研究目的
To investigate the impact of dopant diffusion on random dopant fluctuation in Si nanowire field-effect transistors (FETs) using statistical quantum transport simulations.
研究成果
The statistical quantum transport simulations demonstrate that dopant diffusion into the channel region significantly increases variability in Si nanowire FETs, particularly in OFF-state current and threshold voltage. As the diffusion length increases (modeled by higher standard deviations in Gaussian doping profiles), the yield decreases dramatically, with total yield dropping from 65% for no diffusion (σ=0.0 nm) to only 2% for significant diffusion (σ=2.0 nm). This highlights the importance of minimizing thermal budget during fabrication to reduce variability problems.
研究不足
The study is based on simulations and does not involve experimental validation. The models used, such as the effective mass approximation and Gaussian doping profiles, may have simplifications that do not capture all real-world complexities. The specific device dimensions and parameters may limit generalizability to other nanowire FET designs. The rejection scheme for generating random discrete dopants assumes a Poisson distribution, which might not fully represent actual dopant distributions in fabricated devices.
1：Experimental Design and Method Selection:
The study uses statistical quantum transport simulations based on the Non-Equilibrium Green's Function (NEGF) formalism. An effective mass Hamiltonian is employed, with effective masses extracted from tight-binding band structure calculations. Dopant diffusion is modeled using Gaussian doping profiles with different standard deviations. Random discrete dopants are generated using a rejection scheme considering the 3D atomic arrangement of the nanowire structures.
2：Sample Selection and Data Sources:
The simulations are performed on n-type gate-all-around Si nanowire FETs with a 3×3 nm2 square cross-section. The devices have specific dimensions: source/drain length of 28 nm, gate length of 9 nm, effective oxide thickness of 1 nm, and doping concentrations of 10^15 cm-3 (p-type) in the channel and 10^20 cm-3 (n-type) in the source/drain regions. Six different standard deviation values (σ) for the Gaussian doping profiles are considered: 0.0, 0.25, 0.5, 1.0, 2.0, and 3.5 nm.
3：0, 25, 5, 0, 0, and 5 nm.
List of Experimental Equipment and Materials:
3. List of Experimental Equipment and Materials: The simulations are conducted using software tools: QuantumATK from Synopsys Quantumwise for tight-binding band structure calculations and Glasgow Nano-Electronic Simulation Software (NESS) for transport simulations. No physical equipment is mentioned.
4：Experimental Procedures and Operational Workflow:
The band structure of the Si nanowire is calculated using the sp3d5s* tight-binding method. Effective masses are extracted and used in the effective mass Hamiltonian. Poisson and coupled mode-space NEGF transport equations are solved self-consistently to calculate electron density and current. For each σ value, 100 devices with random discrete dopants are simulated, and approximately 84 well-converged devices are used for statistical analysis. Drain current vs. gate voltage characteristics, threshold voltage, ON-state current, OFF-state current, and yields are computed.
5：Data Analysis Methods:
Statistical analysis is performed on the simulation results, including calculation of mean and median values, correlation plots between parameters (e.g., ION vs. VTH, ION vs. IOFF), and yield calculations with a 25% margin to target values. The variability is assessed in terms of OFF-state current, threshold voltage, and ON-state current.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
