Integral of a 1-form along a path #
In this file we define the integral of a 1-form along a path indexed by [0, 1]
and prove basic properties of this operation.
The integral ∫ᶜ x in γ, ω x is defined as $\int_0^1 \omega(\gamma(t))(\gamma'(t))$.
More precisely, we use
Path.extend γ tinstead ofγ t, because both derivatives andintervalIntegralexpect globally defined functions;derivWithin γ.extend (Set.Icc 0 1) t, notderiv γ.extend t, for the derivative, so that it takes meaningful values att = 0andt = 1, even though this does not affect the integral.
The argument ω : E → E →L[𝕜] F is a 𝕜-linear 1-form on E taking values in F,
where 𝕜 is ℝ or ℂ.
The definition does not depend on 𝕜, see curveIntegral_restrictScalars and nearby lemmas.
However, the fact that 𝕜 = ℝ is not hardcoded
allows us to avoid inserting ContinuousLinearMap.restrictScalars here and there.
Main definitions #
curveIntegral ω γ, notation∫ᶜ x in γ, ω x, is the integral of a 1-formωalong a pathγ.CurveIntegrable ω γis the predicate saying that the above integral makes sense.
Main results #
We prove that curveIntegral behaves well with respect to
- operations on
Paths, seecurveIntegral_refl,curveIntegral_symm,curveIntegral_transetc; - algebraic operations on 1-forms, see
curveIntegral_addetc.
We also show that the derivative of fun b ↦ ∫ᶜ x in Path.segment a b, ω x
has derivative ω a at b = a.
We provide 2 versions of this result: one for derivative (HasFDerivWithinAt) within a convex set
and one for HasFDerivAt.
Implementation notes #
Naming #
In literature, the integral of a function or a 1-form along a path is called “line integral”, “path integral”, “curve integral”, or “curvilinear integral”.
We use the name “curve integral” instead of other names for the following reasons:
for many people whose mother tongue is not English, “line integral” sounds like an integral along a straight line;
we reserve the name "path integral" for Feynman-style integrals over the space of paths.
Usage of ContinuousLinearMaps for 1-forms #
Similarly to the way fderiv uses continuous linear maps
while higher order derivatives use continuous multilinear maps,
this file uses E → E →L[𝕜] F instead of continuous alternating maps for 1-forms.
Differentiability assumptions #
The definitions in this file make sense if the path is a piecewise $C^1$ curve.
Poincaré lemma (formalization WIP, see #24019) implies that for a closed 1-form on an open set U,
the integral depends on the homotopy class of the path only,
thus we can define the integral along a continuous path
or an element of the fundamental groupoid of U.
Usage of an extra field #
The definitions in this file deal with 𝕜-linear 1-forms.
This allows us to avoid using ContinuousLinearMap.restrictScalars
in HasFDerivWithinAt.curveIntegral_segment_source
and a future formalization of Poincaré lemma.
The function t ↦ ω (γ t) (γ' t) which appears in the definition of a curve integral.
This definition is used to factor out common parts of lemmas
about CurveIntegrable and curveIntegral.
Instances For
A 1-form ω is curve integrable along a path γ,
if the function curveIntegralFun ω γ t = ω (γ t) (γ' t) is integrable on [0, 1].
The actual definition uses Path.extend γ,
because both interval integrals and derivatives expect globally defined functions.
Instances For
Integral of a 1-form ω : E → E →L[𝕜] F along a path γ,
defined as $\int_0^1 \omega(\gamma(t))(\gamma'(t))$.
The actual definition uses curveIntegralFun which uses Path.extend γ
and derivWithin (Path.extend γ) (Set.Icc 0 1) t,
because calculus-related definitions in Mathlib expect globally defined functions as arguments.
Instances For
Integral of a 1-form ω : E → E →L[𝕜] F along a path γ,
defined as $\int_0^1 \omega(\gamma(t))(\gamma'(t))$.
The actual definition uses curveIntegralFun which uses Path.extend γ
and derivWithin (Path.extend γ) (Set.Icc 0 1) t,
because calculus-related definitions in Mathlib expect globally defined functions as arguments.
Instances For
curve integral is defined using Bochner integral, thus it is defined as zero whenever the codomain is not a complete space.
Operations on paths #
Alias of the reverse direction of curveIntegrable_cast_iff.
If ‖ω z‖ ≤ C at all points of the segment [a -[ℝ] b],
then the curve integral ∫ᶜ x in .segment a b, ω x has norm at most C * ‖b - a‖.
If a 1-form ω is continuous on a set s,
then it is curve integrable along any $C^1$ path in this set.
Algebraic operations on the 1-form #
Eta-expanded form of CurveIntegrable.zero
Eta-expanded form of CurveIntegrable.neg
Derivative of the curve integral w.r.t. the right endpoint #
In this section we prove that the integral of ω along [a -[ℝ] b], as a function of b,
has derivative ω a at b = a.
We provide several versions of this theorem, for HasFDerivWithinAt and HasFDerivAt,
as well as for continuity near a point and for continuity on the whole set or space.
Note that we take the derivative at the left endpoint of the segment.
Similar facts about the derivative at a different point are true
provided that ω is a closed 1-form (formalization WIP, see #24019).
The integral of ω along [a -[ℝ] b], as a function of b, has derivative ω a at b = a.
This is a HasFDerivWithinAt version assuming that ω is continuous within a convex set s
in a neighborhood of a within s.
The integral of ω along [a -[ℝ] b], as a function of b, has derivative ω a at b = a.
This is a HasFDerivWithinAt version assuming that ω is continuous on s.
The integral of ω along [a -[ℝ] b], as a function of b, has derivative ω a at b = a.
This is a HasFDerivAt version assuming that ω is continuous in a neighborhood of a.
The integral of ω along [a -[ℝ] b], as a function of b, has derivative ω a at b = a.
This is a HasFDerivAt version assuming that ω is continuous on the whole space.