Neuromorphic computing takes inspiration from the brain to create energy-efficient hardware for information processing, capable of highly sophisticated tasks. Systems built with standard electronics achieve gains in speed and energy by mimicking the distributed topology of the brain. Scaling-up such systems and improving their energy usage, speed and performance by several orders of magnitude requires a revolution in hardware. We discuss how including more physics in the algorithms and nanoscale materials used for data processing could have a major impact in the field of neuromorphic computing. We review striking results that leverage physics to enhance the computing capabilities of artificial neural networks, using resistive switching materials, photonics, spintronics and other technologies. We discuss the paths that could lead these approaches to maturity, towards low-power, miniaturized chips that could infer and learn in real time. Neuromorphic computing takes inspiration from the brain to create energy-efficient hardware for information processing, capable of highly sophisticated tasks. Including more physics in the algorithms and nanoscale materials used for computing could have a major impact in this field.

Physics for neuromorphic computing

Conventional content addressable memory (BCAM and TCAM) uses specialized 10T/16T bit cells that are significantly larger than 6T SRAM cells. A new BCAM/TCAM is proposed that can operate with standard push-rule 6T SRAM cells, reducing array area by 2–5× and allowing reconfiguration of the SRAM as a CAM. In this way, chip area and overall capacitance can be reduced, leading to higher energy efficiency for search operations. In addition, the configurable memory can perform bit-wise logical operations: “AND” and “NOR” on two or more words stored within the array. Thus, the configurable memory with CAM and logical function capability can be used to off-load specific computational operations to the memory, improving system performance and efficiency. Using a 6T 28 nm FDSOI SRAM bit cell, the 64×64 (4 kb) BCAM achieves 370 MHz at 1 V and consumes 0.6 fJ/search/bit. A logical operation between two 64 bit words achieves 787 MHz at 1 V.

A 28 nm Configurable Memory (TCAM/BCAM/SRAM) Using Push-Rule 6T Bit Cell Enabling Logic-in-Memory

The invention includes semiconductor constructions containing vertically-extending pillars, and methods for forming such constructions. The vertically-extending pillars can be incorporated into transistor devices, and can contain vertically-extending channel regions of the transistor devices. The transistor devices can be incorporated into integrated circuitry, and in some aspects are incorporated into memory constructions, such as, for example, dynamic random access memory (DRAM) constructions.

Methods of forming semiconductor constructions

Artificial intelligence (AI) has the ability of revolutionizing our lives and society in a radical way, by enabling machine learning in the industry, business, health, transportation, and many other fields. The ability to recognize objects, faces, and speech, requires, however, exceptional computational power and time, which is conflicting with the current difficulties in transistor scaling due to physical and architectural limitations. As a result, to accelerate the progress of AI, it is necessary to develop materials, devices, and systems that closely mimic the human brain. In this work, we review the current status and challenges on the emerging neuromorphic devices for brain-inspired computing. First, we provide an overview of the memory device technologies which have been proposed for synapse and neuron circuits in neuromorphic systems. Then, we describe the implementation of synaptic learning in the two main types of neural networks, namely the deep neural network and the spiking neural network (SNN). Bio-inspired learning, such as the spike-timing dependent plasticity scheme, is shown to enable unsupervised learning processes which are typical of the human brain. Hardware implementations of SNNs for the recognition of spatial and spatio-temporal patterns are also shown to support the cognitive computation in silico. Finally, we explore the recent advances in reproducing bio-neural processes via device physics, such as insulating-metal transitions, nanoionics drift/diffusion, and magnetization flipping in spintronic devices. By harnessing the device physics in emerging materials, neuromorphic engineering with advanced functionality, higher density and better energy efficiency can be developed.

https://iopscience.iop.org/article/10.1088/1361-6528/ab554b/pdf

Emerging neuromorphic devices.

Increased process variability presents a major challenge for future SRAM scaling. Fast and accurate validation of SRAM read stability and writeability margins is crucial for estimating yield in large SRAM arrays. Conventional SRAM read/write metrics are characterized through test structures that are able to provide limited hardware measurement data and cannot be used to investigate cell bit fails in functional SRAM arrays. This work presents a method for large-scale characterization of read stability and writeability in functional SRAM arrays using direct bit-line measurements. A test chip is implemented in a 45 nm CMOS process. Large-scale SRAM read/write metrics are measured and compared against conventional SRAM stability metrics. Results show excellent correlation to conventional SRAM read/write metrics as well as VMIN measurements near failure.

Large-Scale SRAM Variability Characterization in 45 nm CMOS

A 1T cell for high-density eDRAM has been successfully developed on bulk silicon substrate for the first time. The device architecture is fully compatible with CMOS logic process integration, allowing very low chip cost for SoC applications. Experimental results show a retention time over 1s at 25/spl deg/C and 100ms at 85/spl deg/C, which is compatible with eDRAM requirements. Non-destructive readout is experimentally demonstrated at 85/spl deg/C. The integration of the memory cell in a matrix arrangement is evaluated. Gate and drain disturb are characterized, showing enough disturb margins for memory operations.

A one transistor cell on bulk substrate (1T-Bulk) for low-cost and high density eDRAM

A capacitor-less DRAM cell on very thin film (Tsi=16nm) and short gate length (Lg=75nm) fully depleted (FD) device is demonstrated for the first time. Memory operations mechanisms are presented and retention time compatible to eDRAM requirements is measured at 85/spl deg/C. Nondestructive reading is demonstrated at 25/spl deg/C and disturb margins are deeply investigated, showing the possibility of matrix integration. This study is then extended to another type of FD device: the very promising double gate architecture.

A capacitor-less DRAM cell on 75nm gate length, 16nm thin fully depleted SOI device for high density embedded memories

Mechanisms of charge modulation in the floating body of triple-well nMOSFET capacitor-less DRAMs

In this paper, we report on parasitic bipolar conduction occurring in floating-body effect based capacitor-less DRAMs. A way to include these effects into a previously developed model is presented. The enhanced model is then compared with electrical data realized on triple-well nMOSFET devices within the 26/spl deg/C-100/spl deg/C temperature range.

Further insight into the physics and modeling of floating-body capacitorless DRAMs

A one transistor DRAM cell realized on bulk substrate (lT-Bulk) with CMOS 90nm platform is presented for the first time. The device fabrication is fully compatible with logic process integration and includes only few additional steps, thus making this IT cell very attractive for low-cost embedded memories. Very scaled devices were fabricated with a gate length down to 80nm and several gate oxide thicknesses: their performances in terms of memory effect amplitude, retention time and disturb margins are very promising for future high density eDRAM.

R. Ranica

Papers

A one transistor cell on bulk substrate (1T-Bulk) for low-cost and high density eDRAM

A capacitor-less DRAM cell on 75nm gate length, 16nm thin fully depleted SOI device for high density embedded memories

Mechanisms of charge modulation in the floating body of triple-well nMOSFET capacitor-less DRAMs

Further insight into the physics and modeling of floating-body capacitorless DRAMs

Scaled IT-Bulk devices built with CMOS 90nm technology for low-cost eDRAM applications