Fundamental theorem of calculus for a logarithmic-derivative integrand #
For a function f : ℝ → ℂ continuous on [a, b], differentiable off a countable set P, and
staying in Complex.slitPlane on [a, b], the principal Complex.log ∘ f is a single-valued
antiderivative of the logarithmic-derivative integrand f' t / f t, so
∫ t in a..b, f' t / f t = Complex.log (f b) - Complex.log (f a).
Specializing to f t = (γ t - w) / (γ a - w) for a curve γ avoiding w gives the contour form
used downstream, ∫ t in a..b, γ' t / (γ t - w) = Complex.log ((γ b - w) / (γ a - w)), where the
normalization by γ a - w is what keeps the ratio in the slit plane and makes the value at the
basepoint a equal to Complex.log 1 = 0. The exceptional set P accommodates the finitely many
breakpoints of a piecewise-C¹ contour, and the oriented interval [a, b] needs no a ≤ b
assumption.
Specializing the other way, to f = h ∘ γ for a piecewise-C¹ curve γ and an analytic h,
gives the curve form of the same principle: if h takes its values along γ in
Complex.slitPlane, then Complex.log ∘ h ∘ γ is a single-valued primitive of the
argument-principle integrand deriv γ • (logDeriv h ∘ γ), so that integral is an endpoint
difference — in particular it vanishes on a closed curve. This is the corner-tolerant replacement
for circleIntegral.integral_eq_zero_of_hasDerivWithinAt, which needs a genuinely differentiable
contour.
That same integrand is what the winding number of the image curve computes: if h is moreover
zero-free along γ, then h ∘ γ misses the origin and its index integral
(2πi)⁻¹ ∮_{h ∘ γ} dw / w is, after the substitution w = h z, exactly
(2πi)⁻¹ ∫_a^b γ' • (logDeriv h ∘ γ). Formally the substitution is the chain rule
(h ∘ γ)' = γ' · h'(γ), available off the breakpoints of γ and so almost everywhere, which is all
an integral sees. This bridge is what turns every logarithmic-derivative integral identity — the
argument principle above all — into a statement about how often an image curve winds.
Main results #
TauCeti.Contour.analyticAt_logDeriv_of_analyticAt—logDeriv fis analytic whereverfis analytic and nonzero; the regularity input shared by the results below and by the argument principle.TauCeti.Contour.intervalIntegrable_deriv_div_and_integral_deriv_div_eq_log_sub_log_of_im_nonnegandintervalIntegrable_deriv_div_and_integral_deriv_div_eq_log_neg_sub_log_neg_of_im_nonpos— the boundary-tolerant comparison FTCs: the logarithmic integral ofgis integrable and evaluates to an endpoint-log difference through a comparisonhthat is continuous on the closed interval, differentiable on the open one off a countable set, has integrable logarithmic integrand, is confined to a closed half-plane, is nonvanishing at the two endpoints and slit-plane-valued strictly between them, and agrees withgboth on the open interval and at each endpoint — so endpoint values may sit on the negative-real boundary.intervalIntegrable_deriv_div_and_integral_deriv_div_eq_log_sub_log_of_mem_slitPlane(inTauCeti.Contour, unqualified here only to stay inside the line limit) — the third member of that family, for a comparison confined to the slit plane on the whole closed interval rather than held off the cut by a half-plane condition. Its conclusion could be reached inline fromintegral_deriv_div_eq_log_sub_logplusIntervalIntegrable.congr_uIooandintervalIntegral.integral_congr_uIoo, but that transport is performed at nine call sites across the winding developments, so it is factored here.- the
_of_leforms of the slit-plane and upper comparisons — the ordered-interval versions callers usually have:a ≤ b, hypotheses read onSet.Icc a bandSet.Ioo a brather than throughminandmax, and integrability derived from a continuous derivative. TauCeti.Contour.integral_deriv_div_eq_log_sub_log— the slit-plane logarithmic-derivative FTC in generalf' / fform.TauCeti.Contour.integral_deriv_div_sub_eq_log— its contour specialization tof t = (γ t - w) / (γ a - w), the per-segment step for evaluating the winding-number integral as a sum ofComplex.logargument increments.TauCeti.Contour.integral_inv_sub_mul_deriv_eq_log— thederiv γform with the winding-integral integrand(γ t - w)⁻¹ * deriv γ t, ready for the downstream winding sum.TauCeti.Contour.intervalIntegrable_deriv_smul_logDeriv— interval-integrability of the argument-principle integrandderiv γ • (logDeriv h ∘ γ)along a piecewise-C¹curve.TauCeti.Contour.integral_deriv_smul_logDeriv_eq_zero_of_mem_slitPlane— that integral vanishes along a closed piecewise-C¹curve on whichhis analytic and slit-plane-valued.TauCeti.Contour.windingNumber_comp_eq_integral_logDeriv— the bridge to the image curve: the winding number ofh ∘ γabout the origin is(2πi)⁻¹ ∫_a^b γ' • (logDeriv h ∘ γ).
Provenance #
Adapted from segment_log_FTC in WindingInteger.lean of the AINTLIB LeanModularForms
development, split from the argument-lift PR (#759) as an independent contour prerequisite.
The boundary-tolerant comparison forms port the corner FTC pieces of AINTLIB's
valence-formula winding-weight development
(ForMathlib/ValenceFormula/WindingWeights/I.lean, Rho.lean, RhoPlusOne.lean)
onto the current Mathlib pin.
Logarithmic-derivative FTC on the slit plane. For f continuous on [a, b], differentiable
off a countable set P, taking values in Complex.slitPlane throughout [a, b], and with f' / f
interval-integrable, the integral of f' t / f t over a..b telescopes through the single-valued
branch Complex.log ∘ f:
∫ t in a..b, f' t / f t = Complex.log (f b) - Complex.log (f a).
FTC for a contour logarithmic-derivative integrand. The f t = (γ t - w) / (γ a - w)
specialization of integral_deriv_div_eq_log_sub_log: for γ continuous on [a, b] and
differentiable off a countable set P, with the normalized ratio in Complex.slitPlane throughout
[a, b] and t ↦ γ' t / (γ t - w) interval-integrable, the integral of that integrand over a..b
equals Complex.log ((γ b - w) / (γ a - w)).
Winding-integrand form of the contour log-derivative FTC. The γ' = deriv γ specialization
of integral_deriv_div_sub_eq_log, stated with the winding-integral integrand
(γ t - w)⁻¹ * deriv γ t in both the integrability hypothesis and the conclusion (matching the
g (γ t) * deriv γ t shape of integral_comp_mul_deriv_eq_sub_of_hasDerivAt and the winding API),
so a downstream winding sum can apply it per segment without rearranging the integrand.
At a point where a function is analytic and non-vanishing, its logarithmic derivative
logDeriv f = deriv f / f is analytic. This is the regularity input for every integrability and
residue statement about a logarithmic-derivative integrand, from the interval-integrability lemma
below to the residue form of the argument principle
(TauCeti.Contour.residue_logDeriv_eq_meromorphicOrderAt).
The argument-principle integrand deriv γ • (logDeriv h ∘ γ) is interval-integrable along a
piecewise-C¹ curve γ on which h is analytic and zero-free. This supplies the integrability
hypothesis of the logarithmic-derivative contour results — IntervalIntegrable is what
integral_deriv_div_eq_log_sub_log and its specializations ask of the integrand.
A slit-plane-valued function has a single-valued logarithm along a closed curve. If h is
analytic along a closed piecewise-C¹ curve γ and takes its values there in
Complex.slitPlane, then Complex.log ∘ h is a primitive of logDeriv h along γ, so the
contour integral of logDeriv h vanishes.
This is the corner-tolerant replacement for circleIntegral.integral_eq_zero_of_hasDerivWithinAt:
γ need only be differentiable off the countably many breakpoints, which is what
integral_deriv_div_eq_log_sub_log asks for. Nothing is required of h off the curve — in the
Rouché application h = g / f has both zeros and poles inside.
The winding number of the image curve is the logarithmic-derivative integral. If h is
analytic and zero-free along a piecewise-C¹ curve γ, the composite h ∘ γ avoids the origin
and
n_0(h ∘ γ) = (2πi)⁻¹ ∫_a^b γ' t · (logDeriv h) (γ t).
This is the substitution w = h z in the index integral (2πi)⁻¹ ∮_{h ∘ γ} dw / w, carried out by
the chain rule. The chain rule is available only off the breakpoints of γ, a finite set, so the
two integrands agree merely almost everywhere — enough for both the integrals and the
interval-integrability to transfer. No regularity of h ∘ γ beyond continuity and this
almost-everywhere derivative is needed, so nothing has to be said about h ∘ γ being
piecewise C¹ (it is, but windingNumber does not ask).
The analyticity hypothesis is local: AnalyticAt ℂ h (γ t) asks only for some neighbourhood of
each point of the curve on which h is analytic, never for one ambient open set carrying the whole
curve, and imposes no condition beyond those neighbourhoods. So the lemma applies verbatim to a
function with zeros and poles inside the region the curve encloses; that is what the argument
principle then counts.
The boundary-tolerant logarithmic FTC, upper form: for a comparison function h
continuous on the closed interval, confined to the closed upper half-plane, nonvanishing at the
endpoints, slit-plane-valued strictly between them, differentiable there off a countable set, and
with integrable logarithmic integrand, and for a g agreeing with h on the open interval and
at both endpoints, the logarithmic integral of g is integrable and evaluates to the difference
of its endpoint logarithms — even when the endpoint values sit on the slit-plane boundary.
The conclusion is about g, which is otherwise unconstrained: it is the agreement with h, at
the endpoints as well as inside, that carries the regularity across.
The boundary-tolerant logarithmic FTC, lower form: for a comparison function h
continuous on the closed interval, differentiable on the open one off a countable set, with
integrable logarithmic integrand, confined to the closed lower half-plane, nonvanishing at the
endpoints and whose negation is slit-plane-valued strictly between them, and for a g agreeing
with h on the open interval and at both endpoints, the logarithmic integral of g is
integrable and evaluates to the difference of the endpoint logarithms of the negations.
The comparison logarithmic FTC on the slit plane: for a comparison function h that
stays in the slit plane on the whole oriented closed interval — endpoints included — is
continuous there, is differentiable strictly inside off a countable exceptional set P, and
has interval-integrable logarithmic derivative, and for a g agreeing with h on the open
interval and at both endpoints, the logarithmic integral of g is integrable and evaluates
to the difference of its endpoint logarithms.
Integrability of deriv h / h is assumed rather than derived: callers that have a continuous
derivative and nonvanishing h get it from ContinuousOn.div plus
ContinuousOn.intervalIntegrable.
This is the sibling of the two half-plane forms above: there the comparison may touch the branch cut at an endpoint and is held off it by a half-plane condition, whereas here it is slit-plane-valued throughout and no half-plane hypothesis is needed.
The comparison logarithmic FTC on the slit plane, ordered-interval form. The form
callers actually have to hand: an oriented interval a ≤ b, a comparison function h
continuous with continuous derivative on Icc a b, slit-plane-valued there and
differentiable strictly inside, and a g agreeing with h on Ioo a b and at both
endpoints.
Integrability of deriv h / h is derived rather than assumed — slit-plane confinement
already forces h to be nonvanishing — and the interval hypotheses read on Set.Icc a b
and Set.Ioo a b instead of through min and max. Everything else is
intervalIntegrable_deriv_div_and_integral_deriv_div_eq_log_sub_log_of_mem_slitPlane, which
remains the general statement to reach for when the interval is unoriented, the
exceptional set is nonempty, or integrability comes from somewhere other than a continuous
derivative.
The boundary-tolerant logarithmic FTC on an ordered interval, upper form. The form
callers usually have: an oriented interval a ≤ b, a comparison function h continuous with
continuous derivative and nonvanishing on Icc a b, confined to the closed upper half-plane
there and slit-plane-valued strictly inside, and a g agreeing with h on Ioo a b and at
both endpoints.
Integrability of deriv h / h is derived from the continuous derivative and the
nonvanishing rather than assumed, and the interval hypotheses read on Set.Icc a b and
Set.Ioo a b instead of through min and max. Everything else is
intervalIntegrable_deriv_div_and_integral_deriv_div_eq_log_sub_log_of_im_nonneg, which
stays the statement to reach for on an unoriented interval or a nonempty exceptional set.