Integrating Dirac Delta Gives Two Different Answers

164 Views Asked by At

I am trying to solve the following differential equation

$$\frac{dx}{dt} = x \delta(t-\tfrac{1}{2})$$

from $t=0$ to $t=1$ where $\delta$ denotes the typical Dirac function. If I separate variables and integrate, I get

$$ \int_{x_0}^{x_t} \frac{1}{x} dx = \int_0^t \delta (\tau-\tfrac{1}{2}) d\tau$$

leading to

$$ \ln (x_t) - \ln(x_0) = \begin{cases} 0, & t \leq 1/2 \\ 1, & t > 1/2 \end{cases} $$

On the other hand, if I rewrite the original equation as

$$ \int_{x_0}^{x_t} dx = \int_0^t x(\tau) \delta (\tau-\tfrac{1}{2}) d \tau$$

and simply apply the definition of the Dirac function, I get

$$ x_t - x_0 = \begin{cases} 0, & t \leq 1/2 \\ x\left(\tfrac{1}{2}\right), &t > 1/2 \end{cases} $$ Why do I get two different answers? Is either of these right or is the original equation ill-defined in some sense ... perhaps the Radon-Nikodym derivative does not exist since the LHS is wrt to Lebesgue and the RHS is wrt Dirac? Thanks in advance!

1

There are 1 best solutions below

0
On BEST ANSWER

The second way is not very well defined, but there's not really a contradiction between them. The only problem is that $x$ is not really the same for $t < 1/2$ as it is for $t = 1/2$, it's tricky to define $x$ at this point, since there's a discontinuity in the function here.

If you look at the first equation, you have

$$ \ln (x_t) - \ln(x_0) = \begin{cases} 0, & t < 1/2 \\ 1, & t > 1/2 \end{cases} $$

Using logarithm rules to rewrite it,

$$ \ln (x_t/x_0) = \begin{cases} 0, & t < 1/2 \\ 1, & t > 1/2 \end{cases} $$ $$ x_t/x_0 = \begin{cases} e^0, & t < 1/2 \\ e^1, & t > 1/2 \end{cases} $$ $$ x_t = \begin{cases} x_0, & t < 1/2 \\ x_0 e, & t > 1/2 \end{cases} $$

And the second equation is

$$ x_t = \begin{cases} x_0, & t < 1/2 \\ x_0 + x(1/2), & t > 1/2 \end{cases} $$

So the definitions are the same for $t < 1/2$ and if you pretend $x(1/2) = x_0(e-1)$, then they are the same for $t > 1/2$ as well.