Under the strategic imperatives of the “dual carbon” goals and the relentless push for lightweighting in advanced equipment manufacturing, Austempered Ductile Iron (ADI) has garnered significant and sustained attention. Its unique strength-toughness synergy, derived from a multiphase microstructure of bainitic ferrite and carbon-enriched retained austenite, positions it as an ideal candidate for critical components in high-speed rail brake discs, new energy vehicles, and multi-megawatt wind turbine gears. The foundational material, nodular cast iron, is renowned for its castability and the graphite nodules that confer ductility.
However, the conventional research and development paradigm for ADI is fraught with significant challenges. The final mechanical properties are governed by a complex, nonlinear interplay of multiple parameters: the base composition of the nodular cast iron (particularly C and Si), the strategic addition of alloying elements like Mn, Cu, and Ni, and the precise control of the austempering temperature and time. The synergistic control mechanisms between austenite stability, bainitic transformation kinetics, and the critical martensite content remain incompletely understood, creating a strong coupling problem. Furthermore, the traditional trial-and-error approach, requiring full cycles of melting, casting, heat treatment, and mechanical testing for each parameter set, suffers from severe efficiency bottlenecks. Exploring the vast combinatorial parameter space becomes prohibitively expensive and time-consuming. Therefore, there is an urgent need to develop a more scientific and efficient design methodology for ADI.
The advent of computational thermodynamics, particularly the CALPHAD (Calculation of Phase Diagrams) method, offers a powerful solution. It is currently the only methodology capable of precisely calculating phase equilibria in multicomponent systems with the accuracy required for practical engineering applications. This approach has fundamentally altered the material development landscape, moving it away from reliance on empirical trial-and-error towards a more predictive, knowledge-driven paradigm. By leveraging thermodynamic equilibrium calculations and transformation kinetics simulations, high-throughput CALPHAD calculations enable the automated screening of billions of potential composition and process combinations, dramatically accelerating the discovery and optimization of new materials.
In this work, we employ a high-throughput CALPHAD framework to systematically investigate the microstructural evolution of ADI under various austempering conditions. The core of this study is the construction of a three-stage phase transformation model to simulate the key steps in ADI heat treatment. We perform batch calculations to analyze the content of bainitic ferrite, retained austenite, and martensite. Building upon this computational analysis, we identify and develop key microstructural feature descriptors that correlate with mechanical performance, ultimately enabling the intelligent optimization of ADI composition and processing parameters. This work aims to pioneer a shift from an “experience-driven” to a “computation-driven” design paradigm for high-performance ADI based on engineered nodular cast iron.
The Three-Stage Phase Transformation Model
The heat treatment of Austempered Nodular Cast Iron (ADI) fundamentally involves three critical stages: austenitization, austempering (isothermal holding), and final cooling. We have constructed a computational model to simulate the microstructural outcomes of each stage based on CALPHAD thermodynamics and transformation criteria.
Stage 1: Austenitization. The first step involves heating the nodular cast iron to a temperature typically between 840–950°C and holding for 1–2 hours to allow the matrix to fully transform into carbon-saturated austenite. For our high-throughput calculations, we assume this process reaches equilibrium. Using the phase equilibrium module in Thermo-Calc software, we calculate the two-phase equilibrium state (Austenite + Graphite) at 900°C for 2 hours. This calculation provides the initial conditions for the subsequent stages: the volume fraction of graphite, the volume fraction of austenite, and, crucially, the carbon content of this austenite ($C_{\gamma, initial}$).
Stage 2: Austempering (Bainitic Transformation). After austenitization, the casting is rapidly quenched to an isothermal holding temperature ($T_{AT}$) between 230°C and 400°C. During this isothermal hold, part of the austenite transforms into bainitic ferrite (acicular ferrite). This transformation is displacive; the growing ferrite plates reject excess carbon into the surrounding austenite. The reaction proceeds until the carbon concentration in the remaining austenite reaches a critical value, often associated with the $T_0$ curve (where the Gibbs free energy of austenite and ferrite of the same composition are equal), halting further diffusional transformation. We simulate this stage using property calculation routines to determine the volume fraction of bainitic ferrite ($V_{f}^{BF}$), the volume fraction of untransformed austenite ($V_{f}^{\gamma, untrans}$), and its corresponding carbon content ($C_{\gamma, enriched}$) after a specified holding time (2.5 hours in this study).
Stage 3: Final Cooling & Potential Martensite Transformation. Following the austempering hold, the casting is cooled to room temperature. Whether the enriched, untransformed austenite transforms into martensite depends entirely on its stability, which is primarily a function of its chemical composition (especially $C_{\gamma, enriched}$) and the cooling rate. For this model, we assume fast cooling. We calculate the martensite-start temperature ($M_s$) of the enriched austenite composition. If the $M_s$ temperature is above room temperature, a certain volume fraction of martensite ($V_{f}^{M}$) will form. The remaining phase is then the stable retained austenite ($V_{f}^{RA}$). The model calculates these final phase fractions.
The entire workflow, integrating these three stages, was automated using the TC-Python interface within the Thermo-Calc software, enabling high-throughput computation across wide compositional and processing ranges. The schematic logic is as follows:
Input: Full alloy composition (Fe-C-Si-Mn-Cu-Ni…) + Austenitizing parameters (900°C, 2h) + Austempering Temperature ($T_{AT}$).
Stage 1 Compute: $V_{f}^{Gr}$, $V_{f}^{\gamma}$, $C_{\gamma, initial}$.
Stage 2 Compute: $V_{f}^{BF}$, $V_{f}^{\gamma, untrans}$, $C_{\gamma, enriched}$ at $T_{AT}$.
Stage 3 Compute: $M_s$ of enriched austenite. Determine final $V_{f}^{M}$ and $V_{f}^{RA}$.
Output: Complete final microstructure: $V_{f}^{Gr}$, $V_{f}^{BF}$, $V_{f}^{RA}$, $C_{RA}$, $V_{f}^{M}$.
High-Throughput Calculations and Microstructural Analysis
We focused our investigation on the key alloying elements known to critically influence the behavior of nodular cast iron during austempering: C, Si, Mn, Cu, and Ni. The high-throughput study was conducted in two sequential parts to manage complexity and reveal primary effects.
Part 1: Influence of C, Si, Mn and Austempering Temperature
The composition ranges for the first set of calculations, along with the austempering temperatures ($T_{AT}$), are detailed in Table 1. Copper was fixed at 0.5 wt.% and Nickel at 0 wt.% for this phase. This resulted in 864 unique composition-process combinations.
| Element/Parameter | Min (wt.%) | Max (wt.%) | Step (wt.%) | Notes |
|---|---|---|---|---|
| C | 3.0 | 4.0 | 0.2 | – |
| Si | 2.0 | 3.0 | 0.2 | – |
| Mn | 0.2 | 0.6 | 0.2 | – |
| $T_{AT}$ (°C) | 230 | 400 | Variable* | *Steps: 30°C (230-320°C), 20°C (350-390°C), 10°C (390-400°C) |
The correlation analysis between the calculated microstructural features and the input variables (C, Si, Mn, $T_{AT}$) is summarized in a Pearson correlation matrix. The most striking observation is the dominant influence of the austempering temperature ($T_{AT}$). It shows the highest absolute correlation coefficients with all phases except graphite. Specifically, $T_{AT}$ is positively correlated with bainitic ferrite content, retained austenite content, and its carbon concentration, but negatively correlated with martensite content. This underscores $T_{AT}$ as the primary process lever for controlling the ADI microstructure derived from nodular cast iron.
To elucidate the role of base composition, we analyzed results at two characteristic austempering temperatures: 290°C (representing lower-temperature austempering) and 370°C (representing higher-temperature austempering). The influence of C, Si, and Mn at $T_{AT}$ = 290°C is profound and systematic:
- Graphite ($V_{f}^{Gr}$): Almost exclusively controlled by carbon content, with a near-perfect positive correlation ($\rho \approx 0.99$). Silicon has a minor positive effect, while Mn’s influence is negligible. This is consistent with the metallurgy of the base nodular cast iron.
- Bainitic Ferrite ($V_{f}^{BF}$): Primarily influenced by Si and Mn, both showing strong negative correlations ($\rho_{Si} \approx -0.8, \rho_{Mn} \approx -0.6$). Higher Si/Mn suppresses the bainitic transformation kinetics, leading to less ferrite.
- Martensite ($V_{f}^{M}$): Strongly positively correlated with Si and Mn ($\rho_{Si} \approx 0.8, \rho_{Mn} \approx 0.6$). By inhibiting bainite formation, Si and Mn result in more untransformed austenite that is less enriched in carbon and thus more prone to transform to martensite upon cooling.
- Retained Austenite ($V_{f}^{RA}$) and its Carbon ($C_{RA}$): Both show strong negative correlations with Si and Mn. Higher Si/Mn leads to less bainite, but the austenite that remains is also less effectively enriched with carbon, making it less stable.
The behavior at $T_{AT}$ = 370°C reveals a significant shift in the influence of alloying elements:
- Graphite: The dominance of carbon remains unchanged.
- Bainitic Ferrite: The influence of Si becomes negligible ($\rho \approx 0$), while Mn retains a strong negative correlation ($\rho \approx -0.99$). At higher $T_{AT}$, the transformation driving force is lower, and Mn’s effect on austenite stability becomes the overriding factor.
- Martensite: Correlates strongly and negatively with Si ($\rho \approx -0.95$), but weakly with Mn and C. High Si strongly stabilizes austenite against martensite transformation at this temperature.
- Retained Austenite: Shows a strong positive correlation with Si ($\rho \approx 0.95$) and a moderate positive one with Mn. High Si increases the amount and stability of retained austenite.
- $C_{RA}$: Is almost perfectly negatively correlated with Mn ($\rho \approx -1.0$), indicating Mn severely retards carbon enrichment during the slower transformation at high $T_{AT}$.
This stark contrast between the low and high $T_{AT}$ regimes highlights the complex, temperature-dependent coupling between composition and phase transformation behavior in ADI processed from nodular cast iron.
Part 2: Influence of Cu and Ni Additions
Following the analysis of the base nodular cast iron composition, we investigated the effects of Cu and Ni. Five fixed base compositions of C, Si, Mn were selected, and Cu and Ni were varied as per Table 2, leading to 480 additional calculations.
| Element/Parameter | Min (wt.%) | Max (wt.%) | Step (wt.%) | Notes |
|---|---|---|---|---|
| Cu | 0.5 | 1.5 | 0.5 | – |
| Ni | 0.0 | 2.0 | 0.5 | – |
| Base Compositions | 5 fixed sets of (C, Si, Mn) | |||
| $T_{AT}$ (°C) | 230 | 400 | As in Table 1 | |
The analysis at $T_{AT}$ = 290°C indicates that, within the context of the dominant C-Si-Mn system, Nickel exerts a more pronounced influence than Copper on most microstructural features. Ni strongly reduces bainitic ferrite content and the carbon level in retained austenite (correlation coefficients ~ -0.6), while its effect on increasing martensite content is also significant. Copper’s effects are milder, showing a weak tendency to promote bainite and reduce martensite.
At $T_{AT}$ = 370°C, the roles partially shift. Nickel maintains a strong influence on reducing $C_{RA}$ and martensite content, while Copper shows a clearer effect on increasing retained austenite volume fraction. The influence of both elements on bainitic ferrite is moderate and opposite in sign (Cu positive, Ni negative). These results confirm that Cu and Ni are powerful tools for fine-tuning the microstructure and stability of austenite in ADI, but their effectiveness is modulated by the base nodular cast iron composition and the chosen austempering temperature.
Intelligent Optimization of ADI Composition and Process
The high-throughput calculations generated a vast dataset linking composition, process ($T_{AT}$), and final microstructure. To translate this into an actionable design strategy, we defined three key microstructural feature parameters based on physical metallurgy principles that are critically linked to the final mechanical properties of ADI.
Feature 1: Carbon Equivalent (CE) Proximity to Eutectic. The fluidity and soundness of the initial nodular cast iron melt are governed by its carbon equivalent. For optimal castability, minimizing shrinkage and graphite flotation, the CE should be near the eutectic point (~4.3-4.7). We use a standard simplified formula for nodular cast iron:
$$ CE = C + \frac{1}{3}Si – 0.03Mn $$
We define our first feature parameter ($F_1$) as the absolute deviation from a target value of 4.5:
$$ F_1 = | CE – 4.5 | $$
A lower $F_1$ indicates better castability of the base iron.
Feature 2: Retained Austenite Stability. The stability of retained austenite is paramount for the TRIP (Transformation Induced Plasticity) effect, which enhances ductility and toughness. A common metric for stability is the product of the retained austenite volume fraction and its carbon content ($V_{f}^{RA} \times C_{RA}$). Higher values indicate more stable austenite. We define our second feature parameter as the inverse of this product, so that a lower value indicates higher stability:
$$ F_2 = \frac{1}{V_{f}^{RA} \times C_{RA}} $$
Feature 3: Martensite Content. While martensite increases strength, its presence, especially in high volumes, is detrimental to ductility and toughness. For a balanced high-performance ADI, martensite should be minimized or eliminated. Therefore, our third feature parameter is simply the martensite volume fraction:
$$ F_3 = V_{f}^{M} $$
A lower $F_3$ is desired.
An ideal ADI composition and process should simultaneously minimize $F_1$, $F_2$, and $F_3$. To enable a single-figure ranking of all computed combinations, we construct a Comprehensive Criterion ($CC$) by first normalizing each feature ($\hat{F_i}$) over the entire dataset and then calculating the Euclidean norm:
$$ \hat{F_i} = \frac{F_i – min(F_i)}{max(F_i) – min(F_i)} \quad \text{(for i = 1, 2, 3)} $$
$$ CC = \sqrt{ (\hat{F_1})^2 + (\hat{F_2})^2 + (\hat{F_3})^2 } $$
A lower $CC$ value represents a more optimal combination for achieving excellent castability, high austenite stability, and low martensite content—all hallmarks of superior ADI performance.
A partial correlation analysis of the $CC$ with all input variables confirms that the austempering temperature ($T_{AT}$) is the most influential factor, followed by the carbon content. Sorting the entire high-throughput dataset by the $CC$ value allowed us to identify the Pareto-optimal combinations. The top-ranking composition and process identified was:
Optimal Composition: Fe – 3.6C – 2.6Si – 0.6Mn – 1.0Ni – 0.5Cu (wt.%)
Optimal Austempering Temperature ($T_{AT}$): 320°C
This optimized nodular cast iron composition, when subjected to the specified austempering cycle, is predicted to yield a microstructure with a near-ideal carbon equivalent, a high volume of stable retained austenite, and negligible martensite. Literature reports on ADI with similar compositions (e.g., Fe-3.59C-2.72Si-0.18Mn-0.47Ni-0.68Cu) processed at 320°C exhibit excellent mechanical properties (tensile strength > 1280 MPa, elongation > 6%), providing strong validation for the predictive capability of our CALPHAD-driven optimization framework.
Conclusions
This work successfully established a high-throughput, computation-driven design framework for Austempered Ductile Iron (ADI) based on the CALPHAD method. The key outcomes are:
- We developed a three-stage phase transformation model that effectively simulates the microstructural evolution of nodular cast iron through austenitization, isothermal bainitic transformation, and final cooling with potential martensite formation.
- High-throughput calculations revealed the austempering temperature ($T_{AT}$) as the most dominant factor controlling phase fractions. The influence of key alloying elements (C, Si, Mn, Cu, Ni) is strong but highly dependent on $T_{AT}$, with Si and Mn playing critical but temperature-sensitive roles in regulating bainite formation and austenite stability.
- By defining and integrating three key microstructural feature parameters—carbon equivalent deviation ($F_1$), inverse retained austenite stability ($F_2$), and martensite content ($F_3$)—we constructed a Comprehensive Criterion ($CC$) for holistic performance evaluation.
- Optimization via the $CC$ identified an optimal composition (Fe-3.6C-2.6Si-0.6Mn-1Ni-0.5Cu) and process (320°C austempering) predicted to yield an ideal ADI microstructure conducive to superior mechanical properties.
This study demonstrates a powerful paradigm shift from empirical trial-and-error to a targeted, knowledge-guided design process for advanced nodular cast iron derivatives like ADI. The integrated CALPHAD high-throughput and feature-based optimization approach significantly accelerates the development cycle and provides deep insights into the complex composition-process-property relationships, paving the way for the next generation of high-performance cast iron components.