Solution to Integral Equation

140 Views Asked by At

Which of the following is a solution of $u(x) = x + \int_{0}^{x} (t-x)u(t)dt$??

(A) $\sin x$

(B) $x \cos x$

(C) $\ln (x+1)$

(D) $x e^{-x}$

(E) $xe^x$

Since all of the choices are twice differentiable, the first thing that came to mind my was to differentiate and then use Leibniz' rule; then differentiate again and then use the FTOC. Doing this gave me $\frac{u''(x)}{u(x)} = x-1$, but I don't remember how to solve differential equations. Does anyone know the solution to this?

Also, what is the more orthodox way of solving this? Keep in mind that this is a practice problem for the GRE, so it has to be done within two minutes, which is rather absurd in my estimation.

I also realized, after having already typed this up, that none of the choices appear to satisfy the differential equation. What is wrong with my method? I probably made a stupid calculation error, which I fear I may do quite a lot on the GRE (I hate these sorts of tests...)

EDIT: Here is my calculation. First, the partial with respect to $x$ of the integrand is $\frac{\partial}{\partial x} tu(t) - \frac{\partial}{\partial x} x u(t) = - u(t)$. Taking the first derivative, we get

$$u'(x) = 1 + (x-x)u(x) + \int_{0}^{x} (-u(t))dt= 1 - \int_{0}^{x} u(t)dt$$

Hence the second derivative is

$$u''(x) = -u(x)$$.

2

There are 2 best solutions below

2
On BEST ANSWER

Quicker kill, with initial condition $u(0) = 0$: the Laplace transform.

Transforming both sides yields

$$ U(s) = \frac{1}{s^2} - \frac{U(s)}{s^2} $$

(I rewrote the integral as $-\int_0 ^x u(t)(x-t) \ dt $ to properly use the convolution formula.)

Now, rearrange to get

$$ U(s) \left( 1 + \frac{1}{s^2} \right) = \frac{1}{s^2} $$

and some division gives

$$ U(s) = \frac{1}{s^2 + 1} $$

giving the desired solution $u(x) = \sin x$.

3
On

If you had applied Cauchy's repeated integral formula, you could've differentiated the integral more easily, giving

$$u''(x)=-u(x)$$

Since

$$\int_0^x(x-t)u(t)~\mathrm dt=\int_0^x\int_0^\sigma u(t)~\mathrm dt~\mathrm d\sigma$$

And it is obvious then that A is the correct answer.