Chemical Short-Range Order in Alloys

• Chemical Short-Range Order (CSRO) is the local deviation from a random solid solution, where atomic arrangements over a few coordination shells exhibit preferential bonding and finite-range correlations.
• CSRO is quantified using descriptors like the Warren–Cowley parameter and motif-based metrics, linking local chemical environments to observable scattering and thermodynamic behavior.
• CSRO significantly influences alloy properties such as defect kinetics, corrosion resistance, and mechanical response, providing a tunable parameter for advanced material design.

Searching arXiv for recent CSRO papers to ground the article in published work. Chemical short-range order (CSRO) is the departure of a crystalline solid solution from ideal random site occupancy over only a few coordination shells. In a CSRO state, the average lattice may remain single-phase and crystallographically disordered, yet the probability of finding species $j$ near species $i$ differs from the random value implied by bulk composition. CSRO therefore occupies the conceptual space between a perfectly random solid solution and long-range order: it is local, chemistry-sensitive, and often nanoscale, but not a separate superlattice phase. Across concentrated alloys, medium- and high-entropy alloys, and classical ordering systems, CSRO has emerged as a state variable that regulates local energetics, defect behavior, transport, passivation, and the pathway from disorder to ordered or transformed states (Chen et al., 7 Apr 2026, Islam et al., 2024, Bacurau et al., 29 May 2026).

1. Definition, scope, and physical meaning

In the standard random-solid-solution approximation, atoms occupy lattice sites with probabilities fixed only by global composition. CSRO denotes the local failure of that approximation: some motifs are statistically enriched and others depleted, even though the alloy remains chemically homogeneous on larger scales. In concentrated FCC CoNiV, for example, CSRO was identified as a pronounced V-centered ordering in which V atoms prefer Co and Ni as nearest neighbors while avoiding V-V nearest-neighbor contacts (Chen et al., 7 Apr 2026). In FCC CrCoNi, the same general phenomenon appears as non-random chemical mixing involving Cr, with favored unlike pairs and suppressed Cr-Cr nearest neighbors, despite preservation of a single-phase FCC average structure (Hsiao et al., 2022). In equiatomic CoCrNi, neutron diffraction showed that the average structure remains FCC while a broad diffuse feature near $Q \approx 1.85\,\mathrm{\AA}^{-1}$ reveals finite-range chemical correlations rather than long-range periodic order (Bacurau et al., 29 May 2026).

The distinction between CSRO and long-range order is fundamental. Long-range order breaks translational symmetry and produces sharp superlattice reflections. CSRO produces broad diffuse intensity in reciprocal space because the correlations are finite in extent. This is why CoCrNi can exhibit a diffuse neutron-scattering peak and nanoscale Ni-rich domains while remaining a single-phase FCC alloy with essentially unchanged Bragg peaks (Bacurau et al., 29 May 2026). The same distinction underlies the interpretation of local L1$_1$-type and L1$_2$-type nanoclusters in CrCoNi: they are chemically ordered motifs embedded in a chemically mixed matrix, not a fully developed ordered intermetallic (Hsiao et al., 2022).

A second conceptual distinction is between CSRO as a pairwise occupancy bias and CSRO as a many-body statistical state. Several recent studies argue that pair correlations are often insufficient because different many-body motifs can share the same nearest-neighbor composition or Warren–Cowley signature while differing strongly in probability, local distortion, and physical consequence (Sheriff et al., 2023, Sheriff et al., 2024). This suggests that CSRO is most rigorously understood as a distribution over local chemical motifs plus a spatial correlation structure over those motifs, rather than only as a matrix of pair parameters.

2. Quantification and descriptors

The most widely used CSRO descriptor in the alloy literature remains the Warren–Cowley short-range-order parameter. In one standard first-shell form,

$\alpha_{ij}(1)=1-\frac{P_{ij}(1)}{c_j},$

where $P_{ij}(1)$ is the conditional probability that a first-nearest neighbor of species $i$ is species $j$, and $c_j$ is the overall concentration of $i$0 (Chen et al., 7 Apr 2026). Equivalent shell-resolved forms appear throughout recent work on multicomponent alloys, including MoNbTaVW, CoCrNi, CoCrNi corrosion studies, and percolation-based models of binary passivation (Liu et al., 16 Jul 2025, Anber et al., 11 Jun 2025, Roy et al., 2024). The interpretation is consistent across these studies: $i$1 indicates random statistics, negative values indicate preferred unlike-neighbor association, and positive values indicate avoidance or clustering, depending on pair type and convention (Chen et al., 7 Apr 2026, Liu et al., 16 Jul 2025).

Warren–Cowley parameters are physically transparent and directly tied to diffuse scattering, pair probabilities, and Monte Carlo equilibrated configurations. They are also sufficient to expose strong local motifs. In CoNiV, strongly negative $i$2 and $i$3 together with strongly positive $i$4 define the V-centered CSRO that suppresses V-V clustering (Chen et al., 7 Apr 2026). In Cu$i$5Au, negative first-shell and positive second-shell values encode the incipient L1$i$6 ordering tendency and agree closely with classic diffuse-scattering measurements (Morris et al., 2022). In CrCoNi- and CoCrNi-based passivation studies, shell-resolved Cr-Cr CSRO parameters are used to interpret how thermal treatment alters the local chromium arrangement relevant to oxide formation (Anber et al., 11 Jun 2025, Blades et al., 2024).

At the same time, pair descriptors have well-defined limitations. In fcc CrCoNi, the nearest-neighbor motif space contains $i$7 raw motifs and $i$8 symmetry-distinct motifs after quotienting by Euclidean symmetry (Sheriff et al., 2023). Motifs that are identical under nearest-neighbor Warren–Cowley parameters can still differ by orders of magnitude in probability and by large differences in local lattice distortion. One example given is a Cr-centered nearest-neighbor composition with 7 Ni, 2 Cr, and 3 Co neighbors: it contains 182 unique motifs that are indistinguishable to first-shell pair statistics, yet their probabilities differ by two orders of magnitude and their local distortions span from the 39th to the 97th percentile (Sheriff et al., 2023).

This limitation motivated motif-based and information-theoretic descriptors. One such measure is the Jensen–Shannon divergence $i$9, introduced as a scalar measure of the distance between the actual local-motif distribution and that of a chemically random alloy, with $Q \approx 1.85\,\mathrm{\AA}^{-1}$0 for complete randomness (Islam et al., 2024). A related multiscale approach uses the Kullback–Leibler divergence

$Q \approx 1.85\,\mathrm{\AA}^{-1}$1

to quantify how equilibrium motif populations deviate from random-solid-solution statistics (Sheriff et al., 2023). In this formulation, CSRO has two irreducible components: motif population bias and the spatial organization of motifs, the latter represented by motif-resolved correlation functions $Q \approx 1.85\,\mathrm{\AA}^{-1}$2 and characteristic fluctuation lengths $Q \approx 1.85\,\mathrm{\AA}^{-1}$3 (Sheriff et al., 2023, Sheriff et al., 2024).

A different line of work proposes scalar bond-count descriptors such as the $Q \approx 1.85\,\mathrm{\AA}^{-1}$4-parameter and the OPERA framework for engineering prescribed chemical order in canonical or semi-canonical ensembles. There, $Q \approx 1.85\,\mathrm{\AA}^{-1}$5 denotes a perfectly random bond-count state, $Q \approx 1.85\,\mathrm{\AA}^{-1}$6 indicates enhanced like-bond clustering, and $Q \approx 1.85\,\mathrm{\AA}^{-1}$7 indicates enhanced unlike-bond ordering (Anand et al., 2022). This suggests a broader methodological trend: CSRO is increasingly treated not only as a descriptive pair statistic, but as a high-dimensional structural variable that can be compressed into alternative scalar measures when inverse design or high-throughput exploration is required.

3. Experimental observation and direct imaging

Direct experimental observation of CSRO has historically been difficult, particularly in alloys such as CoCrNi or CrCoNi where neighboring transition metals offer limited contrast in conventional diffraction and microscopy. Recent work has nevertheless established multiple direct routes. In equiatomic CoCrNi, neutron diffraction provided direct evidence of CSRO via a diffuse peak centered at $Q \approx 1.85\,\mathrm{\AA}^{-1}$8, whose intensity increases under aging conditions favorable to local ordering and remains substantial even in the gas-atomized state (Bacurau et al., 29 May 2026). Because coherent neutron scattering lengths differ strongly among Cr, Co, and Ni, neutron total scattering resolves local chemical order where X-ray diffraction does not (Bacurau et al., 29 May 2026).

That neutron result was complemented by small-angle neutron scattering, which identified Ni-rich, disk-shaped domains with radii of approximately $Q \approx 1.85\,\mathrm{\AA}^{-1}$9 and thicknesses of about $_1$0, consistent with a nanoscale CSRO length scale (Bacurau et al., 29 May 2026). Fast-Fourier-transform analyses of simulated projections further linked the observed diffuse reflections to local motif families associated with D1a-, Pt$_1$1Mo-, and D0$_1$2-like arrangements, while explicitly stopping short of claiming bulk long-range ordered intermetallic formation (Bacurau et al., 29 May 2026). This is an important clarification: motif analogies extracted from diffuse scattering are not equivalent to identifying a fully developed ordered phase.

Electron-based approaches reveal complementary aspects. In CrCoNi, a data-mining workflow built on energy-filtered scanning electron nanodiffraction identified two distinct classes of nanoscale ordered clusters: L1$_1$3-type nanoclusters associated with alternating $_1$4 plane segregation and L1$_1$5-type nanoclusters associated with suppression of Cr-Cr nearest neighbors (Hsiao et al., 2022). The water-quenched condition was dominated by $_1$6 nm L1$_1$7-type clusters, whereas a 1000 °C, 120 h heat treatment produced smaller $_1$8 nm L1$_1$9-type clusters and a larger prevalence of L1$_2$0-type order (Hsiao et al., 2022).

Atom probe tomography has also moved from indirect composition fluctuation analysis toward 3D imaging of ordering itself. In CoCrNi, machine-learning-enhanced APT used local $_2$1-direction spatial distribution maps and a 1D convolutional neural network to classify voxels as random FCC or CSRO, thereby reconstructing 3D maps of ordered domains (Li et al., 2023). This revealed multiple coexisting CSRO configurations—L1$_2$2-type domains along $_2$3 and Ni-related L1$_2$4/DO$_2$5-type domains along $_2$6—with sizes of roughly $_2$7–$_2$8 nm and number densities on the order of $_2$9 (Li et al., 2023). Annealing at 1273 K for 120 h increased the overall number density of multiple CSROs by about $\alpha_{ij}(1)=1-\frac{P_{ij}(1)}{c_j},$0 and raised room-temperature electrical resistivity by 17%, establishing a direct processing–structure–property relation (Li et al., 2023).

Local-structure probes such as EXAFS add another piece. In aged equiatomic CoCrNi, EXAFS showed shortened effective first-shell bond lengths relative to the homogenized state, consistent with a changed local chemical environment, while atomistic MC–MD simulations assigned the corresponding shell-specific signature to stronger low-temperature CSRO and especially stronger second-nearest-neighbor Cr-Cr ordering (Anber et al., 11 Jun 2025). Because EXAFS in ternary CoCrNi cannot cleanly separate all pair types, this is an example where experiment establishes the existence of local rearrangement and simulation resolves its likely chemical content (Anber et al., 11 Jun 2025).

4. Computational and thermodynamic frameworks

Atomistic and statistical models play a central role in CSRO research because local order is both subtle and thermodynamically nontrivial. A common route is hybrid Monte Carlo/molecular dynamics using machine-learned interatomic potentials. In CoNiV, a moment tensor potential trained on 6,394 DFT configurations with a 5.0 Å cutoff enabled hybrid MC/MD sampling of a $\alpha_{ij}(1)=1-\frac{P_{ij}(1)}{c_j},$1 FCC supercell with 2,048 atoms, revealing V-centered ordering over 300–1000 K (Chen et al., 7 Apr 2026). In MoNbTaVW, a neuroevolution potential reproduced known ordering tendencies and threshold displacement energies, then enabled million-atom cascade simulations and long MC/MD anneals inaccessible to more expensive models (Liu et al., 16 Jul 2025). In CrCoNi, chemically sampled moment tensor potentials were shown to reproduce DFT-predicted short-range order much better than conventional solid-solution potentials trained without explicit ordering coverage (Sheriff et al., 2023).

Beyond pair statistics, recent work has pushed motif-based machine learning to formalize CSRO as a many-body state. An $\alpha_{ij}(1)=1-\frac{P_{ij}(1)}{c_j},$2-equivariant graph-neural-network framework defines each local motif as a central atom plus its first coordination polyhedron, converts that motif into a graph, and maps it to a symmetry-aware embedding $\alpha_{ij}(1)=1-\frac{P_{ij}(1)}{c_j},$3 or $\alpha_{ij}(1)=1-\frac{P_{ij}(1)}{c_j},$4, depending on implementation (Sheriff et al., 2024, Sheriff et al., 2023). This makes it possible to enumerate symmetry-distinct motifs, measure motif distributions directly, correlate per-atom properties such as lattice strain with motif identity, and compute motif-based correlation lengths. In MoTaNbTi, this framework was used to identify B2-like local motifs and extract temperature-dependent chemical fluctuation length scales (Sheriff et al., 2024).

A parallel development occurs at the thermodynamic level. Conventional CALPHAD, built around Bragg–Williams mean-field assumptions, cannot intrinsically represent CSRO. A hybrid CVM–CALPHAD framework, FYL-CVM, addresses this by combining the cluster variation method with the Fowler–Yang–Li transform to reduce the number of variational variables from order $\alpha_{ij}(1)=1-\frac{P_{ij}(1)}{c_j},$5 to order $\alpha_{ij}(1)=1-\frac{P_{ij}(1)}{c_j},$6 while retaining intrinsic short-range order in multicomponent free-energy minimization (Fu, 9 Aug 2025). In this model, Warren–Cowley parameters and multipoint cluster SRO parameters follow directly from equilibrium cluster probabilities. Applied to Cu–Au and Cu–Au–Ag, FYL-CVM produces phase diagrams and SRO diagrams in composition–temperature space, thereby treating CSRO as an intrinsic thermodynamic degree of freedom rather than a post hoc correction (Fu, 9 Aug 2025).

CSRO also changes the thermodynamics of point defects. For equilibrium vacancies in FCC CrCoNi, an exact statistical-mechanical expression for the dilute monovacancy concentration was derived as

$\alpha_{ij}(1)=1-\frac{P_{ij}(1)}{c_j},$7

with the average taken over a fully equilibrated dense alloy in which CSRO has already formed (Tang et al., 26 Dec 2025). The key implication is that vacancy thermodynamics depends not on a single formation energy, but on the full distribution of vacancy-formation free energies over chemically ordered local environments. In CrCoNi, stronger low-temperature CSRO increases vacancy formation energy and suppresses equilibrium vacancy concentration to the ppb scale over 600–900 K, which in turn affects the kinetics of further short-range-order formation (Tang et al., 26 Dec 2025).

5. Nonequilibrium formation, processing routes, and dynamic evolution

A major revision in the modern understanding of CSRO is that it is not solely an equilibrium annealing product. Conventional thinking places CSRO on an equilibrium line between random solid solutions and more ordered states, with lower temperature producing more order. Recent atomistic work on CrCoNi instead shows that manufacturing routes can generate nonequilibrium steady states of SRO that are not equivalent to equilibrium states at any temperature (Islam et al., 2024). In that framework, remnant SRO emerges during solidification or plastic deformation because nonequilibrium atomic rearrangements possess an inherent ordering bias rather than acting as chemically neutral randomizers (Islam et al., 2024).

In tensile deformation of CrCoNi at 300 K and $\alpha_{ij}(1)=1-\frac{P_{ij}(1)}{c_j},$8, all initial conditions—including a chemically random solid solution—converged to a path-independent steady state of finite $\alpha_{ij}(1)=1-\frac{P_{ij}(1)}{c_j},$9, governed by

$P_{ij}(1)$0

with fitted parameters $P_{ij}(1)$1 and $P_{ij}(1)$2 (Islam et al., 2024). In this interpretation, the $P_{ij}(1)$3 term represents dislocation-driven destruction of existing order, whereas the constant $P_{ij}(1)$4 term represents continuous dislocation-driven creation of order. This directly contradicts the common assumption that plasticity merely erases local chemical order.

Solidification generates a different nonequilibrium CSRO class. In CrCoNi, steady-state growth under undercoolings spanning interface velocities from $P_{ij}(1)$5 at $P_{ij}(1)$6 to above $P_{ij}(1)$7 for $P_{ij}(1)$8 produced finite remnant SRO under all conditions, but these as-cast states remained far from the equilibrium SRO manifold (Islam et al., 2024). The authors interpret this as inheritance of chemical order from the liquid-like solid–liquid interface rather than bulk vacancy-mediated equilibration in the solid. The broader implication is that conventional manufacturing processes provide access to a broader nonequilibrium SRO design space than equilibrium annealing alone (Islam et al., 2024).

CSRO is also dynamic under irradiation. In MoNbTaVW, pre-existing local order improves defect recombination during individual cascades, but cumulative irradiation rapidly degrades the ordering, driving Warren–Cowley parameters below 0.3 by about 0.03 dpa after 1000 cascades (Liu et al., 16 Jul 2025). In Fe-Ni-Cr alloys with Cr-rich CSRO, irradiation drives low-initial-CSRO states toward higher CSRO and high-initial-CSRO states toward lower CSRO, approaching a nontrivial steady-state $P_{ij}(1)$9 rather than complete disorder (Arkoub et al., 2023). These results indicate that CSRO can be both a beneficial kinetic modifier and a dynamically evolving state variable whose stability under service conditions is itself a materials-design problem.

6. Effects on properties and design implications

The strongest recent evidence for the importance of CSRO comes from property-specific studies in which local order is varied while overall composition and crystal structure are held nearly fixed. In CoNiV, V-centered CSRO suppresses V-rich octahedral interstitial motifs and raises the average hydrogen solution energy from about $i$0 eV in a random alloy to about $i$1 eV in the ordered alloy, while reducing the fraction of sites with negative hydrogen solution energy from 7.8% to 3.1% (Chen et al., 7 Apr 2026). This reduces bulk hydrogen uptake and helps explain the alloy’s resistance to hydrogen embrittlement. Hydrogen still segregates to tensile partial-dislocation core regions, but the effective trap depth remains shallow, about $i$2 eV in the random alloy and $i$3 eV in the ordered state, much weaker than classical vacancy traps (Chen et al., 7 Apr 2026).

In refractory multicomponent alloys, CSRO strongly modifies defect-mediated mechanical response. In MoNbTi and TaNbTi, increasing CSRO raises the mean unstable stacking-fault energy while decreasing its coefficient of variation; in the coupled phase-field dislocation-dynamics simulations, the mean unstable stacking-fault energy controls the critical stress for glide, whereas the distribution width controls stochastic glide-plane hardening (Zheng et al., 2022). The result is not a simple “ordering strengthens” statement: CSRO can simultaneously raise the stress required to initiate glide and reduce the extent of pinning/depinning hardening during subsequent motion (Zheng et al., 2022).

In nanocrystalline HfNbTaTiZr, the contrast between random solid-solution strengthening and CSRO is even sharper. Relative to the random state, the MC-relaxed CSRO state reduces modulus, yield strength, ultimate strength, and flow stress, but improves strain hardening and failure resistance, shifts the dominant plasticity toward transformation-induced plasticity, and suppresses the Hall–Petch to inverse Hall–Petch transition by reducing the critical grain size from roughly 13–15 nm to roughly 10–11 nm, depending on the model used (Wu et al., 28 Jun 2025). This suggests that CSRO can be used to tune not only strength levels but also the competition among slip, twinning, phase transformation, and grain-boundary processes.

Corrosion and passivation provide another prominent application. In CoCrNi, aging modifies local bonding and increases second-nearest-neighbor Cr-Cr ordering, which correlates with faster passive-film formation and markedly improved protection in deaerated $i$4 mol/L $i$5 mol/L NaCl (Anber et al., 11 Jun 2025). The aged alloy dissolves only about 100 monolayers before passivation stabilizes, compared with about 280 monolayers for the homogenized state, and its long-term passive-film impedance at 1 mHz is roughly three orders of magnitude higher (Anber et al., 11 Jun 2025). More generally, a percolation model for binary FCC alloys shows that clustering of the passivating element lowers the 3D site-percolation threshold from about 0.2088 at $i$6 to about 0.1873 at $i$7, promoting formation of a dominant spanning cluster and reducing the thin-film thickness required for percolation crossover (Roy et al., 2024). This provides a geometric framework for why short-range clustering of Cr can improve passivation.

The same logic underlies low-Cr stainless-like behavior in FeCoNiCrAl-based alloys. Adding only 0.03–0.06 Al to $i$8 changes the magnitude and, for some shells, the sign of the Cr-Cr CSRO parameter, increasing local Cr-Cr adjacency and enabling passivation behavior similar to 304L stainless steel despite only about 10 at.% Cr (Blades et al., 2024). This is a particularly clear illustration that bulk composition alone does not set the passivation threshold; local chemical topology matters.

In radiation environments, CSRO affects defect kinetics, not merely initial damage production. In MoNbTaVW, CSRO lowers the interstitial migration barrier from $i$9 eV to $j$0 eV while raising the vacancy migration barrier from $j$1 eV to $j$2 eV, thereby increasing recombination efficiency during cascade recovery (Liu et al., 16 Jul 2025). In Fe-Ni-Cr alloys, high CSRO leads to Cr-rich interstitial clusters and loops that preferentially reside in or near Cr-rich domains, while random solid solutions exhibit the strongest tendency toward radiation-induced mixing (Arkoub et al., 2023). These results indicate that CSRO acts as a kinetic modifier of defect migration, clustering, and recovery pathways.

Taken together, these studies suggest a consistent design principle. CSRO is neither a negligible correction to random-alloy models nor a synonym for incipient precipitation. It is a local thermodynamic and kinetic field that can be tuned by composition, heat treatment, deformation path, irradiation history, or manufacturing route, and that field can substantially alter hydrogen uptake, dislocation glide, passivation, point-defect transport, vacancy thermodynamics, and radiation tolerance (Chen et al., 7 Apr 2026, Liu et al., 16 Jul 2025, Blades et al., 2024). A plausible implication is that future alloy design will increasingly treat CSRO alongside composition, grain size, and phase constitution as an independent microstructural variable rather than as a hidden byproduct of processing.