Jacobi Corollary and solving an ODE

75 Views Asked by At

I'm working through Jacobi's formula, and I'm really stuck with solving an ODE and would appreciate someone helping me out, as some of my basic maths is a bit rusty.

Essentially proving this corollary. Taking

$$\frac{d}{dt} \det e^{tB} =\operatorname{tr}(B) \det e^{tB}$$

to the solution

$$\det e^{tB} = e^{\operatorname{tr} \left(tB\right)}$$

I assume this boils down to a simple first order ODE

$$\dfrac {d}{dt}\left( y\left( t\right) \right) = q\left( t\right) y\left( t\right)$$

but can't seem to make it work.

1

There are 1 best solutions below

3
On BEST ANSWER

Let $f: \Bbb{R} \to \Bbb{R}$ be the function defined by $f(t) = \det e^{tB}$. Suppose you already manage to establish that $f$ satisfies the ODE \begin{align} f'(t) &= \text{tr}(B)\cdot f(t) \end{align} Solving this is pretty easy; in fact this is the protoypical example of a separable ODE. The solution to this ODE is just \begin{align} f(t) &= f(0) \cdot e^{\text{tr}(B) \cdot t} \\ &= \det(I) \cdot e^{\text{tr}(tB)} \\ &= e^{\text{tr}(tB)} \end{align}


And just to answer your other question, the solution to \begin{align} y'(t) = q(t)\cdot y(t) \end{align} subject to the initial conditions $y(t_0) = y_0$ is given by \begin{align} y(t) = y_0\exp\left( \int_{t_0}^t q(s) \, ds \right) \end{align}

The "usual" way of justifying this is to divide both sides by $y(t)$ and then integrate: \begin{align} \dfrac{y'(t)}{y(t)} &= q(t) \\\\ \int \dfrac{y'(t)}{y(t)} \, dt &= \int q(t) \, dt \\ \ln \big( y(t) \big) &= \int q(t) \, dt \end{align} Now exponentiate both sides. I've purposely been sloppy with this method regarding constants of integration, because there is already a problem with this approach: it implicitly assumes $y(t)$ is no-where vanishing (because otherwise we cannot divide by $y(t)$). But hopefully you see where the solution "comes from". A more complete way which I like for solving this ODE is to just recognise that the formula for $y(t)$ I gave above is actually a solution, and then use a uniqueness argument to conclude there aren't anymore.