Grain Condition: Multi-Domain Diagnostics

Updated 7 July 2026

Grain condition is a state defined by measurable descriptors, capturing key characteristics such as packing angles, kernel defects, microstructural topology, and mantle composition.
In granular discharge, geometric profiles and packing state measurements reveal that despite different conditions, discharge rates remain stable while particle roughness influences slope contrasts.
Across domains—from polycrystalline materials to interstellar chemistry—grain condition underpins diagnostic workflows that enable targeted predictions and control based on explicit, calibrated observables.

“Grain condition” denotes a domain-specific state of grains, grain assemblies, or grain-bearing interfaces, and its meaning depends on the physical system being studied. In granular discharge, it denotes the packing state of a bed and is diagnosed from the evolving surface depression; in cereal science, it denotes kernel quality, defect status, completeness, and chalkiness; in polycrystalline materials, it denotes the geometric, topological, crystallographic, and stress state of grains and grain-boundary networks; in interstellar chemistry, it denotes the physical and chemical state of grain mantles set by accretion and freeze-out. Across these literatures, the term is operational rather than generic: it is defined through measurable descriptors such as volume fraction $\phi$ , discharge angles $\theta_1$ and $\theta_2$ , hyperspectral signatures, anomaly scores, grain size and orientation distributions, triple-junction force balance, or mantle composition and thickness (Pacheco-Vazquez et al., 2017, Karmakar et al., 2024, Fan et al., 2023, Qin et al., 2024, Qiu et al., 12 Mar 2026, Das et al., 2010).

1. Conceptual scope and domain-specific meanings

The cited literature uses “grain condition” to identify the state of a system through observables that are intrinsic to that system. The term therefore has a family of technical meanings rather than a single universal definition.

Domain	Operational meaning	Primary descriptors
Granular discharge	Packing state of a granular bed	$\phi$ , $\theta_1$ , $\theta_2$ , $Q$ , $R(t)$
Cereal inspection	Kernel quality and defect state	variety/type, defects, completeness, chalkiness, spectra
Polycrystalline materials	Microstructural state of grains and interfaces	size, aspect ratio, orientation, stress, topology, SRO
Interstellar grain mantles	Chemical and physical mantle state	O/CO ratio, cloud density, composition, thickness

This breadth is not accidental. In each field, “condition” is tied to a diagnostic workflow. For granular media, the relevant state variable is packing; for agricultural kernels, it is quality relative to grading or anomaly criteria; for materials science, it is the coupled state of grains, boundaries, and junctions; for astrochemistry, it is the mantle composition reached under specified accretion and density histories (Pacheco-Vazquez et al., 2017, Fan et al., 2022, Qiu et al., 12 Mar 2026, Das et al., 2010).

2. Granular packing state in silo discharge

In silo discharge of close-packed grains, grain condition is the packing state of the bed, quantified by the volume fraction $\phi$ and diagnosed through the geometry of the surface depression. Standard packing is denoted $\phi_s$ , increased packing obtained by vibration or tapping is denoted $\theta_1$ 0, and the fluidized central channel has a lower volume fraction $\theta_1$ 1. For sand in the reported experiments, $\theta_1$ 2 and $\theta_1$ 3 can reach $\theta_1$ 4. Standard packing produces a conical depression with a single slope, whereas increased packing produces a double-angle depression with a lower angle $\theta_1$ 5 over the fluidized channel and a larger angle $\theta_1$ 6 in the stagnant periphery. For rough sand, $\theta_1$ 7 can reach about $\theta_1$ 8 early in discharge, while $\theta_1$ 9, with $\theta_2$ 0 for the sand used (Pacheco-Vazquez et al., 2017).

The geometric transition occurs at the boundary between a densely packed stagnant peripheral region and the central flowing funnel-flow channel above the aperture. This boundary can be tracked by a transition coordinate $\theta_2$ 1, and the cavity radius $\theta_2$ 2 follows model-dependent growth laws. For standard packing, the 2D and 3D single-angle models give

$\theta_2$ 3

with measured fits $\theta_2$ 4 and $\theta_2$ 5. For close-packed beds, the mass balance explicitly contains both the emptied depression volume and the fluidized-zone volume,

$\theta_2$ 6

and the resulting $\theta_2$ 7 depends on the evolving $\theta_2$ 8, $\theta_2$ 9, and $\phi$ 0.

A central result is that the discharge rate remains effectively independent of the initial packing condition. In 2D sand, $\phi$ 1 g/s across packing conditions; in 3D sand, $\phi$ 2 g/s for standard packing and $\phi$ 3 g/s for high packing. The interpretation given is that all material must pass through the fluidized central zone before exiting, so the initial peripheral packing does not control $\phi$ 4, consistent with Beverloo-type dense-flow behavior. Grain condition is therefore readable from the surface profile even when throughput is unchanged.

Particle roughness sharply modulates the diagnostic strength of the geometry. For smooth glass beads of the same size and density, the flowing layer is roughly twice as thick as for sand, the transition from moving to static zones is diffuse, and the slope contrast is much smaller: $\phi$ 5 for beads, versus up to about $\phi$ 6 for sand. This establishes that high volume fraction and friction must combine to produce the pronounced double-angle geometry.

3. Cereal grains: quality, defect state, and machine inspection

In cereal inspection, grain condition is a quality concept tied to type, damage, contamination, and market grade. GrainSpace formulates grain appearance inspection as three computer-vision tasks—fine-grained recognition, domain adaptation, and out-of-distribution recognition—and grounds them in ISO5527-aligned taxonomies for wheat, maize, and rice. The dataset contains 5.25 million single-kernel images, with wheat, maize, and rice sampled from five countries and more than 30 regions. Wheat and maize labels distinguish NORMAL from multiple damaged and unsound classes, while rice labels correspond to eight sub-types used for commercial stratification (Fan et al., 2022).

A complementary anomaly-detection formulation defines grain condition at the kernel level as normal versus anomalous. AI4GrainInsp treats healthy and edible kernels as the normal distribution and damaged grains or foreign objects as anomalies. Its prototype device uses dual vertically aligned industrial cameras at 860 DPI with a transparent moving plate and vibration bands, producing UP and DOWN images that jointly capture 92–98% of the superficial area per kernel. AD-GAI is trained only on normal samples and uses patch-aware features plus synthetic image-level and feature-level anomalies. On OOD-GrainSet, the AD-GAI ensemble reaches AUROC values of 95.9 for Wheat(set1), 94.1 for Wheat(set2), 88.2 for Maize(set1), and 82.8 for Maize(set2). In a practical wheat inspection comparison, the device processes a 60 g sample of about 1600 kernels in roughly 73 s, versus roughly 1550 s for senior inspectors, a speedup of about $\phi$ 7 (Fan et al., 2023).

Hyperspectral imaging extends grain condition beyond appearance to spectral cues related to protein, moisture, physical defects, and contamination. In the reported HSI system, bulk-grain images are acquired with a Specim FX17 line-scan camera over 900–1700 nm. Of 224 channels, the first and last 10 are excluded, leaving 204 channels. The best-performing model uses metric-based few-shot learning with Prototypical Networks, a modified ResNet-18 backbone, and a squeeze-and-excitation block before spectral downsampling. For the 8-way seen-class task, 204 channels with SE attention achieve 97.75% accuracy, and pre-computed collective class prototypes improve inference from $\phi$ 8 to 97.75%. For unseen-class generalization, a two-class support-set strategy reaches 98.33% (Karmakar et al., 2024).

Rice quality evaluation makes grain condition explicitly standards-based. Under GB/T 1354–2018, the cited system evaluates variety identification, completeness grading, and chalkiness. Whole grains are defined as kernels with length not less than three-quarters of the average whole-grain length from the same batch; broken rice is split into sizeable broken and tiny broken according to the 2.0 mm and 1.0 mm sieve criteria. Chalkiness is quantified at the kernel level by

$\phi$ 9

and batch-level chalk content is computed as

$\theta_1$ 0

Using about 20,000 images from six rice varieties, the proposed real-time mechanism reports 99.14% mAP for object detection, 97.89% accuracy for the classification task, and an average completeness-grading accuracy of 97.56% (Xia et al., 19 Feb 2025).

4. Grain condition in polycrystalline microstructures

In rapid solidification and grain-growth modeling, grain condition denotes the state of the microstructure itself. GrainGNN defines it through geometric and crystallographic descriptors at each solid–liquid interface height $\theta_1$ 1: grain size, aspect ratio, orientation distribution, and elimination statistics. The grain size descriptor is the volume-equivalent diameter

$\theta_1$ 2

and the orientation condition is summarized by the volume-weighted average misorientation

$\theta_1$ 3

Using a dynamic graph representation with grain and junction vertices, GrainGNN reproduces phase-field quantities of interest with 80–90% pointwise accuracy, KS values mostly in 0.03–0.06, and inference speedups of $\theta_1$ 4– $\theta_1$ 5 relative to high-fidelity phase-field simulation for a 100-grain problem; it also modeled the formation of 11,600 grains in 220 s on a single CPU core (Qin et al., 2024).

Under shear-coupled grain growth, grain condition additionally includes internal stress, faceting, and stress-dependent kinetics. The coupling law is

$\theta_1$ 6

and the simulations show that introducing shear coupling produces a more heterogeneous, less equiaxed microstructure than curvature-driven growth. The number-of-sides distribution shifts from a peak at $\theta_1$ 7 under curvature flow to a peak at $\theta_1$ 8 when internal stress is included. Roundness, defined as $\theta_1$ 9, shifts from a peak near 0.88 to a broader distribution centered near 0.84. Mean grain area follows $\theta_2$ 0 with $\theta_2$ 1 for curvature flow, $\theta_2$ 2 with internal stress, $\theta_2$ 3 with internal plus external shear, and $\theta_2$ 4 with internal plus external tension. The statistical conclusion is explicit: highly stressed grains shrink faster, while lightly stressed grains grow faster (Qiu et al., 12 Mar 2026).

Thin-film grain growth introduces a further distinction between universal structural statistics and non-universal kinetics. Large-scale phase-field crystal simulations reproduce the experimentally observed excess of small grains (“ear”) and excess of very large grains (“tail”) in the normalized size distribution, in contrast to the standard Mullins model. The same simulations show that the growth exponent is not universal. Reported cases give $\theta_2$ 5 with $\theta_2$ 6 crossing over to $\theta_2$ 7, or remaining near $\theta_2$ 8, as well as $\theta_2$ 9 and $Q$ 0 for other parameter sets. The structural statistics are well fit by a log-normal form and remain self-similar in time (Backofen et al., 2013).

Amorphous grain-boundary complexions add local short-range order to the definition of grain condition. In Cu–Zr bicrystals and a random polycrystal, ordered fcc-like Voronoi polyhedra appear only in the amorphous–crystalline transition regions, while the amorphous interior is dominated by icosahedral-like motifs. The ACTR thickness is about 0.6 nm, and the difference in ordered-motif density between the two ACTRs, $Q$ 1, is inversely related to the grain incompatibility difference $Q$ 2, with $Q$ 3. Symmetric grain boundaries have symmetric ACTR order; asymmetric grain pairs do not. This identifies grain condition as a coupled grain–film–grain equilibrium state rather than a property of one adjoining grain alone (Garg et al., 2022).

5. Grain boundaries, triple junctions, and energetic stability

At triple junctions, grain condition often means local equilibrium or its controlled violation. In the stochastic grain-boundary framework, the generalized Herring condition at equilibrium is

$Q$ 4

with the dynamic version

$Q$ 5

The same model derives explicit angle relations in terms of the misorientation-dependent tensions and gives a Fokker–Planck steady state of Boltzmann form,

$Q$ 6

Here grain condition is the joint state of triple-junction position and misorientation, together with the equilibrium or mobility-limited force balance that links them (Epshteyn et al., 2021).

The same stability idea reappears in global grain-boundary energy modeling. For fixed misorientation $Q$ 7, the inclination-dependent energy is $Q$ 8, and Herring’s condition is expressed as h-convexity of the reciprocal plot:

$Q$ 9

The cited analysis argues that many recent global models violate this stability condition, often because of cusp constructions with infinite slope such as the Read–Shockley–Wolf form. A remedy is h-convexification by constructing a preliminary interpolant and convexifying its $R(t)$ 0 plot. In this usage, grain condition is the stability-admissible energy landscape of boundary-plane orientations at fixed misorientation (Morawiec, 2 Jan 2025).

Thin films place limits on such equilibrium reconstructions. In nanocrystalline tungsten and aluminum films analyzed by PED, Herring-based extraction of relative grain-boundary energy distributions from triple-junction geometry does not reproduce molecular-dynamics energetic trends and generally does not inversely correlate with the measured grain-boundary character distribution, except for the most anisotropic and highest-population Al $R(t)$ 1 boundaries after twin adjustment. The paper therefore argues that triple-junction geometry in these films is not governed solely by boundary energies and proposes a generalized force balance with additional surface, strain, geometric, and triple-junction-drag contributions. Under these conditions, grain condition at a junction cannot be reduced to the conventional Herring equation alone (Patrick et al., 2022).

Because many grain-condition metrics depend on accurate boundary-network extraction, segmentation quality becomes part of the measurement problem. In electron-microscopy images of polycrystalline oxides, the CRF-based refinement method treats grain condition as a property of boundary continuity, connectivity, triple-junction integrity, and grain enclosure. Relative to the initial segmentation, validation IoU improves from 0.833 to 0.925, DSC from 0.909 to 0.961, and grain alignment from 4062.7 to 2007.7, a 50.6% decrease. These gains matter because small gaps in thin masks can merge grains, remove junctions, and bias downstream size and connectivity statistics (Aksoy et al., 2023).

6. Specialized environments and comparative perspective

In radiation materials science, grain condition includes the sink efficiency of grain boundaries for radiation-induced point defects and the evolving coupling between defect transport and grain-boundary motion. The coarse-grained jump Robin-type boundary condition is

$R(t)$ 2

and for low-angle tilt grain boundaries

$R(t)$ 3

The paper also gives the sink-strength scaling $R(t)$ 4 and reports that internal point-defect sources accelerate grain shrinkage, while radiation tends to trigger the extension of twin boundary sections (Zhu et al., 2018).

In sediment transport, the relevant grain condition is grain shape and the hydrodynamic boundary condition acting at the grain surface. The critiqued shape-corrected bedload relation modifies the Shields number through $R(t)$ 5, where $R(t)$ 6 depends on a settling drag coefficient $R(t)$ 7. The cited analysis shows that approximating Navier slip by shifting the no-slip condition to a virtual surface inside the grain is valid only if the boundary-layer thickness satisfies $R(t)$ 8; for the simulations under discussion, this criterion is not met, and the procedure substantially overestimates $R(t)$ 9 for a Navier-slip sphere. A null-hypothesis model based on virtual-grain size explains the data without invoking a grain-shape correction (Chen et al., 21 Jan 2026).

In graphene nanoribbons, grain condition is set by grain-boundary type, misorientation, support condition, and loading direction. Two 5–7 grain boundaries are considered: LAGBI with $\phi$ 0 and LAGBII with $\phi$ 1. The buckling strain ranges from approximately 3.0% for LAGBI under free boundary conditions with loading parallel to the grain boundary, down to approximately 0.01% for LAGBII under free boundary conditions with loading perpendicular to the grain boundary, the latter effectively indicating a relaxed state that is already buckled. Large-angle boundaries therefore weaken buckling resistance, while small-angle boundaries can either raise or lower it depending on the loading/support geometry (Neek-Amal et al., 2012).

In interstellar chemistry, grain condition refers to the physical and chemical state of grain mantles as set by accretion, reaction, and freeze-out. With only H, O, and CO accreting at 10 K, the mantle thickness grows anywhere between 60 and 500 layers in two million years. Methanol is always over-produced when the number density of accreting O is less than three times that of CO, water exceeds methanol when the O accretion rate is about $\phi$ 2 the CO rate, and including freeze-out of O and CO leads to an almost steady state on the grain surface after about $\phi$ 3 yr. In this usage, grain condition is a chemically evolving state set by initial gas-phase ratios and cloud density rather than by mechanics or morphology (Das et al., 2010).

Taken together, these usages suggest that “grain condition” is best understood as a family of state descriptions tied to the governing physics of each field. In every case, the term becomes meaningful only when paired with explicit observables, calibration procedures, and stability criteria: packing angles for granular discharge, spectra and defect taxonomies for cereal kernels, topology and internal stress for polycrystals, force balance for triple junctions, sink efficiency for irradiated interfaces, or mantle composition for interstellar grains. A plausible implication is that the unifying feature of the term is not a shared material ontology, but a shared methodological role: it designates the measurable state on which diagnosis, prediction, and control are based.