RASP: Reliability ab initio simulation package of MOSFETs based on all-state model

The paper introduces RASP, an ab initio simulation package that utilizes an all-state model to accurately predict defect-induced reliability degradation in scaled MOSFETs by systematically accounting for all possible defect configurations and transition pathways in amorphous gate dielectrics.

The Gist

Here is an explanation of the paper "RASP: Reliability ab initio simulation package of MOSFETs based on all-state model," translated into simple, everyday language with creative analogies.

The Big Picture: The Tiny Transistor's "Aging" Problem

Imagine your smartphone is a bustling city made of billions of tiny switches called transistors. These switches turn electricity on and off to process your photos, texts, and videos. As technology advances, these switches are getting smaller and smaller—now they are smaller than a virus.

However, as these switches shrink, they become fragile. Over time, they start to "age" and get tired. This aging causes the switch to turn on slower or not as strongly as it used to. In the world of electronics, this is called reliability degradation. If too many switches get tired, your phone slows down, crashes, or stops working.

The main culprit? Tiny defects (flaws) in the insulating layer (the "gate oxide") that sits on top of the switch. Think of these defects as holes in a dam. Sometimes water (electricity) leaks through them, messing up the pressure (voltage) needed to run the city.

The Old Way of Thinking: The "Two-Door" House

For a long time, scientists tried to predict how these defects would age the chips using simple models. They imagined a defect as a house with only two doors:

1. Door A: The house is empty (neutral).
2. Door B: The house has a guest (charged).

They thought the defect just walked back and forth between these two doors. If a guest entered, the switch slowed down. If the guest left, it recovered.

The Problem: Real life is messy. In the chaotic, disordered world of modern materials (like amorphous glass), a defect isn't just a house with two doors. It's a mansion with hundreds of rooms, secret passages, and different furniture arrangements. A defect can be in a "ground state" (the main living room) or a "metastable state" (a cozy attic, a basement, or a sunroom). It can jump between these rooms in complex ways that the old "two-door" model completely missed.

Because the old models were too simple, they often told engineers, "Don't worry, this defect is harmless," when in reality, it was a ticking time bomb.

The New Solution: RASP (The "All-Rooms" Simulator)

The authors of this paper developed a new software package called RASP (Reliability Ab initio Simulation Package).

Think of RASP as a super-smart, high-tech tour guide for that defect mansion. Instead of assuming there are only two doors, RASP maps out every single room, every hallway, and every secret passage the defect can possibly take.

Here is how RASP works, broken down into four simple steps:

1. The Map Maker (Device Electrostatics Module)

Before the tour begins, RASP draws a perfect map of the city (the transistor). It calculates exactly how the "pressure" (voltage) and "traffic" (electrons) are moving around at any given moment. It knows exactly where the walls are and how strong the electric fields are.

2. The Speed Calculator (Transition Rate Module)

Now, RASP asks: "How fast can the defect move from one room to another?"

• Thermal Transitions: Sometimes the defect just wanders from the living room to the attic because it's warm (thermal energy).
• NMP Transitions: Sometimes the defect needs to catch a "guest" (an electron or hole) to move. This is like a complex dance where the defect has to vibrate in just the right way to grab the guest.
• The Magic Trick: Calculating these dance moves for millions of defects usually takes forever. RASP uses a clever math trick (Fourier transforms and interpolation) to do the math super fast. It's like having a GPS that predicts traffic jams in milliseconds instead of hours.

3. The Crowd Controller (Defect Occupation Module)

Now RASP simulates the chaos. It asks: "If I turn on the switch for 1 second, where is the defect? If I turn it off for 10 seconds, where does it go?"
It uses a method called a Markov Chain (think of it as a giant flowchart of probabilities). It tracks every single defect, calculating the chance it is in the attic, the basement, or the living room at any specific moment. It solves a massive puzzle where every piece is constantly moving.

4. The City Planner (Device Reliability Module)

Finally, RASP looks at the whole picture. "Okay, if 10,000 defects are currently hiding in the attic, how much does that slow down the whole city?"
It calculates the Threshold Voltage Shift (how much the switch has slowed down). It can do this in two ways:

• Level 1 (Weak Link): Assumes the defects are few and don't change the city's layout much.
• Level 2 (Strong Link): Assumes there are so many defects they actually change the city's layout, which changes how the defects move. RASP loops back and forth to get the perfect answer.

The Big Discovery: Oxygen Vacancies are Villains

The researchers used RASP to study a specific type of defect called an Oxygen Vacancy (a missing oxygen atom in the glass).

• The Old View: Previous models said, "These oxygen vacancies are too deep underground to ever catch a guest. They are harmless."
• The RASP View: By looking at all the rooms (configurations) and all the paths, RASP found that these vacancies are actually very active. They have secret shortcuts (metastable states) that allow them to catch guests easily and cause the transistor to age.

The Result: RASP proved that oxygen vacancies are a major cause of NBTI (Negative Bias Temperature Instability), a specific type of aging that kills PMOS transistors. The old models missed this because they didn't look at the "attic" and "basement" rooms.

Why This Matters

Imagine you are building a bridge.

• Old Model: You check the main pillars and say, "Looks good."
• RASP Model: You check every bolt, every crack, and every vibration pattern. You realize, "Oh, the wind makes a tiny crack in the foundation wiggle, which loosens a bolt, which eventually collapses the bridge."

RASP allows engineers to predict exactly how long a chip will last before it fails. This helps them design better, longer-lasting phones and computers. It moves us from guessing to knowing, ensuring that the technology in our pockets doesn't break down before we're ready to upgrade.

In short: RASP is the ultimate simulator that stops us from underestimating the tiny, chaotic flaws in our electronics, ensuring our devices stay reliable for years to come.

Technical Summary

Here is a detailed technical summary of the paper "RASP: Reliability ab initio simulation package of MOSFETs based on all-state model."

1. Problem Statement

As transistor technology scales into the sub-10 nm regime, device reliability has become a critical bottleneck. Key degradation mechanisms such as Negative Bias Temperature Instability (NBTI), Random Telegraph Noise (RTN), and Time-Dependent Dielectric Breakdown (TDDB) are driven by defect generation and charge trapping/detrapping in gate dielectrics.

Current simulation approaches face significant limitations:

• Oversimplified Models: Conventional reliability models (e.g., the two-state and four-state models) rely on a "bistability" assumption, considering only a limited number of defect configurations (ground and metastable states).
• Amorphous Complexity: Real gate dielectrics (like amorphous SiO₂) possess long-range disorder and low symmetry. Defects (e.g., Oxygen Vacancies, $V_O$) in these materials exhibit a vast array of structural configurations and transition pathways that simple bistable models fail to capture.
• Accuracy Gap: Previous studies using limited models often excluded $V_O$ as a major source of NBTI, leading to potential misidentification of defect origins and inaccurate predictions of circuit aging.
• Computational Challenge: There is a lack of a unified tool that can efficiently integrate ab initio defect parameters with device-level electrostatics to simulate the complex, time-dependent competition among all possible defect states.

2. Methodology: The RASP Package and All-State Model

The authors developed RASP (Reliability Ab initio Simulation Package), a software framework designed to implement the "all-state model." This model systematically considers all possible defect configurations and all transition pathways (both Nonradiative Multiphonon (NMP) and thermal transitions) in amorphous gate dielectrics.

RASP consists of four integrated modules:

1. Device Electrostatics Module:

   • Solves the gate voltage equation and Poisson equation to determine band diagrams, electrostatic potential distributions, and carrier concentrations under various bias conditions.
   • Accounts for the coupling between charged defects and the internal electric field (self-consistent iteration).

2. Transition Rate Module:

   • Carrier Tunneling: Calculates tunneling probabilities using the WKB approximation.
   • NMP Transitions: Computes carrier capture/emission rates using Fermi's Golden Rule. To overcome the computational cost of calculating lineshape functions (LSF) for thousands of defects, RASP employs a Fourier transform method combined with 2D interpolation over the ($\Delta E$, $\Delta Q$) parameter space. This allows for rapid, high-accuracy rate calculations (10,000 defects in ~87 ms).
   • Thermal Transitions: Calculates structural transitions between configurations of the same charge state using Transition State Theory (TST).

3. Defect Occupation Module:

   • Models defect kinetics as a Continuous-Time Markov Chain (CTMC).
   • Solves the Master Equations to determine the time-dependent probability of a defect occupying each specific state (charge and configuration) under competing transition pathways.
   • Handles both steady-state and transient operating conditions using matrix exponentiation.

4. Device Reliability Module:

   • Quantifies the impact of defect occupation on macroscopic parameters, specifically the threshold voltage shift ($\Delta V_{th}$).
   • Offers two simulation schemes:
     • LEVEL 1 (Weak Coupling): Assumes low defect concentration; calculates $\Delta V_{th}$ directly from trapped charge without re-solving Poisson's equation at every step.
     • LEVEL 2 (Strong Coupling): For high defect concentrations; performs self-consistent iteration between defect charge states and electrostatic potential distributions to capture non-linear feedback effects.

3. Key Contributions

• The All-State Model: A paradigm shift from limited-state models to a comprehensive framework that accounts for the full structural diversity of defects in amorphous materials, including multiple metastable configurations and complex transition networks.
• Computational Efficiency: Development of a Fourier transform-based LSF calculation method with 2D interpolation, enabling the simulation of tens of thousands of defects with high accuracy and speed, making device-level simulation feasible.
• Site-Resolved Simulation: Unlike previous statistical approaches, RASP evaluates the contribution of defects at each specific oxygen site individually, capturing the site-dependent variations in energy levels and relaxation parameters inherent to amorphous structures.
• Unified Framework: Integration of ab initio defect physics (via DASP/doped inputs) with macroscopic device electrostatics in a single package.

4. Results

The authors applied RASP to simulate NBTI in Si/SiO₂ MOSFETs caused by Oxygen Vacancies ($V_O$):

• Re-evaluation of $V_O$ Role: Contrary to previous conclusions based on four-state models (which deemed $V_O$ inactive for NBTI), the all-state model simulation reveals that $V_O$ is a non-negligible source of NBTI.
• Mechanism Discovery: The simulation showed that while ground-state transitions might have high barriers, metastable configurations provide alternative pathways with lower activation energies. These pathways allow for efficient hole capture and emission, significantly contributing to $\Delta V_{th}$.
• Defect Classification: $V_O$ defects were categorized into three types based on their impact:
  1. Inactive Traps: Deep levels with negligible capture rates.
  2. Quasi-permanent Traps: Moderate capture but very slow emission, causing long-term drift.
  3. Fast Transient Traps: Shallow levels with fast capture/emission, causing rapid initial shifts and recovery.
• Experimental Validation: The total simulated $\Delta V_{th}$ (summing contributions from all defect sites) showed excellent agreement with experimental data, validating the necessity of the all-state approach.
• Kinetic Analysis: The model successfully reproduced the distinct time constants and voltage dependencies observed in Time-Dependent Defect Spectroscopy (TDDS) experiments, distinguishing between "fixed" and "switching" trap behaviors through the interplay of thermal and NMP transitions.

5. Significance

• Correcting Historical Misconceptions: The work overturns the long-held belief that $V_O$ is not a primary driver of NBTI in SiO₂, demonstrating that the exclusion of metastable states in previous models led to erroneous conclusions.
• Design Tool for Reliability: RASP provides a powerful, physics-based tool for the semiconductor industry to predict circuit aging and reliability at advanced technology nodes (GAAFETs, CFETs) where defect variability is extreme.
• General Applicability: While demonstrated on $V_O$, the all-state model and RASP framework are applicable to other complex defects (e.g., hydrogen-related defects) and other amorphous dielectrics (e.g., HfO₂, Al₂O₃), offering a universal approach to reliability physics.
• Bridging Scales: It successfully bridges the gap between atomic-scale ab initio calculations and circuit-level reliability assessment, enabling a more accurate understanding of the physical origins of device degradation.