🔬 materials science

State aligned graph wavelet fingerprints of electron trapping in amorphous Si $_3$ N $_4$

This paper introduces state-aligned graph-wavelet fingerprints to quantitatively resolve and classify electron trapping mechanisms in amorphous Si $_3$ N $_4$ by simultaneously linking electronic identity, multiscale structural reorganization, and energetic stabilization across diverse trapping scenarios.

Original authors: Jonathon Cottom, Emilia Olsson

Published 2026-10-06

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Jonathon Cottom, Emilia Olsson

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Inside the tiny, invisible world of computer chips, there are materials that act as the guardians of our digital memories. One such material is amorphous silicon nitride, a glass-like substance used to store data in devices that must keep their information even when the power is turned off. For these devices to work reliably, the material must hold onto electrical charges in a stable way. However, the atomic structure of this glass is messy and disordered, lacking the neat, repeating patterns found in crystals. This disorder creates hidden traps where electrons can get stuck. Sometimes these traps are harmless, but if they are too deep or too numerous, they can shift the voltage needed to run the chip, causing the memory to fail over time. Scientists have long known these traps exist, but because the material is so irregular, it has been incredibly difficult to understand exactly how they form, what they look like, or why some are far more dangerous than others.

To solve this puzzle, researchers Jonathon Cottom and Emilia Olsson developed a new way to look at the problem, treating the atomic network not just as a collection of atoms, but as a connected map. They focused on a specific question: when an electron gets caught in this glassy material, what happens to the atoms around it? Does the electron simply settle into a pre-existing hole, or does its arrival force the atoms to rearrange themselves into a new shape? By simulating hundreds of different scenarios in a computer, they tracked the journey of an electron from the moment it is free to the moment it is trapped, observing how the surrounding atomic network responded. They found that the answer is not a single story, but a mix of three distinct behaviors, and that the most dangerous traps are created by a dramatic chemical change triggered by the electron itself.

The researchers began by creating a large collection of computer models representing the disordered structure of silicon nitride. They simulated 415 different pairs of these models: one version with no extra charge and another with an extra electron added. For each pair, they carefully matched the electronic state before and after the electron was captured, ensuring they were comparing the same "identity" of the electron as it moved from a free state to a trapped one. This alignment allowed them to see exactly how the electron's position changed and how the atoms around it reacted. They discovered that the electron trapping process falls into three main categories. In some cases, the electron simply settles into a spot where it was already somewhat confined, tightening its grip without changing the surrounding atoms much. In other cases, the electron lands on a pre-existing flaw in the structure, a spot where an atom was already missing a neighbor. But the third category, which turned out to be the most significant, involves the electron actively breaking a chemical bond to create a new flaw.

The most striking finding was that the deepest and most problematic traps are almost always created by this third mechanism. When an electron is captured at a site where a silicon atom is normally bonded to four nitrogen neighbors, the arrival of the charge causes one of those bonds to snap. This breaks the silicon atom's connection to the network, leaving it with only three neighbors instead of four. This transformation, where a stable four-bonded atom becomes an unstable three-bonded one, creates a deep energy well that holds the electron very tightly. The researchers found that this bond-breaking event happens in more than half of the cases where a stable silicon atom captures an electron. Furthermore, these newly created, broken-bond traps are disproportionately responsible for the most severe energy shifts. In the group of traps that caused the deepest energy changes, nearly 70 percent were the result of this bond-breaking process, even though they made up less than half of all the traps observed.

To understand why this matters, the team used a mathematical tool they called a "graph-wavelet fingerprint." Imagine the atomic network as a web of connections. When an electron gets trapped, it sends a ripple through this web. The fingerprint measures how far and how strongly that ripple spreads. For the traps where the electron simply settled into an existing flaw, the ripple was small and stayed close to the trap. But for the traps where the electron broke a bond, the ripple was massive, spreading across a wide area of the network and causing a significant reorganization of the atoms. This large-scale rearrangement is what stabilizes the deep trap, locking the electron in place. The researchers could use the size and shape of this fingerprint to predict exactly how much the material would stabilize and how much the atoms would move, linking the invisible electronic behavior directly to the visible physical changes in the material.

The study also clarified that not all traps are created equal. While the bond-breaking traps are the most dangerous, they are not the only ones. The researchers identified that traps formed by electrons settling into pre-existing flaws or by tightening their grip on existing structures are also common, but they tend to be shallower and less disruptive. By separating these different behaviors, the new method provides a clear way to distinguish between a harmless trap and a destructive one. This distinction is crucial for engineers designing the next generation of memory devices, as it allows them to predict which parts of the material are likely to fail and which are stable. The work demonstrates that in disordered materials, the act of capturing a charge can fundamentally alter the local chemistry, turning a stable atom into a defect. This insight moves the field beyond guessing which traps exist to actually understanding how they are born and how they evolve, offering a precise roadmap for improving the reliability of the technology that powers our digital lives.

Technical Summary: State-Aligned Graph Wavelet Fingerprints of Electron Trapping in Amorphous Si $_3$ N $_4$

Problem Statement
Classifying charge traps in amorphous dielectrics, specifically amorphous silicon nitride (a-Si $_3$ N $_4$ ), presents a significant challenge due to the absence of a single stable defect archetype. In these disordered networks, intrinsic-polaronic trapping, coordination-driven trapping, and charge-driven conversion coexist within the same ensemble. Traditional classification methods struggle because energetic proximity, frontier orbital rank, and final-state localisation alone fail to establish a correspondence between the precursor neutral state and the trapped charged state. Furthermore, charge capture can induce complex network reorganisations—ranging from further localisation of intrinsic precursors to the creation of undercoordinated centres via bond breaking—that are difficult to resolve without a unified framework linking electronic identity to multiscale structural response.

Methodology
The authors develop and apply a "state-aligned graph-wavelet fingerprinting" framework to a dataset of 415 pairs of relaxed neutral and negatively charged a-Si $_3$ N $_4$ cells. The methodology proceeds through three integrated stages:

State Alignment: To resolve the correspondence between precursor and trapped states, the authors align the electronic states before and after localisation. This involves mapping the neutral precursor state (specifically the $\alpha$ /mo641 state) to the occupied charged-state HOMO on a common energy scale referenced to the N 2s centroid. This alignment accounts for spectral reordering and shifts in the electronic centre position, ensuring that changes in localisation and energy are tracked for the same specific electronic identity.
Graph-Wavelet Construction: The atomic network is represented as a periodic Si–N contact graph. Using diffusion-wavelet operators ( $H_t = D^{-1/2}e^{-tL}D^{1/2}$ ) across dimensionless scales $t \in \{0.5, 1, 2, 4, 8\}$ , the authors generate "fingerprints" that describe the local environment. These fingerprints are derived from seven structural channels (coordination deficit/excess, bond lengths, angular deviations, etc.), as well as potential and spin-polarisation fields.
Paired Analysis: The framework compares the "precursor amplitude" ( $T^0$ ) of the neutral state with the "relative fingerprint change" ( $\eta$ ) induced by charge capture. This paired readout distinguishes the initial chemical environment from the magnitude and scale of the network reorganisation. The analysis is grounded against physical observables, including localisation gain (measured by inverse participation ratio), local ionic relaxation (RMS displacement), and corrected stabilisation energy.

Key Results
The application of this framework to a-Si $_3$ N $_4$ yields the following findings:

Classification of Trapping Routes: The 412 classified traps are resolved into three distinct routes based on precursor amplitude and fingerprint change:
- Polarons (Si $_4 \to$ Si $_4$ ): Intrinsic-like localisation where the coordination remains unchanged.
- Induced K Centres (Si $_4 \to$ Si $_3$ ): Charge-driven conversion where a Si–N bond breaks, creating an undercoordinated centre.
- Native K Centres (Si $_3 \to$ Si $_3$ ): Capture at pre-existing dangling bonds.
  The paired fingerprints correctly classify 95.7% of these traps under cross-validation, with the coordination contribution channels being essential for separating polarons from induced K centres.
Multiscale Structural Response:
- Induced K Centres exhibit the largest relative structural fingerprint change ( $\eta$ ) and the most extensive reorganisation across graph scales, consistent with bond breaking and the creation of a new defect.
- Polarons show significant localisation gain but smaller structural reorganisation compared to induced K centres.
- Native K Centres show the smallest fingerprint changes, indicating capture at pre-existing defects with minimal additional network rearrangement.
Dominance of Charge-Driven Conversion: The study identifies charge-induced Si–N bond breaking as the dominant mechanism for the deepest spectral traps. Among the 42 traps in the upper corrected-stabilisation decile, 71.4% are induced K centres, and 69.0% of the deepest spectral traps are formed via this route. Although induced K centres account for only 44.2% of the total classified traps, they are heavily enriched in the most energetically stable (deepest) trap population.
Physical Grounding: The framework successfully links electronic and structural descriptors to physical quantities:
- Fixed-nuclei occupation response ( $-\delta\epsilon_{occ}$ ) tracks localisation gain.
- Structural fingerprint change ( $\eta$ ) tracks local ionic RMS displacement.
- Collective graph deformation ( $Q_D$ ) tracks corrected stabilisation energy.

Significance and Claims
The paper claims to provide a quantitative, multiscale, and multimetric framework for resolving charge-trap formation in amorphous materials. By aligning the trap state before and after localisation, the method connects electronic identity directly to the multiscale response of the surrounding atomic network.

The authors assert that this approach resolves the ambiguity of coexisting trapping modes in a-Si $_3$ N $_4$ , demonstrating that charge-induced bond breaking creates the undercoordinated centres that dominate the deepest trap population. The fingerprints unify electronic identity, chemically interpretable network reorganisation, and energetic stabilisation. The paper positions this framework as a generalizable tool for identifying precursor environments and their responses to charge capture in other disordered materials, offering a route-resolved structure-property understanding of how coupled electronic and atomic transformations create and stabilise charge traps. The authors explicitly state that the work establishes a common means of identifying these transformations without inventing new applications beyond the scope of the described methodology.

Technical Summary: State-Aligned Graph Wavelet Fingerprints of Electron Trapping in Amorphous Si3_33​N4_44​

More like this

Technical Summary: State-Aligned Graph Wavelet Fingerprints of Electron Trapping in Amorphous Si $_3$ N $_4$