Experimental and computational diffusion analysis in Ni-X binary and Ni-Al-X (X = Cr, Mo, Ta, W, Re) ternary systems

This study presents a comprehensive experimental and computational analysis of diffusion in Ni-X binary and Ni-Al-X ternary systems, revealing that while main interdiffusion coefficients remain comparable to their binary counterparts, cross-diffusion effects significantly influence fluxes, and establishing a robust framework that combines first-principles calculations with physics-informed neural networks to accurately model composition-dependent diffusion across the full range.

The Gist

Imagine a giant, high-temperature kitchen where the main ingredient is Nickel (Ni). To make this kitchen durable enough to withstand extreme heat (like in jet engines), chefs add special "spices" like Aluminum (Al), Chromium (Cr), Molybdenum (Mo), Tantalum (Ta), Tungsten (W), and Rhenium (Re).

The problem? When you heat these alloys up, the atoms start to move around, or diffuse. If they move too fast or in the wrong way, the material's structure can break down, and the engine fails. Scientists need to know exactly how fast each "spice" atom moves and how they interact with each other.

This paper is like a detailed map and a new set of rules for predicting how these atoms move in a Nickel-based kitchen. Here is the breakdown of their findings in simple terms:

1. The "Solo" vs. "Group" Dance (Binary vs. Ternary Systems)

First, the researchers looked at Binary systems (Nickel + just one spice, like Nickel + Chromium). They measured how fast the spice atoms moved on their own.

• The Finding: Some spices move very fast (like Aluminum), while others are slow and stubborn (like Rhenium). They found that the "slowness" of Rhenium is mostly because it takes a lot of energy for it to jump into an empty spot (a vacancy) in the metal grid. It's like trying to push a heavy boulder up a hill versus rolling a marble.

Then, they looked at Ternary systems (Nickel + Aluminum + a third spice). This is more like a dance floor with three partners.

• The Finding: When Aluminum and a third spice are both present, they don't just move independently. They influence each other.
  • The "Traffic" Effect: If Aluminum and the third spice are trying to move in the same direction, they help each other speed up.
  • The "Braking" Effect: If they are trying to move in opposite directions, they slow each other down.
  • The Surprise: In the past, scientists only looked at the "average" speed of the group. This paper shows that looking at the average can be misleading. You have to look at the specific interactions (the "cross-diffusion") to understand what's really happening. For example, in the Nickel-Aluminum-Rhenium mix, the average data suggested a strong negative interaction (like a fight), but the real data showed they barely interact at all.

2. The "Rhenium" Problem

Rhenium is a special spice that moves incredibly slowly. Because it moves so slowly, when scientists tried to measure how it interacts with Aluminum, the two "paths" of diffusion barely crossed each other. It was like trying to find the exact spot where two slow-moving snails met; the data was too fuzzy to trust.

• The Solution: Instead of trying to find where two paths crossed, they used a clever trick involving a "Kirkendall marker" (a tiny line of inert particles that marks the center of the dance floor). This allowed them to calculate the speeds accurately even with just one diffusion path.

3. The "Smart Calculator" (PINN)

Usually, to figure out how fast atoms move at every possible concentration (not just the specific spots they tested), scientists use math models. However, the researchers found that if you just feed a computer the diffusion profiles (the pictures of where atoms ended up) and ask it to guess the speeds, the computer can come up with a mathematically correct answer that is physically wrong. It's like a student guessing the right answer to a math problem but using the wrong formula.

• The Innovation: They used a Physics-Informed Neural Network (PINN). Think of this as a super-smart calculator that knows the laws of physics (the rules of the dance) and is also forced to check its work against real-world measurements.
• The Key Rule: They discovered that for the calculator to give a reliable answer, you must give it some real, measured data points as "anchors" (constraints). If you don't give it these anchors, the calculator might fit the curve perfectly but get the physics completely wrong. By anchoring it with real data, they could accurately predict how the atoms move across the entire range of concentrations.

4. The "Serpentine" Paths

When they plotted the movement of these atoms on a triangular map (called a Gibbs triangle), the paths didn't go in straight lines. They curved like snakes.

• Why? This happens because the different atoms move at different speeds. If Aluminum is a sprinter and Rhenium is a tortoise, the path of the mixture bends to compensate for who is getting ahead. The researchers showed that the shape of these "snake paths" perfectly matches the speed differences they calculated, proving their data is accurate.

Summary

This paper didn't just measure how fast atoms move; it built a robust framework to understand how they influence each other in complex mixtures.

1. Rhenium is the slowest mover, and its slowness is due to high energy barriers.
2. Cross-interactions matter: Atoms can speed up or slow down their neighbors depending on which way they are moving.
3. Averages can lie: You can't just look at the average speed; you need to look at the specific interactions between elements.
4. Smart AI needs anchors: To use advanced AI (PINN) to predict diffusion, you must feed it real experimental data as "truth checks," or the results will be unreliable.

The result is a much clearer, more accurate map for designing better, longer-lasting superalloys for high-temperature applications.

Technical Summary

Technical Summary: Experimental and Computational Diffusion Analysis in Ni-X Binary and Ni-Al-X Ternary Systems

Problem Statement
Diffusion-controlled processes govern the microstructural evolution, high-temperature properties, and service life of Ni-based superalloys. While extensive diffusion data exists for binary Ni-X systems (where X = Cr, Mo, Ta, W, Re) and the Ni-Al binary system, studies on ternary Ni-Al-X systems remain limited. A critical gap exists in understanding how the presence of Aluminum (Al) modifies the diffusion behavior of alloying elements X, specifically regarding diffusional interactions (cross-diffusion coefficients). Previous literature often reports only interdiffusion coefficients ($\tilde{D}_{ij}^n$), which represent an average of intrinsic and tracer coefficients, potentially obscuring the specific roles of individual elements. Furthermore, existing studies often lack reported activation energies for ternary systems, and numerical inverse methods used to extract composition-dependent coefficients have been shown to yield non-unique or unreliable results if not constrained by experimental data.

Methodology
The study employs a multi-faceted approach combining experimental diffusion couple analysis, first-principles calculations, and advanced numerical optimization:

1. Experimental Diffusion Couples: Binary (Ni-X) and ternary (Ni-Al-X) diffusion couples were synthesized and annealed at temperatures ranging from 1100°C to 1250°C for 50 hours. Composition profiles were analyzed using Wavelength-Dispersive Spectroscopy (WDS).
2. Coefficient Estimation:
   • Binary Systems: Interdiffusion coefficients ($\tilde{D}$) were calculated using the Sauer-Freise relation.
   • Ternary Systems: For most systems (Ni-Al-Cr, Mo, Ta, W), coefficients were estimated at intersecting diffusion paths. For the Ni-Al-Re system, where intersections occurred near end-members (causing gradient uncertainties), a single-profile method utilizing the Kirkendall marker plane was employed.
   • Types of Coefficients: The study calculated tracer ($D_i^*$), intrinsic ($D_{ij}^n$), and interdiffusion ($\tilde{D}_{ij}^n$) coefficients.
3. First-Principles Calculations: Density Functional Theory (DFT) was used to calculate vacancy formation energies and migration barriers for impurity diffusion in Ni. These values were combined to determine activation energies, helping to distinguish between thermodynamic and kinetic contributions to diffusion trends.
4. Physics-Informed Neural Network (PINN): A PINN-based inverse optimization method was applied to extract composition-dependent diffusion coefficients across the full range of the diffusion profiles. Crucially, this study enforced experimentally estimated diffusion coefficients as equality constraints at specific compositions (intersections or marker planes) to ensure physical reliability.

Key Contributions and Results

• Binary System Trends: The study established a consistent hierarchy of diffusion coefficients in binary Ni-X systems at 2.5 at.%: $D_{Al} > D_{Ta} \approx D_{Cr} > D_{Mo} > D_{W} > D_{Re}$. Re exhibited the highest activation energy (~339 kJ/mol), while Ta showed the lowest. First-principles calculations revealed that variations in activation energy are primarily driven by differences in migration energies rather than vacancy formation energies.
• Ternary System Interactions:
  • Main Coefficients: The main interdiffusion coefficients of X in ternary Ni-Al-X systems were found to be comparable to their binary counterparts, with similar activation energies.
  • Cross-Diffusion Effects: Cross-diffusion coefficients were shown to significantly influence fluxes. Depending on the sign of the cross-term and the relative diffusion directions of Al and X, fluxes can be enhanced or reduced. For instance, in Ni-Al-Re, the interdiffusion coefficient $\tilde{D}_{ReAl}^{Ni}$ suggested a strong negative interaction, whereas the intrinsic coefficient $D_{ReAl}^{Ni}$ revealed a negligible positive interaction, highlighting the potential for misinterpretation when relying solely on averaged interdiffusion data.
  • Diffusion Paths: The "serpentine" nature of diffusion paths on Gibbs triangles was rationalized by the relative diffusion rates of the elements. The depth of deviation correlated with the diffusivity of X, following the sequence: Ta < Cr < Mo < W < Re.
• Validation of PINN with Constraints: A pivotal finding was that optimizing diffusion profiles without experimental equality constraints leads to non-unique solutions that may fit the profile data but yield physically incorrect diffusion coefficients (e.g., incorrect relative magnitudes of $D_i^*$). Imposing experimentally determined coefficients as constraints was essential to recover reliable, composition-dependent parameters.
• Transferability: The optimization parameters derived from ternary systems were successfully validated by predicting binary interdiffusion coefficients for Ni-Cr and Ni-Al, which matched well with experimental data.

Significance and Claims
The paper claims to provide the first comprehensive, temperature-dependent diffusion analysis comparing binary Ni-X and ternary Ni-Al-X (X = Cr, Mo, Ta, W, Re) systems. Its primary significance lies in:

1. Clarifying Diffusional Interactions: Demonstrating that cross-diffusion coefficients play a critical role in ternary fluxes and that relying exclusively on interdiffusion coefficients ($\tilde{D}_{ij}^n$) can lead to misleading conclusions about element interactions.
2. Methodological Robustness: Establishing a framework where physics-informed neural networks (PINN) must be constrained by experimental data to yield physically reliable, composition-dependent diffusion coefficients. The study explicitly states that unconstrained optimization, even with excellent profile fitting, fails to guarantee reliable diffusion parameters.
3. Comprehensive Dataset: Providing a consistent set of diffusion coefficients and activation energies across multiple alloying elements, filling a gap in the literature regarding the specific influence of Al on the diffusion of refractory elements in Ni-based superalloys.

The authors maintain a modest tone regarding the complexity of activation energy determination in alloys, noting that experimental values represent bulk averages over various atomic configurations, while first-principles calculations provide insight into specific migration mechanisms. The work is presented as a robust framework for future diffusion studies in multicomponent Ni-based systems.