Understanding Steel Corrosion: Surface Chemistry and Defects Explored Through DFT Modelling—A Review

Abstract

Corrosion poses a critical challenge to the durability and performance of metals and alloys, particularly steel, with significant economic, environmental, and safety implications. The corrosion susceptibility of steel is influenced by aggressive chemical species, intrinsic material defects, and environmental factors. Understanding the atomic-scale mechanisms governing corrosion is essential for developing advanced corrosion-resistant materials. Density functional theory (DFT) has become a powerful computational tool for investigating these mechanisms, providing insight into the adsorption, diffusion, and reaction of corrosive species on iron surfaces, the formation and stability of metal oxides, and the influence of defects such as vacancies and grain boundaries in localised corrosion. This review presents a comprehensive analysis of recent DFT-based studies on iron and steel surfaces, emphasising the role of solvation effects and van der Waals corrections in improving model accuracy. It also explores defect-driven corrosion mechanisms and the formation of protective and reactive oxide layers under varying oxygen coverages. By establishing accurate DFT modelling approaches, this review provides up-to-date literature insights that support future integration with machine learning and multiscale modelling techniques, enabling reliable atomic-scale predictions.

1. Introduction

Corrosion refers to the gradual degradation of metals or alloys due to chemical reactions at the surface with their environment, leading to material deterioration, structural compromise, and ultimately necessitating costly remediation efforts, such as replacement or extensive repair [1]. Its impact is profound across industries, causing significant economic, environmental, and safety concerns [1,2,3]. Economically, the International Measures of Prevention, Application, and Economics of Corrosion Technologies (IMPACT) report by National Association of Corrosion Engineers (NACE) International estimates that the global cost of corrosion is approximately US$2.5 trillion, which represents 3.4% of the global GDP [4]. Importantly, the report identifies that by implementing current corrosion control practices, this substantial cost could be reduced by 15–35%, which equates to a potential savings of US$375–US$875 billion each year [4]. Corrosion can lead to significant environmental damage by causing infrastructure failures, including bridge collapses, degradation of ships and docks, leaks in pipelines and storage tanks, rusting of industrial equipment, building and railway materials, and even the falling of traffic signage [1,3,5,6].

One of the primary materials impacted by corrosion is iron (steel), widely used in national infrastructure. Steels are iron-carbon alloys further enhanced by the addition of various elements to tailor their properties [7]. They are commonly categorised based on both their chemical composition and microstructure. With respect to chemical composition, steels are categorised as unalloyed, alloyed, and high-alloyed [7,8]. Microstructurally, steels are classified into ferritic, austenitic, duplex (austenitic–ferritic), and martensitic types, each offering distinct properties suited for specific applications and industrial use [7,9]. However, steel is highly susceptible to environmental degradation [10,11,12]. For example, carbon steel, mild steel [13,14], steel rebar [10,15,16], and stainless steel [11,12,17] are all susceptible to corrosion, despite stainless steel being highly regarded for its corrosion resistance.

When steel is exposed to environmental conditions, spontaneous reactions occur between its surface and corrosive species. These reactions typically result in the formation of corrosion products, primarily iron oxides and hydroxides, commonly known as rust [18]. As depicted in Figure 1, the corrosion mechanism is highly dependent on the surrounding environment. Depending on the conditions, other minor products like iron sulphides, carbides, or nitrides can also form, though these are more common in specialised conditions. Therefore, corrosion can proceed through various mechanisms, including chlorination, sulfidation, nitridation, and carburisation [19].

The corrosion of steel, unlike other materials, is largely driven by an electrochemical process in the presence of moisture. This process involves the formation of a localised electrochemical cell, where distinct anodic and cathodic reactions take place. As shown in Figure 1a, the anode is the site of iron oxidation, leading to the dissolution of Fe²⁺ ions into the electrolyte (Fe → Fe²⁺ + 2e⁻). Meanwhile, at the cathode, oxygen and water are reduced to form hydroxide ions (O₂ + 2H₂O + 4e⁻ → 4OH⁻) [10,21]. These reactions collectively drive the corrosion process, facilitating the degradation of steel over time. Figure 1b illustrates a schematic representation of the corrosion process affecting rebar in a reinforced concrete environment [15]. Chloride ions (Cl⁻) from the surrounding concrete are known to play a significant role in promoting rebar corrosion. Numerous studies have demonstrated that Cl⁻ ions can compromise the passive oxide layer that naturally forms on the rebar surface, thereby accelerating the corrosion process [15,22]. This highlights the importance of the corrosive environment in influencing and driving the corrosion process.

While experimental studies are essential for validating corrosion mechanisms, real-world corrosion processes often occur over extended timescales, making direct observations challenging [23]. Accelerated experimental techniques are commonly used to speed up corrosion testing; however, these methods may not always accurately replicate long-term degradation under natural conditions. As a result, a significant portion of corrosion research increasingly relies on ‘computer experiments’ [22]—advanced simulations that offer atomic-scale insights into corrosion dynamics, complementing experimental findings and enabling predictive modelling of material behaviour [22]. Computational methods are invaluable in the study of metal corrosion, and each method provides unique strengths and limitations, as shown in Table 1. These methods allow researchers to explore atomic-scale interactions and mechanisms that are difficult to observe experimentally. However, the complexity of real-world corrosion processes requires insights spanning from electronic interactions to macroscopic behaviour. This has led to multiscale modelling approaches that integrate different computational methods across multiple scales, linking quantum mechanical calculations with molecular simulations, kinetic models, and continuum mechanics to address interconnected corrosion processes that cannot be understood through single-scale methods alone. Figure 2 illustrates the recent application of multi-scale modelling in steel corrosion.

Within these multiscale modelling frameworks, techniques like density functional theory (DFT) simulations provide insights into the role of surface chemistry, defects, and environmental factors in corrosion processes, enabling the prediction of material performance and the design of more corrosion-resistant materials. These approaches also reduce the cost and time associated with experimental investigations by complementing and guiding laboratory studies [23]. A general corrosion mechanism seems to be simply described; its real scenario is profoundly complex, as it is influenced by multiple factors, including the material’s microstructure, characteristics of its environment, and localised conditions. Thus, computer simulations at the atomic levels are indispensable for elucidating how these variables interact to shape the corrosion process. Among the computation techniques, DFT has become a valuable tool for studying surface properties such as reactivity, adsorption behaviours, electronic structure, charge transfer process, reaction mechanism, defects and other atomic scale processes to understand corrosion mechanisms. However, while static DFT provides detailed energetic and structural information at 0 K, it cannot capture time-dependent atomic behaviour under thermal conditions [28,29]. Ab initio molecular dynamics (AIMD) extends DFT capabilities by enabling finite-temperature simulations that reveal thermally activated processes—bond breaking and conformational transitions at solid–liquid interfaces [28,29,30]. AIMD uniquely captures fluctuation-driven mechanisms, reaction events, and free energy barriers under realistic conditions, providing dynamic insights inaccessible to static approaches [29]. This becomes critical in aqueous corrosion systems where thermal fluctuations, conformational changes, and activation processes govern explicit water–surface interactions.

While numerous recent reviews have been published on computational approaches to metal corrosion [19,23,31,32,33,34], this review explores the application of DFT to unravel corrosion mechanisms on steel surfaces with defects, with iron (Fe) as the focal material, which has been studied less. It reveals how DFT simulations elucidate the adsorption and diffusion of corrosive species and investigate the impacts of structural defects, such as vacancies and grain boundaries, on Fe surface chemistry and the development of localised corrosion. Finally, we highlight the strengths of current DFT approaches in advancing the understanding of corrosion while acknowledging areas for further refinement, and we outline promising future research directions to enhance their applicability and effectiveness.

Table 1. Key strengths, primary limitations, main outputs, and recent applications of computational methods used in corrosion studies.

Method	Key Strengths	Primary Limitations	Main Outputs	Recent Applications	Refs.
Density Functional Theory (DFT)	•Provide a good balance between accuracy and computational cost •Excellent for surface adsorption and reaction energetics, and electronic structures •Reliable prediction of thermodynamic properties	•Fail to capture long-range dispersion forces accurately • Less feasible for large systems •Restricted to short simulation durations	Adsorption energies, defect formation energies, and interlayer relaxation Electronic Properties: Work functions, density of state, Mulliken population, and Bader charge analysis Magnetic Properties	Adsorption and passivation mechanisms, identifying reactive sites and surface stability, and inhibitor design	[23,35,36]
Ab Initio Molecular Dynamics (AIMD)	•Forces on atoms are derived from quantum calculations, yielding highly accurate structures and bonding •Provide kinetic and thermodynamic insights •Ideal for understanding reaction pathways and electronic interactions. •Highly transferable across different surface types, adsorbates, and environmental conditions	•Computationally expensive, limiting its application to large systems and long simulation times •Convergence issues for complex systems	Diffusion coefficients and reaction rates (short-time scale) Reaction pathways and barriers Dissolution energies, chemical binding, and charge transfer	Water–metal interfaces, oxide nucleation, and dissolution corrosion	[34,37]
Reactive Force Fields (ReaxFF)	•Enable the study of chemical reactions with bond formation/breaking • Capability to model the effects of temperature, pH, and aggressive ions on corrosion rate and mechanism • Enables modelling of complex reactive systems over a long time	• Depends on empirical parameters, limiting its accuracy and transferability across different systems • Needs reparameterization when applied to a new chemical environment (e.g., introducing new alloy elements, different pH, or complex oxidizers)	Diffusion coefficient and reaction rates Oxide layer thickness	Growth of oxide layers, pit formation studies, and stress corrosion	[34,38]
Monte-Carlo and Kinetic Modelling	• Computationally efficient for large systems and high-dimensional phase spaces. • Effectively samples equilibrium configurations and capable of simulating short-range order in high-entropy or multi-element alloy surfaces exposed to corrosive environments • Successfully simulate the kinetics of corrosion processes at different scales	• Relies on empirical or fitted potential energy models, which limits its transferability and accuracy for systems with complex chemistries	Phase diagrams Order–disorder transitions in surface adsorbates or alloy atoms Corrosion mass gain and corrosion rates Equilibrium surface coverage and segregation profiles	Kinetics of passive film formation, alloy segregation behaviour, and modelling of corrosion inhibitors	[39,40]
Finite Element method	• Provide macroscale structural analysis • Capability to handle complex geometries, and multi-field coupling	• Lacks the capability to capture atomistic-scale phenomena •Depends on precise material properties and constitutive models to generate reliable results	Stress distributions Mechanical strength and structural integrity Life span of materials	Pipeline corrosion assessment, coating failure analysis, and galvanic corrosion modelling	[23,41]
Machine learning and AI modelling	• Can integrate experimental and simulation data to predict the severity of corrosion • Can be tailored to different data characteristics, balancing between accuracy, computational cost, and robustness to noise/outliers • Can handle and analyse very large volumes of data efficiently	• The effectiveness of the outcome relies on having ample, reliable, and varied data to learn from.	Corrosion rate Corrosion-fatigue crack growth Corrosion defect growth	Automated corrosion detection by image processing, corrosion severity prediction, and defect depth estimation in infrastructure	[23,42,43]
Multiscale Modelling	• Bridges multiple length/time scales • Provide a comprehensive modelling approach • Provide synergistic accuracy enhancement • Reduces computational cost vs. full quantum	• Complex implementation and validation • Error propagation between scales • High computational demands • Interface definition challenges	Hierarchical predictions Validated mechanisms Corrosion inhibitors and alloy design guidelines	Design of corrosion-resistant alloys, modelling corrosion at grain boundaries and interfaces	[19,44]

Table 1. Key strengths, primary limitations, main outputs, and recent applications of computational methods used in corrosion studies.

Method	Key Strengths	Primary Limitations	Main Outputs	Recent Applications	Refs.
Density Functional Theory (DFT)	•Provide a good balance between accuracy and computational cost •Excellent for surface adsorption and reaction energetics, and electronic structures •Reliable prediction of thermodynamic properties	•Fail to capture long-range dispersion forces accurately • Less feasible for large systems •Restricted to short simulation durations	Adsorption energies, defect formation energies, and interlayer relaxation Electronic Properties: Work functions, density of state, Mulliken population, and Bader charge analysis Magnetic Properties	Adsorption and passivation mechanisms, identifying reactive sites and surface stability, and inhibitor design	[23,35,36]
Ab Initio Molecular Dynamics (AIMD)	•Forces on atoms are derived from quantum calculations, yielding highly accurate structures and bonding •Provide kinetic and thermodynamic insights •Ideal for understanding reaction pathways and electronic interactions. •Highly transferable across different surface types, adsorbates, and environmental conditions	•Computationally expensive, limiting its application to large systems and long simulation times •Convergence issues for complex systems	Diffusion coefficients and reaction rates (short-time scale) Reaction pathways and barriers Dissolution energies, chemical binding, and charge transfer	Water–metal interfaces, oxide nucleation, and dissolution corrosion	[34,37]
Reactive Force Fields (ReaxFF)	•Enable the study of chemical reactions with bond formation/breaking • Capability to model the effects of temperature, pH, and aggressive ions on corrosion rate and mechanism • Enables modelling of complex reactive systems over a long time	• Depends on empirical parameters, limiting its accuracy and transferability across different systems • Needs reparameterization when applied to a new chemical environment (e.g., introducing new alloy elements, different pH, or complex oxidizers)	Diffusion coefficient and reaction rates Oxide layer thickness	Growth of oxide layers, pit formation studies, and stress corrosion	[34,38]
Monte-Carlo and Kinetic Modelling	• Computationally efficient for large systems and high-dimensional phase spaces. • Effectively samples equilibrium configurations and capable of simulating short-range order in high-entropy or multi-element alloy surfaces exposed to corrosive environments • Successfully simulate the kinetics of corrosion processes at different scales	• Relies on empirical or fitted potential energy models, which limits its transferability and accuracy for systems with complex chemistries	Phase diagrams Order–disorder transitions in surface adsorbates or alloy atoms Corrosion mass gain and corrosion rates Equilibrium surface coverage and segregation profiles	Kinetics of passive film formation, alloy segregation behaviour, and modelling of corrosion inhibitors	[39,40]
Finite Element method	• Provide macroscale structural analysis • Capability to handle complex geometries, and multi-field coupling	• Lacks the capability to capture atomistic-scale phenomena •Depends on precise material properties and constitutive models to generate reliable results	Stress distributions Mechanical strength and structural integrity Life span of materials	Pipeline corrosion assessment, coating failure analysis, and galvanic corrosion modelling	[23,41]
Machine learning and AI modelling	• Can integrate experimental and simulation data to predict the severity of corrosion • Can be tailored to different data characteristics, balancing between accuracy, computational cost, and robustness to noise/outliers • Can handle and analyse very large volumes of data efficiently	• The effectiveness of the outcome relies on having ample, reliable, and varied data to learn from.	Corrosion rate Corrosion-fatigue crack growth Corrosion defect growth	Automated corrosion detection by image processing, corrosion severity prediction, and defect depth estimation in infrastructure	[23,42,43]
Multiscale Modelling	• Bridges multiple length/time scales • Provide a comprehensive modelling approach • Provide synergistic accuracy enhancement • Reduces computational cost vs. full quantum	• Complex implementation and validation • Error propagation between scales • High computational demands • Interface definition challenges	Hierarchical predictions Validated mechanisms Corrosion inhibitors and alloy design guidelines	Design of corrosion-resistant alloys, modelling corrosion at grain boundaries and interfaces	[19,44]

2. Surface Slab Setup for Simulation

Steel is a polycrystalline material consisting of numerous individual crystal grains, with its structure primarily governed by the arrangement of iron atoms. Iron exists in two main forms: body-centred cubic (BCC) and face-centred cubic (FCC).

Figure 3a–d show the target slabs (e.g., finite portion of the crystal structure that represents the bulk metal truncated to expose specific surfaces) and the corresponding atomic arrangements for BCC and FCC crystal surfaces of Fe, respectively [7,45]. Experimentally, it is well known that the low-temperature BCC phase of iron, α-Fe, undergoes a structural phase transition to the face-cantered cubic (fcc) γ-Fe phase at approximately 1185 K, followed by a second transition to the high-temperature bcc δ-Fe phase at around 1667 K [46]. These structural transformations are closely coupled with magnetic transitions which influence the thermodynamic behaviour of iron [47,48].

In both BCC and FCC crystal surfaces, crystal planes are described by Miller indices, designated as (hkl) [45,50]. Seven most densely packed surfaces in BCC iron are Fe(100), Fe(110), Fe(111), Fe(210), Fe(211), Fe(310), and Fe(321), as in Figure 3e [49]. The stability trend typically follows: Fe(110) > Fe(100) > Fe(111) [51,52], driven by the Fe atomic packing densities and the corresponding surface energies. For example, the Fe(110) surface exhibits the highest atomic density, with atoms more closely packed, compared to Fe(100) and Fe(111) [53,54,55]. This dense packing leads to a lower surface energy, which in turn translates into a more thermodynamically stable configuration, making Fe(110) the most stable surface among the planes of BCC iron [13,51,56]. Low-index Fe surfaces are commonly used as prototypical substrates in various studies for surface properties. Among them, Fe(110) is the most frequently studied due to its superior stability, while Fe(100) is often included because it is the most reactive surface among the low-index planes [54]. For FCC iron surfaces, the Fe(111) surface is the most stable, following the stability order Fe(111) > Fe(100) > Fe(110) [5,57].

3. DFT Based Approaches in Corrosion Studies

This section reviews studies that employ DFT calculations to investigate metal corrosion, with a particular focus on those using the Vienna Ab initio Simulation Package (VASP) [58,59], a widely utilised tool for surface studies. While VASP is the primary software in this field, other computational packages such as CASTEP [60], Quantum ESPRESSO [61], WIEN2k [62], ABINIT [63,64], and CP2K [65] are also available for similar purposes. Table 2 comprehensively compares the most used DFT software packages in corrosion research, highlighting their key characteristics, including basis set types, computational requirements, licensing models, and recent applications in the field. This overview helps researchers select the most appropriate software based on their specific computational resources, budget constraints, and research requirements.

VASP is a planewave-based code for first-principles DFT calculations, optimised for efficiently resolving the electronic structure of surfaces and delivering accurate results for large atomic systems [66,67]. Its significance in Fe surface simulations is well established, providing valuable insights into key surface phenomena through computational experiments [68]. Additionally, VASP facilitates the modelling of bulk systems with built-in functions for generating various lattice structures, including BCC, FCC, and hexagonal lattices, enabling users to construct and manipulate complex crystal geometries with ease [69].

Table 2. Overview of widely used DFT software in corrosion science, categorised by basis set type, computational demand, licensing, and recent applications in corrosion.

Software	Basis Set	Computational Demand	Licensing	Recent Applications in Corrosion
VASP	Plane-wave	High	Commercial	[2,5,70,71]
CASTEP	Plane-wave	Moderate	Open source	[10,72]
Quantum ESPRESSO	Plane-wave	Moderate	Open source	[73,74]
WIEN2k	Full-potential LAPW	High	Commercial	[75,76]
Gaussian	Localised/flexible	Flexible	Commercial	[77]
ABINIT	Plane-wave	Moderate	Open source	[78]
CP2K	Hybrid (Gaussian + PW)	Lower	Open source	[52]

Table 2. Overview of widely used DFT software in corrosion science, categorised by basis set type, computational demand, licensing, and recent applications in corrosion.

Software	Basis Set	Computational Demand	Licensing	Recent Applications in Corrosion
VASP	Plane-wave	High	Commercial	[2,5,70,71]
CASTEP	Plane-wave	Moderate	Open source	[10,72]
Quantum ESPRESSO	Plane-wave	Moderate	Open source	[73,74]
WIEN2k	Full-potential LAPW	High	Commercial	[75,76]
Gaussian	Localised/flexible	Flexible	Commercial	[77]
ABINIT	Plane-wave	Moderate	Open source	[78]
CP2K	Hybrid (Gaussian + PW)	Lower	Open source	[52]

To simulate corrosion processes on an iron surface, selecting appropriate computational parameters is essential to ensure accuracy. This includes choosing a suitable exchange-correlation functional, vdw corrections, pseudopotentials, spin polarisation settings, and other critical parameters. For the exchange-correlation functional (Vxc), Perdew’s ‘Jacob’s ladder’ provides a hierarchical framework for understanding the evolution and complexity of density functionals in DFT, as illustrated in Figure 4. For corrosion studies, the generalised gradient approximation (GGA) [79,80] is most commonly applied, with the Perdew–Burke–Ernzerhof (PBE) [81] functional being the predominant choice for several reasons. It provides a significantly improved description of chemisorption energetics for adsorbates on transition-metal surfaces, correcting the severe over-binding observed in LDA [82], and balanced computational cost and accuracy for the transitional metal system [83]. For example, Jonathon et al. benchmarked the performance of various Vxc by comparing properties such as calculated interatomic distances, cohesive energies, and bulk moduli of 3d, 4d, and 5d transition metals with experimental data [83]. Their results showed that the PW91 and PBE Vxc functionals produced highly accurate results across the transition metal series, consistently yielding the smallest deviations from experimental values [83]. Other Vxc functionals, such as hybrid functionals, meta-GGA functionals, and correction schemes like the Hubbard-U scheme (GGA + U) are also employed to address specific limitations of conventional GGA Vxc functionals, especially for systems with strongly localised electrons (e.g., metal oxides) [32,79].

Exchange functionals inadequately represent long-range dispersion forces, necessitating the introduction of van der Waals (vdW) corrections to properly model physisorption effects, surface–water interfaces, and interactions with organic molecules such as corrosion inhibitors [80,85,86]. The inclusion of vdW corrections has been shown to significantly influence the predicted properties of metallic surfaces, including iron [87,88,89,90,91,92]. For example, Patra et al. demonstrated that neglecting vdW interactions in PBE calculations results in substantial errors in surface energies and work function values of metallic surfaces [87]. Several vdW correction methods exist, including semi-empirical approaches (DFT-D2, DFT-D3, and DFT-D4), non-local functionals (vdW-DF, vdW-DF2, and DF-cx), and many-body dispersion methods. Among these, the DFT-D3 (Grimme-D3) method performs well in predicting magnetic moments [93], and for systems where atoms are densely packed or highly interactive [80]. This strength is notable, as accurate magnetic moment prediction has a great impact in modelling certain materials such as Fe. Further, studies confirmed that PBE + D3 accurately models water adsorption on Fe(110) surfaces [37] while maintaining the electronic structure integrity [54,93].

Non-local functionals (vdW-DF, vdW-DF2, and DF-cx) incorporate dispersion interactions by adding a correlation term calculated from electron densities at different spatial points simultaneously [94]. For Fe surfaces, DF-cx excels in predicting structural and mechanical properties, such as lattice constants, cohesive energy, elastic constants, and bulk modulus. DF-cx underestimates the lattice constant by only 1.39% and overestimates cohesive energy 19.2% [93], making it a preferred choice for studies focused on structural and mechanical properties of bulk and surface iron. The optPBE functional also shows promising results for iron systems, with performance closely aligned with DF-cx [95]. The SCAN + rVV10 method, which incorporates long-range vdW interactions, often provides accurate predictions for many metals. However, it has significant limitations when applied to iron (Fe) systems, particularly in reproducing magnetic moments [96]. Incorporating long-range vdW corrections does not universally improve accuracy, as their effectiveness can be system-specific [93].

Table 3. Comparison of DFT and available experimental data for crystallographic orientations of BCC iron from related literature.

Method	Orientation	Surface Energy (J/m2)	Magnetic Moment (µB)	Work Function (eV)	Refs.
PBE	Fe(100)	2.59	2.97	3.92	[2]
Expt	Fe(100)	2.36 a	-	4.67 b	a—[97], b—[98]
PBE	Fe(110)	2.41	2.60	4.75	[99]
PBE DF CX	Fe(110)	2.46	2.13	5.06	[93]
Expt	Fe(110)	2.41 c	2.90–3.10 d	4.80 e	c—[100], d—[101], e—[102]
PBE	Fe(111)	2.76	2.87	3.87	[2]

^a–e—superscript letters correspond to specific reference numbers as listed in the reference section.

presents a summary of structural, electronic, and magnetic properties of low-index Fe surfaces calculated with different functionals compared to experimental values. Therefore, selecting an appropriate vdW correction method is essential in DFT calculations to accurately model metal properties, as different corrections can influence structural, electronic, and magnetic characteristics in distinct ways [80].

Pure iron (Fe) exhibits ferromagnetism at temperatures below its Curie temperature (Tc ~ 1043 K) [103], which must be explicitly accounted for in DFT calculations [104]. Spin-polarised settings are therefore essential when studying iron surfaces, allowing the incorporation of ferromagnetic (FM) ordering that characterise Fe systems [105]. However, certain iron compounds and alloys can exhibit antiferromagnetic (AFM) or non-magnetic behaviour [106,107]. When performing DFT calculations for such systems, it is important to initialise the calculation with the correct magnetic structure to obtain accurate results for electronic properties, surface energetics, and reaction mechanisms [70]. Meanwhile researchers sometimes need to consider both FM and AFM configurations, especially when dealing with iron oxides or other corrosion products that form on the surface [106,107].

A pseudopotential is an approximation used to simplify the treatment of core electrons while accurately describing valence electron interactions. It replaces the full Coulomb potential of the atomic nucleus and core electrons with a smoothed, effective potential, reducing computational cost without significantly affecting accuracy. VASP primarily employs PAW pseudopotentials, which retain all-electron accuracy while improving efficiency for plane-wave calculations [60,108]. A reliable PAW dataset for iron is typically provided within VASP’s built-in library, ensuring it accurately represents the valence electrons (e.g., Fe-cv for including semi-core p states). The superiority of PAW over other pseudopotential types has been demonstrated in various studies. For example, in DFT simulations of iron, comparisons between PAW and Vanderbilt ultrasoft pseudopotentials have shown that PAW yields local magnetic moments that are in closer agreement with experimental values, whereas ultrasoft pseudopotentials tend to overestimate these values [108]. Therefore, the PAW method, introduced by Blöchl, offers a hybrid approach that retains the advantages of pseudopotentials while enabling the recovery of full-core electron wave functions, making it suitable for applications involving complex materials, like transition metals [66,108,109].

Based on the comparative analysis in

Table 4. Comparison of recent DFT applications on Fe surface–oxidation studies.

Study	Surface	DFT Vxc	Software	Details *	Basic Set #	Ref.
Impact of oxygen coverage, diffusion, and defects on oxidation	Fe(100), (110)	PBE	VASP	3 × 3 9 a 15 Å	500, 551 −0.01	[99]
Dissociative adsorption of O2 and co-adsorption of H2O and O2	Fe(100)	PW91	CASTEP	2 × 2 5 a 15 Å	340, 441 −0.03	[10]
Properties of the oxide/metal interface	Fe(001)	PBE + U	VASP	2 × 2 9 a 27 Å b	500, 441 d −0.01	[70]
Adsorption of O on stepped iron surfaces	Fe(110), (210), (211)	PBE DF-CX	VASP	- 7(3), 13(5), 21(7) 15 Å	520, 2π × 0.04 Å−1 −0.001	[93]
Cl adsorption: α-Fe2O3 vs. α-Cr2O3 (0001) surfaces	α-Fe2O3 (0001), Cr2O3 (0001)	PBE + U	VASP	2 × 1 18 c 15 Å	500, 361 −0.02	[115]
Cl induced depassivation (interactions with Cl, O, OH, and H2O)	Fe(100)	PBE + D3	VASP	2 × 2 5 a 15 Å	400, 441 −0.01	[113]
Sequential and simultaneous adsorption of chloride and sulphate	Fe(100)	PW91	CASTEP	4 × 4 5 a 15 Å	340, 441 −0.03	[22,116]
Properties of Ni/Fe interface	Fe(100)	PBE	CASTEP	2 × 2 6 a 15 Å	400, 551,−0.001	[117]

* Details of the lab model Super cell, total layers (fixed layers), and vacuum size. ^# Basis set information including three parameters (cut-off energy in eV, k-points, and EDIFFG in eV/Å). ^a 3 bottom fixed. ^b Min of 14 Å after adding FeO layers. ^c 4 bottom fixed. ^d Optimisation and 12121 Static calculations.

, several key computational requirements emerge for corrosion studies, which often involve charge transfer and adsorbate interactions. Accurate modelling of the Fe surface requires a sufficiently high plane-wave energy cutoff (usually at least 400–500 eV) and a dense k-point mesh (e.g., 4 × 4 × 1 for an Fe(100) surface with a (3 × 3) supercell), depending on the system size [2]. It is also important to optimise the surface structure and account for slab thickness and vacuum spacing to avoid spurious interactions [110]. Typically, slab models consist of 4–9 atomic layers, sufficient to accurately represent surface phenomena and adsorption behaviour [99,111,112]. For systems with adsorbates on a single surface, bottom layers are fixed to maintain bulk properties. When adsorbates are present on both surfaces, central layers are fixed instead to prevent artificial symmetry effects. Usually, a vacuum region of 15–20 Å is maintained between periodic images to eliminate undesired interactions across boundaries [113,114]. Figure 5 presents a comprehensive framework for DFT studies in iron corrosion, summarising the key methodological considerations discussed throughout this section. The flowchart illustrates the interconnected nature of DFT computational decisions, from selecting appropriate functionals and corrections to defining system modelling parameters and computational settings.

4. Corrosive Agents and Their Chemistry

4.1. Chemistry of Oxidation with Oxygen

Corrosion fundamentally involves the oxidation of metals (iron), where the metal atoms lose electrons and transform into metal cations (e.g., Fe → Fe²⁺ + 2e⁻). This oxidation process is the core chemical change in corrosion, often driven by the electrochemical potential difference and facilitated by oxygen or other oxidising agents. Recent studies for properties related to oxygen adsorption on iron surfaces, such as preferred adsorption site, coverage dependence, O₂ dissociative adsorption, dissolution and diffusion of oxygen, and oxidation corrosion, will be updated.

The role of oxygen in enabling or accelerating the oxidation process closely links it to the chemistry of oxidation and oxygen reactivity. When an oxygen atom (O) encounters an Fe surface, it initially undergoes physical adsorption, followed by chemisorption at specific sites. Hence, it is important to identify these preferred adsorption sites for information of the reactivity and stability of iron surfaces under various conditions. The Fe(110) surface features several possible adsorption sites, including long-bridge (LB), short-bridge (SB), and on-top (OT) positions, as well as pseudo-threefold hollow (TH) sites, as shown in Figure 6a [118]. In contrast, oxygen atom adsorption on the Fe(100) surface on different sites, such as four-fold hollow (FH), bridge (BR), and on-top (OT) [118], as indicated in Figure 6b [118]. Each site (position) represents a unique spatial relationship with the vicinity iron atoms of surface, and the stability of adsorption sites on Fe surface can be determined by energy order of the configurations.

The most stable adsorption site for O on Fe(110) is now consistently identified as the TH site across a range of coverages, resolving what had previously been a subject of debate [119,120,121]. On the Fe(100) surface, the FH is the most stable for O atom adsorption [99,118,122]. In addition, studies have explored O adsorption on various Fe surfaces and considering different orientations, including Fe(001) [70], Fe(210), and Fe(211) surfaces [93]. The selection of the preferred adsorption site is determined by the adsorption energy, with more negative values indicating greater stability. This adsorption energy, in turn, is strongly influenced by the geometry and structure of the adsorption site [93,99,118]. For example, White et al. identified two key coordination numbers (CNs) that influence the adsorption energy of O [93]. The first is the CN of the adsorbed oxygen atoms (CNO), which represents the number of surface iron atoms directly bonded to each oxygen atom. The next is the CN of the surface iron atoms in a clean surface, representing the number of neighbouring Fe atoms each surface atom is bonded to prior to adsorption [93]. As the CNO increases, the adsorption energies tend to decrease, suggesting a negative correlation between the CN of surface Fe atoms interacting with the adsorbed O and the adsorption properties [93]. Importantly, incorporating vdW corrections in DFT studies does not alter the most stable site for O, but it does reduce the calculated adsorption energies. For instance, on Fe(110), the adsorption energy shifts from −3.98 eV (PBE) [99] to −3.59 eV (DF-cx) [93], and on Fe(100), from −4.03 eV (PBE) [99] to −3.41 eV (PBE + D3) [113]. Thus, dispersion effects primarily moderate the strength of O binding while leaving site preference unchanged.

Surface coverage refers to the fraction or density of adsorption sites on a surface that are occupied by adsorbed species, such as atoms, molecules, or ions. Coverage dependence describes how a material’s properties or behaviour—such as adsorption energy, reaction kinetics, electronic properties, and structural changes—vary with surface coverage. Extensive investigations that adsorption energy changes with oxygen coverage on Fe surfaces have been conducted using DFT calculations [99,118,119]. As oxygen coverage increases from 0.2 to 1.0 ML on Fe(100) and Fe(110), adsorption energies become more negative, favouring higher coverage [99]. At high coverage, repulsive O–O interactions make further adsorption less favourable, with this effect more pronounced on Fe(110) due to its higher atomic density [99]. Oxygen exhibits a preference for dissociative adsorption on Fe surfaces, where O₂ molecules split into atoms upon contact, forming stronger bonds with iron [10,118]. For instance, Chen et al. investigated a single O₂ molecule in various configurations on the Fe(100) surface using DFT calculations [10]. The O₂ configuration induces a competition between O₂ dissociation and Fe-O bond formation on Fe(100) surface, highlighting Fe(100)’s strong affinity for dissociative adsorption [10]. Furthermore, it was demonstrated that O₂ dissociation on the Fe(100) surface is non-activated, meaning it occurs effortlessly without energy barriers—even at high coverages [118]. On the Fe(110) surface, however, adsorption behaviour is more coverage-dependent. At lower coverages, molecular adsorption is more common, while at higher coverages, dissociative adsorption becomes the dominant mechanism [118].

In an oxygen-rich environment, oxygen atoms initially adsorb and diffuse on the Fe surface. At lower coverages, these oxygen atoms remain stably bound to the surface [118]. As coverage increases, repulsive interactions drive O into the near-surface region, requiring a 2.74 eV barrier to penetrate, compared to 0.57 eV for bulk diffusion [99]. In both FCC and BCC structures, this penetration is driven by the availability of a tetrahedral and/or octahedral interstitial site [45]. Figure 6c,d illustrate the positions of these subsurface interstitial sites in the BCC structure with low formation energy. The octahedral interstitial sites emerge as the most favourable locations for oxygen atoms, making the sites energetically advantageous for oxygen dissolution [99,123,124]. Studies revealed that the diffusion of oxygen in perfect BCC Fe, suggesting that oxygen migrates between adjacent octahedral interstitial sites, passing through a tetrahedral site as the saddle point. Reported jump energy barriers for this migration vary slightly, with values of 0.526 eV [125], 0.512 eV [126], 0.46 eV [124], and 0.48 eV [127]. Comparatively, the dissolution of oxygen atoms into the near-surface region is more challenging on the Fe(110) surface than on Fe(100) [99].

Upon contact with iron, oxygen adsorbed on the surface undergoes dissociative chemisorption, creating a monolayer of oxygen atoms [17]. The electrons are then transferred from iron atoms to the oxygens, which results in the nucleation and development of thicker oxide films on the metal surfaces [17]. As oxygen coverage on iron surface increases, iron oxide formation initiates, with oxygen atoms aggregating and penetrating below the surface to form an ultrathin layer of iron monoxide (FeO) [17]. This process is critical, as the presence of additional oxygen leads to the gradual transformation of FeO into more stable phases, such as magnetite (Fe₃O₄) and hematite (α-Fe₂O₃) [17,70]. Notably, a residual FeO region remains at the buried iron/iron-oxide interface [70], highlighting the complex interplay between oxide composition and stability, as well as the significance of initial oxidation states in influencing the overall corrosion behaviour of iron. In materials containing chromium, such as stainless steel, atomic oxygen is more likely to be trapped by chromium (Cr) [128,129]. The dense chromium oxide barrier formed by the reaction of Cr with oxygen prevents further oxidation of Fe. This mechanism is generally recognised as a key factor contributing to the corrosion resistance of stainless steel [17]. However, one downside is the depletion of chromium content at the surface in oxygen-rich environments [17]. While the exact chemical composition of this passive film is complex, hematite (α-Fe₂O₃) and chromia (α-Cr₂O₃) are considered the dominant components of the outer and inner layers, respectively [115]. Studies have shown that hematite and chromia exhibit differing protective capacities. While both oxides can form protective layers, α-Cr₂O₃ is characterised by a wider bandgap and superior corrosion resistance compared to α-Fe₂O₃. This difference highlights the significant role of chromium in enhancing the corrosion resistance of stainless steels [115].

A recent study reveals passive layer formation on Fe surfaces [70]. In this study, the impact of FeO film thickness on the structural and electronic properties of the FeO/Fe(001) interface was investigated using DFT calculations [70]. It focuses on FeO films ranging from one to five monolayers compressed onto a relaxed Fe(001) substrate as shown in Figure 7. The DFT study finds strong adhesion between the FeO film and the Fe substrate, peaking at 4 monolayers of oxide film before decreasing beyond 4 monolayers, and the optimal energy configuration of the system was determined [70]. The results suggest that thinner (~4) oxide layers adhere well, while thicker films may be more susceptible to detachment. Notably, the bond strength between FeO and Fe substrate at the FeO/Fe interface is significantly weaker than that of intrinsic adhesion between Fe(001) slabs, indicating a complex relationship between film thickness and interface stability. Significant charge transfer from the Fe substrate to the FeO layer is observed next, with the Bader charge analysis showing variations of ±0.3–0.4e per atom [70]. A reduction in work function with increasing film thickness, except for a slight increase observed at 2 monolayers, indicates a transition to a true FeO/Fe interface [70].

4.2. Other Oxidation Agent

The breakdown of the protective oxide layer on Fe surfaces, known as de-passivation, is a critical issue in corrosion science. The stability and growth kinetics of this passive layer is governed by several factors: ion migration within the oxide film, oxygen concentration at the interface, and environmental conditions, including radiation exposure, temperature, pH, and mechanical stress [130,131,132]. Irradiation not only affects the thickness of the oxidation layer but also induces the structural modifications of the oxidation layer, further influencing its protective properties [132]. However, several factors—such as high chloride ion concentrations, mechanical damage, and pH fluctuations—can compromise this passive layer, exposing the underlying Fe surface to aggressive environments and accelerating corrosion [33]. The breakdown of passivity typically occurs through three primary mechanisms [115]: the (1) penetration of aggressive ions via an ion exchange process, leading to localised dissolution, (2) mechanical rupture of the oxide film due to stress or external forces, and (3) formation of cation vacancies, which destabilize the oxide layer and facilitate further degradation [115].

Chloride ions are particularly notorious for compromising the stability of the passive film, leading to localized corrosion [22,24,113]. Chloride ions (Cl⁻) can infiltrate the protective oxide layers, such as the passive layers of α-Fe₂O₃ and α-Cr₂O₃ of stainless steel [115]. Energy barrier calculations for chloride diffusion through oxide layers reveal that α-Fe₂O₃ has lower barriers than α-Cr₂O₃, indicating that chromia provides enhanced protection against chloride-induced corrosion [24,115]. This protective behaviour is further validated by the CDD analysis, as illustrated in Figure 8I [24]. The analysis demonstrates that Fe₂O₃ undergoes more substantial electron density redistribution compared to Cr₂O₃, indicating extensive charge reorganisation within the iron oxide structure during chloride interaction [24]. This pronounced charge rearrangement in Fe₂O₃ correlates with its reduced ability to resist Cl⁻ penetration. Although Cr₂O₃ is generally regarded as a protective oxide, it exhibits poor corrosion resistance under anodic conditions, especially when exposed to strong electric fields. The DFT-PBE results show that the work function of Cr₂O₃ drops significantly from 4.64 eV to 0.80 eV as the electric field increases to 3.0 V/nm. This reduction in work function leads to a stronger adsorption of H, O, and F ions and an increased loss of electrons from the Cr₂O₃ passive layer, ultimately resulting in severe corrosion. The corrosion resistance of Cr₂O₃ can be improved through doping; elements such as Ni, Zn, and Cd raise the work function and provide electronic stabilisation, while Ti, W, Al, and Zr are particularly effective at suppressing corrosion both kinetically and thermodynamically, offering improved protection under harsh conditions [133].

DFT studies show that the doping of Cr and Ni on Fe surfaces significantly modifies the adsorption properties and stability of the Fe surface, directly impacting corrosion resistance [117,134]. For example, increasing Cr content on the Fe(110) surface raises the work function and enhances the surface’s resistance to corrosion, while moderate Ni addition can be beneficial, but excessive Ni leads to a decrease in work function and potentially compromises the surface’s stability [134]. This composition-dependent susceptibility to chloride attack is also observed in advanced alloy systems. In dual-phase high-entropy alloys, such as Fe_1.125Ni_1.06CrAl, both the DFT modelling and experimental results reveal that the NiAl-rich phase is more prone to chloride adsorption and electron gain by chloride ions compared to the FeCr-rich phase. These findings underscore the importance of considering both the chemical composition and microstructural phase distribution when designing alloys for enhanced corrosion resistance in chloride-rich environments [135].

Localised corrosion, particularly pitting corrosion, is a significant concern in steel structures, especially when exposed to aggressive environments, such as marine conditions [22,136]. This phenomenon is commonly observed in the presence of chloride and/or sulphate ions, which can compromise the stability of the protective passive oxide layer typically formed on steel surfaces, ultimately leading to the breakdown of this protective film and initiating localised corrosion [22,136]. Understanding the co-adsorption behaviours of these species on iron and iron-based alloys is vital for elucidating the underlying mechanisms of localised corrosion [121].

Recent studies have employed various theoretical approaches to investigate these interactions. For instance, joint density functional theory (JDFT) has been used to examine the interactions between O atoms and Cl⁻ on Fe(110) surfaces, with and without Cr doping at different concentrations [121]. This study revealed that the presence of Cl⁻ weakens the oxygen–metal interaction, hindering the formation of a corrosion-resistant passivation film [121]. It is further demonstrated by the local density of states (LDOS) analysis, as illustrated in Figure 8II. It shows a significant decrease in the hybridisation between the O 2p and Fe 3d states on surfaces pre-adsorbed with chlorine (Cl). This reduction reflects a weakened adsorption strength of oxygen (O) in the presence of Cl. In contrast, chromium (Cr) doping enhances this hybridisation, indicating a stronger bond between O and the surface. The LDOS confirmed that Cl disrupts the passivation process of the surface. Conversely, the presence of Cr alloying effectively counteracts this detrimental effect, thereby enhancing the corrosion resistance of the material [121].

Other DFT studies have shed light on the competitive interactions between chloride (Cl⁻) and sulphate ions (SO₄²⁻) [22,116]. SO₄²⁻ exhibit stronger surface interactions compared to Cl⁻, as evidenced by more negative adsorption energies and supported by project density of analysis (PDOS). The PDOS results demonstrate that SO₄²⁻ adsorption induces a more pronounced broadening of the Fe 3d electronic states compared to Cl⁻, confirming the superior binding affinity of SO₄²⁻ ions to the iron substrate, as illustrated in Figure 8III [22]. This stronger interaction leads to competitive adsorption effects that mitigate chloride-induced corrosion during the simultaneous adsorption take place [22,116]. However, Cl⁻ retains a higher catalytic efficiency in promoting corrosion once adsorbed [116]. Importantly, both experimental results and DFT calculations confirm that higher pH environments significantly suppress the adsorption of Cl⁻ and SO₄²⁻ ions on the iron surface, thereby reinforcing the stability of the passivation film and enhancing overall corrosion resistance [116].

4.3. The Role of Water in Corrosion Processes

Water plays a pivotal role in the corrosion of iron, acting as both a medium for electrochemical reactions and a reactant in the formation of corrosion products. This interaction of water with iron surfaces is necessary to understanding processes such as corrosion mechanism, electrochemistry, and catalysis as water serves as a catalyst in the process of iron corrosion (rusting) [137]. Under moist conditions, water molecules seep into small cracks and pits (defects) in iron, where they react with oxygen and other elements to form rust, which contributed significantly to material degradation.

Studies have revealed that water molecules preferentially adsorb at top sites in a parallel orientation on Fe(110) [51,54,138] surfaces. For Fe(100) surfaces, Chen et al. initially reported that the most stable configuration for molecular water is also parallel at the top site [10]. However, more recent grand canonical DFT studies have identified the H-down orientation, where one hydrogen atom points toward the surface, as the most stable adsorption mode compared to the other two configurations, as shown in Figure 9 [137]. Importantly, the strength of H₂O adsorption energy varies depending on the DFT functional and computational environment, as summarised in

Table 5. Comparison of water adsorption energies and configuration data on Fe surfaces.

Surface	DFT Vxc	Vacuum/Solvation	Eads (eV)	d(Fe-O) (Å)	Ref.
Fe(110)	PBE	Vacuum	−0.38	2.183	[138]
	PBE	Vacuum	−0.296	2.27	[54]
	PBE + D3	Vacuum	−0.50	2.25	[54]
	opt-PBE	Vacuum	−0.43	2.24	[95]
	PBE	Vacuum	−0.29	1.78	[51]
	PBE	Implicit solvation (Linear PCM)	−0.21	2.11	[51]
Fe(100)	PW91	Vacuum	−0.36	2.248	[10]
	PBE	Vacuum	−0.42	2.25	[139]
	PBE a	Vacuum	−0.50	2.19	[139]
	opt-PBE	Hybrid solvation (explicit-implicit)	−0.69	2.92 [d(Fe-H2O)]	[137]

^a 10% stretched Fe(100) surface.

. As shown in this table, van der Waals corrections tend to enhance the adsorption energy, while solvation effects have minimal impact on adsorption energetics. However, these variations do not alter the preferred adsorption site or geometry.

The strength of the interaction varies between surfaces. Studies show that water molecule adsorption on Fe(100) is relatively weak [10], while on Fe(110), water molecules exhibit remarkable mobility, easily migrating between adjacent top sites with minimal energy barriers (0.10 eV) [138]. Additionally, water molecules on Fe(110) can rotate perpendicular to the surface with a low energy barrier 30 meV, maintaining a stable adsorption energy [51,54].

The behaviour of water adsorption is highly dependent on surface coverage. At low coverage, water behaves as isolated adsorbates with consistent Fe–O bond lengths, and dissociation is more favourable due to minimal intermolecular interactions [10,138,140,141]. As coverage increases, adsorption energies become more negative, but hydrogen bonding between water molecules weakens water–surface interactions and favours molecular adsorption over dissociation [95,138,140,141]. Thus, water dissociation is most pronounced at low coverage, while molecular adsorption dominates at higher coverage on both Fe(110) and Fe(100). It should be noted that reported adsorption energies and energy barriers for water dissociation on Fe surfaces vary across studies, depending on the chosen functional, surface model, and computational settings. For Fe(110), calculations with the PBE functional report a dissociative adsorption energy of −1.66 eV with an energy barrier of +0.68 eV [138]. The opt-PBE functional yields a slightly more negative adsorption energy of −1.68 eV, likely due to the inclusion of vdW interactions [95]. For Fe(100), earlier studies using the PW91 functional reported an adsorption energy of −1.05 eV [10]. However, a subsequent study that incorporated vdW corrections in a hybrid explicit/implicit solvation model found a slightly lower adsorption energy, with an associated energy barrier of +0.59 eV [137]. A comparative summary of these key findings for Fe(110) and Fe(100) is provided in

Table 6. Comparative adsorption and diffusion properties of O₂ and H₂O on BCC Fe(110) and Fe(100) surfaces.

Agents	Behaviour/Property	Fe(110)	Fe(100)
O/O2	Preferred adsorption site	3-fold hollow	4-fold hollow
	Coverage dependence	With increasing coverage, O–O repulsion makes adsorption less favourable, due to the high atomic density Low coverage—molecular adsorption is common High coverage—dissociative adsorption is favoured	Dissociative adsorption is favoured
	Diffusion	Surface to subsurface barrier: 2.74 eV Bulk diffusion: 0.57 eV	Surface to subsurface diffusion, easier than Fe(110)
H2O	Adsorption	Stable adsorption with high mobility	Weak adsorption, enhanced with pre-adsorbed O
	Coverage dependence	Low coverage: molecular adsorption is preferred High coverage: dissociative adsorption is preferred	Low coverage: molecular adsorption is preferred High coverage: dissociative adsorption is preferred

.

The presence of pre-adsorbed species on Fe surfaces can significantly influence the stability and dissociation behaviour of subsequently adsorbed H₂O. For low-index Fe surfaces: Fe(110) [140], Fe(100) [112,137], and Fe(111) [112], oxygen pre-coverage significantly enhances H₂O dissociation by lowering the energy barriers, promoting the formation of OH and H species. However, it was found that after a certain limit of oxygen coverage, H₂O desorption becomes more favourable than dissociative adsorption [138,140]. Impurities or alloying atoms on Fe surfaces can significantly influence H₂O dissociation by lowering energy barriers, with both the preferred dissociation site and ease of dissociation depending on the specific impurity [142,143]. Understanding the corrosion mechanism requires considering the co-adsorption of both O₂ and H₂O, as these species are simultaneously abundant in real-world corrosion environments [10,144]. Recent findings indicate that when both O₂ and H₂O co-adsorb on the bare Fe(100) surface, the introduction of an additional oxygen atom promotes its interaction with Fe–OH species, resulting in the formation of the more stable FeOOH group [10]. This phase, as depicted in Figure 10I, is recognised as a predominant corrosion product under such conditions [10]. In contrast, Chew et al. investigated the role of Fe dimer surface defects in initiating and propagating corrosion on Fe(100) surfaces, focusing on the co-adsorption of O₂ and H₂O [144]. As shown in Figure 10II Fe dimer defects serve as active surface sites where O₂ and H₂O molecules interact, resulting in the formation of the Fe₂(OH)₂ + 2OH intermediate through successive reactions [144].

The solvation effect in DFT simulations of corrosion accounts for the influence of the surrounding environment, particularly the electrolyte solution, on surface reactivity and adsorption behaviour. In real-world corrosion scenarios, metal surfaces interact not only with water but also with dissolved ions, oxygen, and other species in the solution [145]. These interactions alter the electronic structure, charge distribution, and adsorption energies, significantly impacting reaction pathways such as oxidation, hydrogen evolution, or ion adsorption. Neglecting solvation effects can lead to inaccuracies in predicting corrosion rates, reaction mechanisms, or passivation behaviour [145,146,147]. Advanced solvation models, like implicit or explicit solvation schemes, provide a more realistic representation of the electrochemical environment, improving the reliability of DFT simulations in corrosion studies. To gain a clearer and more accurate understanding of corrosion mechanisms, it is crucial to consider solid/liquid interfaces [145].

There are two primary approaches for modelling solid/liquid interfaces in DFT studies: the explicit and the implicit solvation models (SM), or a hybrid of both (hybrid implicit/explicit model) [148,149,150]. The explicit SM treats each solvent molecule individually, providing a detailed depiction of solute–solvent interactions by optimising both electronic and ionic configurations of the solute and solvent molecules. Although this model offers high accuracy, it is computationally intensive due to the need for numerous solvent molecules and extensive averaging over solvent configurations, making it less feasible for large systems [51,148,150]. Tobon et al. investigated the explicit solvation effects on low-index Fe surfaces, particularly Fe(111), and small Fe particles, focusing on their role as adsorbents for arsenic species. Their results revealed that adding explicit water molecules enhances adsorption energies, particularly on hydroxylated surfaces, suggesting stronger van der Waals (vdW) and hydrogen bonding interactions [146].

The implicit SM simplifies the treatment of the solvent by representing it as a continuum dielectric medium, characterised by a specified dielectric constant [148,149,151]. This model allows the solute to be treated quantum-mechanically while the solvent effects are captured through charge redistribution [151,152], making it computationally more efficient. The implicit solvation method is particularly useful for systems where electrostatic interactions are critical, such as for polar or ionic solutes in polar solvents [148]. Yin et al. applied JDFT coupled with the CANDLE approach to study the effect of solvation on the adsorption of O, H, OH, and H₂O on the Fe(100) surface. They also explored the interactions between O, Cl, and Fe(110) alloy surfaces with and without Cr doping. Their results showed that solvation had minimal impact on surface energy, adsorption energy, and the preferred adsorption sites, but it did reduce the work function of the system [51,121].

4.4. Hydrogen Embrittlement

Hydrogen embrittlement is a phenomenon where metals become brittle and prone to fracture due to the presence of hydrogen atoms within their structure. It typically occurs in high-strength metals like steel when they are exposed to hydrogen-containing environments, such as during electroplating, chemical reactions, or corrosion processes [153]. This begins with the diffusion of hydrogen into the metal lattice, causing internal stresses that weaken its mechanical properties. When the hydrogen concentration exceeds the critical threshold, embrittlement occurs, significantly increasing the material’s susceptibility to failure [154].

The stability of the H-adsorption on the surface depends on the site orientation and the surface. For example, in the BCC system, the most preferred adsorption site is the 3FH site and the second most stable site is the LB site, (with a slightly shorter distance from the surface) on the Fe(110) surface [51,56,140]. The top (T) site is found to be the least stable adsorption site on the same Fe(110) surface [51]. However, the bridge (B) site found to be the most stable followed by hollow (H) site, while the T site is not stable for H adsorption on the Fe(100) surface [155]. In the FCC system, the most stable adsorption sites are 4FH site on the Fe(100) surface, SB site on the Fe(110) surface, and 3FH on the Fe(111) surface [156]. Both FCC and BCC structures have been considered in hydrogen embrittlement studies because Fe and its alloys can exist in either phase, depending on temperature and application, such as in nuclear reactors or hydrogen storage systems, where high temperatures are common.

Hydrogen prefers to remain on the surface of Fe rather than diffusing into the bulk, due to the small interstitial spaces in Fe and its high mobility [157]. This diffusion property varies based on the lattice type and crystallographic orientation [156,157]. For example, Xing et al. conducted a DFT study to investigate the crystallographic orientation dependence of BCC and FCC Fe lattices on H diffusion [155]. The obtained energy barrier as a function of the Bader charge is shown in Figure 11I. The hydrogen diffuses more slowly on the Fe(111) surface than on the Fe(001) and Fe(101) surfaces in FCC-Fe, whereas hydrogen diffuses faster in BCC-Fe under the same conditions, due to different diffusion energy barriers and charge uniformity [155]. In FCC-Fe, the Fe(001) and Fe(101) surfaces exhibit smaller values for these factors compared to the Fe(111) surface. In contrast, in BCC-Fe, the Fe(001) and Fe(101) surfaces show larger values than the Fe(111) surface [155]. The diffusion behaviour of H also depends on H concentration, temperature, and isotope. Both molecular dynamics and DFT studies show that hydrogen diffusivity increases with temperature and decreases with increasing H concentration, especially above 1%, where H–H interactions and cluster formation hinder mobility and can induce local structural changes in the Fe lattice [158,159]. Isotope effects are also significant: lighter isotopes (H) diffuse faster than heavier ones (D, T), primarily due to quantum tunnelling, which is especially pronounced below the crossover temperature (e.g., ~190 K for H) [158].

Passive oxide films that typically inhibit H absorption can be compromised under mechanical strain from gravitational, thermal, and dislocation-induced stresses [160]. These strains are categorised into hydrostatic, uniaxial (tensile/compressive), and shear strains [161]. Electronic properties of α-Fe₂O₃ films, are highly sensitive to the type of strain, magnitude, as well as the surface termination [162]. For example, compressive and tensile strains have opposite effects on the work function of Fe-terminated α-Fe₂O₃ films. Compressive strain significantly decreases the work function, from 6.91 eV at +3% strain to 4.04 eV at 6% strain, which lowers the corrosion potential and makes the material more susceptible to corrosion. In contrast, tensile strain increases the work function, with values rising above the unstrained state as the strain becomes more positive [162]. Strain directly impacts H adsorption at oxide interfaces, where tensile strain weakens Fe–O bonds, enhancing H adsorption, while compressive strain reduces H adsorption up to 3% compression [160]. This scenario is illustrated in Figure 11II, which demonstrates how tensile stress influences various properties, including adsorption energy, distortion energy, binding energy, and others [160].

The influence of mechanical stress on H diffusion extends beyond conventional hydrostatic or uniaxial strain. Álvarez et al. showed that while hydrostatic stress has negligible effects on bulk H diffusion, uniaxial and especially shear stresses can dramatically alter H diffusion barriers. Shear stress can reduce barriers by up to 100%, potentially doubling H mobility even at moderate stress levels [161]. Additionally, mechanical stress significantly alters iron surface chemistry, where a 10% tensile strain on Fe(100) strengthens H₂O adsorption, lowers H₂O dissociation barriers, and stabilises reactive OH and H fragments [139]. Therefore, the proper consideration of strain effects in corrosion models is important for understanding and preventing stress corrosion cracking failures in load-bearing infrastructure operating in moisture-rich service environments.

5. Modelling Defects on the Iron Surface

Defects in materials are imperfections or irregularities in the atomic structure of a solid, which can range from point defects to larger-scale defects [45]. In the context of corrosion, defects often serve as initiation sites for chemical reactions, influencing the material’s susceptibility to degradation [19,163]. For instance, vacancies and dislocations can enhance the diffusion of corrosive species like oxygen, water, and chloride ions, while grain boundaries may act as pathways for accelerated corrosion [125,164,165]. As a result, the macroscopic behaviour of metallic materials, particularly iron (Fe) and its alloys, is heavily influenced by the presence of defects within their structure. Defects are commonly categorised based on their geometry and dimensionality: point defects, line defects (one-dimensional), planar defects (two-dimensional), and volume defects (three-dimensional) [45,166].

5.1. Modelling the Effects of Point Defects

Point defects are zero-dimensional defects that consist of vacancies, where atoms are missing, or impurities, which are additional atoms located at non-regular sites within the lattice [45]. Figure 12 illustrates a visual representation of point defects commonly found in crystalline solids [167]. Among these, vacancy defects are considered fundamental, which occur when an atom is absent from its normal lattice site [45].

The minimum energy required to create a vacancy within the crystal structure is called the vacancy formation energy (E_x) [168]. It is an intrinsic characteristic of vacancy defects that determines their equilibrium concentration at finite temperatures [169]. The vacancy formation energy is generally lower at the surface (~0.9 eV) than in the bulk (~2.0 eV), as reported in many theoretical studies [54,99,170]. This is because surface atoms have fewer bonding interactions compared to bulk atoms, making it easier to remove an atom from the surface. DFT studies using the PBE functional yield a bulk vacancy formation energy of 2.15 eV [171], which shows reasonable agreement with positron annihilation spectroscopy experiments that report formation energies of 1.5 eV for pure iron and 2.0 eV for carbon steels. The energetic favourability of vacancy formation promotes the migration of vacancies from the bulk to the surface or adjacent voids [105], where they can aggregate and form vacancy clusters or super vacancies [163,170,172]. For example, Ahlawat et al. found that the formation of vacancy clusters at nearest neighbour is energetically feasible in both Fe-lattices. However, they noted that the ability of vacancies to trap foreign atoms is significantly stronger in the BCC-Fe lattice compared to the FCC-Fe lattice [171].

Vacancy clusters act as trapping centres for impurity atoms, such as C, Cr, H, and O, significantly affecting their dissolution and diffusion behaviour [54,99,173]. Figure 13 illustrates how vacancies influence the migration path of oxygen atoms on the Fe(110) surface [99]. Due to their strong binding energies, vacancies form stable interstitial atom–vacancy pairs [99,125,174]. This interaction is believed to be the underlying mechanism for the high solubility of interstitial atoms in vacancy-defected lattices. For instance, Domain et al. found that vacancies lower the solution energy for oxygen, facilitating its dissolution and promoting its accumulation at defect sites [173]. However, studies on the effects of vacancies on Fe surfaces have yielded contradictory results. Shang et al. reported that vacancies increase the energy barrier for oxygen migration, thereby reducing its diffusion coefficient [125]. In contrast, Wang et al. demonstrated a strong mutual dependence between vacancies and oxygen, where vacancies enhance oxygen diffusion [175]. These findings highlight the complex role of vacancies in modulating both the solubility and mobility of oxygen in metallic lattices [175]. Further studies in this direction are warranted.

Chloride and sulphate-induced depassivation mechanisms are fundamentally governed by the formation and migration of point defects, particularly iron and oxygen vacancies, as described by the point defect and ion exchange models [176,177]. According to the point defect model, aggressive anions like Cl⁻ and SO₄²⁻ lower the formation energies of these vacancies, facilitating defect accumulation at the oxide surface [178,179,180] and ultimately leading to passive film breakdown, leading to passive film degradation [179]. For example, Pang et al. found that Fe vacancy formation on a pristine α-Fe₂O₃ (0001) surface is highly endothermic, but Cl adsorption lowers this energy by approximately 18%, making vacancy formation more favourable [179]. Conversely, according to the ion exchange model, Cl insertion into an O vacancy is energetically more favourable than cation vacancies [115]. Thus, O vacancies in the oxide layer provide diffusion pathways for Cl, further destabilising the passive film [115,181].

Experimental studies increasingly support the applicability of these DFT findings to real-world corrosion phenomena. Electrochemical measurements (OCP, EIS, and polarisation) reveal that passive film breakdown occurs only when Cl⁻ concentration exceeds a critical threshold [182,183]. Mott-Schottky analysis reveals a significant increase in donor density (related to Fe and Ovacancies) after depassivation [184], while XPS and SEM confirm thinning of the passive film and the onset of pitting only above the threshold [182]. DFT explains this by demonstrating that Cl⁻ lowers the energy barrier for vacancy formation, making the film more susceptible to localised degradation, as mentioned above. For SO₄²⁻ X-ray photoelectron spectroscopy (XPS) revealed that increasing sulphate concentration leads to a higher abundance of O vacancies, as indicated by the intensified O vacancy peak in the O 1s spectra. Additionally, a decrease in the Fe³⁺/Fe²⁺ ratio was observed in the outer layer of the passive film, signifying preferential dissolution of Fe³⁺ species. These experimental findings are consistent with DFT results, which show that SO₄²⁻ adsorption on α-Fe₂O₃ surfaces weakens Fe–O bonds, reduces vacancy formation energies, and facilitates the generation of Fe–O pair vacancies, thereby destabilising the passive layer [177]. Similarly, in the hydrogen charging study, TEM and Raman show thicker, more porous oxides with more vacancies, and DFT explains this by showing that hydrogen weakens Fe–O bonds and promotes vacancy formation [180]. Furthermore,

Table 7. Key Findings: Effects of Vacancy and Grain Boundary Defects on Iron Corrosion.

Type of Defects	Study	Lattice/Surface	DFT Functional	Key Findings	Ref.
Vacancy	Effect of vacancies on oxidation	Fe(100), Fe(110) *	PBE	Vacancies form more easily at surfaces, enhance O dissolution and diffusion, and create stable V–O pairs	[99]
	Interactions of foreign interstitial atoms with vacancy defects	BCC, FCC	PBE	Vacancies trap foreign atoms much more strongly in BCC-Fe than in FCC-Fe	[171]
	Interactions of Cl with the α-Fe2O3 (0001) surface	α-Fe2O3 (0001)	DFT + U, PBE	Cl adsorption makes Fe vacancy formation more favourable; Fe vacancies are stable in the bulk, while O vacancies are stable on the surface	[179]
	Interaction of Cl− with hematite and chromia	α-Fe2O3 (0001), α-Cr2O3 (0001)	DFT + U, PBE	Cl insertion into an O vacancy is more favourable than into Fe vacancies or interstitials	[115]
GB	Σ3 and Σ5 grain boundaries of Fe	$\begin{array}{c}\mathsf{\Sigma }3\left(112\right)\left[\stackrel{-}{11}0\right]\end{array}$$,\begin{array}{c}\mathsf{\Sigma }5\left(310\right)\left[\stackrel{-}{00}1\right]\end{array}$*	PBE	Higher Σ GBs have higher energy and are more prone to degradation and adsorbate diffusion	[185]
	Dissolution and diffusion properties of O at GBs	Σ3<110>(111), Σ5<001>(310) *	PBE	GBs enhance O dissolution and diffusion; oxygen segregation is greater at Σ3(111) than at Σ5(310)	[186]
	Segregation of elements on the GB	FCC Fe3O4, Fe(001) #	DFT + U	GB strength varies with segregating element; Sn weakens, and Nb strengthens GB cohesion	[187]

* Lattice structure—BCC. ^# Lattice structure—FCC.

summarizes key findings from DFT-based vacancy studies.

5.2. Planar Defects and Their Modelling

Grain boundary, a type of planar defect, represents the interface that separates two adjacent grains or crystals with different crystallographic orientations in a polycrystalline material [45]. When comparing commercial-grade iron with high-purity iron, it is evident that the grain boundaries in commercial-grade iron are significantly more corroded. In contrast, high-purity iron, due to its reduced impurities and defects, demonstrates greater resistance to corrosion at these boundaries [144]. However, surface defects still serve as critical initiation sites for corrosion and other redox reactions, even in high-purity iron. Furthermore, chemical properties of fine-grained iron vary significantly from coarse-grained iron [188], highlighting the need for more information on grain boundaries (GBs). The grain boundary characteristics in metallic materials vary depending on the coincidence site lattice (CSL) value of the grain boundary [185,186]. Therefore, careful selection of the CSL value must be undertaken when modelling corrosion at grain boundaries. A higher value of Σ exhibits a higher grain boundary energy and lower work of separation compared to a lower value of Σ grain boundary, making it more susceptible to structural changes and degradation [185,186,189]. For example, Huang et al. found that hydrogen diffuses more easily at the Σ5 grain boundary compared to the Σ3 grain boundary in the same metallic material [185].

GB similarly act as sinks for corrosive species, serving as preferential sites for adsorption and dissolution and rapid diffusion along these boundaries, which contributes to localised oxidation and accelerates the corrosion process [186,190,191]. In their recent study, Liu et al. used first-principles calculations to study the dissolution and diffusion of interstitial O atoms near grain boundaries in BCC Fe, focusing on Σ3<110>(111) and Σ5<001>(310) boundaries [186]. They found that grain boundaries act as strong sinks for oxygen, with lower solution energies than the bulk octahedral sites, making them energetically favourable for oxygen accumulation, as illustrated in Figure 14 [186]. Diffusion behaviour at GB varies: Σ5(310) has higher energy barriers, causing oxygen to concentrate around the GB, while Σ3(111) allows easier diffusion towards the boundary, leading to greater oxygen segregation at Σ3(111) [186]. This serves as evidence for how the GB CSL value impacts against this sinking behaviour as well. However, scientists modelling such systems must carefully consider the extended nature of this sink effect, as oxygen accumulation extends 3–5 atomic layers from the GB plane, requiring appropriately sized supercells to capture the full segregation profile.

The physical and chemical properties of the Fe lattice are significantly influenced by the interactions of alloying metals with GBs, with the extent of segregation playing a central role in these interactions [192,193,194]. The extent of segregation is influenced by the affinity of these elements for GB sites [193], which in turn affects cohesion, local composition changes, surface energy, and corrosion resistance [187,192,194]. Depending on the type of segregating element, GBs can either be strengthened or weakened, altering the material’s overall performance [187,194]. For instance, Yuan et al. demonstrated that elements such as tin (Sn) and niobium (Nb) can segregate at the GBs of oxide films, influencing both cohesion and corrosion resistance. While Sn segregation weakens GB cohesion and reduces the protective properties of the oxide film, Nb strengthens GB cohesion, enhancing the film’s protective capabilities [187]. Moreover, Figure 15I provides a three-dimensional evaluation framework that maps various alloying elements based on their GB segregation tendency, interaction strength with hydrogen, and embrittling or strengthening effect on GBs [191]. When modelling grain boundary (GB) behaviour, it is important to account for both the varying segregation patterns of alloying elements at different concentrations [194] and the multiple diffusion pathways available to corrosive species [191], as demonstrated in Figure 15II,III. Additionally, Table 7 provides key findings from DFT GB studies.

6. Corrosion Prevention Strategies: Organic Corrosion Inhibitors

Organic corrosion inhibitors are widely used to protect iron and steel surfaces due to their ease of application, cost-effectiveness, and high efficiency, especially compared to traditional inorganic inhibitors, which often pose toxicity and environmental risks [195,196,197]. These organic compounds operate primarily by adsorbing onto the metal surface, where they form a protective layer that hinders the activity of corrosive species and disrupts the electrochemical reactions responsible for corrosion. DFT simulations enhance our understanding of organic corrosion inhibitors by predicting their performance using a range of molecular and electronic parameters. These include the energies of the highest occupied and lowest unoccupied molecular orbitals (E_HOMO and E_LUMO), the energy gap (ΔE), chemical hardness (ƞ) and softness (σ), electronegativity (χ), electrophilicity (ω), nucleophilicity (ε), and electron charge transfer (ΔN) [198,199]. These quantum chemical descriptors enable the establishment of reliable structure–activity relationships, allowing researchers to screen and predict the corrosion inhibition efficiency of organic molecules on Fe surfaces prior to the experimental validation. Although DFT has proven to be effective in quantitatively elucidating the structure–activity relationships, its application is limited by high computational demands, making it impractical for large organic molecules. This is a significant drawback, as large inhibitors often show superior performance due to multiple active sites and enhanced surface coverage. To address this, the DFTB (density functional tight binding) method has been developed, offering a practical alternative that enables qualitative analysis of large inhibitor–metal systems and the study of stable configurations and electronic properties with a much lower computational cost [198]. Many recent corrosion inhibition studies have successfully used DFTB to overcome DFT’s limitations for large molecules [198,199,200].

The effectiveness of organic inhibitors is closely linked to their molecular structure, particularly the presence of heteroatoms (such as N, O, and S), polar functional groups, extended π-conjugated systems like aromatic rings, and the length of hydrocarbon chains. These features facilitate strong interactions with the Fe surface through electron donation and back-donation mechanisms, enhancing adsorption and protective film formation [199,201,202,203]. For example, DFT reported that molecules like 1H-benzotriazole (with N atoms of triazole ring) exhibit a high inhibition efficiency due to their ability to strongly adsorb onto Fe surfaces through donor–acceptor interactions, as evidenced by their negative adsorption energies and small HOMO-LUMO gaps [203]. Similarly, DFT + AIMD studies showed that a large π-network of coumaric acid enables strong π–d interactions and covalent bonding to Fe(110), resulting in the highest adsorption energy (371.275 kJ/mol) among the studied green inhibitors [202]. Further, Figure 16 illustrates recent DFT-based studies on organic corrosion inhibitors, highlighting the structural diversity and adsorption mechanisms uncovered through computational approaches.

7. Conclusions

This review provides a comprehensive analysis of corrosion mechanisms on iron (and steel) surfaces, with a particular focus on the application of density functional theory (DFT) to investigate the electronic structure, adsorption behaviour, and interaction of corrosive species with Fe surfaces. Key topics include adsorption, co-adsorption, pre-coverage states, and diffusion processes. The geometry and electronic structure of adsorption sites play a crucial role in determining the adsorption behaviour of corrosive species, such as oxygen coverage on Fe surfaces, FeO layer formation, and the impact of oxygen concentration. The review also highlights the significance of point and planar defects—such as vacancies and grain boundaries—in facilitating the diffusion and accumulation of corrosive species, thereby accelerating localised corrosion. Alloying strategies, particularly the incorporation of chromium and other elements, are shown to enhance corrosion resistance by stabilising protective oxide films and mitigating the effects of aggressive ions like chloride.

Recent DFT-based atomistic simulations of gas–iron interfaces have been reviewed, with a focus on Fe(100), Fe(110), and Fe(111) surfaces. The accuracy of these simulations is discussed in the context of exchange-correlation (Vxc) functionals and long-range van der Waals (vdW) dispersion corrections. The role of surface chemical agents in catalysing oxidation and corrosion processes is also explored. Furthermore, the impact of surface defects, particularly vacancies and grain boundaries, is analysed, highlighting their role in enhancing the dissolution of corrosive species like oxygen, which leads to localised corrosion. Given the complexity of corrosion, a comprehensive model should incorporate defect effects, solvation interactions, and vdW corrections to accurately simulate environmental and electrochemical influences at the atomic scale.

Despite significant advancements in understanding iron corrosion, the mechanisms governing corrosion under defect-rich conditions and in the presence of moisture remain insufficiently explored. To address these gaps and establish best practices for future research, the following directions are recommended. The generalised gradient approximation (GGA) with the Perdew–Burke–Ernzerhof (PBE) functional is widely used for iron systems, with PBE + D3, optPBE, and DF-cx offering improved accuracy for dispersion and structural properties. Meta-GGA or GGA + U schemes should be considered for systems with strong electron localisation. To simulate solid–liquid environments, employ explicit, implicit, or hybsolvation models to accurately capture the effects of water molecules, electrolyte solutions, and aggressive ionic species on corrosion processes.

Future research should integrate multi-scale modelling approaches, such as coupling DFT with molecular dynamics (MD), to bridge the gap between atomic-scale mechanisms and macroscopic corrosion phenomena. Additionally, machine learning techniques present promising opportunities for predicting corrosion behaviour based on large datasets (to be established) and optimising materials for enhanced durability. A holistic approach that incorporates defect modelling, solvation effects, and advanced computational techniques will provide deeper insights into corrosion mechanisms, paving the way for the development of more corrosion-resistant materials and improved mitigation strategies.