Weak Slope: Concepts & Applications

• Weak slope is a multifaceted concept that generalizes traditional slope definitions by relaxing strict conditions across various mathematical and applied domains.
• In nonsmooth variational analysis, it is defined as the lower semicontinuous envelope of a local descent rate, governing the dynamics of optimization much like a gradient norm.
• Applications extend to unilateral metric slopes in gradient flows, generalized inequalities in fibred surfaces, and weak stability in tangent sheaves, showcasing its broad relevance.

Searching arXiv for recent and foundational uses of “weak slope” across mathematics and related fields. Weak slope is a polysemous technical expression whose meaning depends strongly on domain. In the supplied literature it denotes, in different settings, the lower semicontinuous envelope of a local descent rate in nonsmooth variational analysis, a unilateral metric slope for constrained gradient flows, a generalized or non-sharp slope inequality in the theory of fibred surfaces, and a weaker algebraic stability notion such as $(-K)$-slope stability for tangent sheaves of Fano or weak del Pezzo varieties. Other appearances are terminologically adjacent rather than identical: weak solutions for PDEs whose unknown is a slope variable, weak-coupling expansions of a slope function in integrability, and weak or partial recovery phenomena around SLOPE-type estimators (Drusvyatskiy et al., 2012, 1908.10111, Sun et al., 2016, Chen et al., 26 Jan 2026).

1. Terminological scope

The supplied papers use “weak slope” in several non-equivalent ways. The common feature is a weakening of a more rigid slope concept: lower semicontinuous replacement of a local slope, one-sided restriction of admissible directions, generalized lower bounds replacing sharp universal constants, or an algebraic stability notion weaker than $K$-stability.

Domain	Object	Weakening mechanism
Nonsmooth variational analysis	Limiting slope $|f|$	Lower semicontinuous envelope of local slope
Constrained gradient flows	Unilateral slope $|\partial F|_{L^2_+}$	Only nonnegative directions are admissible
Fibred surfaces	“Weak slope inequalities”	Lower bounds that are generalized, non-sharp, or proved under weaker hypotheses
Fano and del Pezzo geometry	$(-K)$-slope stability	Weaker but more algebraic than $K$-stability

This breadth matters because the underlying objects differ: functions on metric spaces, energies on Hilbert spaces with unilateral constraints, numerical invariants of fibrations, and slopes of torsion-free sheaves. The term therefore has no field-independent definition.

2. Weak slope in nonsmooth variational analysis

In metric variational theory, the basic local slope of an extended-real-valued function $f:\mathcal{X}\to\overline{\mathbb{R}}$ on a complete metric space $(\mathcal{X},d)$ is

$$|\nabla f|(\bar{x}) := \limsup_{x\to \bar{x},\, x\neq \bar{x}} \frac{(f(\bar{x})-f(x))^+}{d(\bar{x},x)}.$$

Because this quantity is not lower semicontinuous, the relevant stabilized object is the limiting slope

$$|f|(\bar{x}) := \liminf_{x \xrightarrow[f]{} \bar{x}} |\nabla f|(x),$$

where $K$0 means $K$1. The paper explicitly identifies this $K$2 as the metric counterpart of the weak slope used in variational methods; a point $K$3 with $K$4 is a lower-critical point (Drusvyatskiy et al., 2012).

This weak slope governs descent dynamics through curves of near-maximal slope. A curve $K$5 is such a curve when it is absolutely continuous, satisfies

$K$6

and its energy decay obeys

$K$7

The role of weak slope is therefore analogous to the norm of the gradient in smooth steepest descent: it sets both the speed and the dissipation rate (Drusvyatskiy et al., 2012).

In Euclidean spaces, the weak slope is tied directly to subdifferential calculus. For lower semicontinuous $K$8,

$K$9

Hence $|f|$0 if and only if $|f|$1. For semi-algebraic, locally Lipschitz functions, the paper proves a chain rule and shows that curves of near-maximal slope are precisely solutions of the subgradient differential inclusion

$|f|$2

a.e. Along bounded trajectories, semi-algebraicity further yields finite length and convergence to lower-critical points (Drusvyatskiy et al., 2012).

3. Unilateral weak slope for monotonicity-constrained gradient flows

A distinct notion appears in gradient flows under the constraint that the state be nondecreasing in time. On $|f|$3, the unilateral metric is defined by

$|f|$4

so finite distance is allowed only for nonnegative increments. This induces the quasi-metric

$|f|$5

The associated unilateral weak slope of a time-dependent energy $|f|$6 is

$|f|$7

It is the metric local slope computed with respect to the unilateral metric rather than the symmetric Hilbert norm (1908.10111).

The paper gives an exact variational representation: $|f|$8 where

$|f|$9

A central identification is that $|\partial F|_{L^2_+}$0 exactly when $|\partial F|_{L^2_+}$1 is a Radon measure with $|\partial F|_{L^2_+}$2, in which case

$|\partial F|_{L^2_+}$3

Thus the unilateral weak slope is the $|\partial F|_{L^2_+}$4-norm of the positive part of the distributional gradient, reflecting the fact that only nonnegative directions are admissible (1908.10111).

Weak solutions are then formulated as unilateral gradient flows satisfying an energy inequality, and in fact an energy identity,

$|\partial F|_{L^2_+}$5

The same solutions satisfy the PDE characterization

$|\partial F|_{L^2_+}$6

for a.e. $|\partial F|_{L^2_+}$7, so the velocity equals the unilateral slope pointwise in time. The paper proves existence, uniqueness, comparison, and continuous dependence, and in the autonomous case identifies the constrained flow with the obstacle problem for

$|\partial F|_{L^2_+}$8

Here “weak slope” is therefore neither a lower semicontinuous envelope nor a sheaf-theoretic slope, but a one-sided metric slope adapted to irreversible evolution (1908.10111).

4. Weak slope inequalities for fibred surfaces

In algebraic geometry, the slope of a relatively minimal fibration $|\partial F|_{L^2_+}$9 with general fiber of genus $(-K)$0 is the numerical invariant

$(-K)$1

where

$(-K)$2

The classical Xiao–Cornalba–Harris slope inequality gives the universal bound

$(-K)$3

In this setting, the expression “weak slope inequality” is used broadly for lower bounds that are generalized, easier to obtain, hold under weaker structural assumptions, or are not sharp for all fibrations (Sun et al., 2016).

The paper studies exactly such a generalized inequality. If $(-K)$4 is a relative nef divisor on $(-K)$5 whose restriction to a general fiber $(-K)$6 is globally generated and special, and if

$(-K)$7

then

$(-K)$8

For $(-K)$9, this specializes to the classical slope inequality. For general $K$0, the right-hand side depends on fiberwise data such as $K$1, so the result is weaker in universality while stronger in scope. The paper explicitly describes this as a kind of weak slope inequality (Sun et al., 2016).

The same work clarifies three overlapping senses in which “weak slope” arises. First, divisor-based inequalities for $K$2 are weak because they do not yield a genus-only constant. Second, the classical lower bound $K$3 becomes weak when restricted to non-hyperelliptic fibrations, because it is not optimal there. Third, one may call a result weak when it is obtained under weaker hypotheses, such as without semistability assumptions in positive characteristic (Sun et al., 2016).

Using Langer’s theory of strongly semistable sheaves, the paper extends Xiao’s approach to any characteristic and rederives the classical slope inequality without semistability assumptions. It also proves an improved bound for relatively minimal non-hyperelliptic fibrations,

$K$4

thereby showing that the classical lower bound is sharp for hyperelliptic fibrations but only a weaker threshold in the non-hyperelliptic range (Sun et al., 2016).

5. Weak slope stability of tangent sheaves

A further, conceptually different usage occurs in the slope theory of torsion-free sheaves. For a normal projective variety $K$5 polarized by an ample, or more generally nef and big, divisor $K$6, the slope of a torsion-free sheaf $K$7 is

$K$8

When $K$9, one speaks of $f:\mathcal{X}\to\overline{\mathbb{R}}$0-slope stability. For the tangent sheaf,

$f:\mathcal{X}\to\overline{\mathbb{R}}$1

The paper on Fano varieties explicitly describes $f:\mathcal{X}\to\overline{\mathbb{R}}$2-slope stability as a weaker but more algebraic concept than $f:\mathcal{X}\to\overline{\mathbb{R}}$3-stability, and studies the slope instability of $f:\mathcal{X}\to\overline{\mathbb{R}}$4 through its maximal destabilizing subsheaf (Chen et al., 26 Jan 2026).

This context has its own historical line. A conjecture attributed to Iskovskikh asserted that the tangent bundle of a Picard rank one Fano manifold should be slope stable. Peternell–Wiśniewski and Hwang proved this up to dimension five, but Kanemitsu disproved it in 2021. The paper responds by analyzing the maximal destabilizing sheaf $f:\mathcal{X}\to\overline{\mathbb{R}}$5 using modern foliated minimal model program techniques (Chen et al., 26 Jan 2026).

For weak $f:\mathcal{X}\to\overline{\mathbb{R}}$6-Fano varieties with canonical singularities, the maximal destabilizing sheaf is shown to be an algebraically integrable foliation with rationally connected leaves, and its canonical divisor $f:\mathcal{X}\to\overline{\mathbb{R}}$7 is not pseudo-effective. This makes it natural to run a $f:\mathcal{X}\to\overline{\mathbb{R}}$8-MMP. In dimension two, that analysis leads to a complete classification of $f:\mathcal{X}\to\overline{\mathbb{R}}$9-slope unstable weak del Pezzo surfaces with canonical singularities. For nonsingular weak del Pezzo surfaces, the unstable cases are precisely $(\mathcal{X},d)$0 with $(\mathcal{X},d)$1 and the explicitly constructed surfaces $(\mathcal{X},d)$2, $(\mathcal{X},d)$3, and $(\mathcal{X},d)$4; in every case, the maximal destabilizing foliation is induced by the canonical fibration $(\mathcal{X},d)$5 (Chen et al., 26 Jan 2026).

Several consequences are emphasized. The work gives the first conceptual proof that $(\mathcal{X},d)$6 and $(\mathcal{X},d)$7 are the only $(\mathcal{X},d)$8-slope unstable nonsingular del Pezzo surfaces, recovering Fahlaoui’s 1989 result. It also exhibits a phenomenon absent for smooth Fano manifolds: there exists a del Pezzo surface with type $(\mathcal{X},d)$9 singularities admitting a weak Kähler–Einstein metric while its tangent sheaf is slope unstable. Here “weak slope” belongs to a stability-theoretic vocabulary, not to descent theory or slope inequalities (Chen et al., 26 Jan 2026).

6. Specialized and transferred usages

In some PDE literature, “weak slope” is best understood literally as a weak solution for an equation whose unknown is a slope variable. The continuum model

$$|\nabla f|(\bar{x}) := \limsup_{x\to \bar{x},\, x\neq \bar{x}} \frac{(f(\bar{x})-f(x))^+}{d(\bar{x},x)}.$$0

describes the step slope of a vicinal surface in the attachment–detachment-limited regime, where $$|\nabla f|(\bar{x}) := \limsup_{x\to \bar{x},\, x\neq \bar{x}} \frac{(f(\bar{x})-f(x))^+}{d(\bar{x},x)}.$$1 is the step slope as a function of step height. The paper defines a weak solution through regularity of $$|\nabla f|(\bar{x}) := \limsup_{x\to \bar{x},\, x\neq \bar{x}} \frac{(f(\bar{x})-f(x))^+}{d(\bar{x},x)}.$$2, an integral formulation on the positivity set, and two energy-dissipation inequalities; it proves existence of a global weak solution that is positive almost everywhere, long-time convergence to a constant solution, and space-time Hölder continuity. This usage is tied to the weakness of the solution concept, not to a generalized slope functional (Gao et al., 2016).

In high-dimensional statistics, SLOPE and Graph-Slope use “slope” as the name of the estimator rather than as a gradient or intersection-theoretic quotient. Graph-Slope minimizes a quadratic loss plus the ordered $$|\nabla f|(\bar{x}) := \limsup_{x\to \bar{x},\, x\neq \bar{x}} \frac{(f(\bar{x})-f(x))^+}{d(\bar{x},x)}.$$3 norm $$|\nabla f|(\bar{x}) := \limsup_{x\to \bar{x},\, x\neq \bar{x}} \frac{(f(\bar{x})-f(x))^+}{d(\bar{x},x)}.$$4 on graph differences and satisfies a sharp oracle inequality in prediction error; pattern recovery by SLOPE formalizes sign, clustering, and cluster ranking through the SLOPE pattern and gives a necessary and sufficient condition for exact pattern recovery (Bellec et al., 2017, Bogdan et al., 2022). A plausible implication is that “weak slope” in this statistical setting refers only informally to weak or partial pattern recovery, since the supplied papers treat exact recovery, asymptotic probabilities, and approximate clustering rather than introducing a standard object called weak slope (Bogdan et al., 2022).

In planar ABJM theory, the slope function is the coefficient of the linear term in the small-spin expansion of the anomalous dimension in the $$|\nabla f|(\bar{x}) := \limsup_{x\to \bar{x},\, x\neq \bar{x}} \frac{(f(\bar{x})-f(x))^+}{d(\bar{x},x)}.$$5-type sector. The paper computes this observable exactly via the Quantum Spectral Curve and studies both its weak-coupling and strong-coupling expansions. Here “weak” modifies the coupling regime, not the definition of slope; the final exact result is an all-loop formula in terms of the interpolating function $$|\nabla f|(\bar{x}) := \limsup_{x\to \bar{x},\, x\neq \bar{x}} \frac{(f(\bar{x})-f(x))^+}{d(\bar{x},x)}.$$6 (Gromov et al., 2014).

In geomechanics, finally, the phrase “weak slopes” denotes marginally stable slopes that may fail under rainfall-induced changes in pore pressure and saturation. The cited hydro-mechanical model couples Richards’ equation with linear elasticity and evaluates local stability through the Local Factor of Safety. This is again terminologically independent of the weak-slope notions of variational analysis and algebraic geometry, though it shows how “weak slope” can also acquire a direct physical meaning as a fragile slope configuration (Berrone et al., 6 Jul 2026).