Memory Integration Research Articles

Gate-controllable two-dimensional transition metal dichalcogenides for spintronic memory

Rapid technological advancement has increased the demand for high-speed, low-power devices, particularly applied in artificial intelligence (AI) and the Internet of Things (IoT). Limitations of traditional memory devices underscore the urgency for high-performance, energy-efficient memory technologies. On the other hand, developing two-dimensional (2D) material-based devices is indispensable for next-generation memory and three-dimensional integration circuit (3D-IC) systems. Here, we propose a gate-controllable spin valve utilizing transition metal dichalcogenides (TMDs) to meet the urgent need. A high tunneling magnetoresistance (TMR) of over 4000% can be reached in reading with the help of the controlled gate. Moreover, under ungated conditions, a giant spin current density with an ultralow power consumption of 80 μW and a high spin-polarized ratio of 0.9 can be achieved, enabling high-speed and energy-efficient switching during writing. According to our design, the gate-controllable system can prevent undesired mixed reading and writing operations. Our results highlight TMDs as a promising material in spintronic memory devices.

Targeted memory reactivation during sleep improves emotional memory modulation following imagery rescripting

Targeted Memory Reactivation (TMR) during sleep benefits memory integration and consolidation. In this pre-registered study, we investigated the effects of TMR applied during non-rapid eye movement (NREM) sleep following modulation and updating of aversive autobiographical memories using imagery rescripting (ImR). During 2–5 nights postImR, 80 healthy participants were repeatedly presented with either idiosyncratic words from an ImR updated memory during sleep (experimental group) or with no or neutral words (control groups) using a wearable EEG device (Mobile Health Systems Lab-Sleepband, MHSL-SB) [1] implementing a close-loop cueing procedure. Multivariate analysis were conducted to assess change score trajectories in five key emotional memory characteristics (positive and negative valence, emotional distress, arousal, and vividness) across assessments (timepoints, t) and between the study groups (TMR condition). While ImR showed significant effects on all memory characteristics (d = 0.76–1.66), there were significant additional improvements in the experimental group. Memories were significantly less vivid and afflicted with less emotional distress and arousal following ImR-words cueing. TMR during sleep in individuals’ homes was feasible and further improved some ImR’s adaptive memory effects. If replicated in clinical samples, TMR may be utilized to augment the effects of ImR and other clinical memory modulation procedures and create personalized treatment options. Such advances in emotional memory treatments are direly needed, as aversive memories are a salient feature across mental disorders, such as post-traumatic stress disorder (PTSD).

Enhanced performance of self-selective RRAM devices in V/TaOx/Pt structure

Abstract The performance of resistive random-access memory devices with vanadium as the top electrode and different TaOx insulating layer is investigated in this paper. The results indicate that the VO2 oxide layer generated by the oxidation of the vanadium electrode can serve as inherent selector, without the need for additional series selectors required in conventional methods. A large amount of oxygen vacancy migration was limited in the double insulating layers, enabling the device to achieve stable Self-Selective resistive random-access memory performance. This indicated that the advantage of the double insulation layers lay in manipulating the migration of oxygenions and vacancies. Endurance tests showed no degradation over 103 cycles, and the device maintained stability after 20,000 pulse applications. The subthreshold swing of the selector was less than 42 mV/dec, and the switching time was less than 2 μs. This research presents a promising advancement in resistive random-access memory technology with potential for high-density memory integration and reliable performance.

Load Forecasting for Commercial Buildings Using BiLSTM–Transformer Network and Cyber–Physical Cognitive Control Systems

With the widespread adoption of electric vehicles (EVs), their charging and discharging schedules pose new challenges for real-time load forecasting in commercial buildings. This study proposes a prediction model based on the integration of bidirectional long short-term memory (BiLSTM) networks and Transformer architecture, along with the introduction of a cognitive control system and cyber–physical systems (CPS) to address issues such as data loss and excessive computation time during the forecasting process. The BiLSTM–Transformer model significantly improves load-forecasting accuracy and real-time performance by combining time-series modeling with global feature extraction capabilities. Additionally, the cognitive control system includes user-aware cognitive control (UACC) and Microgrid Control Center Cognitive Control (MACC). UACC quantifies information gaps in real time and adaptively adjusts strategies during communication instability, while MACC employs Q-learning algorithms to evaluate the impact of data loss on scheduling and optimize power resource allocation. The synergy between these mechanisms ensures system stability and predictive performance in scenarios involving data loss or communication disruptions. Experimental results demonstrate that the model achieves outstanding predictive accuracy under complete data conditions and significantly reduces errors in scenarios with data loss, validating its superior accuracy and robustness. This provides reliable support for load forecasting in commercial buildings.

Integration of overlapping sequences emerges with consolidation through medial prefrontal cortex neural ensembles and hippocampal-cortical connectivity.

Systems consolidation theories propose two mechanisms that enable the behavioral integration of related memories: coordinated reactivation between hippocampus and cortex, and the emergence of cortical traces that reflect overlap across memories. However, there is limited empirical evidence that links these mechanisms to the emergence of behavioral integration over time. In two experiments, participants implicitly encoded sequences of objects with overlapping structure. Assessment of behavioral integration showed that response times during a recognition task reflected behavioral priming between objects that never occurred together in time but belonged to overlapping sequences. This priming was consolidation-dependent and only emerged for sequences learned 24 hr prior to the test. Critically, behavioral integration was related to changes in neural pattern similarity in the medial prefrontal cortex and increases in post-learning rest connectivity between the posterior hippocampus and lateral occipital cortex. These findings suggest that memories with a shared predictive structure become behaviorally integrated through a consolidation-related restructuring of the learned sequences, providing insight into the relationship between different consolidation mechanisms that support behavioral integration.

Effects of sequential and non-sequential presentation conditions of multiple-stem facts on memory integration and cognitive resource allocation.

What limits the self-generation of new knowledge in the memory integration process? One striking contender is the amount of necessary pieces of information that are dispersed. Specifically, when essential information is scattered across multiple sources/places, it becomes challenging to effectively integrate and generate new knowledge. Most of the studies on memory integration have focused on the study of paired stem facts, but have neglected the exploration of multiple-stem facts. The present study examined college students' performance on memory integration under different conditions of three stem facts. In Experiment 1, participants were exposed to a series of novel, authentic stem facts in which every three relevant ones could be integrated to generate new knowledge. The results of Experiment 1 found that college students could spontaneously generate a new piece of information by integrating two or three separate but related facts. The integration can occur in at least two distinct types due to the different presentation orders of the learning materials: sequential recursive integration and non-recursive integration. College students performed better in sequential recursive integration than in non-sequential recursive integration, and this difference in integration performance is not caused by differences in memory for the stem facts. Based on Experiment 1, Experiment 2 used eye-tracking technology to explore the allocation of internal cognitive resources across different conditions of three stem facts. We found that in non-sequential recursive integration, college students had the longest visual duration and the highest number of fixations on the second stem fact. In sequential recursive integration, there were no other significant differences in the number and duration of visual fixations for the three stem facts. College students paid longer fixations to the second stem fact and the third stem fact in the non-sequential recursive condition than in the sequential recursive condition. Our study suggests that when information is related but cannot be integrated, longer fixation indicates stumbling when dealing with an unresolvable difficulty. When knowledge is presented in a stepwise manner (such as in the sequential recursive integration condition), it results in better semantic memory extension.

Building and Breaking the Chain: A Model of Reward Prediction Error Integration and Segmentation of Memory.

Prediction errors drive reinforcement learning and organize episodic memory into distinct contexts, but do these effects interact? Here, we review the roles of midbrain dopamine, the locus coeruleus, and the hippocampus in event cognition to propose and simulate the theoretical influence of two prediction error signals in integrating versus segmenting events in memory. We suggest that signed reward prediction errors can build mental models of reward environments, increasing the contextual similarity (integration) of experiences with stronger, more stable reward expectations. On the other hand, unsigned reward prediction errors can signal a new model of the environment, generating a contextual shift (segmentation) between experiences that crossed them. We moreover predicted that these differences in contextual similarity give rise to distinct patterns of temporal-order memory. We combined these ideas in a computational model to account for a seemingly paradoxical pattern of temporal-order memory where greater representational distance helps order memory within context but impairs it across contexts. We found that simulating signed reward prediction error integration and unsigned reward prediction error segmentation differentially enabled the model to perform associative chaining, which involved reactivating items between two tested probes to assist with sequential retrieval. In summary, our simulations provide a unifying explanation for the varied ways that neuromodulatory systems may alter event cognition and memory.

Working Meaningfully With Older Adults in Psychotherapy: Memory, Autobiography, Self, and the Conversational Model

As the population of older adults increases, a growing number of older people will seek psychotherapy and counselling in the coming years. Therefore, therapists should deepen their understanding of the unique developmental challenges that accompany the later years. While varying approaches and theoretical frameworks exist for how to understand these developmental challenges and work in a way that promotes wellbeing, therapy with older adults remains an under-explored area. This paper outlines my clinical observations while working with a single man in his mid-70s who started seeing me for long-term psychotherapy. It explores the intersection of memory, autobiography, self, and ego integration and contributes to the discussion about how to work meaningfully with older adults. I hypothesise that working with our clients to compose an autobiographical consciousness that can be narrated and witnessed by an empathic other could be an important aspect of ego integration. This is because narrative is a meaning-making tool that facilitates one’s sense of self and promotes the process of reconciliation with the past and direction towards the future. This exploration is situated within the framework of the Conversational Model of psychotherapy developed by Robert Hobson and Russell Meares.

How Psychedelics Modulate Multiple Memory Mechanisms in Posttraumatic Stress Disorder.

Posttraumatic stress disorder (PTSD) is a psychiatric disorder with defining abnormalities in memory, and psychedelics may be promising candidates for the treatment of PTSD given their effects on multiple memory systems. Most PTSD and psychedelic research has investigated memory with fear conditioning and extinction. While fruitful, conditioning and extinction provide a limited model of the complexity of PTSD and phenomenology of psychedelics, thereby limiting the refinement of therapies. In this review, we discuss abnormalities in fear conditioning and extinction in PTSD and review 25 studies testing psychedelics on these forms of memory. Perhaps the most reliable effect is that the acute effects of psychedelics can enhance extinction learning, which is impaired in PTSD. However, the post-acute effects may also enhance extinction learning, and the acute effects can also enhance fear conditioning. We then discuss abnormalities in episodic and semantic memory in PTSD and review current knowledge on how psychedelics impact these memory systems. Although PTSD and psychedelics acutely impair the formation of hippocampal-dependent episodic memories, psychedelics may acutely enhance cortical-dependent learning of semantic memories that could facilitate the integration of trauma memories and disrupt maladaptive beliefs. More research is needed on the acute effects of psychedelics on episodic memory consolidation, retrieval, and reconsolidation and post-acute effects of psychedelics on all phases of episodic memory. We conclude by discussing how targeting multiple memory mechanisms could improve upon the current psychedelic therapy paradigm for PTSD, thereby necessitating a greater emphasis on assessing diverse measures of memory in translational PTSD and psychedelic research.

Predicting the Deformation of a Concrete Dam Using an Integration of Long Short-Term Memory (LSTM) Networks and Kolmogorov–Arnold Networks (KANs) with a Dual-Stage Attention Mechanism

Predicting the Deformation of a Concrete Dam Using an Integration of Long Short-Term Memory (LSTM) Networks and Kolmogorov–Arnold Networks (KANs) with a Dual-Stage Attention Mechanism

Holding the product of visual working memory integration: The role of attention.

The involuntary integration of discrete fragments into meaningful units (e.g., Gestalt) within visual working memory (VWM) is a crucial process in mind. However, the mechanisms governing the maintenance of these integrated products within VWM have remained largely unexplored. The current study sought to address this gap by investigating whether maintaining such VWM integration products places a greater demand on attention resources compared to discrete representations. We hypothesized that maintenance may be costless or require additional attention, which may be domain-specific or domain-general. To examine these hypotheses, we tested whether the emerged Gestalts by VWM integration can be abolished by an attention consumption task. Participants were required to memorize a sequence of oriented disks with or without Gestalt cues, alongside a secondary task during maintenance, consuming a specific type of attention. We found that a task consuming spatial attention impaired the VWM Gestalts of bar contours (Experiments 1 and 3), but not the Gestalts of square contours (Experiment 2). Moreover, a task consuming domain-general attention did not affect the VWM Gestalts of bar contours (Experiment 4). These findings provide evidence suggesting that maintaining VWM integration products requires more attention than discrete representations and that the type of attention required is domain-specific.

Improvement of charge trapping memory performance by modulating band alignment with oxygen plasma

Metal-oxide charge trapping memory (CTM) integration into amorphous and organic flexible devices encounters challenges due to high-temperature treatment. Our approach enhances memory performance via room-temperature oxygen plasma treatment, subtly adjusting surface band alignment without changing the original material structure and surface roughness. Infiltration of oxygen plasma induces band alignment bending, creating a barrier for charge trapping. The device with oxygen plasma treatment exhibits an impressive 19.06 V memory window and a charge trapping density of 3.58 × 1013/cm2. In comparison, the memory window of untreated device only has 5.56 V, demonstrating that oxygen plasma treatment significantly improves memory characteristics. The charge retention rate exhibits outstanding stability, potentially reaching 94% after a decade. It should be noted that careful control during plasma treatment is crucial to maintaining optimal memory effects. This facile, efficient technique, applicable to various oxide layers, offers a way for future advancements in metal-oxide CTM technology.

Inducing representational change in the hippocampus through real-time neurofeedback.

When you perceive or remember something, other related things come to mind, affecting how these competing items are subsequently perceived and remembered. Such behavioural consequences are believed to result from changes in the overlap of neural representations of these items, especially in the hippocampus. According to multiple theories, hippocampal overlap should increase (integration) when there is high coactivation between cortical representations. However, prior studies used indirect proxies for coactivation by manipulating stimulus similarity or task demands. Here, we induce coactivation in visual cortex more directly using closed-loop neurofeedback from real-time functional magnetic resonance imaging (fMRI). While viewing one object, participants were rewarded for activating the representation of another object as strongly as possible. Across multiple real-time fMRI sessions, participants succeeded in using this neurofeedback to increase coactivation. Compared with a baseline of untrained objects, this protocol led to memory integration in behaviour and the brain: the trained objects became harder for participants to discriminate behaviourally in a categorical perception task and harder to discriminate neurally from patterns of fMRI activity in their hippocampus as a result of losing unique features. These findings demonstrate that neurofeedback can be used to alter and combine memories.This article is part of the theme issue 'Neurofeedback: new territories and neurocognitive mechanisms of endogenous neuromodulation'.

A two-step approach for damage identification in bridge structure using convolutional Long Short-Term Memory with augmented time-series data

This paper presents a novel two-step approach to identifying structural damages in bridge structure through the integration of 1D Convolutional Neural Network (1DCNN) and Long Short-Term Memory (LSTM) networks, enhanced by the augmentation and transformation techniques using Symbolic Aggregate approXimation (SAX) for time-series data analysis. In the first step, the time-series data of the bridge is diversified and quantified by augmentation techniques to make the model more robust and increase its generalization capabilities. After that, SAX is implemented to reduce the volume and categorize time series data through the transformation of continuous time series into discrete symbols, thereby decreasing the size of the data for more efficient training performance. In the second step, an advanced DL model combining 1DCNN and LSTM is proposed to tackle the damage identification problems of the processed data. By leveraging the strengths of CNNs in feature extraction and LSTMs in sequence learning, combined with advanced techniques for data augmentation, our methodology offers a robust solution not only for improving the model's training process but also for enabling it to learn from a more diverse and comprehensive dataset that mimics different damage scenarios, allowing more accurate detection of damages within bridge structures. Validation of the proposed method is conducted using time-series data collected from Chuong Duong Bridge structure. The effectiveness of the proposed method is compared with other models, such as 1DCNN, LSTM, and the combined 1DCNN-LSTM. The results show that the proposed 1DCNN-LSTM-SAX outperforms the other methods in terms of accuracy and, thus, can be used extensively to deal with the damage identification problems of bridges using time-series data.

Active contextualization reduces traumatic memory intrusions via memory integration

Traumatic memory intrusions, the involuntary retrieval of unwanted memories, significantly impact mental health. The dual representation theory proposes that the origin of intrusion lies in the overactivated sensory memory not being integrated with the corresponding contextual memory, highlighting the crucial associations between memory contextualization and intrusion. To test this, our study investigated whether enhancing memory contextualization could effectively reduce intrusion. After experiencing analogue trauma with the trauma film paradigm, 96 healthy participants were randomly allocated to three intervention groups: active contextualization (AC) in which participants actively retrieve and restructure film content, passive contextualization (PC) in which participants passively restudy content-matched pre-contextualized information, and working memory taxation (WM) in which participants performed a working memory dual-task. Diary recordings over the subsequent week revealed a significant reduction in intrusion frequency in the AC group compared to both the PC group and a no-intervention control group. Furthermore, comparing AC with WM, a well-established laboratory intervention on intrusion, established a superior efficacy of the AC intervention in reducing intrusions. Finally, analyses of the explicitly recollected film memories identified the critical element of active contextualization to be memory integration induced by active memory retrieval. Together, our findings suggest that active contextualization causally diminishes intrusions, providing novel insights into the regulation of the contextual memory system in intrusion intervention.

Co-optimizing of H-curing, stress, and thermal annealing—A more effective way for realizing FEOL-BEOL compatible eMRAM chip

Metallization, plasma, and thermal annealing used in magneto-resistive random access memory integration differs significantly from traditional interconnection processes, usually resulting in the device performance deterioration of metal-oxide-silicon field-effect transistors. Commonly used methods, such as hydrogen postalloying, may not be effective to recover performance when magneto-resistive random access memory is integrated. The study focused on the interplay between hydrogen content in the dielectric films, wafer configuration induced by film stress, and thermal annealing during the chip fabrication process. Post-thermal annealing can enhance the release of hydrogen from the dielectric films, which can repair the silicon–hydrogen bond breakage resulting from magneto-resistive random access memory integration. However, this process can also worsen wafer bowing in an undesired direction which is detrimental to the performance of metal-oxide-silicon field-effect transistors. Consequently, utilizing silicon nitride with specific hydrogen content, in conjunction with post-thermal annealing, can improve effectively both p-type and n-type metal-oxide-silicon field-effect transistors. This approach holds significant potential for application in more advanced technological nodes since it is compatible with traditional process and does not require more advanced fabrication equipment. Additionally, the resulting embedded magneto-resistive random access memory chips maintain competitive magnetic tunneling junction reliability in term of endurance and reflow resistivity.

Individual differences diminish the pretest effect under productive memory conditions.

Pretesting, or asking a test question prior to the onset of learning, is a well-established means of enhancing learning. Research on pretesting has focused primarily on direct factual learning outcomes. Yet building a coherent knowledge base also depends on productive memory processes that permit going beyond the information directly given. In the specific productive process of self-derivation through memory integration, individual differences are prominent; verbal comprehension is a consistent predictor. In the current work, we integrated these research trends by testing the extent to which pretesting enhances learning through productive memory processes and the role played by individual differences in verbal comprehension. Across four within-subjects experiments, we assessed the pretest effect after accounting for variability associated with verbal comprehension. In Experiments 1-3, we assessed the productive memory process of self-derivation through memory integration. Adults were more successful on pretest trials compared to control (i.e., no pretest) trials, but this effect was no longer significant after controlling for verbal comprehension. This pattern emerged when we used stem-fact pretests (Experiment 1) and integration-fact pretests (Experiment 2) to probe self-derivation across single-sentence stimuli and replicated when we used stimuli more akin to everyday learning materials (i.e., text passages and photographs; Experiment 3). In Experiment 4, we shifted the test target from productive processes to fact recall and found the pretest effect held even after controlling for verbal comprehension. This research bridges the pretest and productive process literature to provide novel insight into ways of maximizing learning. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

RCLNet: an effective anomaly-based intrusion detection for securing the IoMT system.

The Internet of Medical Things (IoMT) has revolutionized healthcare with remote patient monitoring and real-time diagnosis, but securing patient data remains a critical challenge due to sophisticated cyber threats and the sensitivity of medical information. Traditional machine learning methods struggle to capture the complex patterns in IoMT data, and conventional intrusion detection systems often fail to identify unknown attacks, leading to high false positive rates and compromised patient data security. To address these issues, we propose RCLNet, an effective Anomaly-based Intrusion Detection System (A-IDS) for IoMT. RCLNet employs a multi-faceted approach, including Random Forest (RF) for feature selection, the integration of Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) models to enhance pattern recognition, and a Self-Adaptive Attention Layer Mechanism (SAALM) designed specifically for the unique challenges of IoMT. Additionally, RCLNet utilizes focal loss (FL) to manage imbalanced data distributions, a common challenge in IoMT datasets. Evaluation using the WUSTL-EHMS-2020 healthcare dataset demonstrates that RCLNet outperforms recent state-of-the-art methods, achieving a remarkable accuracy of 99.78%, highlighting its potential to significantly improve the security and confidentiality of patient data in IoMT healthcare systems.

Navigating Complex Process and Design Challenges in Hybrid Bonding for Advanced Semiconductor Integration: A Comprehensive Analysis

Abstract In advanced electronic systems, achieving top-tier interconnect interfaces with fine-pitch integration is of paramount importance. Among interconnect options, hybrid bonding is the preeminent choice given its exceptional capacity for accommodating a high input/output (I/O) count, which facilitates high-density memory integration, increased power delivery, and enhanced signal speed. One key technique for ensuring utmost quality in hybrid bonding involves embedding Cu interconnects within the dielectric layer. Equally pivotal is surface planarization, accomplished through chemical mechanical polishing (CMP), which encompasses a meticulous two-step procedure, commencing with copper bulk CMP and culminating in barrier CMP. The latter step is particularly critical, yielding the indispensable surface finish required for a successful hybrid bonding process. Several crucial surface properties significantly influence overall bond yield, including a copper recess (“dishing”) in the vias, erosion and roughness of the dielectric layer, and surface topography changes from high-density to low-density copper vias. To optimize these parameters, a comprehensive understanding of the interconnect layer's design is essential. Here, we delve into the ramifications of via scaling, ranging from 5 to 1 µm, on dishing and roll-off, and the effects of copper via density variation, spanning from 16% to 13%, on topography. Furthermore, we explore how incorporation of dummy vias impacts the final surface finish, potentially leading to improved bond yie

Spatial memory obviates following behaviour in an information centre of wild fruit bats.

According to the information centre hypothesis (ICH), colonial species use social information in roosts to locate ephemeral resources. Validating the ICH necessitates showing that uninformed individuals follow informed ones to the new resource. However, following behaviour may not be essential when individuals have a good memory of the resources' locations. For instance, Egyptian fruit bats forage on spatially predictable trees, but some bear fruit at unpredictable times. These circumstances suggest an alternative ICH pathway in which bats learn when fruits emerge from social cues in the roost but then use spatial memory to locate them without following conspecifics. Here, using an unique field manipulation and high-frequency tracking data, we test for this alternative pathway: we introduced bats smeared with the fruit odour of the unpredictably fruiting Ficus sycomorus trees to the roost, when they bore no fruits, and then tracked the movement of conspecifics exposed to the manipulated social cue. As predicted, bats visited the F. sycomorus trees with significantly higher probabilities than during routine foraging trips (of >200 bats). Our results show how the integration of spatial memory and social cues leads to efficient resource tracking and highlight the value of using large movement datasets and field experiments in behavioural ecology. This article is part of the theme issue 'The spatial-social interface: a theoretical and empirical integration'.