Computational Physics

(Rick Simeone) #1
7.3 Phase transitions 191
in terms of additional exponents, the so-calledfinite-size scaling exponents:


  • The shift in the position of the maximum with respect to the critical temperature
    is described by
    Tc(L)−Tc(∞)∝L−λ. (7.61)

  • The width of the peak scales as


T(L)∝L−. (7.62)


  • The peak height grows with the system size as


Amax(L)∝Lσm. (7.63)
The behaviour of a system is determined by two length scales:L/aandξ/a, withξ
the correlation length of the infinite system, which in the finite-size scaling region
is larger than the linear system sizeL. As in the critical region, the fluctuations
determining the behaviour of the system extend over large length scales; phys-
ical properties should be independent ofa. This leavesL/ξas the only possible
parameter in the system and this leads to the so-called finite-size scalingAnsatz.
Defining
ε≡

T−Tc
Tc

, (7.64)


we can formulate the finite-size scalingAnsatzas follows:
AL(ε)
A∞(ε)

=f

[


L


ξ∞(ε)

]


. (7.65)


Suppose the exponent of the critical divergence of the quantityAisσ:
A∞∝ε−σ. (7.66)
Using, moreover, the scaling form of the correlation lengthξ∝ε−ν, we can write
the scalingAnsatzas
AL(ε)=ε−σf(Lεν) (7.67)
which can be reformulated as
AL(ε)=Lσ/νφ(L^1 /νε) (7.68)
where we have replaced the scaling function,f, by another one,φ, by extracting a
factor(Lεν)σ/νfromfand then writing the remaining function in terms of(Lεν)^1 /ν
rather than(Lεν). Obviously,φ(x)will have a maximumφmax for some value
x=xmaxwith a peak widthx. From Eq. (7.68) we then see immediately that:


  • The peak height scales asLσ/ν, henceσm=σ/ν.

  • The peak position scales asL−^1 /ν, henceλ= 1 /ν.

  • The peak width also scales asL−^1 /ν, hence= 1 /ν.

Free download pdf