I've encountered an integro-differential equation of the following form: $$ \frac{dx(t)}{dt} = \int_0^t ds\ f_{1}(s) - \int_0^t ds\ f_{2}(s) x(t - s) $$

The functions $f_{1}(t)$ and $f_{2}(t)$ are known explicitly, and the goal is to solve for $x(t)$ given some initial condition $x(0)$.

My question is, what are some approaches to getting information about possible solutions $x(t)$? I've had two ideas:

(a) Fixed points for this equation.

Might there exist a constant solution $x(t) = x^{\star}$? In this case, we'd set the LHS of the above equation to $0$ and solve for a constant $x^{\star}$ which gives: $$ x^{\star} = \frac{\int_{0}^t ds\ f_{1}(s)}{\int_0^t ds\ f_{2}(s)} $$

But this must be wrong since there is time dependence explicitly in this. I wanted to ask is there a method for finding a fixed point solution $x^{\star}$? (maybe the implication of the above is that there aren't any fixed points?)

(b) Getting a second order DE and solving that.

If you differentiate both sides of the integro-differential equation you get: $$ \frac{d^2x(t)}{dt^2} = f_{1}(t) - f_{2}(t) x(0) $$ which looks a lot simpler then the initial equation. This looks like a simple second-order DE, with the only unusual feature being that the DE explicitly depends on the initial condition $x(0)$. I am confused here also because naively integrating the above equation gives a solution like: $$ \frac{dx(t)}{dt} = \int_0^t ds\ f_{1}(s) - x(0) \int_0^t ds\ f_{2}(s) + c $$ Where the constant $c$ is constrained by the original integro-differential equation to be $c=0$. This is weird to me, because my naive integration has a factor $x(0)$ outside the integral, rather than a $x(t-s)$ underneath the integral as in the original DE. What am I getting wrong in this naive integration of $\ddot{x}(t)$?


Your equation can be written as $$ \dot x=F_1(t)+(f_2*x)(t) $$ where the last term is the convolution product. Without knowing more about $f_2$ you can say nothing more about the equation. Its derivative is $$ \ddot x=f_1(t)+(f_2*\dot x)(t)=f_1(t)+(\dot f_2*x)(t) $$ and so on, so that if there is a linear DE with constant coefficients for $f_2$ the last term can be annihilated in some linear product resulting in a linear DE for $x$.

Fixed points, that is constant solutions, have to satisfy $\dot x=0$, which would require $$ 0=F_1(t)+\int_0^tf_2(s)x^*\,ds=F_1(t)+x^*F_2(t), $$ that is, $F_1$ has to be a constant multiple of $F_2$, and thus the same has to hold for their derivatives, $0=f_1(t)+x^*f_2(t)$.


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.