11.07.2015 Views

Feynman Path Integral Formulation

Feynman Path Integral Formulation

Feynman Path Integral Formulation

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

292 8 Numerical Studiesthe scaling properties is the renormalization group behavior of the free energy F =− 1 V logZ F(t,{u j })=F reg (t,{u j })+b −d F sing (b y tt,{b y ju j }) , (8.64)where F sing is the singular, non-analytic part of the free energy, and F reg is the regularpart. b is the block size in the RG transformation, while y t and y j ( j ≥ 2) arethe relevant eigenvalues of the RG transformation, and t the reduced temperaturevariable that gives the distance from criticality. One denotes here by y t > 0therelevanteigenvalue, while the remaining eigenvalues y j ≤ 0 are associated with eithermarginal or irrelevant operators. Usually the leading critical exponent yt−1 is calledν, while the next subleading exponent y 2 is denoted −ω. For more details on themethod we have to refer to the comprehensive review in (Cardy, 1988).The correlation length ξ determines the asymptotic decay of correlations, in thesense that one expects for example for the two-point function at large distances< O(x)O(y) > ∼|x−y|≫ξ e−|x−y|/ξ . (8.65)The scaling equation for the correlation length itselfξ (t) =b ξ (b y tt) , (8.66)implies for b = t −1/y tthat ξ ∼ t −ν with a correlation length exponentν = 1/y t . (8.67)Derivatives of the free energy F with respect to t then determine, after setting thescale factor b = t −1/y t, the scaling properties of physical observables, including correctionsto scaling. Thus for example, the second derivative of the free energy withrespect to t yields the specific heat exponent α = 2 − d/y t = 2 − dν,∂ 2∂t 2 F(t,{u j}) ∼ t −(2−dν) . (8.68)In the gravitational case one identifies the scaling field t with k c − k, where k =1/16πG involves the bare Newton’s constant. The appearance of singularities inphysical averages, obtained from appropriate derivatives of F, is rooted in the factthat close to the critical point at t = 0 the correlation length diverges.The above results can be extended to the case of a finite lattice of volume V andlinear dimension L = V 1/d . The volume-dependent free energy is then written asF(t,{u j },L −1 )=F reg (t,{u j })+b −d F sing (b y tt,{b y ju j },b/L) . (8.69)For b = L (a lattice consisting of only one point) one obtains the Finite Size Scaling(FSS) form of the free energy [see for example (Brezin and Zinn-Justin, 1985) for a

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!