I came across this integro-differential equation to solve $$\frac{du(x;t)}{dt}=-\lambda\int_0^xu(\xi;t)\;d\xi\tag{1}$$ under the initial condition $u(x;0)=f(x)$ where $x$ is a parameter, $\lambda$ is a constant, and $0<t<\infty$.

My first thought is that I can just directly integrate the equation to obtain $$u(x;t)=-\lambda\int_0^t\int_0^xu(\xi;\tau)\;d\xi\,d\tau\;.\tag{2}$$ This equation is highly implicit and an explicit expression is desired. So then I thought of using Laplace transforms instead. Let $U(\cdot)$ be the Laplace transform of $u(\cdot)$, and $s$ be the complex co-variable of the real variable $t$, then $$s\,U(x;s)-f(x)=-\lambda\,\int_0^xU(\xi;s)\,d\xi\tag{3}$$ which can be converted to the differential equation $$s\,U'(x;s)+\lambda\,U(x;s)=f'(x)\tag{4}$$ where the derivative is now with respect to $x$. Equation $(4)$ is easily solvable. Though I am uncertain if I did the following Laplace transform correctly, $$\mathscr{L}\left\{\int_0^xu(\xi;t)\,d\xi\right\}=\int_0^xU(\xi;s)\,d\xi\;.\tag{5}$$ I figured since the integration is over the parameter instead of the transforming variable I could bring it into the integral under the heuristic that the Laplace transform of a sum is the sum of the Laplace transforms, but I am unsure of this.

Any insight on any of this or alternative methods to solve Equation $(1)$ is welcome.


The solution detailed below is :

With $\quad F(s)=$ Laplace transform of $f(x)$.

$$\Phi(s,t)=e^{-\frac{\lambda \:t}{s}}F(s)$$ $$\boxed{u(x,t)= \text{ Inverse Laplace Transform of } \Phi(s,t)}$$

The result cannot be expressed more explicitly until the function $f(x)$ be explicitly given.

enter image description here


Incomplete attempt at a solution but thought I would offer it anyways as food for thought.

Differentiate both sides of the integral equation with respect to $x$.

$$ \frac{\partial ^2 u}{\partial x\partial t}=-\lambda u(x,t) $$

$$ u_{\alpha}(x,t)=A(\alpha)\cos(\alpha x+\frac{\lambda}{\alpha} t)+B(\alpha)\sin(\alpha x+\frac{\lambda}{\alpha}t),\;\;\;\alpha>0 $$

Now take advantage of the linearity of the equation, usually we sum but here we have a continuous parameter $\alpha$.

$$ u(x,t)=\int_{0}^{\infty}u_{\alpha}(x,t)\mathbb{d}\alpha $$

$$ f(x)=u(x,0)=\int_{0}^{\infty}u_{\alpha}(x,0)\mathbb{d}\alpha $$

$$ f(x)=\int_{0}^{\infty}A(\alpha)\cos(\alpha x)\mathbb{d}\alpha+\int_{0}^{\infty}B(\alpha)\sin(\alpha x)\mathbb{d}\alpha $$

I would like to now solve for $A(\alpha),B(\alpha)$ in terms of $f(x)$. Perhaps if one converted the above trigonometric expressions to complex exponentials a certain Laplace/Fourier transform structure would emerge that would allow one to use the established formalism to invert (solve for $A,B$ in terms of $f$).

  • $\begingroup$ This is promising. There is an additional boundary condition, however, arising from putting $x=0$ in the original equation: $$\frac{\partial u}{\partial t}(0,t)=0.$$ This yields a type of Goursat problem. $\endgroup$ – Bob Pego Jul 4 '17 at 9:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.