Documentation

LeanPool.CIVAxisymmetric.Comparison.IntervalIntegralLimit

Interchanging a parameter limit with the outer time integral #

The comparison energy inequality eq:aniso:comparison:positive:energy is assembled from a per-slice statement: for each time s a regularization parameter kap is sent to 0โบ, and the resulting slice limits must then be integrated over the time interval [a, b]. Carrying out that last step means moving the limit across โˆซ s in a..b, ยท, which is dominated convergence along the filter ๐“[>] 0 rather than along a sequence.

Mathlib supplies the two ingredients โ€” MeasureTheory.tendsto_integral_filter_of_dominated_convergence and its interval counterpart intervalIntegral.tendsto_integral_filter_of_dominated_convergence โ€” but both are stated with hypotheses restricted to the half-open interval Set.uIoc a b, while every slice estimate in this development is produced on the closed interval Set.Icc a b. Bridging the two by hand at each call site is exactly the friction this file removes.

Main results #

The two small conversion lemmas intervalIntegrable_of_integrableOn_Icc and intervalIntegral_eq_setIntegral_Icc record the closed/half-open interval bridge separately, since they are useful on their own.

Closed versus half-open interval #

A function integrable on the closed interval Set.Icc a b is interval integrable on a..b. The closed interval is the domain on which the slice estimates of the comparison argument are produced, while IntervalIntegrable is phrased through Set.uIoc a b.

theorem CIV.intervalIntegral_eq_setIntegral_Icc {f : โ„ โ†’ โ„} {a b : โ„} (hab : a โ‰ค b) :
โˆซ (s : โ„) in a..b, f s = โˆซ (s : โ„) in Set.Icc a b, f s

The interval integral over a..b agrees with the set integral over the closed interval Set.Icc a b whenever a โ‰ค b: the endpoint a is a null set for Lebesgue measure. No integrability hypothesis is needed, both sides being 0 when the integrand is not integrable.

Dominated convergence along a filter #

theorem CIV.tendsto_intervalIntegral_of_tendsto_of_dominated_filter {ฮน : Type u_1} {l : Filter ฮน} [l.IsCountablyGenerated] {f : ฮน โ†’ โ„ โ†’ โ„} {g G : โ„ โ†’ โ„} {a b : โ„} (hab : a โ‰ค b) (hmeas : โˆ€แถ  (n : ฮน) in l, MeasureTheory.AEStronglyMeasurable (f n) (MeasureTheory.volume.restrict (Set.Icc a b))) (hlim : โˆ€แต (s : โ„) โˆ‚MeasureTheory.volume.restrict (Set.Icc a b), Filter.Tendsto (fun (n : ฮน) => f n s) l (nhds (g s))) (hdom : โˆ€แถ  (n : ฮน) in l, โˆ€แต (s : โ„) โˆ‚MeasureTheory.volume.restrict (Set.Icc a b), |f n s| โ‰ค G s) (hG : MeasureTheory.IntegrableOn G (Set.Icc a b) MeasureTheory.volume) :
Filter.Tendsto (fun (n : ฮน) => โˆซ (s : โ„) in a..b, f n s) l (nhds (โˆซ (s : โ„) in a..b, g s))

Dominated convergence for the interval integral, with every hypothesis stated on the closed interval Set.Icc a b and along an arbitrary countably generated filter l.

Measurability and the dominating bound are only required eventually along l, which is what the underlying Mathlib lemma provides and what the applications supply: a bound valid for all small positive parameters, not for every parameter.

Dominated convergence for the interval integral as the regularization parameter kap tends to 0 from the right. This is the form the comparison energy argument consumes: the slice limits are known for almost every time s, the slice bound |f kap s| โ‰ค G s holds for all small kap > 0, and G is integrable over the time interval.

The sequential form of tendsto_intervalIntegral_of_tendsto_of_dominated: a sequence of integrands dominated by a single integrable G on Set.Icc a b may be passed to the limit under the interval integral.

Dominated convergence with a constant dominating function. On a bounded interval a constant is automatically integrable, so no integrability hypothesis survives: a uniform slice bound |f kap s| โ‰ค C valid for all small kap > 0 suffices.

Dominated convergence for a nonnegative integrand, where the domination is naturally a one-sided bound 0 โ‰ค f kap s โ‰ค G s rather than a bound on the absolute value. The squared quantities of the energy identity arrive in exactly this shape.

The vanishing corollary: if the slice integrand tends to 0 for almost every time s and is dominated by an integrable G, then the time integral itself tends to 0. This is the shape the error terms of the comparison argument take.

The same limits for the set integral over Set.Icc a b #

The conclusion of tendsto_intervalIntegral_of_tendsto_of_dominated phrased for the set integral over the closed interval, which is how the time integral is written when it is produced by a slice-wise construction rather than by the fundamental theorem of calculus.

The constant-dominating-function form of tendsto_setIntegral_Icc_of_tendsto_of_dominated: on a bounded time interval a uniform slice bound alone licenses the interchange.