High-performance Memory Research Articles

FROM Si NANOWIRES TO Ge NANOCRYSTALS FOR VIS-NIR-SWIR SENSORS AND NON-VOLATILE MEMORIES: A REVIEW

"Nanocrystalline Si and Ge are ofhigh interestfor integrated Si photonics related to light emission, opticul sensors, photodetectors, solar energy harvesting and conversion devices, and also forfloating gate non-volatile memories (NVMs). In this review, we have focused on nanocrystalline porous Si (nc-PS) with extension to Si nanodots, and Ge nanocrystals (NCs)Zquantum dots (QDs)/nanoparticles (NPs) embedded in oxides (SiCh, TiCE, HfCh, AI2O3). The great asset ofnc-PS is its intense photoluminescence in VIS at room temperature (RT), while Ge NCs/NPs embedded in oxides show high photosensitivity in VISNIR-SWIR in the spectral photocurrent up to 1325 nm at RT. Ge NCs/NPs/QDsfloating gate NVMs present high memory performance, the retention characteristics corresponding to the state of the art for NCs floating gate NVMs. We prove the relevance of controlling the preparation parametersfor obtainingfilms with targetedphotoluminescence, photosensitivity and charge storage properties for applications, e.g. VIS-NIR-SWIR optical sensors and photodetectors, and electronic and photoelectric NVMs. We evidence the correlation of preparation conditions, morphology, composition and crystalline structure with optical, electrical, photoelectrical and charge storage properties and also evidence the contribution ofquantum confinement effect, localized States and trapping centers."

N-Type Digital Memory Characteristics of Diketopyrrolopyrrole-Based Narrow Bandgap Polymers

New π-conjugated polymers based on dithienyl-diketopyrrolopyrrole (dtDPP) were synthesized by direct arylation polymerization, that is, poly(dtDPP), poly(dtDPP-FL), and poly(dtDPP-TPA). They revealed extremely low bandgaps, which could be achieved by controlling the donor–acceptor powers among the backbone building blocks. They were all thermally stable up to 250 °C and, furthermore, solution-processible in conventional ways, producing good-quality thin films. Poly(dtDPP) favorably self-assembled, forming an edge-on layer structure in nanoscale films, whereas the others were preferentially oriented within the film plane without any regular packing structures. All the dt-DPP-based polymers as active layers in devices demonstrated high-performance n-type digital memories (low switching voltage, high ON/OFF current ratio, and high stability) because of the narrow bandgaps. Moreover, the memory modes (nonvolatile and volatile) could be tuned by adjusting the donor–acceptor powers in the backbone. Overall, the dtDPP-based polymers are suitable for the low-cost mass production of n-type nonvolatile and volatile high performing memory devices. In addition, this study provides a guideline on how to achieve high-performance n-type digital memories by balancing the donor and acceptor powers among the chemical building blocks.

KLIFF: A framework to develop physics-based and machine learning interatomic potentials

Interatomic potentials (IPs) are reduced-order models for calculating the potential energy of a system of atoms given their positions in space and species. IPs treat atoms as classical particles without explicitly modeling electrons and thus are computationally far less expensive than first-principles methods, enabling molecular simulations of significantly larger systems over longer times. Developing an IP is a complex iterative process involving multiple steps: assembling a training set, designing a functional form, optimizing the function parameters, testing model quality, and deployment to molecular simulation packages. This paper introduces the KIM-based learning-integrated fitting framework (KLIFF), a package that facilitates the entire IP development process. KLIFF supports both physics-based and machine learning IPs. It adopts a modular approach whereby various components in the fitting process, such as atomic environment descriptors, functional forms, loss functions, optimizers, quality analyzers, and so on, work seamlessly with each other. This provides a flexible framework for the rapid design of new IP forms. Trained IPs are compatible with the Knowledgebase of Interatomic Models (KIM) application programming interface (API) and can be readily used in major materials simulation packages compatible with KIM, including ASE, DL_POLY, GULP, LAMMPS, and QC. KLIFF is written in Python with computationally intensive components implemented in C++. It is parallelized over data and supports both shared-memory multicore desktop machines and high-performance distributed memory computing clusters. We demonstrate the use of KLIFF by fitting a physics-based Stillinger–Weber potential and a machine learning neural network potential for silicon. The KLIFF package, together with its documentation, is publicly available at: https://github.com/openkim/kliff. Program summaryProgram Title: KIM-based Learning-integrated Fitting Framework (KLIFF)CPC Library link to program files:https://doi.org/10.17632/fk77gs5b2d.1Licensing provisions: LGPL 2.1Programming language: Python, C++Supplementary material: Example scripts for the demonstrations, which are compatible with KLIFF v0.3.0.Nature of problem: Development of a model called an interatomic potential (IP) representing the potential energy of a system of atoms based on their positions in space and species. This is a complex iterative process involving multiple steps: assembling a training set, designing a functional form, optimizing the function parameters, testing IP quality, and deployment of the fitted IP to molecular simulation packages.Solution method: The fitting process is formulated as an optimization problem where a loss function characterizing the IP error over a training set is minimized to obtain the optimal fitting parameters. KLIFF is designed in a modular fashion providing the user with flexible access to different functional forms, loss functions, optimization algorithms, and analyzers for testing the quality of the fitted IP. KLIFF is built on the Knowledgebase of Interatomic Models (KIM) API standard, which enables immediate deployment of fitted IPs to major materials simulation packages that are compatible with KIM.

Deactivation of the attention-shifting ventromedial prefrontal cortex during the encoding and hold phases predicts working memory performance.

Recent neuroimaging studies have suggested that the deactivation of the ventromedial prefrontal cortex (VMPFC) works with the attention shifting area to facilitate the encoding of behaviorally relevant inputs. These findings have led to the notion that the deactivation of VMPFC substantially contributes to the cognitive control of emotions. Although VMPFC deactivation during working memory tasks is established, whether it contributes to performance in emotionally distracted working memory tasks remains unclear. This study aimed to investigate whether the magnitude of VMPFC deactivation predicts better performance in emotionally distracted working memory tasks. Twenty-nine female participants performed delayed-response working memory tasks with emotional distracters presented during the hold phase of working memory while undergoing functional MRI. A GLM and a paired t-test were used to observe brain responses to emotional distracters. The correlation between brain response and working memory performance was also computed to investigate brain areas that predict working memory performance in emotionally distracted tasks. Three trends in brain activity were strongly correlated with high working memory performance: (1) increased activity in cognitive control areas (dorsolateral prefrontal cortex), (2) lower activity in emotional reactivity areas (fusiform gyrus), and (3) deactivation of the attention shifting area, mainly VMPFC. In addition, all three trends correlated with high working memory performance during the hold phase of working memory, whereas only (2) and (3) correlated with high working memory performance during the encoding phase. These results provide further evidence of the functional importance of VMPFC and demonstrate that VMPFC deactivation is particularly important during the encoding and hold phases of working memory.

Degradable photo-crosslinked starch-based films with excellent shape memory property

With the increasingly serious plastic pollution, people's demand for the multi-functional biodegradable plastics is becoming more and more urgent. Inspired by the crosslinked shape memory polymers, the crosslinked starch films were synthesized by inducing the decomposition of benzophenone into free radical and depriving hydrogen on starch macromolecules under UV irradiation, in order to gain a high shape memory performance. The results showed that a three-dimensional crosslinking network between starch macromolecule chains was formed. Compared with the uncrosslinked starch films, the photo-crosslinked films not only had higher mechanical property (tensile strength increased by 154%), but also had better water resistance (water contact angle from 60° to 87°) due to the reduction of free hydroxyl groups. In addition, the stable covalent bonds serving as netpoints endow photo-crosslinked films with great improvement in shape memory property, with nearly 180° bending recovery. More importantly, the maximum shape memory fixity ratio (Rf) and shape memory recovery ratio (Rr) under stretch deformation were 96.5% and 99.8%, respectively. And the Rf and Rr could reach 94.6% and 79.8% even at higher strain. In all, the excellent shape memory performance and good degradability crosslinked starch films, which have great potential application in disposable heat-shrinkable packaging materials.

Resistive Switching in HfO2-x/La0.67Sr0.33MnO3 Heterostructures: An Intriguing Case of Low H-Field Susceptibility of an E-Field Controlled Active Interface.

High-performance nonvolatile resistive random access memories (ReRAMs) and their small stimuli control are of immense interest for high-speed computation and big-data processing in the emerging Internet of Things (IoT) arena. Here, we examine the resistive switching (RS) behavior in growth-controlled HfO2/La0.67Sr0.33MnO3 (LSMO) heterostructures and their tunability in a low magnetic field. It is demonstrated that oxygen-deficient HfO2 films show bipolar switching with a high on/off ratio, stable retention, as well as good endurance owing to the orthorhombic-rich phase constitution and charge (de)trapping-enabled Schottky-type conduction. Most importantly, we have demonstrated that RS can be tuned by a very low externally applied magnetic field (∼0-30 mT). Remarkably, application of a magnetic field of 30 mT causes RS to be fully quenched and frozen in the high resistive state (HRS) even after the removal of the magnetic field. However, the quenched state could be resurrected by applying a higher bias voltage than the one for initial switching. This is argued to be a consequence of the electronically and ionically "active" nature of the HfO2-x/LSMO interface on both sides and its susceptibility to the electric and low magnetic field effects. This result could pave the way for new designs of interface-engineered high-performance oxitronic ReRAM devices.

Ni-loss compensation and thermomechanical property recovery of 3D printed NiTi alloys by pre-coating Ni on NiTi powder

The approach of pre-coating Ni on NiTi powder was adopted as an effective route to address the thermomechanical property degradation of NiTi alloy caused by Ni-loss during the printing process of laser powder bed fusion. The electroless plating process of Ni on NiTi powder was initiated with the removal of the outermost TiO 2 layer by alkaline solution and the exposure of Ni catalytic nucleation site. The newly exposed Ni was an active catalyst for coating Ni and eventually triggered the deposition of Ni on NiTi powder via self-catalytic reaction. The intrinsic homogeneity of Ni, as a major alloying element in NiTi powder, ensured high uniformity of Ni-coating on the surface of the powder and good sphericity, necessary for good printability. Furthermore, Ni-coating increased laser absorbability, and a higher content of Ni compared to the theoretical value was thereby required to compensate for the Ni-loss. Simultaneously, densification was enhanced even at lower laser energy density on using Ni-coated powder. During the printing process, ultrafine Ni-coating was rapidly homogenized such that the thermomechanical property was completely recovered. As a consequence, it is underscored that Ni-loss can be addressed by compensating Ni on the original powder prior to the printing process, providing a new pathway toward printing of high-performance shape memory alloys.

Flexible and recyclable bio-based transient resistive memory enabled by self-healing polyimine membrane

The recyclable, self-healing and easily-degradable transient electronic technology has aroused tremendous attention in flexible electronic products. However, integrating the above advantages into one single flexible electronic device is still a huge challenge. Herein, we demonstrate a flexible and recyclable bio-based memory device using fish colloid as the resistive switching layer on a polyimine substrate, which affords reliable mechanical and electrical properties under repetitive conformal deformation operation. This flexible bio-based memory device presents potential analog behaviors including memory characteristics and excitatory current response, which undergoes incremental potentiation in conductance under successive electrical pulses. Moreover, this device is expected to greatly alleviate the environmental problems caused by electronic waste. It can be decomposed rapidly in water and well recycled, which is a promising candidate for transient memories and information security. We believe that this study can provide new possibilities to the field of high-performance transient electronics and flexible resistive memory devices.

Facile Fabrication of Eucommia Rubber Composites with High Shape Memory Performance.

We processed a series of shape memory Eucommia rubber (ER) composites with both carbon–carbon and ionic cross-linking networks via a chemical cross-linking method. The influence of the carbon–carbon cross-linking and ion cross-linking degree of ER composites on curing, mechanical, thermal, and shape memory properties were studied by DSC, DMA, and other analytical techniques. Dicumyl peroxide (DCP) and zinc dimethacrylate (ZDMA) played a key role in preparing ER composites with a double cross-linking structure, where DCP initiated polymerization of ZDMA, and grafted ZDMA onto polymer molecular chains and cross-linked rubber molecular chains. Meanwhile, ZDMA combined with rubber macromolecules to build ionic cross-linking bonds in composites under the action of DCP and reinforced the ER composites. The result showed that the coexistence of these two cross-linking networks provide a sufficient restoring force for deformation of shape memory composites. The addition of ZDMA not only improved the mechanical properties of materials, but also significantly enhanced shape memory performance of composites. In particular, Eucommia rubber composites exhibited outstanding mechanical properties and shape memory performance when DCP content was 0.2 phr.

Physically Active Lifestyle Is Associated With Attenuation of Hippocampal Dysfunction in Cognitively Intact Older Adults.

Alterations in hippocampal function have been shown in older adults, which are expressed as changes in hippocampal activity and connectivity. While hippocampal activation during memory demands has been demonstrated to decrease with age, some older individuals present increased activity, or hyperactivity, of the hippocampus which is associated with increased neuropathology and poor memory function. In addition, lower functional coherence between the hippocampus and core hubs of the default mode network (DMN), namely, the posteromedial and medial prefrontal cortices, as well as increased local intrahippocampal connectivity, were also demonstrated in cognitively intact older adults. Aerobic exercise has been shown to elicit neuroprotective effects on hippocampal structure and vasculature in aging, and improvements in cardiorespiratory fitness have been suggested to mediate these exercise-related effects. However, how these lifestyle factors relate to hippocampal function is not clear. Fifty-two cognitively intact older adults (aged 65–80 years) have been recruited and divided into physically active (n = 29) or non-active (n = 23) groups based on their aerobic activity lifestyle habits. Participants underwent resting-state and task-based fMRI experiments which included an associative memory encoding paradigm followed by a post-scan memory recognition test. In addition, 44 participants also performed cardiopulmonary exercise tests to evaluate cardiorespiratory fitness by measuring peak oxygen consumption (Vo2peak). While both groups demonstrated increased anterior hippocampal activation during memory encoding, a physically active lifestyle was associated with significantly lower activity level and higher memory performance in the recognition task. In addition, the physically active group also demonstrated higher functional connectivity of the anterior and posterior hippocampi with the core hubs of the DMN and lower local intra-hippocampal connectivity within and between hemispheres. Vo2peak was negatively associated with the hippocampal activation level and demonstrated a positive correlation with hippocampal-DMN connectivity. According to these findings, an aerobically active lifestyle may be associated with attenuation of hippocampal dysfunction in cognitively intact older adults.

High performance and low power consumption resistive random access memory with Ag/Fe2O3/Pt structure

Due to magnetic field tunability and the abundance of iron in the Earth’s crust, iron oxide-based resistive random access memory (RRAM) is considered to be low cost and potential for multi-level storage. However, the relatively high operation voltage (>1 V) and small storage window (<100) limit its application. In this work, the devices with simple Ag/Fe2O3/Pt structure exhibit typical bipolar resistive switching with ultralow set voltage (V set) of 0.16 V, ultralow reset voltage (V reset) of −0.04 V, high OFF/ON resistance ratio of 103, excellent cycling endurance more than 104 and good retention time longer than 104 s. Each major parameter has about an order of magnitude improvement compared to the previous data. The devices demonstrate outstanding stable low power consumption quality. Based on the analysis of the experimental results, a percolation model of silver ion migration was established and confirmed that low operation voltage is attributed to the amorphous oxide layer with large porosity. During electrical testing, the compliance current (I c) and maximum reset voltage (V max) can also affect the device performance. This discovery suggests Fe2O3 memristor has significant potential for application and provides a new idea for the realization of high-performance low-power RRAM.

Heterostructured Vertical Organic Transistor for High-Performance Optoelectronic Memory and Artificial Synapse

Heterostructured Vertical Organic Transistor for High-Performance Optoelectronic Memory and Artificial Synapse

Memory Packaging Challenges for a Growing Market

Abstract Increasing demands for storage and high-performance memory in applications such as artificial intelligence (AI), machine learning and processing storage are driving long-term growth in the market. This paper will highlight the key challenges encountered in packaging next-generation memory devices and discuss some of the technological developments required to address them while considering performance, cost, and yield. An evaluation is highlighted of a remote microwave plasma process to remove damage caused by laser ablation during the wafer dicing process that shows an improvement in die strength using this process.

Reduced Hippocampal-Striatal Interactions during Formation of Durable Episodic Memories in Aging.

Encoding of durable episodic memories requires cross-talk between the hippocampus and multiple brain regions. Changes in these hippocampal interactions could contribute to age-related declines in the ability to form memories that can be retrieved after extended time intervals. Here we tested whether hippocampal–neocortical– and subcortical functional connectivity (FC) observed during encoding of durable episodic memories differed between younger and older adults. About 48 younger (20–38 years; 25 females) and 43 older (60–80 years; 25 females) adults were scanned with fMRI while performing an associative memory encoding task. Source memory was tested ~20 min and ~6 days postencoding. Associations recalled after 20 min but later forgotten were classified as transient, whereas memories retained after long delays were classified as durable. Results demonstrated that older adults showed a reduced ability to form durable memories and reduced hippocampal–caudate FC during encoding of durable memories. There was also a positive relationship between hippocampal–caudate FC and higher memory performance among the older adults. No reliable age group differences in durable memory–encoding activity or hippocampal–neocortical connectivity were observed. These results support the classic theory of striatal alterations as one cause of cognitive decline in aging and highlight that age-related changes in episodic memory extend beyond hippocampal–neocortical connections.

Custom Memory Design for Logic-in-Memory: Drawbacks and Improvements over Conventional Memories

The speed of modern digital systems is severely limited by memory latency (the “Memory Wall” problem). Data exchange between Logic and Memory is also responsible for a large part of the system energy consumption. Logic-in-Memory (LiM) represents an attractive solution to this problem. By performing part of the computations directly inside the memory the system speed can be improved while reducing its energy consumption. LiM solutions that offer the major boost in performance are based on the modification of the memory cell. However, what is the cost of such modifications? How do these impact the memory array performance? In this work, this question is addressed by analysing a LiM memory array implementing an algorithm for the maximum/minimum value computation. The memory array is designed at physical level using the FreePDK 45nm CMOS process, with three memory cell variants, and its performance is compared to SRAM and CAM memories. Results highlight that read and write operations performance is worsened but in-memory operations result to be very efficient: a 55.26% reduction in the energy-delay product is measured for the AND operation with respect to the SRAM read one. Therefore, the LiM approach represents a very promising solution for low-density and high-performance memories.

Two-dimensional triphenylamine-based polymers for ultrastable volatile memory with ultrahigh on/off ratio

Two-dimensional (2D) memristive materials are of primary interest for resistive memory electronics. Herein, we report a high performance volatile dynamic-random-access memory (DRAM) based on 2D triphenylamine polymer. The 2D polymer was prepared via a solid-liquid interface limited polymerization method, with the merits of large-area, structure stability, and controllable film thicknesses. Utilizing the layered polymer as memristive medium in diode, the device shows a typical volatile DRAM performance. Impressively, the memory has an ultrahigh on/off current ratio up to 107, as well as outstanding thermal stability approaching at 300 °C. Both current ratio and thermal stability are higher than most previous polymer DRAM memories. Our work provides an effective strategy to construct scalable and stable 2D polymers toward high performance memory.

Crystallization mechanism and switching behavior of In–S–Sb phase change thin films

The crystallization mechanism of In2S3-doped Sb thin films is studied in detail to verify the potential application in phase change memory. Here, we observe directly that two different crystallization behaviors can exist in In–S–Sb thin layers by using aberration-corrected scanning transmission electron microscopy. The difference between Sb53.3(In2S3)46.7 and Sb30.9(In2S3)69.1 materials is induced by phase separation. The crystallization mechanism of the Sb53.3(In2S3)46.7 material is related to the formation of the nanocomposite structure with continuous precipitation of Sb nanocrystals. The crystallization characteristic of the Sb30.9(In2S3)69.1 material originates from the diffusion-driven In–S/In–S–Sb interface formation that acts as a “n–p” heterojunction, thereby resulting in the “depletion layer effect” and decreasing the carrier density to 7.42 × 1020 cm−3 at 280 °C. Sb30.9(In2S3)69.1 shows good bipolar-type resistance switching characteristics as the conventional Ge2Sb2Te5. This work provides clear experimental evidence to deepen the understanding of the crystallization mechanism for indium chalcogenides alloyed with Sb films, contributing to the improved control of the phase change behavior to establish high-performance multi-level nonvolatile memory and neuromorphic synaptic systems.

Flexible Graphene-Channel Memory Devices: A Review

There is an increasing importance of memory technologies in our ever-digitalizing society, which is characterized by the generation and use of a tremendous amount of real-time data. Beyond traditional performance requirements, mechanical flexibility of memory systems becomes therefore critical to enabling emerging applications such as the Internet of things. Graphene, now an established nanomaterial platform, is a promising element for building such high-performance memories with an unconventional form factor because of its exceptional electrical conductivities, outstanding mechanical properties, and processing versatility desirable for hybrid integration. Here, we provide an overview of recently developed flexible memory devices based on graphene and its functionalized counterparts. A defining feature of this review is that it exclusively compares and analyzes the devices that meet the following two criteria: (i) an explicit demonstration of working devices on a flexible substrate is reported; (ii) graphene is employed as an active channel rather than as an electrode. Our primary focus is to systematically classify various types of memories, in view of their materials, structural, and functional characteristics. For this, the shapes and compositions of the conductive channel, the key operational mechanisms underlying the electrical functionalities, and the major characteristics of representative device structures and materials systems are carefully evaluated. Furthermore, the applicability of flexible graphene memories to the neuromorphic computing area is discussed. Finally, we address several remaining issues that need to be solved for future technological advancements.

High switching uniformity and 50 fJ/bit energy consumption achieved in amorphous silicon-based memristive device with an AgInSbTe buffer layer

In this work, we demonstrated the high switching uniformity and 50 fJ/bit energy consumption in an amorphous silicon-based resistive switching (RS) device by inserting the AgInSbTe (AIST) layer between the silicon insulating layer and Ag top electrodes. The improved RS performance is attributed to the introduction of an Ag ion reservoir layer, which helps to suppress conducting filament overgrowth. After insertion of the AIST layer, the cumulative probability of low/high resistance states decreased from 176.8%/46.2% to 3.1%/11.9%, respectively. The advantages of promoting Ag dissolution enable the realization of fast switching speed (&amp;lt;50 ns) and low set voltage (∼70 mV), which gives our device low energy consumption (∼50 fJ/bit). Moreover, the multi-step of set/reset analytical model of our dual-layer RS device was developed based on the formation and dissolution of the Ag-ion-based conductive filaments. Our work presents an effective method for obtaining high-performance Si-based memory for practical applications.

COPS: A complete oblivious processing system

The communication link between the processor and the off-chip memory in a modern computing system is vulnerable to hardware-level security attacks. An adversary can snoop the memory access behavior over the communication link to retrieve secret information even when the data transferred over the link are encrypted. The Oblivious RAM (ORAM) mechanism is a promising solution for such a problem. ORAM hides a true memory access with a group of random memory read and write operations. To reliably estimate the cost and performance overhead incurred by ORAM in real scenarios, a complete oblivious processing system where a fully functional ORAM is integrated into a real processor system that can run an Operating System (OS) and execute real-world application workloads is required. In this paper, we present a generic design for ORAM integration into a general-purpose processing system that can be used for the integration of any ORAM design. Our design targets the ORAM integration into high-performance memory bus (AXI). The AXI-compatibility of our design makes it usable with any processor system, thus, making the ORAM integration independent of processor architecture. As a case study, we develop a complete oblivious processing system (we call it COPS) by integrating an ORAM into a general-purpose RISC-V based processing system. We evaluate ORAM performance in real-world execution environment using realistic workloads with full Operating System support. Our elevation demonstrates that our ORAM prototype, on average, can return a 64 bytes requested block to the processor in 1.46μs, which is ∼ 13x faster than the prior-art implementation.