On the directional growth of the resolvent norm
Pith reviewed 2026-05-15 20:59 UTC · model grok-4.3
The pith
For any point in the resolvent set of a closed operator, the resolvent norm either attains a global minimum there or increases at least linearly or quadratically along some straight-line direction inside the set.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
For each z in ρ(A), either there exist z' ≠ z in ρ(A) with the line segment [z, z'] contained in ρ(A) such that ||R_A(ζ)|| ≥ ||R_A(z)|| + C |ζ - z|^δ for all ζ on the segment (with δ equal to 1 or 2 and C > 0), or else ζ ↦ ||R_A(ζ)|| attains its global minimum at ζ = z.
What carries the argument
The straight-line segment [z, z'] inside the resolvent set ρ(A), used to test one-sided directional growth of the norm ||R_A(ζ)||.
If this is right
- The resolvent norm cannot remain constant along any nontrivial line segment inside the resolvent set unless that segment lies at a global minimum.
- Local minima of the resolvent norm, when they exist, must be global.
- In directions where the norm increases, the growth rate is at least linear or quadratic, never slower.
- This directional classification applies uniformly to all points in ρ(A) for any closed densely defined operator with nonempty resolvent set.
Where Pith is reading between the lines
- The result may constrain how pseudospectra can deform near the resolvent set, since the norm controls the distance to the spectrum along rays.
- One could test whether the same growth alternatives hold when the line segment is replaced by a smooth curve inside ρ(A).
- The classification might extend to the joint resolvent norm for several commuting operators.
Load-bearing premise
The resolvent set of the closed densely defined operator A is nonempty.
What would settle it
A concrete closed operator A on a separable Hilbert space together with a point z in ρ(A) where ||R_A(ζ)|| is neither a global minimum at z nor satisfies the linear or quadratic lower bound along any line segment from z that stays inside ρ(A).
read the original abstract
Let $A$ be a closed densely defined operator on a separable Hilbert space $\mathcal{H}$. Assume the resolvent set $\rho(A)$ is non-empty. For $z,z'\in\rho(A)$ let $[z,z']$ denote the straight line segment from $z$ to $z'$. For each $z\in\rho(A)$ we classify the behavior of the resolvent norm $\zeta\mapsto\lVert R_A(\zeta) \rVert$ near $z$. Either there are $z'\in\rho(A)$, $z'\neq z$, $[z,z']\subset\rho(A)$, such that $\lVert R_A(\zeta) \rVert \geq \lVert R_A(z) \rVert + C\lvert \zeta-z \rvert^\delta$ for $\zeta\in[z,z']$ with $\delta=1$ or $\delta=2$, or the function $\zeta\mapsto\lVert R_A(\zeta) \rVert$ has a global minimum at $\zeta=z$.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript classifies the local directional behavior of the resolvent norm ||R_A(ζ)|| on the resolvent set ρ(A) for a closed densely defined operator A on a separable Hilbert space H, assuming ρ(A) is non-empty. For each z ∈ ρ(A), either there exists z' ≠ z with [z, z'] ⊂ ρ(A) such that ||R_A(ζ)|| ≥ ||R_A(z)|| + C |ζ - z|^δ (δ = 1 or 2) for ζ on the segment, or ζ ↦ ||R_A(ζ)|| attains a global minimum at z.
Significance. If the classification holds, it provides a sharp, geometry-based description of possible local minima or growth rates for the resolvent norm, leveraging only the openness of ρ(A) and the continuity of the norm (from analyticity of the resolvent). This could inform studies of pseudospectra and operator norms without additional assumptions on A. The result is parameter-free and rests on standard Hilbert-space axioms.
major comments (2)
- §2 (main theorem statement): the claim that the growth is at least linear or quadratic is not accompanied by an explicit constant C or a construction of the direction z'; the proof sketch in §3 relies on the openness of ρ(A) but does not verify that the segment remains in ρ(A) when the minimum is not global.
- Eq. (2.3) (definition of directional derivative): the reduction to δ=1 or δ=2 appears to follow from the fact that the norm is Lipschitz or differentiable in certain directions, but no explicit computation shows why higher-order terms cannot occur or why the constant C is independent of the choice of segment.
minor comments (2)
- The separability assumption on H is stated but never used; either remove it or cite where it is needed for the existence of the segment.
- Notation: the line segment [z, z'] is introduced without specifying whether it is closed or open; this affects the inequality at the endpoint z'.
Simulated Author's Rebuttal
We thank the referee for the careful reading and constructive comments on the manuscript. We address each major comment below and indicate the changes planned for the revised version.
read point-by-point responses
-
Referee: §2 (main theorem statement): the claim that the growth is at least linear or quadratic is not accompanied by an explicit constant C or a construction of the direction z'; the proof sketch in §3 relies on the openness of ρ(A) but does not verify that the segment remains in ρ(A) when the minimum is not global.
Authors: We appreciate the referee highlighting these points for clarification. In the revised manuscript we will explicitly construct the direction: if z is not a global minimum, there must exist a direction v (unit vector) such that the right directional derivative of the resolvent norm at z in direction v is positive; we select z' = z + t v for sufficiently small t > 0. Because ρ(A) is open, the ball of radius r around z lies in ρ(A) for some r > 0, so the entire segment [z, z'] remains in ρ(A) when t < r. The constant C is taken to be half the directional derivative (for δ = 1) or a positive quantity derived from the second-order difference quotient (for δ = 2). We will expand the proof sketch in §3 to include this construction, the choice of sufficiently small t, and the verification that the segment lies in ρ(A). revision: partial
-
Referee: Eq. (2.3) (definition of directional derivative): the reduction to δ=1 or δ=2 appears to follow from the fact that the norm is Lipschitz or differentiable in certain directions, but no explicit computation shows why higher-order terms cannot occur or why the constant C is independent of the choice of segment.
Authors: We agree that an explicit computation is needed. Along any line in ρ(A) the real-valued function f(t) = ||R_A(z + t v)|| is continuous and locally Lipschitz (by holomorphy of the resolvent). We will add the following analysis: the right directional derivative D_v^+ f(0) exists and is finite. If it is positive for some v, set δ = 1 and C equal to half this derivative; the inequality then holds on [0, t] for all sufficiently small t by the definition of the derivative. If the first derivative vanishes in every direction, we examine the second-order quotient liminf_{t→0^+} [f(t) - f(0)] / t^2. If this liminf is positive for some v, set δ = 2 and C equal to half the liminf. Higher-order vanishing (o(t^2)) forces both the first and second quotients to be zero, which we show implies that z is a local minimum of the norm; a separate argument then establishes that any local minimum is global. The constant C depends only on the chosen direction and the local Lipschitz constant of f near 0, hence is independent of the segment length once the segment is short enough to lie inside the neighborhood where the derivative bounds hold. This computation will be inserted after Eq. (2.3) and referenced in the proof. revision: yes
Circularity Check
No significant circularity
full rationale
The paper's central result is a classification of the local directional behavior of the resolvent norm along line segments inside the resolvent set. It follows directly from the holomorphic dependence of the resolvent on the spectral parameter (hence continuity of its norm) together with the openness of ρ(A), which guarantees the existence of line segments [z,z'] entirely contained in ρ(A). No step in the argument reduces by construction to a fitted parameter, a self-definitional equivalence, or a load-bearing self-citation; the alternatives (linear or quadratic growth, or global minimum) are exhaustive consequences of these standard analytic and topological properties. The derivation is therefore self-contained within classical operator theory.
Axiom & Free-Parameter Ledger
axioms (2)
- domain assumption A is a closed densely defined operator on a separable Hilbert space
- domain assumption The resolvent set ρ(A) is non-empty
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.