Physics 715: The hyperscaling hypothesis.

The divergence of the correlation length \xi signals the formation of a special global state, a state of scale invariance. This is very much unlike the global states of thermodynamics which are homogeneous and non-local. The scale-invariant states of the transitioning ferromagnet are very much non-homogeneous with dramatic local fluctuations. It is a “new” kind of global structure, the kind seen in self-similar constructions in fractal geometry and chaotic dynamics.

Right at the critical point, the correlation functions and other “n-point” functions take on a very special form. We saw the derivation of the Orenstein-Zernicke formula

    \[\langle s(r) \, s(0)\rangle\approx {e^{-r/\xi}\over r^{{d-1\over 2}}}, \qquad \xi=(m\sqrt{d})^{-1}\sim t^{-\nu}\]

    \[ \nu={1\over 2}\quad\mbox{since}\quad m^2\sim t\]

right at the critical point

    \[\langle s(r) \, s(0)\rangle\approx r^{{1-d\over 2}}\]

which has the property that

    \[\langle s(\lambda r) \, s(0)\rangle=\lambda ^{{1-d\over 2}}\langle s(r) \, s(0)\rangle\]

which is the classic definition of a scale-invariant function (which implies full conformal invariance of the field s(r), which can only happen for massless fields); a function f(x) is scale invariant if there is some \Delta (the scaling dimension) such that

    \[f(x)=\lambda^{-\Delta} f(\lambda x)\]

Not every function has this property. Pure monomials f(x)=x^a do, polynomials do not, exponentials do not. The scaling hypothesis is that the (most) divergent or dominant part of a function has this feature near (at) the critical point.

The hyperscaling hypothesis attributes the divergence of various thermodynamic functions at the critical point to the divergence of \xi. This “more fundamental” hypothesis stipulates that
\bullet The correlation length has such a form

    \[\xi(t,h)=t^{-\nu} f_\xi\Big({h\over t^\Delta}\Big)\]

\bullet Close to critical, \xi is the only important length scale and its divergence is entirely responsible for all other quantities becoming singular, if they do so.

Rendered by QuickLaTeX.com

In d-dimensions, how does the free energy F or J scale with the system length \ell (it is in a hypercubic box)? It is extensive so, F\sim PV \sim \ell^d and V\sim \ell^d. Now divide the system into blocks each of which is the size of the correlation length in every dimension, each contributing a constant factor to the free energy at criticality to construct from \xi(h,t) above a scaling function for the free energy

    \[F(t,h)\sim C \, (\ell/\xi)^d=t^{d\nu} f^{-d}_\xi\Big({h\over t^\Delta}\Big)=t^{d\nu} f_f\Big({h\over t^\Delta}\Big)\]

    \[ 2-\alpha=d\nu, \quad \mbox{(Josephson's identity)}\]

and so the hyperscaling hypothesis has as a consequence the scaling hypothesis. Furthermore we can eliminate t in favor of \xi and deduce that since

    \[E\sim t^{1-\alpha}, \qquad E\sim t^{1-\alpha}\sim \xi^{-({1-\alpha\over \nu})}\sim {1\over \xi^{{d\nu-1\over \nu}}}={1\over \xi^{d-{1\over \nu}}}\]

which is a hint that the “fields” become scale-invariant functions of inter-spin distances at the critical point, a prelude to conformal field theory and the conformal bootstrap hypothesis.

Home 2.0
error: Content is protected !!