Microscopic Modeling of Surface Roughness Scattering in Inversion Layers of MOSFETs Based on Ando's Linear Model

This paper proposes a microscopic model for surface roughness scattering in MOSFET inversion layers that accounts for the stochastic nature of roughness position, revealing that the scattering rate is intrinsically nonlocal and that conventional Fermi's golden rule approaches tend to underestimate SR-limited mobility under strong fields and low electron energies.

The Gist

The Problem: The "Bumpy Road" Dilemma

Imagine you are driving a high-performance sports car (an electron) on a highway (the inversion layer of a transistor). To get the best speed and efficiency, you want the road to be perfectly smooth.

However, in the microscopic world of computer chips, the "road" isn't smooth. It has tiny, microscopic bumps called Surface Roughness (SR). These bumps act like potholes that scatter your car, making it wobble, slow down, and lose energy. This is what limits the speed (mobility) of your computer processor.

For decades, scientists have used a mathematical "map" (called Ando’s Linear Model) to predict how these bumps affect the car. But there was a huge problem: The math didn't match reality. When scientists looked at the bumps through super-powerful microscopes (TEM), the bumps looked small and manageable. But when they used the old math to predict speed, the math acted as if the bumps were massive and much more destructive than they actually were. It was like saying a tiny pebble would flip a car over.

---

The Solution: The "Blurry Camera" Approach

Nobuyuki Sano’s paper proposes a new way to look at these bumps. He suggests that the old math failed because it treated the "edge" of the road as a single, sharp, frozen line.

In reality, because of the weird laws of quantum physics, that edge isn't a sharp line—it’s more like a blurry, vibrating zone.

The Analogy: The Shaky Handheld Camera
Imagine you are trying to take a photo of a fence.

• The Old Model: Assumes the camera is on a tripod. The fence is a perfectly sharp line. If the fence moves an inch, the math treats it as a massive, sudden jump.
• The New Model (Sano’s): Assumes you are holding the camera in your hand. The fence isn't a sharp line; it’s a "probability" of where the fence might be. Because the camera is slightly shaky, the "edge" of the fence is a soft, fuzzy blur.

By introducing this "blurriness" (which he calls a probability density of roughness position), the math finally aligns with what the microscopes actually see. Suddenly, the "pebbles" in the math behave like pebbles again, and the discrepancy between theory and experiment disappears.

---

The Big Discovery: The "Quantum Fog" Effect

Sano didn't just fix the map; he discovered a new way the bumps behave.

He used a advanced mathematical framework (the Green’s Function scheme) to show that when electrons are moving slowly or when the electric field is very strong, the scattering isn't just a simple "hit and bounce" event. Instead, it becomes "nonlocal."

The Analogy: The Foggy Night

• Old Way (Fermi’s Golden Rule): Imagine driving through a single, sharp obstacle. You hit a rock, you bounce. Simple.
• New Way (Self-Consistent Scheme): Imagine driving through a thick, heavy fog. You don't just hit one thing; the "environment" itself feels thick and resistant. The scattering becomes a continuous, complex interaction with the "fog" of the surface.

Because the old math ignored this "foggy" effect, it was too pessimistic. It predicted that electrons would be much slower than they actually are. Sano’s new model shows that electrons can actually navigate these bumpy surfaces more effectively than we previously thought.

---

Why does this matter?

As we try to make computer chips smaller and smaller (moving toward "nanowires" and "nanosheets"), these surface bumps become the biggest obstacle to speed.

By providing a more accurate "map" of how these bumps work, Sano has given engineers a better toolkit to design the next generation of ultra-fast, tiny electronics. We can now predict exactly how much "speed" we will lose to the bumps, allowing us to build better "roads" for the future of computing.

Technical Summary

Technical Summary: Microscopic Modeling of Surface Roughness Scattering in Inversion Layers of MOSFETs Based on Ando’s Linear Model

Author: Nobuyuki Sano (University of Tsukuba)
Date: April 27, 2026

---

1. Problem Statement

Surface Roughness (SR) scattering is the primary mechanism limiting electron mobility in bulk-MOSFETs under strong inversion. Historically, theoretical models have relied on Ando’s linear model, which assumes the scattering potential is proportional to the interface deviation $\Delta(r)$.

However, a long-standing discrepancy exists: the roughness parameters ($\Delta_A$ and $\Lambda_A$) required to match experimental mobility data using the linear model are significantly larger than those measured directly via Transmission Electron Microscopy (TEM). While recent "nonlinear" models have attempted to fix this by addressing the discontinuity of electrostatic potential derivatives, they often move away from the simplicity of the linear model. The author identifies three specific gaps in current theory:

1. Scale Mismatch: $\Delta(r)$ is treated as a macroscopic landscape, whereas SR is an atomistic, short-range phenomenon.
2. Mathematical Invalidity: The Taylor expansion used in the linear model is technically invalid due to the discontinuity of spatial derivatives at the semiconductor/dielectric interface.
3. Approximation Limits: Conventional models use Fermi’s Golden Rule (FGR), which fails to account for the nonlocal (nondiagonal) nature of short-range scattering in a quantum-mechanical framework.

2. Methodology

The author proposes a microscopic SR model that reconciles Ando’s linear model with atomistic reality through the following steps:

• Stochastic Interface Modeling: Instead of a continuous macroscopic function, the author introduces a probability density $p_X(r)$ of finding an atomistic roughness at each atomic site. The total roughness $\Delta(r)$ is modeled as the sum of deviations at every atomic site, effectively treating the interface position as a stochastic variable.
• Parameter Mapping: The model establishes a mathematical bridge between the microscopic parameters (averaged deviation $\bar{\Delta}_m$ and standard deviation $\sigma_m$) and the experimental TEM parameters ($\Delta_A$ and $\Lambda_A$). This allows the model to use physically accurate, TEM-derived values.
• Green’s Function Framework: To move beyond the limitations of Fermi’s Golden Rule, the author employs the Nonequilibrium Green’s Function (NEGF) scheme. The retarded self-energy $\hat{\Sigma}^r_{sr}(E)$ is derived self-consistently. This approach accounts for the fact that short-range scattering is intrinsically nonlocal (nondiagonal) regarding subband indices and position.

3. Key Contributions

• Resolution of the Parameter Discrepancy: By treating the interface position stochastically at the atomic level, the model successfully uses TEM-measured roughness parameters to predict realistic electron mobility, eliminating the need for the "inflated" parameters used in previous linear models.
• Self-Consistent Scattering Rate: The derivation of a self-consistent scattering rate formula (Eq. 22 and 28) provides a more rigorous treatment of the time-energy uncertainty principle in high-scattering regimes.
• Identification of Nonlocality: The paper proves that SR scattering is inherently nonlocal, a factor previously overlooked in standard transport calculations.

4. Results and Discussion

• Deviation from Fermi’s Golden Rule: Numerical results show that the self-consistent scattering rate deviates significantly from the FGR prediction, particularly in regimes of strong effective electric fields and low electron energies. In these regimes, the FGR tends to overestimate the scattering rate (and thus underestimate mobility).
• Mobility Predictions:
  • The self-consistent model predicts higher electron mobility than the conventional FGR-based model.
  • When applied to Si-MOSFETs, the model using TEM-consistent parameters ($\bar{\Delta}_m = 1.3$ nm, $\sigma_m = 0.7$ nm) provides an excellent fit to experimental mobility data.
• Integrand Behavior: The study shows that under strong fields, the energy-dependent integrand broadens, reflecting the significant impact of the time-energy uncertainty principle when scattering is intense.

5. Significance

This work provides a theoretically sound and physically consistent bridge between atomistic interface observations (TEM) and macroscopic device performance (mobility). By restoring the validity of the "linear" approach through a microscopic stochastic lens, it offers a computationally efficient yet highly accurate method for modeling MOSFETs. This is particularly critical for the design of next-generation, ultra-scaled devices like nanowires and nanosheets, where surface effects dominate transport properties.