Trace Carbon Strengthens Iron Grain Boundaries By Reconfiguring Bonds At The Crack Tip

By Muhammad OsamaReviewed by Susha Cheriyedath, M.Sc.Oct 7 2026

Machine-learning simulations track carbon atom by atom as a crack advances, probing what static models may miss about fracture at iron grain boundaries.

Paper: Dynamic bond reconfiguration drives crack-tip carbon migration to strengthen iron grain boundaries. AI-generated conceptual image created using ChatGPT/OpenAI

*Important notice: This news reports on an unedited version of an accepted paper and is awaiting final editing. Therefore, the paper should not be regarded as conclusive or treated as established information.

A recent study published in npj Computational Materials used machine-learning-driven atomistic simulations to show that trace carbon can increase the crack-growth resistance of general high-angle grain boundaries in alpha-iron (α-Fe), providing a mechanism consistent with earlier low-temperature experiments in which trace carbon suppressed intergranular fracture and promoted cleavage.

Researchers at Nippon Steel Corporation and Fujitsu Research developed a machine-learning interatomic potential (MLIP) to investigate this behavior and identified a crack-tip bond-reconfiguration process they abbreviate as DBR.

Grain-Boundary Fracture Resistance

Developing lightweight, resource-efficient, and durable metal alloys often requires higher tensile strength, which can increase susceptibility to intergranular fracture, particularly in contexts involving hydrogen embrittlement.

Scientists have widely used the Rice-Wang model to assess how segregating solutes alter grain-boundary cohesion through the ideal work of separation. This represents the energy required to separate a grain boundary into two free surfaces without plastic deformation.

The Rice-Wang model relies heavily on relaxed atomic configurations and does not directly describe non-equilibrium processes that occur as a crack advances, such as local solute enrichment and atomic rearrangement. Carbon is a useful model solute for examining these effects, as even parts-per-million (ppm) concentrations can strongly segregate to α-Fe grain boundaries and alter fracture behavior.

Advanced Techniques in Machine-Learning Simulations

To overcome the limits of conventional first-principles calculations, researchers developed a fast Fe-C MLIP using the atomic cluster expansion (PACE) format. The model was trained on 27,723 atomic structures assembled through the GeNNIP4MD active-learning workflow and containing more than 1.39 million local atomic environments.

The resulting potential ran about an order of magnitude faster than a previous Behler-Parrinello neural network potential while reproducing density functional theory (DFT) results across key validation tests.

The study constructed 12 bicrystal models, each formed by joining two crystals across a general grain boundary. The grains had random crystallographic orientations, with disorientation angles ranging from 20.2° to 57.6° and grain-boundary energies from 1.56 to 1.86 J m-2.

Grand-canonical Monte Carlo with MD (GCMC/MD) segregation simulations were conducted at approximately 373 K to generate carbon-segregation states corresponding to bulk concentrations of 20 and 60 wt ppm (parts per million by weight).

For crack propagation, the bicrystal models were replicated four times along the crack-propagation direction, producing cells approximately 40 nm long, 20 nm high, and 10 nm thick. An edge crack was introduced into each model, and tensile strain was applied normal to the grain-boundary plane at a constant strain rate of about 1 × 10⁹ s-1.

The crack simulations began at 1 K to suppress thermally activated carbon diffusion and thermally assisted dislocation emission, isolating brittle crack growth. A second set at 1 × 10⁸ s-1 preserved the carbon-induced increase. Crack-growth resistance (dU/dA) was calculated from the linear relationship between internal energy (U) and the one-sided crack area (A) during steady crack propagation.

Carbon Segregation and Higher Crack-Growth Resistance

The atomistic simulations reproduced the experimentally measured carbon interfacial excess at general grain boundaries. At the 60 wt ppm validation condition, carbon atoms concentrated within 2.5 Å of the grain-boundary center, with local concentrations reaching about 30 at%.

Structural analysis showed that segregated carbon mainly occupied relatively open, distorted sites with coordination numbers 6 (distorted octahedral sites) and 7 (capped trigonal prisms).

During mechanical loading, a representative clean general grain boundary showed a sharp drop in stress after crack initiation.

Under these deliberately dislocation-free conditions, the 12 clean boundaries had an average crack-growth resistance of nearly 4.71 J m-2, below the ideal Fe(001) cleavage value of 5.08 J m-2. The representative carbon-segregated boundary showed a secondary stress increase after crack initiation, while the 12 carbon-segregated boundaries averaged 7.93 J m-2.

Under these modeled conditions, raising the grain-boundary value above the modeled cleavage value makes boundary fracture less favorable than cleavage through the crystal, helping explain the fracture-mode shift seen in earlier experiments.

Analysis indicated that a static Rice-Wang-type evaluation accounted for about half of the carbon-induced increase in resistance. The rest came from crack-tip interactions that emerged during propagation.

Some carbon atoms stayed on the newly formed fracture surfaces, while others moved with the advancing crack tip. This process, DBR, increased the local carbon concentration near the crack tip from an initial 8.7 at% to approximately 17 at% in the representative grain boundary analyzed in detail.

Implications for Alloy Design

Identifying DBRs suggests a design approach for structural alloys that better resist crack growth. Alloy performance can reflect both static grain-boundary cohesion and how segregated solutes interact with an advancing crack tip. The findings could guide the selection of microalloying elements in steels and other structural alloys.

The Fe-C results also provide a basis for exploring other alloy systems. Solutes that exhibit strong grain-boundary segregation, interstitial rearrangement, and stable bonding with host-metal atoms could show similar crack-tip behavior. The authors identify boron in bcc Fe as one possible strengthening candidate.

Further simulations and experiments are needed to determine whether DBR occurs in these materials and whether grain-boundary chemistry can be adjusted to increase fracture resistance. The paper also cautions that the same type of crack-tip rearrangement could amplify embrittlement; phosphorus is cited as a possible example.

Study Limits and Future Work

The study found that trace carbon makes cracks harder to propagate along general grain boundaries in α-Fe through both changes in static cohesion and DBR. Carbon atoms can move with an advancing crack tip by repeatedly breaking and reforming Fe-C bonds. Together, these results show that solute motion and rebonding at the crack tip can increase the work needed for crack growth beyond that expected from static grain-boundary cohesion alone.

Future studies could examine DBR at finite temperatures and longer timescales, where thermally assisted dislocation emission and plastic relaxation may compete with brittle crack propagation. Such calculations would help determine whether DBR operates in other alloy systems and how it affects the competition between brittle crack growth and plastic dissipation.

Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.