---
title: Strength Inconsistencies in Material Failure Models
url: https://www.emergentmind.com/topics/strength-inconsistencies
type: topic
---

# Strength Inconsistencies in Material Failure Models

Searching arXiv for papers relevant to "strength inconsistencies" across materials, mechanics, and related domains.
arxiv_search.query({"search_query":"all:\"strength inconsistencies\" OR ti:\"strength\" AND ti:\"inconsistencies\"","start":0,"max_results":10})
Searching for specific known papers from the provided data to ground the article with arXiv identifiers.
arxiv_search.query({"search_query":"id:1310.2710 OR id:2012.06099 OR id:2304.12815 OR id:1609.02293 OR id:1706.01591 OR id:2308.05142 OR id:1807.08552","start":0,"max_results":10})
Searching for additional papers from the provided set that frame inconsistency more generally.
arxiv_search.query({"search_query":"id:1109.1755 OR id:1803.06162 OR id:2407.15511 OR id:2506.18781","start":0,"max_results":10})
Strength inconsistencies are non-uniformities or contradictions in the measured, inferred, or predicted strength of a system that persist even when nominal descriptors appear unchanged. In the literature surveyed here, they arise when strength is treated as a single material constant although it depends on hidden internal state, evolving failure mechanism, structural topology, statistical heterogeneity, or an inappropriate governing model. The phenomenon appears in bulk metallic glasses, early-age self-consolidating concrete, nacre-like lamellar networks, polymer-chain ensembles, ideal crystals, truss bridges, and even transition-strength models in nuclear structure. A common theme is that apparent inconsistency is often traceable to omitted state variables or to a mismatch between the object being modeled and the criterion used to characterize its failure or response [1310.2710].

## 1. Conceptual structure of strength inconsistency

In the cited mechanics literature, a “strength inconsistency” is not merely experimental scatter. It is a systematic mismatch between nominally equivalent descriptions of resistance. In bulk metallic glasses, two samples with the same nominal composition and fully amorphous structure can still have different yield strengths because they contain different amounts of free volume, different degrees of structural relaxation, and therefore different elastic constants [1310.2710]. In early-age self-consolidating concrete, the inconsistency is that the material does not possess a single scalar strength during hydration; its resistance depends on loading mode, stress state, and hydration time [1609.02293]. In truss bridges, the inconsistency is member-specific: code live load models reproduce chord-member forces comparatively well, but vertical and diagonal members exhibit force reversals and irregular exceedance rates under real traffic that the same code models do not represent well [2308.05142].

This usage differs from simple random error. The cited papers repeatedly connect inconsistency to a missing dynamical law, a hidden internal state, or an unsuitable reduction of a multiaxial or networked system to a scalar limit. In passive pharmacokinetics, the authors explicitly attribute long-recognized inconsistencies to the lack of a proper evolution equation for the compartment state variable, arguing that quantities such as \(AUC\), \(t_{\max}\), and \(C_{\max}\) become ambiguous if the relation between input, state evolution, and measured concentration is not specified [1109.1755]. A plausible implication is that “strength inconsistency” across fields often signals model-form incompleteness rather than anomalous data.

## 2. Hidden internal states in amorphous and disordered materials

The clearest materials-science formulation appears in work on the fully amorphous alloy Zr\(_{56}\)Co\(_{28}\)Al\(_{16}\), where yield strength is shown to vary with casting current even though all samples remain fully amorphous by XRD [1310.2710]. The paper cites the thermodynamic scaling law
\[
T_y = 3R (T_g - T_r)\, \rho_r / M,
\]
but emphasizes that the actual yield stress is further modulated by internal-state differences such as free volume. Reduced free volume is estimated from density by
\[
x \sim 2\left(\frac{\rho_0}{\rho_r} - 1\right),
\qquad
\rho_r \sim \frac{2\rho_0}{2+x}.
\]
The reported directional relation is unambiguous: as free volume \(x\) increases, yield strength \(\sigma_y\) decreases.

The internal-state correlations are multi-parameter. With increasing casting current, density increases and enthalpy change \(\Delta H\) decreases, indicating a more structurally relaxed glass with less frozen-in free volume [1310.2710]. The paper reports that the lowest-density sample has the largest excess free volume, \(x = 2.31\%\), and that with increasing density the free volume drops from \(1.23\%\) to \(1.16\%\). The same internal-state change is reflected in elastic observables: with increasing strength, the ratio \(\mu/B\), Young’s modulus \(E\), Vickers hardness \(H_V\), and shear modulus \(G\) all increase, while plasticity drops sharply. The sample sequence S1–S4 exhibits critical fracture strains of about \(12.8\%\), \(7.8\%\), \(4.0\%\), and almost no plastic deformation before failure, respectively. The approximate yield-strength range shown is around \(1.90\)–\(2.15\) GPa.

These results identify a specific mechanism behind strength inconsistency in metallic glasses: the same composition and amorphous character do not imply the same frozen-in configurational state. Higher free volume lowers density, increases exothermic enthalpy release on relaxation, facilitates simultaneous plastic shear at different sites, and lowers the stress needed to activate shear transformation zones, while increasing plasticity [1310.2710]. This directly rejects the misconception that BMG yield strength is determined solely by the glass transition.

## 3. Evolving regimes, loading mode, and the non-uniqueness of early strength

Early-age self-consolidating concrete provides a distinct but related case in which inconsistency is temporal and mode-dependent rather than specimen-to-specimen. The material evolves during the first hours after mixing from a thixotropic yield-stress fluid into a brittle cohesive/frictional solid, and the observed resistance depends on whether the test is penetration, punch-through, shear, compression, tension, or four-point bending [1609.02293]. To compare these modes, the paper introduces a reference-time correction based on penetration strength as a state indicator and shifts each data point by
\[
\Delta t_i = t_i^{ref} - t_i^n.
\]
After correction, the first regime of strength growth is represented by
\[
S = a \exp(bt),
\]
with similar \(b\) values across tests and different \(a\) values reflecting mode-dependent scaling.

The reported first-regime fits illustrate both commonality and inconsistency. Penetration gives \(a=0.188\) kPa and \(b=0.033\ \text{min}^{-1}\), punch \(a=0.035\) kPa and \(b=0.031\ \text{min}^{-1}\), shear \(a=0.098\) kPa and \(b=0.028\ \text{min}^{-1}\), compression \(a=0.070\) kPa and \(b=0.033\ \text{min}^{-1}\), and tension \(a=0.161\) kPa and \(b=0.026\ \text{min}^{-1}\) [1609.02293]. In the first regime, the authors collapse tests (a)–(e) onto a single equivalent uniaxial strength by multiplying punch and shear strengths by \(\sqrt{3}\), rescaling penetration by the effective cone-plug area, and leaving compression and tension unscaled. The collapsed first-regime data are fitted by approximately
\[
a = 0.0745 \text{ kPa}, \qquad b = 0.0328\ \text{min}^{-1}.
\]

The inconsistency emerges at the transition to the second regime. In the early stage, tensile and compressive strengths are nearly the same and the material is approximately pressure-insensitive, consistent with a von Mises-type description [1609.02293]. At later times, cracks dominate failure, the response becomes pressure-sensitive, and compressive strength grows faster than tensile strength in absolute terms. The paper states that the large part of the final compressive-tensile difference emerges at the transition from a pressure-insensitive thixotropic fluid to a brittle cohesive-frictional solid; in the second regime, \(\sigma_c/\sigma_t \approx 5.4\), and the compressive transition occurs about 60 minutes later than the tensile one. Transition times are reported as 250.2 min for punch at 81.6 kPa, 219.4 min for shear at 44.1 kPa, 227.4 min for compression at 137.5 kPa, and 167.5 min for tension at 13.1 kPa. The friction angle rises sharply from \(0^\circ\) to \(48^\circ\) during the transition [1609.02293].

This case shows that strength inconsistency can be the signature of a material-state transition rather than an experimental artifact. A single “yield stress” is inadequate because the strength envelope in principal stress space changes shape as hydration proceeds.

## 4. Statistical heterogeneity, topology, and ensemble strength

In polymer networks, inconsistency appears when microscopic heterogeneity prevents simultaneous mobilization of the covalent rupture scale. The parallel-chain model of polymer fracture treats each chain as a freely-jointed chain with the Kuhn–Grün/Langevin relation
\[
\frac{z}{nb} = \mathcal{L}\!\left(\frac{fb}{kT}\right),
\qquad
\mathcal{L}(g)=\coth g-\frac{1}{g},
\]
and a representative covalent breaking force
\[
f_b = 200\,\frac{kT}{b}.
\]
The applied force on rigid plates is the chain-average force
\[
\langle f(z) \rangle = \int_{0}^{\infty} f(z,n)\, p(n)\, dn,
\]
and strength is defined as the maximum applied force divided by the total number of chains [2304.12815]. Chain lengths follow a Weibull distribution, with coefficient of variation \(C_v=\sigma/\mu\). Because shorter chains reach the narrow covalent-force regime first while longer chains remain in the entropic regime, even a small scatter in chain length greatly reduces the strength of the ensemble.

The resulting power-law sensitivity is a central result. The paper reports
\[
\frac{\langle f \rangle_{\max} b}{kT} \approx 2.22\, C_v^{-0.808}
\]
for the Langevin/FJC law,
\[
\frac{\langle f \rangle_{\max} b}{kT} \approx 4.47\, C_v^{-0.931}
\]
for a modified Langevin law, and
\[
\frac{\langle f \rangle_{\max} b}{kT} \approx 56.40\, C_v^{-0.599}
\]
for a linear chain law with \(C_v\ge 0.5\) [2304.12815]. The paper states that even modest scatter can reduce strength by more than an order of magnitude, and that at \(C_v=1\) the reduction is about two orders of magnitude relative to the ideal monodisperse case. The mechanism is not crack-induced stress concentration, but the interaction between a broad chain-length distribution and a very narrow high-force rupture window.

A network-level counterpart appears in nacreous imbricated lamellar materials modeled as a square fishnet pulled diagonally [1706.01591]. The paper argues that these structures do not follow a simple weakest-link strength law because the first local failure does not automatically trigger global failure. The survival probability is expanded as
\[
1-P_f(\sigma)=P_{S_0}(\sigma)+P_{S_1}(\sigma)+P_{S_2}(\sigma)+\cdots,
\]
where \(P_{S_k}\) is the probability that exactly \(k\) links have failed while the structure is still safe. Keeping only \(P_{S_0}\) recovers the classical weakest-link model,
\[
1-P_f(\sigma)=[1-P_1(\sigma)]^N,
\]
but inclusion of \(P_{S_1}\) yields a fishnet correction term that departs from weakest-link behavior [1706.01591]. At low stress, the paper derives
\[
\ln[1-P_f(\sigma)] \simeq -\frac{N(N+1)}{2}[P_1(\sigma)]^2,
\]
so in Weibull coordinates the low-stress slope is doubled relative to the weakest-link model.

Both papers show that ensemble strength is topology-dependent. In polymer chains, heterogeneity destroys synchronized rupture; in fishnets, local failure can be survivable because neighboring links carry redistributed load. This suggests that strength inconsistency in heterogeneous systems is often the macroscopic imprint of a load-sharing architecture.

## 5. Inconsistency of governing criteria and load models

A different class of strength inconsistency arises when a criterion developed for one physical situation is applied outside its domain. The paper on ideal materials argues that Griffith theory is valid for fracture of cracked materials but not for predicting the ideal strength of perfect crystals [2012.06099]. Griffith’s cracked-plate free energy is written as
\[
G(a)= -\frac{\pi \sigma^2 a^2 B}{E'} + 4\gamma(a)Ba,
\qquad
E'= \begin{cases} E, & \text{plane stress} \\ \dfrac{E}{1-\nu^2}, & \text{plane strain} \end{cases}
\]
with critical stress
\[
\sigma_c=\sqrt{\frac{2E'\gamma}{\pi a}}.
\]
If an atomic spacing is taken as an effective crack length, Griffith’s estimate gives the familiar upper bound \(\sigma_{\text{ideal}} \sim E/9\). The paper states that this assumption is physically wrong for perfect materials because there is no pre-existing crack and the ideal strength point does not correspond to creation of a fresh fracture surface.

The replacement proposed is a stability-based thermodynamic criterion. Using
\[
dG=dU-dW=-TdS \le 0 \qquad (\sigma, T=\text{constant}),
\]
the paper expands Gibbs free energy and identifies instability through the second variation:
\[
\delta^2 G > 0 \quad \text{stable}, \qquad \delta^2 G \le 0 \quad \text{unstable},
\]
so the strength criterion is
\[
\delta^2 G = 0.
\]
The corresponding stress is obtained from
\[
\delta G = 0 \quad \Rightarrow \quad \sigma = \frac{\partial U}{\partial \varepsilon}.
\]
First-principles DFT calculations for diamond, c-BN, Cu, and CeO\(_2\) along \([100]\), \([110]\), and \([111]\) show that Griffith theory fails in all four materials; for the \([100]\) direction, the DFT strengths are about 79.8% higher than Griffith prediction for diamond, 19.2% higher for c-BN, 4.3% higher for Cu, and 30.2% higher for CeO\(_2\) [2012.06099]. The ideal strength of diamond along \([100]\) is reported as about 220 GPa.

An analogous criterion mismatch appears in bridge engineering. Code live load models calibrated mainly for girder bridges are compared against three WIM databases for truss bridges, and the paper concludes that current code live load models are not entirely adequate to estimate forces in structural elements of truss bridges [2308.05142]. Chord members are comparatively well represented: top chord forces are compressive, bottom chord forces are tensile, and exceedance rates are practically uniform along the bridge. Vertical and diagonal members are different. Their forces reverse sign under moving traffic, exceedance rates are not uniform along the bridge, and for IMT 66.5 and HL-93 some vertical members show 100% exceedance. The paper attributes this mainly to uniformly distributed lane loads, which impose force patterns unlike discrete axle-by-axle transfer and therefore miss actual load reversals.

Both papers object to transposed criteria. Griffith fracture energetics presupposes a crack; girder-based live-load calibration presupposes global bending and shear as the primary design targets. In both cases, inconsistency appears because the selected criterion is not conjugate to the actual failure mechanism.

## 6. Extended usages and methodological implications

The term “strength” is not restricted to mechanical failure. In a simple tridiagonal matrix Hamiltonian intended to mimic magnetic dipole excitations, transition strength is defined as \(O^2\), where \(O\) is the transition amplitude between eigenstates under a nearest-neighbor operator \(T\) [1807.08552]. The paper finds that ground-state transition strengths tend to fall off approximately exponentially with excitation energy, with semilog fits \(7.38 - 5.72E^*\) for \(v=0.5\), \(5.44 - 3.84E^*\) for \(v=1\), \(3.60 - 2.20E^*\) for \(v=2\), and \(1.42 - 1.34E^*\) for \(v=3\). It also identifies a special case \(E=0\) in which all transition rates vanish because the Hamiltonian becomes proportional to the transition operator. Here the “inconsistency” is exact suppression rather than scatter: seemingly allowed transitions disappear because of matrix structure.

Across the surveyed literature, three methodological lessons recur. First, a proper state equation matters. In passive pharmacokinetics, the proposed one-compartment evolution law
\[
\frac{dN(t)}{dt} = -aN(t) + q(t)
\]
is presented as the missing bridge between input/absorption, state evolution, and concentration-based measures; without it, the relation between absorbed amount and observed concentration remains ambiguous [1109.1755]. Second, scalar summaries may conceal the true variable controlling strength. In BMGs that variable is internal state; in SCC it is the time-dependent failure envelope; in polymer networks it is the chain-length distribution; in fishnets it is survivable damage under redistribution. Third, correction and reconstruction procedures can convert apparent inconsistency into a more coherent representation. The SCC reference-time correction uses penetration response as a batch-specific clock [1609.02293], while the fishnet survival expansion and polymer power-law scaling identify the terms that govern deviations from naive weakest-link or monodisperse expectations [1706.01591; 2304.12815].

A plausible implication is that “strength inconsistency” is best understood not as a single pathology but as a family of model-data mismatches. In some systems it reflects hidden configurational disorder; in others, a change of regime, a nontrivial network topology, or an invalid criterion transfer. The recurring resolution is to restore the missing structure: an explicit evolution equation, an internal-state variable, a topology-aware survival model, or a failure criterion tied to the actual instability mechanism.

Source: https://www.emergentmind.com/topics/strength-inconsistencies