Voorhoeve index¶

An index of total absolute argument variation of an analytic function along a path, used in generalized Rolle and bounded complex-zero estimates.

Version: v1 · 2026-10-25 · History
Domain-specific #: 14454
Domain group: Formal Sciences
Origin domain: Mathematics
Subdomains: Complex Analysis, Analytic Zero Counting → Mathematics

Core Idea¶

The Voorhoeve index measures how much the value of an analytic function turns around zero as its input follows a path.^[1] It adds the absolute changes in argument, so a turn in one direction does not cancel a later turn in the other.^[1] On a real interval \(I=[a,b]\), for a function analytic near \(I\) and nonzero on \(I\), one convention is

\[V_I(f)=\frac{1}{2\pi}\int_a^b\left|\frac{d}{dt}\operatorname{Arg} f(t)\right|dt =\frac{1}{2\pi}\int_a^b\left|\operatorname{Im}\frac{f'(t)}{f(t)}\right|dt.\]

This is the normalized interval definition and its real-path logarithmic-derivative identity.^[1]

The integral is literal only when \(f\) has no zero on the path: \(f'/f\) is undefined at such a zero.^[1] Khovanskii and Yakovenko describe a separate continuity extension of the index for functions whose path contains zeros.^[1] They give the normalized segment inequality \(V_I(f)\leq V_I(f')+\tfrac12\), with its endpoint contribution; it plays an analogous role to the real Rolle comparison between zero counts of a function and its derivative.^[1]

For a parametrized complex curve \(z(t)\), the argument rate is \(\operatorname{Im}((f'/f)(z(t))z'(t))\).^[1]^[2] Some papers omit the \(1/(2\pi)\) normalization.^[2] The index is distinct from signed winding: it records total variation, not net turns.^[1] A closed-boundary index can help bound enclosed zeros only when analyticity, boundary, and domain conditions allow the argument principle to connect the two.^[1]^[2]

Structural Signature¶

Sig role-phrases: analytic function; admissible path; absolute argument change; declared scale; derivative comparison; closed-boundary bridge when counting zeros.

Analytic input. A complex analytic function supplies the values whose argument is followed, and its derivative enters the cited Rolle-type results.^[1] Replacing it with an arbitrary plotted curve may leave a geometric turn measure, but no longer carries the stated analytic derivative or zero theorems.
Admissible path. An interval or complex curve specifies where the values are observed.^[1] Nonvanishing on the path makes the displayed logarithmic-derivative formula valid; a path zero requires the source's extension procedure.^[1] Reversing the path does not change absolute variation.
Absolute variation and scale. Integrating the absolute argument rate is the defining operation.^[1] Dividing by \(2\pi\) reports turns rather than radians; it changes constants, not the underlying variation.^[1]^[2] Removing the absolute value instead changes the quantity to net signed argument change.
Derivative comparison. A Rolle–Voorhoeve inequality connects the index of \(f\) with that of \(f'\) plus a path-dependent term.^[1]^[2] This is a use of the index, not an extra ingredient in its definition: an index can be computed without first proving a zero bound.
Closed-boundary bridge. When \(f\) is analytic and pole-free on a neighborhood of a bounded domain's closure and nonzero on its positively oriented simple closed boundary, total absolute variation on that boundary bounds its signed winding and therefore the count of enclosed zeros.^[1] The one-interval value alone lacks this bridge.

What It Is Not¶

Not signed winding. Winding records net oriented change on a closed curve. Opposite turns can cancel there but add to the Voorhoeve index; equality in a special case does not identify the computations.^[1]
Not a real or complex zero count. The index is a variation quantity. A suitable contour and the argument principle can turn a bound on it into an upper bound on enclosed zeros; the index need not equal the number of zeros.^[1]
Not an automatic neighborhood bound from one interval. For \(f_n(z)=\cosh(nz)\), the argument of \(f_n(x)\) is constant on \([-1,1]\), so its segment index is zero; nonreal zeros \(i\pi(m+\tfrac12)/n\) nevertheless approach that interval as \(n\) grows. This elementary counterexample shows why a closed boundary and further hypotheses matter.
Not the Bernstein index. The papers compare generalized Rolle tools, but the Bernstein index measures a function's modulus growth between nested sets rather than its absolute argument variation along a path.^[1]

Scope of Application¶

The index is literal in complex analysis when an analytic function and a specified interval or curve supply the argument-variation calculation.^[1] On a real segment, Khovanskii and Yakovenko use the \(1/(2\pi)\) convention and obtain the \(+\tfrac12\) derivative comparison.^[1] On the boundary of a bounded convex complex domain, their normalized derivative inequality has a \(+1\) geometric term.^[1] These constants belong to different path settings, even under the same normalization.

One source-worked application is to finite-spectrum quasipolynomials: finite sums of polynomial-weighted exponentials.^[1] The authors use a bounded convex domain, a differential-equation factorization, and repeated contour inequalities to control the Chebyshev excess and hence the possible isolated complex zeros.^[1] Their theorem does not assert a uniform zero count for an entire unbounded strip.

A different application is to analytic power series whose tail coefficients satisfy a bound relative to a chosen coefficient.^[2] Yakovenko uses an unnormalized circle index, coefficient majorants and repeated derivative inequalities to give a sufficiently small centered disc containing at most \(k\) roots for his class \(P_{k,M}\).^[2] Its theorem's hypotheses and radius are specific; not every analytic series receives the same bound.

Clarity¶

A use of “Voorhoeve index” is clear only after it states which path, which function, whether zeros lie on that path, and which normalization is in force. On a real interval \(z(t)=t\), the logarithmic-derivative integrand is \(|\operatorname{Im}(f'/f)|\). Along a general curve it is \(|\operatorname{Im}((f'/f)(z(t))z'(t))|\); the factor \(z'(t)\) cannot be dropped. Yakovenko's inequality bounding this by \(|(f'/f)(z(t))||z'(t)|\) does not make the two integrands equal.^[1]^[2]

The normalization and geometry also determine which constant is being quoted. A \(+\tfrac12\) term belongs to the normalized segment result, a \(+1\) term to the normalized convex-boundary result, and \(+2\pi\) to the unnormalized circle result. A statement that merely says “Rolle's theorem for complex functions” can conceal these differences.^[1]^[2]

Manages Complexity¶

A complex analytic function may trace a curve that repeatedly turns around zero. The index compresses that entire turning history into one nonnegative number. This is useful for upper-bound reasoning: absolute changes cannot cancel, and the derivative comparison lets a problem descend through successive derivatives when the function class permits it.^[1]^[2]

The compression loses information. The scalar does not say where along the path a turn occurred, which direction dominated, or how many zeros lie inside a region. Those questions require the path, a closed-domain argument, and class-specific estimates. Thus the index simplifies a proof's control variable without replacing its geometric and analytic hypotheses.

Abstract Reasoning¶

Specify an analytic function and an interval or curve on which its values are nonzero, or justify the continuity extension at path zeros.
Compute or bound the total absolute argument change, recording whether the reported value is in radians or divided by \(2\pi\).
If a derivative inequality is used, select the version for the actual path geometry. Do not reuse the segment \(+\tfrac12\) term for a circle.
If the goal is an enclosed zero bound, choose a positively oriented simple closed boundary around a bounded domain where \(f\) is analytic and pole-free on a neighborhood of the closure and nonzero on the boundary. Use the argument principle to compare the zero count with the variation bound.^[1]
Supply the additional function-class estimate. For quasipolynomials it comes from the finite exponential spectrum and differential-equation chain; for \(P_{k,M}\) it comes from coefficient bounds on successive derivatives.^[1]^[2]

Knowledge Transfer¶

Within complex analysis, the same absolute-variation calculation can be used on different analytic function classes and paths. What transfers literally is the index calculation and, under its stated assumptions, a Rolle-type derivative comparison. The 1996 quasipolynomial theorem and 2000 power-series theorem show this transfer across unlike proof settings; their domain shapes, constants, coefficient conditions and resulting zero bounds do not transfer automatically.^[1]^[2]

Outside complex analysis, tracking total versus net change is a useful analogy, but the name “Voorhoeve index” brings analytic functions, argument, and complex-zero theorems with it. The broad Prime Measurement describes an instrument and procedure mapping an attribute to a scale with uncertainty; that identity does not contain this analytic index's definition. No cross-domain extension is claimed.

Examples¶

Canonical: exponential quasipolynomials on bounded convex domains¶

Khovanskii and Yakovenko take finite sums \(f(z)=\sum_{\lambda\in\Lambda}p_\lambda(z)e^{\lambda z}\), where \(\Lambda\) is finite and each \(p_\lambda\) is a polynomial.^[1] On a bounded convex complex domain, they use the index along its boundary and a factored differential equation for this function class.^[1] Repeated complex Rolle steps bound the boundary variation, yielding their Chebyshev-excess and isolated-zero estimate.^[1] This is a class-level theorem worked by the paper, not a claimed numerical computation for one selected \(f\).

Mapped back: analytic input → finite-spectrum quasipolynomial; admissible path → suitable boundary of the bounded convex domain; absolute variation and scale → normalized total change of \(\operatorname{Arg} f\) on that boundary; derivative comparison → repeated contour Rolle inequality through the factorization; closed-boundary bridge → the argument principle and Theorem 2's zero-count bound.

Applied: coefficient-bounded power series on small circles¶

Yakovenko's class \(P_{k,M}\) constrains Taylor coefficients after degree \(k\) relative to the \(k\)th coefficient.^[2] On a sufficiently small circle, his Lemma 3 makes the index of a normalized \(k\)th derivative less than \(2\pi\).^[2] Iterating the unnormalized circle inequality and invoking the argument principle, Theorem 1 gives an explicit centered radius whose disc has at most \(k\) roots.^[2] The paper's rational-series sharpness witness shows this class is broader than entire finite exponential quasipolynomials; it is not presented as a separate numerical Theorem 1 computation.^[2]

Mapped back: analytic input → coefficient-bounded convergent Taylor series; admissible path → sufficiently small circle \(C_r\); absolute variation and scale → unnormalized total \(\operatorname{Arg}\) change in radians; derivative comparison → Lemmas 2–4 control successive derivatives; closed-boundary bridge → argument-principle conversion of variation into an at-most-\(k\) root bound.

Structural Tensions¶

The index has no evidenced, constitutive contest between two goals. Absolute variation versus signed winding is a choice of different quantities; segment versus contour is a change in hypotheses; dividing by \(2\pi\) is a scale convention. Each distinction is necessary for correct use, but none is a universal tradeoff that every instance of the index must balance. An application may pose a proof-specific tradeoff between a tractable contour and a tight bound, but the consulted sources do not establish one as part of the named index.

Structural–Framed Character¶

Spectrum placement: strongly structural within mathematics, but tied to complex-analytic primitives rather than free of domain. Evaluative weight: low; the index is a defined nonnegative quantity, not a judgement that more turning is better. Human-practice dependence: the choice of normalization and of which zero problem to study is made by mathematicians, while the value under stated inputs is determined by the formula. Institutional origin: the name and cited theorems have a research history, but the calculation does not depend on an office or convention beyond its explicit scale. Vocabulary travel: “index,” “variation,” and “turning” occur widely; the complete formula travels literally only with complex values and their arguments. Import versus recognition: one recognizes a new instance by calculating the same absolute argument variation, not by calling any oscillation an instance. A broad measurement skeleton may explain the act of assigning a number, but it does not carry the Voorhoeve theorem into another domain. Its character: a reproducible complex-analytic metric with an explicit path, scale, and theorem-limited use.

Structural Core vs. Domain Accent¶

Skeletal relation: take a changing quantity along a path and accumulate its absolute change so reversals do not cancel. That resembles the general mathematical notion of total variation. Whether this skeleton merits a separate Prime across domains is a future Prime question; the named index alone does not establish that breadth. Domain-bound mechanism: the changing quantity is \(\operatorname{Arg} f\) for an analytic complex function; zero-free paths, logarithmic derivatives, endpoint or curvature terms, and the argument principle make the cited applications work. Why domain-specific: removing those conditions leaves an analogy about accumulating change, not the named index or its Rolle–Voorhoeve inequalities.

This entry under conditions is a kind of Total variation.

Measurement specifies an instrument, procedure, frame and uncertainty chain; assigning an analytic index does not alone instantiate that Prime's full signature. Oscillation describes recurrent variation, while the Voorhoeve index can record monotone argument turning as well as back-and-forth turning. Analytic Function names the input type, not a superclass of path-dependent indices. Total Variation is the mathematical parent for accumulated absolute change when the analytic function is nonzero on a piecewise-smooth path and a continuous argument lift exists: then this index is its scaled total variation.^[1] The continuity extension through path zeros is a distinct construction; its value is not asserted to be total variation of an argument lift there.^[1]

Relationships to Other Abstractions¶

Current abstraction Voorhoeve index Domain-specific

Parents (1) — outgoing typed relations

Voorhoeve index is a kind of, conditional Total variation Domain-specific

On a zero-free piecewise-smooth path, the index is scaled total variation of a continuous argument lift.

Hierarchy path (1) — routes to 1 parentless root

Voorhoeve index → is a kind of, conditional Total variation → is a kind of Aggregation → is a kind of Transformation → is a kind of, typical Function (Mapping)

Neighborhood in Abstraction Space¶

Voorhoeve index sits in a sparse region of the domain-specific corpus (77^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (2769 abstractions)

Nearest neighbors

Zeros and Poles — 0.83
Analytic Function — 0.83
Càdlàg Function — 0.83
Incidence (geometry) — 0.82
Asymptote — 0.82

Computed from structural-signature embeddings · 2026-10-25

Not to Be Confused With¶

Signed winding or the argument principle: signed net turns on a closed boundary can count enclosed zeros under hypotheses; Voorhoeve uses absolute total turns, which can be larger.^[1]
Bernstein index: concerns modulus growth between nested sets, not absolute argument change along one path.^[1]
Ordinary real Rolle's theorem: compares zeros of real \(f\) and \(f'\), supplying an analogy and contextual starting point. It is not the definition of the complex index.^[1]
A generic analytic function: the function is the input. The index is the result of a particular path-dependent calculation on it.

References¶

[1] A. Khovanskii and S. Yakovenko, “Generalized Rolle Theorem in ℝⁿ and ℂ,” Journal of Dynamical and Control Systems 2, no. 1 (1996): 103–123, doi:10.1007/BF02259625. Author-hosted complete paper inspected: printed p. 103 for the contextual ordinary real Rolle comparison; pp. 104–108 for the normalized interval and contour indices, zero-free condition, derivative inequalities and quasipolynomial theorem; p. 113 for the segment endpoint term; p. 116 for the quasipolynomial proof. It attributes the earlier index to Voorhoeve (1976), whose full paper was not inspected here. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r ↩s ↩t ↩u ↩v ↩w ↩x ↩y ↩z ↩27 ↩28 ↩29 ↩30 ↩31 ↩32 ↩33 ↩34 ↩35 ↩36 ↩37 ↩38

[2] Sergei Yakovenko, “On Zeros of Functions from Bernstein Classes,” Nonlinearity 13, no. 4 (2000): 1087–1094, doi:10.1088/0951-7715/13/4/306. Author-hosted complete published paper inspected: printed p. 1089 for the coefficient class; pp. 1091–1093 for the unnormalized circle index, Rolle–Voorhoeve inequality, derivative majorant, at-most-\(k\) radius theorem, Corollary 3 and sharpness illustration. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p