of associative storage and recall. We found that this model provides valuable insight into the mechanisms of cognitive error and we hope it will prove useful in medical education.

Quantitative Brain MRI in Congenital Adrenal Hyperplasia: In Vivo Assessment of the Cognitive and Structural Impact of Steroid Hormones.

ContextBrain white matter hyperintensities are seen on routine clinical imaging in 46% of adults with congenital adrenal hyperplasia (CAH). The extent and functional relevance of these abnormalities have not been studied with quantitative magnetic resonance imaging (MRI) analysis.ObjectiveTo examine white matter microstructure, neural volumes, and central nervous system (CNS) metabolites in CAH due to 21-hydroxylase deficiency (21OHD) and to determine whether identified abnormalities are associated with cognition, glucocorticoid, and androgen exposure.Design, Setting, and ParticipantsA cross-sectional study at a tertiary hospital including 19 women (18 to 50 years) with 21OHD and 19 age-matched healthy women.Main Outcome MeasureRecruits underwent cognitive assessment and brain imaging, including diffusion weighted imaging of white matter, T1-weighted volumetry, and magnetic resonance spectroscopy for neural metabolites. We evaluated white matter microstructure by using tract-based spatial statistics. We compared cognitive scores, neural volumes, and metabolites between groups and relationships between glucocorticoid exposure, MRI, and neurologic outcomes.ResultsPatients with 21OHD had widespread reductions in white matter structural integrity, reduced volumes of right hippocampus, bilateral thalami, cerebellum, and brainstem, and reduced mesial temporal lobe total choline content. Working memory, processing speed, and digit span and matrix reasoning scores were reduced in patients with 21OHD, despite similar education and intelligence to controls. Patients with 21OHD exposed to higher glucocorticoid doses had greater abnormalities in white matter microstructure and cognitive performance.ConclusionWe demonstrate that 21OHD and current glucocorticoid replacement regimens have a profound impact on brain morphology and function. If reversible, these CNS markers are a potential target for treatment.

Learning better discourse representation for implicit discourse relation recognition via attention networks

Humans comprehend the meanings and relations of discourses heavily relying on their semantic memory that encodes general knowledge about concepts and facts. Inspired by this, we propose a neural recognizer for implicit discourse relation analysis, which builds upon a semantic memory that stores knowledge in a distributed fashion. We refer to this recognizer as SeMDER. Starting from word embeddings of discourse arguments, SeMDER employs a shallow encoder to generate a distributed surface representation for a discourse. A semantic encoder with attention to the semantic memory matrix is further established over surface representations. It is able to retrieve a deep semantic meaning representation for the discourse from the memory. Using the surface and semantic representations as input, SeMDER finally predicts implicit discourse relations via a neural recognizer. Experiments on the benchmark data set show that SeMDER benefits from the semantic memory and achieves substantial improvements of 2.56\% on average over current state-of-the-art baselines in terms of F1-score.

Memory matrix theory of the dc resistivity of a disordered antiferromagnetic metal with an effective composite operator

We perform the calculation of the dc resistivity as a function of temperature of the “strange-metal” state that emerges in the vicinity of a spin-density-wave phase transition in the presence of weak disorder. This scenario is relevant to the phenomenology of many important correlated materials, such as, e.g., the pnictides, the heavy-fermion compounds and the cuprates. To accomplish this task, we implement the memory-matrix approach that allows the calculation of the transport coefficients of the model beyond the quasiparticle paradigm. Our computation is also inspired by the ϵ=3−d expansion in a hot-spot model embedded in d-space dimensions recently put forth by Sur and Lee (2015), in which they find a new low-energy non-Fermi liquid fixed point that is perturbatively accessible near three dimensions. As a consequence, we are able to establish here the temperature and doping dependence of the electrical resistivity at intermediate temperatures of a two-dimensional disordered antiferromagnetic metallic model with a composite operator that couples the order-parameter fluctuations to the entire Fermi surface. We argue that our present theory provides a good basis in order to unify the experimental transport data, e.g., in the cuprates and the pnictide superconductors, within a wide range of doping regimes.

Calculation of the magnetotransport for a spin-density-wave quantum critical theory in the presence of weak disorder

We compute the Hall angle and the magnetoresistance of the spin-fermion model, which is a successful phenomenological theory to describe the physics of the cuprates and iron-based superconductors within a wide range of doping regimes. We investigate both the role of the spin-fermion interaction that couples the large-momentum antiferromagnetic fluctuations to the so-called “hot-spots” at the Fermi surface and also of an effective higher-order composite operator in the theory. The latter operator provides a scattering mechanism such that the momentum transfer for the fermions close to the Fermi surface can be small. We also include weak disorder that couples to both the bosonic order-parameter field and the fermionic degrees of freedom. Since the quasiparticle excitations were shown in recent works to be destroyed at the “hot-spots” in the low-energy limit of the model, we employ the Mori-Zwanzig memory matrix approach that permits the evaluation of all transport coefficients without assuming well-defined Landau quasiparticles in the system. We then apply this transport theory to discuss universal metallic-state properties as a function of temperature and magnetic field of the cuprates from the perspective of their fermiology, which turn out to be in qualitative agreement with key experiments in those materials.

Do children with grapheme‐colour synaesthesia show cognitive benefits?

Grapheme-colour synaesthesia is characterized by conscious and consistent associations between letters and colours, or between numbers and colours (e.g., synaesthetes might see A as red, 7 as green). Our study explored the development of this condition in a group of randomly sampled child synaesthetes. Two previous studies (Simner & Bain, 2013, Frontiers in Human Neuroscience, 7, 603; Simner, Harrold, Creed, Monro, & Foulkes, 2009, Brain, 132, 57) had screened over 600 primary school children to find the first randomly sampled cohort of child synaesthetes. In this study, we evaluate this cohort to ask whether their synaesthesia is associated with a particular cognitive profile of strengths and/or weaknesses. We tested our child synaesthetes at age 10-11years in a series of cognitive tests, in comparison with matched controls and baseline norms. One previous study (Green & Goswami, 2008, Cognition, 106, 463) had suggested that child synaesthetes might perform differently to non-synaesthetes in such tasks, although those participants may have been a special type of population independent of their synaesthesia. In our own study of randomly sampled child synaesthetes, we found no significant advantages or disadvantages in a receptive vocabulary test and a memory matrix task. However, we found that synaesthetes demonstrated above-average performance in a processing-speed task and a near-significant advantage in a letter-span task (i.e., memory/recall task of letters). Our findings point to advantages for synaesthetes that go beyond those expected from enhanced coding accounts and we present the first picture of the broader cognitive profile of a randomly sampled population of child synaesthetes.

Holographic incoherent transport in Einstein-Maxwell-dilaton gravity

Recent progress in the holographic approach makes it more transparent that each conductivity can be decomposed into the coherent contribution due to momentum relaxation and the incoherent contribution due to intrinsic current relaxation. In this paper we investigate this decomposition in the framework of Einstein-Maxwell-dilaton theory. We derive the perturbation equations, which are decoupled for a large class of background solutions, and then obtain the analytic results of conductivity with the slow momentum relaxation in low frequency approximation, which is consistent with the known results from memory matrix techniques.