Dynamic Relaxation

So far we have treated only the behavior of the muon polarization at fixed locations where the local magnetic fields are static. This is often not the case, and situations in which the muons are hopping between different sites and/or the local fields at those sites are fluctuating comprise many of the most interesting µSR studies.

Therefore it behooves us to model the resultant relaxation functions.

The simplest cases are when either the muon hops between sites with different static magnetic fields and/or the muon is at a fixed site with a randomly fluctuating local magnetic field. In either case we usually assume that the field at each hop or after each fluctuation is selected randomly from an easily modeled distribution <math>{\cal D}(\vec{B})</math> -- e.g. a Gaussian distribution of field magnitudes about an average <math>B_0</math> with a width <math>\Delta</math> :

<math>{\cal D}(|B|) = {1 \over \sqrt{2\pi} \; \Delta} \exp\left[- {1\over2} \left(B-B_0 \over \Delta \right)^2\right] </math>,

with a direction randomly selected from an isotropic distribution. (In ZF, <math>B_0 = 0</math> is usually assumed.)

It is easy to imagine many plausible situations in which these assumptions are violated; but one has to start somewhere. The trick is not to get stuck there.

Next we assume that the muon hops and/or the field fluctuates at a rate <math>\nu</math> in a history-independent way.

One approach, at least in ZF, is to assume that the static relaxation function <math>G_{zz}^{\rm stat}(t) = g(\Delta, t)</math> simply restarts at every hop/fluctuation, using the muon polarization at the time of the hop as a new starting point. This is called the strong collision model and can be modeled using Kehr's recursion relation:

<math>G^{\rm dyn}(\Delta,\nu,t) = g(\Delta,t) \; e^{-\nu t} + \nu \int_0^t G^{\rm dyn}(\Delta,\nu, t-\tau) \; g(\Delta, \tau) \; e^{-\nu \tau} d\tau </math> ,

which is sometimes solvable using Laplace transforms. (Numerical methods work too.)

In the limit of <math>\nu \to 0</math>, the Gaussian distribution above gives a static function

<math>g_G(\Delta, t) = {1\over3} + {2\over3} e^{-{1\over2}\Delta^2 t^2}} \left(1 - \Delta^2 t^2\right)</math>