Leibniz integral

Integration by substitution

\[\int_{a}^{b} f(x)dx=\int_{\alpha }^{\beta } f(\varphi(t)){\varphi}' (t)dt\]

Leibniz integral rule

  1. Leibniz integral rule

  2. How to take the derivative of the integral?

General form: differentiation under the integral sign

\[\int_{a(x)}^{b(x)} f\left ( x,t \right ) dt\]

Theorem — Let f(x,t) be a function such that both \(f\left ( x,t \right )\) are continuous it t and x in some region of the xt-plane, including \(a(x)\le t\le b(x),x_{0} \le x\le x_{1}\). Also suppose that the functions \(a(x)\) and functions \(b(x)\) are both continuous and both have continuous derivatives for \(x_{0} \le x\le x_{1}\).Then for \(x_{0} \le x\le x_{1}\),

\[{\displaystyle {\frac {d}{dx}}\left(\int _{a(x)}^{b(x)}f(x,t)\,dt\right)=f{\big (}x,b(x){\big )}\cdot {\frac {d}{dx}}b(x)-f{\big (}x,a(x){\big )}\cdot {\frac {d}{dx}}a(x)+\int _{a(x)}^{b(x)}{\frac {\partial }{\partial x}}f(x,t)\,dt.}\]

The right hand side may also be written using Lagrange’s notation as:

\[{\textstyle f(x,b(x))\,b^{\prime }(x)-f(x,a(x))\,a^{\prime }(x)+\displaystyle \int _{a(x)}^{b(x)}f_{x}(x,t)\,dt.}\]

another form

\[\Phi(x)=\int_{\alpha(x) }^{\beta(x)} f(x, y) dy\]
\[\Phi(t)=\int_{\alpha(t) }^{\beta(t)} f(t, x) dx\]
\[\Phi(t)=\int_{\alpha(t) }^{\beta(t)} F(x, t) dx\]
\[\begin{split}\begin{align} \frac{\mathrm{d} \Phi(t)}{\mathrm{d} t}&=\frac{\Phi(t+\Delta t)-\Phi(t)}{\Delta t}\\ &=\int_{\alpha(t) }^{\beta(t)}\frac{ F(x, t+\Delta t)-F(x, t)}{\Delta t} dx\\ &+\frac{1}{\Delta t}\int_{\beta(t) }^{\beta(t+\Delta t)}{ F(x, t+\Delta t)} dx \\ &-\frac{1}{\Delta t}\int_{\alpha(t) }^{\alpha(t+\Delta t)}{ F(x, t+\Delta t)} dx \end{align}\end{split}\]
\[\begin{split}\begin{align} \frac{\mathrm{d} \Phi(t)}{\mathrm{d} t}&=\frac{\mathrm{d} }{\mathrm{d} t}\int_{\alpha(t) }^{\beta(t)}F(x, t) dx\\\\ &=\int_{\alpha(t) }^{\beta(t)}\frac{\partial F(x, t)}{\partial t} dx +[ F(\beta(t), t)\dot{\beta}(t) - F(\alpha(t), t)\dot{\alpha}(t) ]\\ \end{align}\end{split}\]

This can also be written as

\[\cfrac{\mathrm{d}}{\mathrm{d}t}{\int_{g(t)}^{h(t)}F(x,t)\mathrm{d}x }={\int_{g(t)}^{h(t)}\frac{\partial F(x,t)}{\partial t}\mathrm{d}x }+\left \{ F[h(t),t]\dot{h}(t)-F[g(t),t]\dot{g}(t) \right \}\]

One proof runs as follows, modulo precisely stated hypotheses and some analytic details. Set

\[\Phi(u, v, t)=\int_{u}^{v} F(x, t) d x\]

\(u = g(t)\), and \(v = h(t)\). By the chain rule

\[\frac{d}{d t} \Phi[g(t), h(t), t]=\left(\frac{\partial \Phi}{\partial u} \dot{g}+\frac{\partial \Phi}{\partial v} h\right)+\frac{\partial \Phi}{\partial t}\]

The first two terms are bracketed because they measure all changes due to variation of the interval of integration [g(t), h(t)], and they are evaluated by applying the Fundamental Theorem to \(\Phi(u, v, t)=\int_{u}^{v} F(x, t) d x\). The third term measures change due to variation of the integrand. If enough smoothness is assumed to justify interchange of the integration and differentiation operators, then

\[\frac{\partial \Phi}{\partial t}=\frac{\partial}{\partial t} \int_{u}^{v} F(x, t) d x=\int_{u}^{v} \frac{\partial F(x, t)}{\partial t} d x .\]

Another proof

\[\begin{split}\begin{aligned} \phi_{t}(u) & =x(u, t), \\ \phi_{t}:[a, b] & \rightarrow\left[\phi_{t}(a), \phi_{t}(b)\right]=[g(t), h(t)]=C_{t} . \end{aligned}\end{split}\]

By the formula for change of variable in a simple integral

\[\int_{g(t)}^{h(t)} F(x) d x=\int_{\phi_{t}(a)}^{\phi_{t}(b)} F(x) d x=\int_{a}^{b} F[x(u, t)] \frac{\partial x}{\partial u} d u .\]

This transition is excellent, because it has changed the integral over a moving domain to one over a fixed domain. We pay for this fixed domain with a time-varying integrand. No matter, we like it; we thrive on differentiation under the integral sign:

\[\begin{split}\begin{aligned} \frac{\mathrm{d} }{\mathrm{d} t} \int_{g(t)}^{h(t)} F(x) d x &=\frac{\mathrm{d} }{\mathrm{d} t} \int_{a}^{b} F[x(u, t)] \frac{\partial x(u, t)}{\partial u} d u \\ &=\int_{a}^{b}\frac{\partial}{\partial t}\left \{ F[x(u, t)] \frac{\partial x(u, t)}{\partial u}\right \}d u \\ &=\int_{a}^{b}\left \{\frac{\partial F[x(u, t)]}{\partial x}\frac{\partial x(u, t)}{\partial t} \frac{\partial x(u, t)}{\partial u}+F[x(u, t)]\frac{\partial x^{2} (u, t)}{\partial u\partial t}\right \}d u \\ \end{aligned}\end{split}\]

The fixed domain has done its job, and we return to the moving domain. The instantaneous velocity is \(v = v(u, t) = \partial x(u, t)/ {\partial t}\), which we also consider as a function of \(x\) and \(t\) via the transformation \((u, t) \leftrightarrow (x, t)\). When \(t\) is fixed,

\[\begin{split}\begin{aligned} \frac{\partial x^{2} (u, t)}{\partial u\partial t}&=\frac{\partial v(u, t)}{\partial u}\\ &=\cfrac{\cfrac{\partial v(u, t)}{\partial u}}{\cfrac{\partial x(u, t)}{\partial u}} {\frac{\partial x(u, t)}{\partial u}} \\ &={\cfrac{\partial v(u, t)}{\partial x(u, t)}}{\cfrac{\partial x(u, t)}{\partial u}} \end{aligned}\end{split}\]

hence

\[\begin{split}\begin{aligned} &\int_{a}^{b}\left \{\frac{\partial F[x(u, t)]}{\partial x}\frac{\partial x(u, t)}{\partial t} \frac{\partial x(u, t)}{\partial u}+F[x(u, t)]\frac{\partial x^{2} (u, t)}{\partial u\partial t}\right \}d u \\ =&\int_{a}^{b}\left \{\frac{\partial F[x(u, t)]}{\partial x}\frac{\partial x(u, t)}{\partial t} \frac{\partial x(u, t)}{\partial u}+F[x(u, t)]{\cfrac{\partial v(u, t)}{\partial x(u, t)}}{\cfrac{\partial x(u, t)}{\partial u}}\right \}d u \\ =&\int_{a}^{b}\left \{\frac{\partial F[x(u, t)]}{\partial x}\frac{\partial x(u, t)}{\partial t} +F[x(u, t)]{\cfrac{\partial v(u, t)}{\partial x(u, t)}}\right \}{\cfrac{\partial x(u, t)}{\partial u}}d u \\ =&\int_{\phi_{t}(a)}^{\phi_{t}(b)}\left \{\frac{\partial F[x(u, t)]}{\partial x}\frac{\partial x(u, t)}{\partial t} +F[x(u, t)]{\cfrac{\partial v(u, t)}{\partial x(u, t)}}\right \}d x \\ \end{aligned}\end{split}\]

that is

\[\begin{split}\begin{aligned} &\int_{\phi_{t}(a)}^{\phi_{t}(b)}\left \{\frac{\partial F[x(u, t)]}{\partial x}\frac{\partial x(u, t)}{\partial t} +F[x(u, t)]{\cfrac{\partial v(u, t)}{\partial x(u, t)}}\right \}d x \\ =&\int_{\phi_{t}(a)}^{\phi_{t}(b)}\frac{\partial }{\partial x} [F[x(u, t)]v(u, t)]d x \\ =&\int_{g(t)}^{h(t)}\frac{\partial }{\partial x} [F[x(u, t)]v(u, t)]d x \end{aligned}\end{split}\]

Two-dimensional, time-dependent case

We are also given a function \(F(x, y, t)\). The problem is to find

\[\frac{\mathrm{d} }{\mathrm{d} t}\iint_{D(t)}^{} F(x,y,t)dxdy\]

Certainly our first move should be separation of boundary variation from integrand variation. This is easy enough by the chain rule device in the first section and results in

\[\begin{split}\begin{aligned} &\frac{\mathrm{d} }{\mathrm{d} t}\iint_{D(t)}^{} F(x,y,t)dxdy{\Bigg|}_{t=t_{0} }\\ =&\frac{\mathrm{d} }{\mathrm{d} t}\iint_{D(t)}^{} F(x,y,t_{0})dxdy{\Bigg|}_{t=t_{0} }+ \iint_{D(t_{0})}^{} \frac{\partial F(x,y,t)}{\partial t}{\Bigg|}_{t=t_{0} }dxdy \end{aligned}\end{split}\]

This is routine. The essence of the problem is to find

\[\frac{\mathrm{d} }{\mathrm{d} t}\iint_{D(t)}^{} F(x,y)dxdy\]

Let \(\mathbf{v}=\mathbf{v}(x, y, t)\) denote the velocity vector at a boundary point \((x, y)\) of \(Dt\) and let \(n\) denote the outward unit normal. In the difference

\[\iint_{D(t+\Delta t )}^{} F(x,y)dxdy-\iint_{D(t)}^{} F(x,y)dxdy\]

everything in the overlap of \(D(t)\) and \(D(t+\Delta t)\) cancels; only the thin boundary strip makes a contribution. From the detail, this contribution is

\[F(x,y)(\mathbf{v}\Delta t)\cdot(\mathbf{n}ds)\]

up to higher order differentials, where \(ds\) is the element of arc length. Hence

\[\begin{split}\begin{aligned} &\lim_{\Delta t \to 0} \cfrac{1}{\Delta t}\left \{ \iint_{D(t+\Delta t )}^{} F(x,y)dxdy-\iint_{D(t)}^{} F(x,y)dxdy\right \}\\ =&\lim_{\Delta t \to 0}\cfrac{1}{\Delta t}\int_{\partial D(t)}^{} F(x,y)(\mathbf{v}\Delta t)\cdot(\mathbf{n}ds) \end{aligned}\end{split}\]
\[\cfrac{\mathrm{d} }{\mathrm{d} t}\iint_{D(t)}^{} F(x,y)dxdy = \int_{\partial D(t)}^{} F(x,y){\mathbf{v}}\cdot{\mathbf{n}}ds\]

where \(\partial\) a denotes boundary. Before taking limits, we compute \({\mathbf{v}}\cdot{\mathbf{n}}ds\). We rotate the unit tangent \((dy/ds,-dx/ds)\) backwards through a right angle to obtain \(\mathbf{n}=(dy/ds,-dx/ds)\), hence

\[\mathbf{n}=(dy/ds,-dx/ds)=(u,v)\cdot(dy,-dx)=udy-vdx\]

Therefore

\[\begin{split}\begin{aligned} \cfrac{\mathrm{d} }{\mathrm{d} t}\iint_{D(t)}^{} F(x,y)dxdy =& \int_{\partial D(t)}^{} F(x,y){\mathbf{v}}\cdot{\mathbf{n}}ds \\ =&\int_{\partial D(t)}^{} F(x,y)(u,v)\cdot(dy,-dx) \\ =&\int_{\partial D(t)}^{} F(x,y)(udy-vdx)\\ =&\int_{\partial D(t)}^{} (-F(x,y)vdx+F(x,y)udy) \end{aligned}\end{split}\]

We can transform the boundary integral into an integral over D, by Green’s Theorem.

\[\begin{split}\begin{aligned} P(x,y)&\equiv -F(x,y)v(x,y)\\ Q(x,y)&\equiv F(x,y)u(x,y) \end{aligned}\end{split}\]
\[\begin{split}\begin{aligned} (\cfrac{\partial Q(x,y)}{\partial x}-\cfrac{\partial P(x,y)}{\partial y})= &(\cfrac{\partial [F(x,y)u(x,y)]}{\partial x}-\cfrac{\partial [-F(x,y)v(x,y)]}{\partial y})\\ =&(\cfrac{\partial [F(x,y)u(x,y)]}{\partial x}+\cfrac{\partial [F(x,y)v(x,y)]}{\partial y})\\ =&(\cfrac{\partial [F(x,y)u(x,y)]}{\partial x}+\cfrac{\partial [F(x,y)v(x,y)]}{\partial y})\\ \end{aligned}\end{split}\]
\[\begin{split}\begin{aligned} \cfrac{\partial [F(x,y)u(x,y)]}{\partial x}-\cfrac{\partial [-F(x,y)v(x,y)]}{\partial y}=&\\ \cfrac{\partial [F(x,y)u(x,y)]}{\partial x}+\cfrac{\partial [F(x,y)v(x,y)]}{\partial y}=\mathrm{div} \left \{ F(x,y)\mathbf{v(x,y)} \right \} \\ \cfrac{\partial (F u)}{\partial x}+\cfrac{\partial (F v)}{\partial y}=\mathrm{div} ( F\mathbf{v} ) \end{aligned}\end{split}\]
\[\cfrac{\partial (F u)}{\partial x}+\cfrac{\partial (F v)}{\partial y}=\mathrm{div} ( F\mathbf{v} )= \nabla \cdot( F\mathbf{v} )\]

Here

\[\begin{split}\begin{aligned} \mathrm{div} ( F\mathbf{v} )\equiv\nabla \cdot( F\mathbf{v} )=&\cfrac{\partial (F u)}{\partial x}+\cfrac{\partial (F v)}{\partial y}\\ =&F \nabla \cdot \mathbf{v}+\mathbf{v} \cdot \nabla F \\ =&\nabla F \cdot \mathbf{v} + F \nabla \cdot \mathbf{v} \end{aligned}\end{split}\]
\[\nabla \cdot(\phi \mathbf{v})=\phi \nabla \cdot \mathbf{v}+\mathbf{v} \cdot \nabla \phi\]
\[\nabla \cdot(F \mathbf{v})=F \nabla \cdot \mathbf{v}+\mathbf{v} \cdot \nabla F\]
\[\begin{split}\begin{aligned} &\frac{\mathrm{d} }{\mathrm{d} t}\iint_{D(t)}^{} F(x,y,t)dxdy\\ \end{aligned}\end{split}\]
\[\begin{aligned} \frac{\mathrm{d} }{\mathrm{d} t}\iint_{D(t)}^{} F(x,y,t)dxdy =\int_{\partial D(t)}^{} F(x,y){\mathbf{v}}\cdot{\mathbf{n}}ds+ \iint_{D(t)}^{} \frac{\partial F(x,y,t)}{\partial t}dxdy \end{aligned}\]
\[\begin{split}\begin{aligned} \frac{\mathrm{d} }{\mathrm{d} t}\iint_{D(t)}^{} F(x,y,t)dxdy =&\iint_{D(t)}^{} \nabla \cdot(F \mathbf{v})dxdy+ \iint_{D(t)}^{} \frac{\partial F(x,y,t)}{\partial t}dxdy\\ =&\iint_{D(t)}^{} \left \{\nabla \cdot(F \mathbf{v})+\cfrac{\partial F(x,y,t)}{\partial t}\right \}dxdy \end{aligned}\end{split}\]

A space formula.

\[\begin{split}\begin{aligned} \frac{d}{d t} \iiint_{D_{t}} F(\mathbf{x}, t) d x d y d z & =\iint_{\partial D_{t}} F \mathbf{v} \cdot \mathbf{n} dS+\iiint_{D_{t}} \frac{\partial F}{\partial t} d x d y d z \\ \end{aligned}\end{split}\]
\[\begin{split}\begin{aligned} \frac{d}{d t} \iiint_{D_{t}} F(\mathbf{x}, t) d x d y d z & =\iint_{\partial D_{t}} F \mathbf{v} \cdot d\mathbf{S} +\iiint_{D_{t}} \frac{\partial F}{\partial t} d x d y d z \\ \end{aligned}\end{split}\]
\[\begin{split}\begin{aligned} \frac{d}{d t} \iiint_{D_{t}} F(\mathbf{x}, t) d x d y d z & = \iiint_{D_{t}}\left[\operatorname{div}(F \mathbf{v})+\frac{\partial F}{\partial t}\right] d x d y d z\\ &=\iiint_{D_{t}}\left[\nabla \cdot(F \mathbf{v})+\frac{\partial F}{\partial t}\right] d x d y d z\\ \end{aligned}\end{split}\]

Green’s theorem

  1. Green’s theorem

Let C be a positively oriented, piecewise smooth, simple closed curve in a plane, and let D be the region bounded by C. If L and M are functions of (x, y) defined on an open region containing D and have continuous partial derivatives there, then

\[\oint_{C}^{} (Ldx+Mdy)=\iint_{D}^{} (\cfrac{\partial M}{\partial x}-\cfrac{\partial L}{\partial y})dxdy\]

where the path of integration along C is anticlockwise.

This formula can also be written as

\[\begin{split}\oint_{C}^{} (P(x,y)dx+Q(x,y)dy)=\iint_{D}^{} (\cfrac{\partial Q}{\partial x}-\cfrac{\partial P}{\partial y})dxdy\\\end{split}\]

A Leibniz integral rule for a two dimensional surface moving in three dimensional space is[3][4]

\[{\displaystyle {\frac {d}{dt}}\iint _{\Sigma (t)}\mathbf {F} (\mathbf {r} ,t)\cdot d\mathbf {A} =\iint _{\Sigma (t)}\left(\mathbf {F} _{t}(\mathbf {r} ,t)+\left[\nabla \cdot \mathbf {F} (\mathbf {r} ,t)\right]\mathbf {v} \right)\cdot d\mathbf {A} -\oint _{\partial \Sigma (t)}\left[\mathbf {v} \times \mathbf {F} (\mathbf {r} ,t)\right]\cdot d\mathbf {s} ,}\]