Peking University Analog-Switching Resistive Random Access Memory (RRAM) Verilog-A model 3.1.0
The Peking University Analog-switching RRAM physical model can capture the pulse conductance updates of analog RRAM devices rapidly and accurately. The model is described by Verilog-A and can be embedded in SPICE and Cadence for circuit simulations.
Listed in Compact Models
Additional materials available
Version 3.1.0 - published on 01 Feb 2021 doi:10.21981/MRFT-C373 - cite this
Licensed under NEEDS Modified CMC License according to these terms
Description
Different from the previously published RRAM_v_2_1 from Peking University (https://nanohub.org/publications/284/2), RRAM_v_3_1 is designed for fast and accurate simulation of analog metal-oxide based RRAM devices with intermediate modulation layer [1]. The underlying influences of electrical and thermal properties of IML on the analog switching behavior is dominant in improving the linearity [2] [3]. The model captures the pulse conductance update of analog RRAM devices with microscopic physical descriptions including the changes of oxygen vacancies (Vo) and oxygen ions (O2−) .
The proposed physical model of filamentary analog RRAM including the resistive switching (RS) model and conduction model. The RS behavior involves the processes of oxygen vacancy (Vo) generation and recombination, oxygen ions (O2−) absorption and release from IML. The conductive filament (CF) consists of one RS region and one Vo rich (VR) region. The electrical properties of the filament are determined by the dimension of the filament [4] [5]. For SET process, the increased percentage of Vo in RS regions is attributed to the Vo generation, where CV is the Vo percentage denoting the number of Vo divided by the total number of lattice oxygen. For RESET process, the release of O2− from IML and the recombination of Vo and O2− lead to the decrease of CV . The changes of CV are directly associated with the conductance of CF, especially associated with the conductance of RS region. Parameter extraction method can be referred to [6].
In VR region, the conductivity is assumed as metallic-like conduction.
In RS region, the effective conductivity is deduced by adopting the Maxwell-Garnett model [7] based on effective medium theory (EMT), which depends on three parameters, namely, the conductivity of intrinsic HfO2, the conductivity of VR region, and Vo percentage (CV ). Here the CF region is regarded as the intrinsic RS material with the changeable Vo. The main conduction mechanism of the intrinsic RS material without traps involved is assumed as Fowler-Nordheim tunnelling [8] and the con ductivity is a function of electric field.
In IML region, The conductivity of IML is calculated by considering the O2− percentage of IML according to EMT.
Model Release Components
Peking University Analog-Switching Resistive Random Access Memory (RRAM) Verilog-A model 3.1.0 Verilog-A(VA | 8 KB)
Peking University Analog-Switching Resistive Random Access Memory (RRAM) Verilog-A model 3.1.0 Benchmarks(ZIP | 1 KB)
Peking University Analog-Switching Resistive Random Access Memory (RRAM) Verilog-A model 3.1.0 Parameters(PDF | 85 KB)- Peking University Analog-Switching Resistive Random Access Memory (RRAM) Verilog-A model 3.1.0 Manual(PDF | 435 KB)
- License terms
Cite this work
Researchers should cite this work as follows:
- Han, L.; Cai, L.; Kang, J.; Liu, X.; Huang, P. (2021). Peking University Analog-Switching Resistive Random Access Memory (RRAM) Verilog-A model. (Version 3.1.0). nanoHUB. doi:10.21981/MRFT-C373