TCAD modeling of neuromorphic systems based on ferroelectric tunnel junctions
- PDF / 1,313,778 Bytes
- 6 Pages / 595.276 x 790.866 pts Page_size
- 104 Downloads / 309 Views
TCAD modeling of neuromorphic systems based on ferroelectric tunnel junctions Yu He1 · Wei‑Choon Ng1 · Lee Smith1
© Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract A new compact model for HfO2-based ferroelectric tunnel junction (FTJ) memristors is constructed based on detailed physical modeling using calibrated TCAD simulations. A multi-domain configuration of the ferroelectric material is demonstrated to produce quasi-continuous conductance of the FTJ. This behavior is demonstrated to enable a robust spike-timing-dependent plasticity-type learning capability, making FTJs suitable for use as synaptic memristors in a spiking neural network. Using both TCAD–SPICE mixed-mode and pure SPICE compact model approaches, we apply the newly developed model to a crossbar array configuration in a handwritten digit recognition neuromorphic system and demonstrate an 80% successful recognition rate. The applied methodology demonstrates the use of TCAD to help develop and calibrate SPICE models in the study of neuromorphic systems. Keywords TCAD · Ferroelectric tunnel junction · Synapse · Memristor · Spiking neural network
1 Introduction Neuromorphic computing has emerged as one of the most effective options in achieving both low-power and highly parallel neural network-type tasks [1]. A typical neuromorphic system is built on neuron- and synapse-interconnected circuits. To realize an electrical synapse that emulates a biological synapse, a continuous conductance is required. The memristor, a nanoscale nonvolatile memory element, is a natural device for the synapse. Many attempts in realizing memristor and neuromorphic systems with various nonvolatile memory devices, such as phase change memory (PCM) [2, 3], resistive memory (RRAM) [4–6], spin transfer torque memory (STT–RAM) [7], ferroelectric memory (FeRAM) [8, 9], and ferroelectric tunnel junction (FTJ) [10–12], have been reported. Among these, the FTJ [13] memristor has the advantages of low power consumption, high endurance,
* Yu He [email protected] Wei‑Choon Ng [email protected] Lee Smith [email protected] 1
Synopsys Inc., Mountain View, CA 94043, USA
high operation speed, high ON/OFF ratio, and a simple sandwiched structure. The recent demonstration of CMOSfriendly HfO2-based FTJ memory applications [11] makes it a promising candidate as a memristor in neuromorphic system. Neuromorphic system simulation typically adopts analytical or SPICE-level models to describe the behavior of the synapses [14–17]. These models aim to relate the conductance change to the physics and chemistry behind various types of memristors [18–21]. Despite many successes, the lack of actual process and physical structure makes them less helpful in exploring new device options or device variations. TCAD-level modeling, on the other hand, offers the capability to simulate device operation and semiconductor processing accurately. It has been widely used in industry and has shown great success in exploring new device structures and optimizing device performa
Data Loading...
