I'm stuck with the following problem I found (without a proof) in the Peskin and Schroeder textbook on quantum field theory (the differential equation mentioned below is equivalent to the Callan-Symanzik equation describing the evolution of the n-point correlation functions under variation of the energy scale at which the theory is defined):

Show that the solution of the equation $$ [\partial_t + v(x)\partial_x - \rho(x)] D(t,x) = 0 $$ has the form $$ D(t,x) = D(0,X_t(x)) \exp\left(\int_0^t d t' \rho(X_{t'}(x)) \right), $$ where $X_t(x)$ is the solution of the equation $$ \partial_t X_t(x) = - v(X_t(x)) $$ with the initial condition $$ X_0(x) = x. $$

Let's compute e.g.

$$ \partial_t \left[D(0,X_t(x)) \exp\left(\int_0^t d t' \rho(X_{t'}(x)) \right) \right]= \left[\partial_t D(0,X_t(x))\right] \exp\left(\int_0^t d t' \rho(X_{t'}(x)) \right) + D(0,X_t(x)) \rho(X_t(x)) \exp\left(\int_0^t d t' \rho(X_{t'}(x)) \right) $$

I don't see any possible simplifications with other terms.

I wonder how to show that the solution of the mentioned differential equation is indeed of the above form.

  • $\begingroup$ What is the question? $\endgroup$ – Jacky Chong Oct 4 '16 at 19:36
  • $\begingroup$ @JackyChong: I don't know how to show that the solution of the pde is really of the form written in the post. $\endgroup$ – user72829 Oct 4 '16 at 19:40

This is a simple transport type equation. We shall solve the PDE using the methods of characteristics. Suppose we could rewrite the PDE as follows \begin{align} \partial_t D + \nu(x) \partial_x D- \rho(x)D =&\ \partial_t D+ X'(t)\partial_x D -\rho(x)D \\ =&\ \frac{d}{dt}D(t, X(t)) -\rho(x)D = 0 \end{align} where $X'(t) = \nu(X)$. Solving for the characteristic $X$ leads to the solution \begin{align} \frac{X'}{\nu(X)} = 1 \ \ \Rightarrow \ \ \int \frac{dX}{\nu(X)} = t+ C \end{align} which really depends on $\nu(X)$ (nevertheless it exists). Hence let us assume that we found $X(t)$. Back to the PDE, we have \begin{align} \frac{d}{dt}D(t, X(t)) = \rho(X(t)) D(t, X(t)) \ \ \Rightarrow& \ \ \frac{d}{dt}\log|D(t, X(t))| =\rho(X(t))\\ \Rightarrow&\ \ \ \log|D(t, X(t))| -\log|D(0, X(0))| = \int^t_0 dt' \rho(X(t'))\\ \Rightarrow&\ \ D(t, X(t)) = D(0, X(0))\exp\left( \int^t_0 dt' \rho(X(t'))\right). \end{align} We are almost done, but not done. If we impose the initial condition that $X(0) = x$, then we have a unique curve that pasts through the point $(0, x)$ which will dictate the values of $D(t, x)$ when $(t, X)$ lies on the trace of $(t, X(t))$. Hence, we shall use the notation $X= X_t(x)$ to indicate the curve. Thus, your solution becomes \begin{align} D(t, X) = D(0,x)\exp\left( \int^t_0 dt' \rho(X_t(x)) \right). \end{align}


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.