National Standards Institute Research Articles

Phytochemical Composition and Antimicrobial Efficacy of Salvadora persica Root Extracts Against Carbapenem-Resistant Acinetobacter baumannii.

Background Carbapenem-resistant Acinetobacter baumannii (CRAB) are difficult to eradicate from the environment and are virtually immune to all antibiotics. Consequently, CRABmay culminate in severe outbreaks and fatal infections among people attending hospitals and nursing homes. Salvadora persicahas been used as an herbal remedy and chewing sticks for dental cleansing.Evaluating S.persica's efficacy against CRABmay provide an alternative approach to treating CRAB infections in healthcare environments, considering its traditional application in dental hygiene. Employing S. persica as an herbal remedy could be a part of a more sustainableapproach to control CRAB infections. Aim To investigate the phytochemical composition of S. persica and evaluate its antimicrobial properties. Materials and methods The roots were extracted by Soxhlet apparatus using n-hexane, chloroform, and methanol. Each extract was analyzed using gas chromatography-mass spectrometry (GCMS) and characterized using WN908.Land National Institute of Standards and Technology (NIST) libraries.The antimicrobial activity of each extract against CRAB was evaluated using a broth microdilution assay to determine the minimum inhibitory concentration (MIC) and minimum bactericidal concentration (MBC). Results The GCMS analysis of different solvent extracts of S. persica roots showed the presence of various phytochemical compounds such as steroids, phenolic compounds, fatty acids, alcohols, terpenoids, and vitamin E. Both chloroform and hexane extracts showed the most effective antimicrobial activity with a MIC valueof 3.13 mg/mL and an MBC value of 12.50 mg/mL, respectively. Benzoic acid was the major phytochemical compound identified from S. persica extract. N-hexane, chloroform, and methanol extracts exhibited maximum antimicrobial activity due to the presence of active compounds in them. Conclusion Chloroform and hexane extracts showed the most potent antibacterial activities against CRAB.

Convolution neural network hyperparameter optimization using modified particle swarm optimization

Based on the literature review, a convolutional neural network (CNN) is one of the deep learning techniques most often used for classification problems, especially image classification. Various approaches have been proposed to improve accuracy performance. In CNN architecture, parameter determination is very influential on accuracy performance. Particle swarm optimization (PSO) is a type of metaheuristic algorithm widely used for hyperparameter optimization. PSO convergence is faster than genetic algorithm (GA) and attracts many researchers for further studies such as genetic algorithms and ant colony. In PSO, determining the value of the weight parameter is very influential on accuracy. Therefore, this paper proposes CNN hyperparameter optimization using modified PSO with linearly decreasing randomized weight. The experiments use the modified National Institute of Standards and Technology (MNIST) dataset. The accuracy of the proposed method is superior, and the execution time is slower to random search. In epoch 1, epoch 3, and epoch 5, the proposed method is superior to baseline CNN, linearly decreasing weight PSO (LDWPSO), and RL-based optimization algorithm (ROA). Meanwhile, the accuracy performance of the proposed method is superior to previous studies, namely LeNet-1, LeNet-2, LeNet-3, PCANet-2, RANDNet-2, CAE1, CAE-2, and bee colony. Otherwise, lost to PSO-CNN, distributed PSO (DPSO), recurrent CNN, and CNN-PSO. However, the four methods have a complex architecture and wasteful execution time.

Improved differential fault analysis of Grain128-AEAD

The number of smart devices connected to the Internet has been constantly increasing, and as a result, lightweight cryptography (LWC) has become more important in the past decade. The Lightweight Cryptography (LWC) Project is an initiative taken by the National Institute of Standards and Technology (NIST) to standardize such LWC algorithms. Grain128-AEAD, which was submitted to the NIST LWC project, is an encryption algorithm that provides both confidentiality and integrity assurance. Third-party security analysis of the submitted ciphers is an important aspect of the evaluation of the submission to the NIST LWC project. Although several pieces of existing research, such as the bit-flipping attack, random fault attack, and deterministic random fault attack, have examined the security of Grain128-AEAD, there is still room for improvement in the fault attack models of these studies. This work aims to fill this research gap by analyzing the security margin of Grain128-AEAD against a series of improved differential fault attacks. In this study, we developed a probabilistic random fault attack and applied it to Grain128-AEAD. As an improvement of the existing research, a probabilistic approach can be applied to a more relaxed moderate control attack model. The existing moderate control model assumes the fault to be injected within any bit of a given byte, whereas the faults in our improved approach can be injected within any bits of a two-byte/four-byte segment, thereby relaxing the fault precision. The results indicate that the improved moderate control requires 388 keystreams for the two-byte model and 279 for the four-byte model to identify the target fault locations for implementing a state recovery attack. The relaxed fault attack models presented in this work are more practical to implement; hence, the findings of this research have improved the existing studies and narrowed the current research gap on the fault attack models of Grain128-AEAD. % To maintain consistency in terminology, "Grain-128AEAD" has been revised to "Grain128-AEAD" in both the abstract and the main text. Please confirm this revision.

Blood-inspired random bit generation using microfluidics system

The development of random number generators (RNGs) using speckle patterns is pivotal for secure encryption key generation, drawing from the recent statistical properties identified in speckle-based imaging. Speckle-based RNG systems generate a sequence of random numbers through the unpredictable and reproducible nature of speckle patterns, ensuring a source of randomness that is independent of algorithms. However, to guarantee their effectiveness and reliability, these systems demand a meticulous and rigorous approach. In this study, we present a blood-inspired RNG system with a microfluidics device, designed to generate random numbers at a rate of 5.5 MHz and a high-speed of 1250 fps. This process is achieved by directing a laser beam through a volumetric scattering medium to procure speckle patterns. Additionally, designed microfluidic device requires only a minimal blood sample of 5 µl to capture these speckle patterns effectively. After implementing the two-pass tuple-output von Neumann debiasing algorithm to counteract statistical biases, we utilized the randomness statistical test suite from the National Institute of Standards and Technology for validation. The generated numbers successfully passed these tests, ensuring their randomness and unpredictability. Our blood-inspired RNG, utilizing whole blood, offers a pathway for affordable, high-output applications in fields like encryption, computer security, and data protection.

Noise-resilient single-pixel compressive sensing with single photon counting

The fast expansion of photon detection technology has fertilized the rapid growth of single-photon sensing and imaging techniques. While promising significant advantages over their classical counterparts, they suffer from ambient and quantum noises whose effects become more pronounced at low light levels, limiting the quality of the acquired signal. Here, we study how photon-counting noises degrade a single-pixel optical classifier via compressive sensing, and how its performance can be restored by using quantum parametric mode sorting. Using modified National Institute of Standards and Technology (MNIST) handwritten digits as an example, we examine the effects of detector dark counts and in-band background noises and demonstrate the effectiveness of mode filtering and upconversion detection in addressing those issues. We achieve 94% classification accuracy in the presence of 500 times stronger in-band noise than the signal received. Our results suggest a robust and efficient approach to single photon sensing in a practical environment, where sunlight, ambient, and multiscattering noises can easily dominate the weak signal.

Password authenticated key exchange-based on Kyber for mobile devices.

In this article, a password-authenticated key exchange (PAKE) version of the National Institute of Standards and Technology (NIST) post-quantum cryptography (PQC) public-key encryption and key-establishment standard is constructed. We mainly focused on how the PAKE version of PQC standard Kyber with mobile compatibility can be obtained by using simple structured password components. In the design process, the conventional password-based authenticated key exchange (PAK) approach is updated under the module learning with errors (MLWE) assumptions to add password-based authentication. Thanks to the following PAK model, the proposed Kyber.PAKE provides explicit authentication and perfect forward secrecy (PFS). The resistance analysis against the password dictionary attack of Kyber.PAKE is examined by using random oracle model (ROM) assumptions. In the security analysis, the cumulative distribution function (CDF) Zipf (CDF-Zipf) model is also followed to provide realistic security examinations. According to the implementation results, Kyber.PAKE presents better run-time than lattice-based PAKE schemes with similar features, even if it contains complex key encapsulation mechanism (KEM) components. The comparison results show that the proposed PAKE scheme will come to the fore for the future security of mobile environments and other areas.

Probing the Chemistry of Sulfurous Pollutants: Accurate Thermochemistry Determination of Extensive Sulfur-Containing Species.

Sulfur-containing fuels, such as petroleum fuels, natural gas, and biofuels, produce SO2, SO3, and other highly toxic gases upon combustion, which are harmful to human health and the environment, making it essential to understand their thermochemical properties. This study used high-level quantum chemistry calculations to determine thermodynamic parameters, including entropy, enthalpy, and specific heat capacity for an extensive set of sulfur-containing species. The B3LYP/cc-pVTZ level of theory was used for geometry optimization, vibration frequency, and dihedral scan calculations. To determine an appropriate ab initio method for energy calculation, the Bland-Altman diagram, a statistical analysis method, was employed to visualize the 298 K enthalpy value between experimental data and three sets of ab initio methods: G3, CBS-QB3, and the average of G3 plus CBS-QB3. The CBS-QB3 method exhibited the highest accuracy and was eventually selected for the energy calculation in this study. Thermochemical property parameters were then calculated with the MultiWell program suite for all these sulfur-containing species, and the results were in good agreement with the thermochemical data of organic compounds and the National Institute of Standards and Technology Chemistry WebBook databases. The thermochemical property database established in this study is essential to studying sulfur-containing species in desulfurization.

Control-Etched Ti3C2Tx MXene Nanosheets for a Low-Voltage-Operating Flexible Memristor for Efficient Neuromorphic Computation.

Hardware neural networks with mechanical flexibility are promising next-generation computing systems for smart wearable electronics. Overcoming the challenge of developing a fully synaptic plastic network, we demonstrate a low-operating-voltage PET/ITO/p-MXene/Ag flexible memristor device by controlling the etching of aluminum metal ions in Ti3C2Tx MXene. The presence of a small fraction of Al ions in partially etched MXene (p-Ti3C2Tx) significantly suppresses the operating voltage to 1 V compared to 7 V from fully Al etched MXene (f-Ti3C2Tx)-based devices. Former devices exhibit excellent non-volatile data storage properties, with a robust ∼103 ON/OFF ratio, high endurance of ∼104 cycles, multilevel resistance states, and long data retention measured up to ∼106 s. High mechanical stability up to ∼73° bending angle and environmental robustness are confirmed with consistent switching characteristics under increasing temperature and humid conditions. Furthermore, a p-Ti3C2Tx MXene memristor is employed to mimic the biological synapse by measuring the learning-forgetting pattern for ∼104 cycles as potentiation and depression. Spike time-dependent plasticity (STDP) based on Hebb's Learning rules is also successfully demonstrated. Moreover, a remarkable accuracy of ∼95% in recognizing modified patterns from the National Institute of Standards and Technology (MNIST) data set with just 29 training epochs is achieved in simulation. Ultimately, our findings underscore the potential of MXene-based flexible memristor devices as versatile components for data storage and neuromorphic computing.

Neuromorphic Computing of Optoelectronic Artificial BFCO/AZO Heterostructure Memristors Synapses.

Compared with purely electrical neuromorphic devices, those stimulated by optical signals have gained increasing attention due to their realistic sensory simulation. In this work, an optoelectronic neuromorphic device based on a photoelectric memristor with a Bi2FeCrO6/Al-doped ZnO (BFCO/AZO) heterostructure is fabricated that can respond to both electrical and optical signals and successfully simulate a variety of synaptic behaviors, such as STP, LTP, and PPF. In addition, the photomemory mechanism was identified by analyzing the energy band structures of AZO and BFCO. A convolutional neural network (CNN) architecture for pattern classification at the Mixed National Institute of Standards and Technology (MNIST) was used and improved the recognition accuracy of the MNIST and Fashion-MNIST datasets to 95.21% and 74.19%, respectively, by implementing an improved stochastic adaptive algorithm. These results provide a feasible approach for future implementation of optoelectronic synapses.

Neural Network-based Pattern Recognition in the Framework of Edge Computing

Neural network (NN) model has been widely used in pattern recognition (PR), speech recognition, image processing and other fields, but its application in edge computing (EC) environment faces performance and energy consumption problems. This article first introduced the basic structure and training process of NN, including backpropagation algorithms. Then, this article presented a NN modeling approach based on EC, including NN model compression, distributed NN model and knowledge distillation approach. Finally, this article implemented a PR model for the MNIST (Mixed National Institute of Standards and Technology database) dataset and analyzed the experimental results. The experimental outcomes indicated that the presented approach can significantly enhance the performance of the NN model in the EC environment, while ensuring a high recognition accuracy. The NN modeling approach based on EC can reduce the amount of computation and storage of the NN, thus improving the operating efficiency of the NN in the EC environment by 6%-12%. The NN modeling approach based on EC can optimize the performance and efficiency of the NN model in the EC environment, and provide new ideas and approaches for the application of NN in the EC environment.

Comparative analysis of different model depths on convolutional neural network for handwritten digit recognition

In the rapidly advancing field of artificial intelligence, Convolutional Neural Networks (CNNs), as a representative method, have risen to prominence as a pivotal instrument for handling visual data. However, despite their widespread use, the impact of CNN depth on performance remains under-explored. This study delves into this aspect, evaluating the performance of CNN architectures with different depths - two-layer, four-layer, and five-layer - on the MNIST dataset, a version from the National Institute of Standards and Technology, a well-known benchmark dataset for handwritten digit recognition. Experimental results reveal that the four-layer model achieved the highest average accuracy of 99.76%, while the five-layer model, despite its additional complexity, only slightly trailed behind with a 99.73% accuracy rate. However, the five-layer model required a significantly longer training time. In conclusion, while deeper networks can increase accuracy, they can also introduce computational inefficiencies without significant gains in performance. This research provides a better understanding of CNN depth, guiding optimal model selection for image classification tasks.

Abstract 7024: Development of quality control methods for reliable NGS measurement of methylation in cfDNA

Abstract Background and Purpose: Reliable measurement of circulating free genomic DNA (cfDNA) methylation patterns promises a means for cancer early detection as well as treatment response. Use of Next Generation Sequencing (NGS) methods to measure the cfDNA methylome across a large number of potentially methylated CpG sites may enable earlier detection by overcoming stochastic sampling associated with the small specimens obtainable by a blood draw. Wide-spread adoption of these methods for clinical diagnostic applications will be facilitated through improved quality controls that increase reliability. The goal of this study is to assess the performance of the SNAQ-SEQ™ method as a means to establish a limit of detection (LOD) for measurement of methylation at each targeted CpG site. For this purpose, we used test materials that were developed by the National Institute for Standards and Technology (NIST) to mimic methylated cfDNA. Methods: NIST test materials were formulated by physically shearing cell line gDNA then subjecting it to size separation to mimic a size distribution similar to that observed in cfDNA, then mixing sheared DNA in its “native” state of methylation with sheared DNA that was in vitro methylated. Non-sheared NIST test materials also were assessed in these studies. Synthetic spike-in internal standards (IS) were combined with NIST test materials and the combined samples were bisulfite converted, then subjected to PCR-amplicon library preparation. Libraries were sequenced on an Illumina MiSeq instrument and bioinformatic analysis was conducted using the Qiagen CLC Workbench. Results and Conclusions: In initial testing of an off-the-shelf IS for a region of the SOX2 gene, average methylation (preserved C) across 14 targeted CpG sites was 16.4% (0.5%-50%) in the “native” NIST non-sheared sample. This was consistent with the 11.6% methylation across all of SOX2 measured by the NIST. In contrast, in IS corresponding to the targeted SOX2 region the average preserved C was 0.32% (0.03%-0.5%). Use of synthetic internal standards enabled determination of LOD for each individual methylated site, controlling for all known sources of technical error, including inefficient bisulfite conversion, nucleotide substitution error, and alignment error. Studies are now underway that use the NIST test materials to measure methylation at five targets known to be altered in cfDNA from lung cancer subjects, SOX2, CDO1, TAC1, SOX17, and HOXA7. Citation Format: James C. Willey, Erin L. Crawford, Daniel J. Craig, Thomas M. Morrison, Hua-Jun He, James G. Herman. Development of quality control methods for reliable NGS measurement of methylation in cfDNA [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 7024.

Multilevel Ferroelectric Domain Wall Memory for Neuromorphic Computing

AbstractLow‐power and parallel processing Neuromorphic computing that imitates on the human brain's extraordinary data processing and learning capabilities is promising in non‐von Neumann application platforms in the post‐Moore era. Ferroelectric domain walls appearing as nanoscale topological defects in solids exhibit coexisting ordering parameters besides the spontaneous polarization, leading to the emergence of rich physics and electronic effects in the development of neuromorphic electronics that go beyond the conventional CMOS. In this study, the four‐state domain wall memory is developed on LiNbO3 ferroelectric thin films integrated with Si substrates, where each state can be stably manipulated at a specific low voltage, showcasing outstanding fatigue resistance and retention performance. The constructed crossbar array that functions as a kernel for image convolution can be used to implement processing operations like image filtering, sharpness enhancement, and edge detection. Additional simulation from a convolutional neural network model shows 96.98% accuracy on the Modified National Institute of Standards and Technology handwritten digits dataset. The stable multi‐state domain wall memory provides a new approach for computing and promotes research in domain wall electronics to meet future enormous demands of big data.

Development of a novel seized drug screening method utilizing DART-MS and used weigh paper

The conventional analytical scheme for seized drug analysis utilizes quick and cost-effective presumptive testing, such as color tests, followed by more discriminatory techniques, such as gas chromatography-mass spectrometry (GC–MS). Before the seized drug evidence is analyzed, the evidence is weighed using an appropriate weighing matrix and balance. Typically, the weighing matrix is discarded as chemical waste; however, the process of weighing the evidence leads to the deposition of drug residue onto the weighing matrix, which is then amenable to rapid analysis using ambient ionization approaches, such as direct analysis in real time mass spectrometry (DART-MS). This study demonstrates a DART-MS method for the rapid screening of seized drug evidence using filter paper and glassine paper weighing matrices commonly found in forensic laboratories. Validation experiments were performed using filter paper and glassine paper and compared to the results obtained using glass capillaries to assess the effectiveness of each weighing matrix. A total of 40 authentic samples, 20 filter papers and 20 glassine papers, were analyzed and the resulting spectra were searched using the National Institute of Standards and Technology (NIST) DART-MS Forensics Database and Data Interpretation Tool (DIT). When compared to the ground truth GC–MS results, the developed approach had a 90% correct identification rate, with all misidentifications being due to the presence of tablets rather than powder. The developed DART-MS screening method provides the seized drug community with a method to rapidly screen seized drug residue remaining from the weighing process and to improve the overall laboratory efficiency.

Adjustable artificial neuron based on vortex magnetic tunnel junction

In this Letter, we demonstrate an adjustable artificial neuron based on vortex magnetic tunnel junction (MTJ). By applying a bias current to vortex MTJ, the device exhibits splendid characteristics of stochastic switching and nonlinear rectification. The stochastic switching probability induced by spin transfer torque as a function of bias current can simulate sigmoid activation functions. The nonlinear spin-torque microwave rectification through injection locking is similar to a ReLU-like activation function. These two behaviors further are used to perform the recognition of handwritten digits in the Mixed National Institute of Standards and Technology database, with a produced accuracy of up to 93.56% and 93.25%, respectively. Our work provides a potential way for the construction of artificial neuron based on vortex MTJ.

Design of a novel hybrid quantum deep neural network in INEQR images classification

We redesign the parameterized quantum circuit in the quantum deep neural network, construct a three-layer structure as the hidden layer, and then use classical optimization algorithms to train the parameterized quantum circuit, thereby propose a novel hybrid quantum deep neural network (HQDNN) used for image classification. After bilinear interpolation reduces the original image to a suitable size, an improved novel enhanced quantum representation (INEQR) is used to encode it into quantum states as the input of the HQDNN. Multi-layer parameterized quantum circuits are used as the main structure to implement feature extraction and classification. The output results of parameterized quantum circuits are converted into classical data through quantum measurements and then optimized on a classical computer. To verify the performance of the HQDNN, we conduct binary classification and three classification experiments on the MNIST (Modified National Institute of Standards and Technology) data set. In the first binary classification, the accuracy of 0 and 4 exceeds 98%. Then we compare the performance of three classification with other algorithms, the results on two datasets show that the classification accuracy is higher than that of quantum deep neural network and general quantum convolutional neural network.

Enhancing cryptographic robustness with dual key chaining

In this paper, we introduce an advanced mode of operation for block ciphers, named Dual Key Chaining Mode (DKC), aimed at bolstering cryptographic security for safeguarding sensitive information. Building upon the foundations laid by established modes while adhering to guidelines set by the National Institute of Standards and Technology (NIST), DKC innovates through a dual-key mechanism and the generation of highly unpredictable values. This novel approach markedly enhances security, particularly against chosen plaintext attacks, a common vulnerability in traditional modes. Through rigorous mathematical analysis, we demonstrate DKC’s superiority, proving its indistinguishability under chosen plaintext attacks (IND-CPA) and showing that an adversary cannot practically distinguish DKC-encrypted ciphertexts from those produced by a random permutation. Our security proof employs a structured approach, contrasting DKC with conventional modes to highlight its robust defense mechanisms and its capacity to mitigate error propagation, reduce chain dependency, and resist pattern recognition attacks. The DKC mode not only surpasses existing standards in cryptographic security but also offers significant improvements in efficiency and security complexity, making it particularly suited for environments demanding stringent data protection. This study’s findings underscore DKC’s potential as a leading candidate for securing communication channels, financial transactions, and cloud storage services against an array of cryptographic attacks.

An innovative image encryption algorithm enhanced with the Pan-Tompkins Algorithm for optimal security

This study introduces a cutting-edge image encryption algorithm aimed at elevating security standards. The Pan-Tompkins Algorithm (PTA) for key generation is proposed for the first time in this study. Additionally, employing steganography through the Least Significant Bit (LSB) method for embedding keys within the encrypted image enhances secure key distribution, thereby fortifying the encryption process. On the other hand, the integration of advanced algorithms, such as Zigzag scanning, the Affine Image Encryption Algorithm (AA), and the Vigenere Image Encryption Algorithm (VA), constitutes the fundamental innovation of the proposed image encryption algorithm. The proposed algorithm is named PanAAVA:Affine Algorithm and Vigenere Algorithm Encryption with PTA-Based Key Generation. The PanAAVA algorithm ensures unparalleled security by encrypting the positions and values of pixels using AA and VA. Notably, using PTA for key generation marks a distinctive and new key generation method feature of the algorithm. To assess the effectiveness of the PanAAVA, a comprehensive comparative analysis is conducted against well-established encryption methodologies, including Lena, Baboon, Airplane, and Pepper.The PanAAVA demonstrates exceptional proficiency in histogram analysis. The PanAAVA demonstrates a Unified Average Changing Intensity (UACI) of 33.4044%. Additionally, the Number of Pixels Change Rate (NPCR) is measured at 99.7442%, showcasing the algorithm’s effectiveness in inducing significant pixel changes. The proposed algorithm’s Mean Square Error (MSE) is calculated at 3.20679E5%. The proposed algorithm’s Peak Signal to Noise Ratio (PSNR) is recorded at 9.512475. The Key Space Size of the proposed algorithm is measured at 2209. Regarding correlation analysis, the PanAAVA achieves a high correlation score of 7.9996. The proposed algorithm successfully passes the National Institute of Standards and Technology (NIST) analysis, demonstrating a remarkably strong correlation close to 0 and a Structural Similarity Index Measure (SSIM) of 0.9977. Furthermore, regarding quantum communication, the proposed algorithm maintains stable key rates of 47.5 ± 0.8 kHz during the day and 50.9 ± 0.7 kHz at night. Additionally, PanAAVA achieves low Quantum Bit Error Rate (QBER) values of 4.77 ± 0.02, ensuring reliable and secure communication. The PanAAVA also demonstrates robust asymmetries at 49.81 ± 0.02 and 50.14 ± 0.03 for a crystal length of 20 mm. highlighting PanAAVA’s adaptability and effectiveness in different scenarios. PanAAVA outperforms other encryption algorithms concerning performance measurements and comparisons. In conclusion, the PanAAVA emerges as a beacon of superior security capabilities and innovation in image encryption, showcasing the potential to redefine standards in the field.

Spin‐Transfer‐Torque Magnetic Tunnel Junction Nonlinear In‐Sensor Computing Synapse for Improving the Performance of the Feedforward Neural Network

In‐sensor computing architecture has a great advantage especially in massive data sampling, transfer, and processing compared with the separated intelligent sensor systems. However, most of the in‐sensor computing device is proposed based on the traditional neural network model, where the synapse performs linear multiplication of input and weight. This approach fails to make the most use of the nonlinearity of in‐sensor computing devices. Therefore, in this article, first a modified feedforward neural network model with the nonlinear in‐sensor computing synapse (NSCS) located at the input layer is presented, and the backpropagation (BP) algorithm is modified to train the network. Then, the nonlinear characteristics of the NSCS composed of the spin‐transfer‐torque magnetic tunnel junction (STT–MTJ) devices and simple complementary metal‐oxide‐semiconductor (CMOS) circuit are analyzed. Based on the nonlinear response of STT–MTJ NSCS, the small‐scale network with NSCS synapse is experimented on the Modified National Institute of Standards and Technology dataset and compared with the traditional network of the same network size. In the simulation result, it is shown that better performance can be achieved with the STT–MTJ NSCS, including a 2–15 times improvement in convergence speed and a 2.5%–5.1% increase in accuracy.

Impact of HfO2 Dielectric Layer Placement in Hf0.5Zr0.5O2‐Based Ferroelectric Tunnel Junctions for Neuromorphic Applications

AbstractThe use of Hf0.5Zr0.5O2 (HZO) films within hafnia‐based ferroelectric tunnel junctions (FTJ) presents a promising avenue for next‐generation non‐volatile memory devices. HZO exhibits excellent ferroelectric properties, ultra‐thinness, low power consumption, nondestructive readout, and compatibility with silicon devices. In this study, Mo/HZO/n+ Si devices are investigated, incorporating a 1 nm HfO2 dielectric layer at the top and bottom of the HZO ferroelectric layer. Comparing the FTJ device configurations, it is observed that the metal‐ferroelectric‐dielectric‐semiconductor (MFIS) outperforms the metal‐dielectric‐ferroelectric‐semiconductor (MIFS) in terms of ferroelectricity, displaying a high 2Pr value of ≈69 µC cm−2. Additionally, MFIS exhibits lower leakage current, higher tunneling electro‐resistance ratio, and a thin dead layer during short pulse switching, as confirmed through DC double sweeping of I−V characteristics. The modified half‐bias scheme demonstrates a maximum array size of 191 for MFIS, showcasing its superior performance over MIFS. Synaptic characteristics, including potentiation, depression, paired‐pulse facilitation, spike‐rate‐dependent plasticity, and excitatory postsynaptic current, are measured using MFIS, highlighting its outstanding ferroelectric properties. As a physical reservoir, the FTJ device implements 16 states of 4 bits in reservoir computing. Finally, pattern recognition using a deep learning neural network achieves high accuracy with using the Modified National Institute of Standards and Technology dataset.