Evaluación de la calidad del mantenimiento en la empresa

To better understand the link between self-similarity and fractality, let us consider a system occupying a finite region of spaceΣ ⊂ RD with a local density ρ(r)_.

Following Pietronero (1990), we define the pair correlation function,

д(r)=hρ(r +r0)ρ(r0)ir0, (4.1)

which gives the probability that two points separated byr both belong to the regionΣ. For simplicity, we now assume that the correlation function is isotropic, д(r) = д(r). In the absence of length scales,д obeyshomogeneity laws (orscale- invariance) with respect to aresolutionorcoarse-grainingλ. More specifically, if we

4.2. MULTIFRACTALS 53

rescale lengths asr →r0=λr we have that

д(r0)=λκд(r), (4.2)

whereκ is a homogeneity exponent. The solution to this equation is given by a power-law behaviour,д(r) ∝rκ. If we then fixr as the reference length scale, and д(r)=1,

д(λ)=λκ (4.3)

translates self-similarity into a mathematical relation.

We can finally define fractal objects as self-similar structures whose observed spatial extent (e.g. volume) depends, with a power-law behaviour, on the resolu- tion at which we look at it. For fractals originating from a mathematical relation, the dependence on the resolution can extend over an infinite range. For fractals appearing in physical systems, instead, the range ofλis usually limited by macro- and/or microscopic scales. A very comprehensive list of examples can be found in Malcai et al. (1997).

4.2.1 Measures, fractals and multifractals

Letψ(r)be the wave function of an electron in aL×L×Lvolume. The modulus square|ψ(r)|2defines a normalised measure on this volume, which we partition in boxes of linear sizel =λL. The number of boxes will then beλ−d, whered=3 is the Euclidean dimension of the support of the system. The probability of finding the electron in boxi, then, is thebox-probability

µi =

∫

boxi

|ψ(r)|2_dr. _(4.4)

We can then define thefractal dimensionDof the system by counting the number of boxes where the box-probability does not vanish:¹ N(λ) ∼ λ−D. Because the electron can access any portion of the volume, i.e. there is no region of space with vanishing probability, we conclude that the fractal dimensionD =d =3, which is not very interesting.

1_{It is customary to use∼to indicate that the proportionality constant is independent of the}

resolution and can thus be ignored. This constant might appear, for instance, when the boxes, whichever their shape, do not perfectly cover the system. Since we are covering boxes with boxes, most of the relations in this section are actually equalities.

Compared to the fractal dimension, more insightful is actually the study of the powers of the box-probabilityµ_iq, which is the idea behindmultifractalanalysis.If the wave function is a multifractal, we expect to see the power-law behaviour of (4.3):

hµqi_L ∼λD+τq_{, (4.5)}

whereh. . .i_{Ldenotes the average over all boxes in the volume. Equivalently, we}

can introduce thepartition sumRq(λ)(also thegeneralised inverse participation ratio) ashµqiL=λDRq(λ)and write

Rq(λ)=

Õ

i

µ_iq ∼λτq_{. (4.6)}

Themass exponentsτq describe the scaling behaviour of the moments and do not depend onλ.

Let us stress again that multifractality holds if, in the power-law relation of Eq. (4.5),τq , 0 for a finite range ofλ: the box sizel should be smaller than the

system size, but also larger than the microscopic scalea. At the same time, for critical states at the Anderson transition, the system size is much smaller than the correlation lengthξ, such that

al <L ξ (4.7)

Additionally, the wave function is a truly critical (and hence multifractal) only in the thermodynamic limit, where bothL andξ diverge, henceτq is uniquely defined in the limit λ → 0. For finite systems, instead, we choose states and coarse-grainings that satisfy (4.7). In this case, we canestimateτq by fitting the slope of logRq(λ)versus logλ. We are assuming here that multifractality survives in finite systems (Cuevas and Kravtsov, 2007), and postpone the discussion of this non-trivial assumption to Ch. 5.

From (4.5) and the normalisation of the wave function, it is possible to show thatτ0 = −D andτ1 = 0. This implies that we can generalise the definition of

the fractal dimension to a functionDq such thatD0 =D andτq = Dq(q−1). In the case of a simple fractalDq ≡ D, while for a multifractalDq has a non-trivial dependence onq. The deviation from the simple-fractal case is captured by the anomalous scaling exponent∆q =(Dq−d)(q−1)=τq −d(q−1).

4.2. MULTIFRACTALS 55

4.2.2 The multifractal spectrum

The scaling of the momentsRq, yieldingτq, is enough to fully characterise the multifractal nature of the wave function. Now we present an equivalent descrip- tion of the multifractal that will be useful, in the sections that follow, to validate our results and compare them to the 3D Anderson model. This description is founded on amultifractal measure(Frisch and Parisi, 1985), a distribution such that, around each boxi, µi = λαi. The set of boxes withαi ∈ [α,α +dα], then, con- stitutes a simple fractal with dimensionf(α), such that the number of said boxes is

Nλ(α) ∼λ−f(α) and αi =

logµi

logλ . (4.8)

This is the formalisation of the idea of Castellani and Peliti (1986), where the multifractal is composed of different simple fractals.

We re-express the partition sum of Eq. (4.6) as Rq(λ)= Õ i µq_i =Õ i λqαi = ∫ N(α)λqαdα ∼ ∫ λqα−f(α)dα. (4.9)

For smallλ, we can use the saddle point approximation and find that the biggest contribution in the integral (4.9) comes from the value ofα that maximises (since λ <1) the argument of the exponential, i.e. theαq such that f0(αq)=q. We can then write, from (4.6),τq =qαq−f(αq). If we identifyfq = f(αq)we can see that

(q,τq)and(αq,fq)are related by a Legendre transformation:

fq =q αq−τq and αq =

dτq

dq . (4.10)

It can be proven, e.g. in Janssen (1994), thatτq is a monotonically increasing function inq, which implies thatαq >0,∀q.

We can combinesingularity strengthsαq and thesingularity spectrumfq to ob- tain themultifractal spectrum f(α). This function is equivalent to the generalised dimensionsDq in characterising the multifractal, and in the case of a simple frac- tal analogously reduces to the point(D,D) _{in a}(α,f(α))_{plot. As shown in the}

example of Fig. 4.1, f(α)is a convex function reaching its maximum atα0with a

value f0=τ0 =D. From (4.10) we further notice that f1 =α1, sinceτ1 =0. The

f1 f0 fq 0.5 1.0 1.5 2.0 2.5 3.0 3.5 αq 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

Figure 4.1: Multifractal spectrum f(α) _{for the ONETEP prototype described in}

Ch. 2, computed forqfrom−2 to 5 in steps of 0.1 (increasing from right to left). Dashed lines indicate the functionsf0 ≡D andf1(α)=α.

4.2.3 Symmetry of the multifractal spectrum

Using the nonlinearσ model, Mirlin et al. (2006) have analytically proven that at criticality the multifractal exponents (4.10) satisfy the exact symmetry relation

αq +α1−q =2d f1−q = fq+d−αq . (4.11)

Assuming the universality of the critical properties at the Anderson transition, this result is expected to generally hold for the Wigner-Dyson symmetry classes (see Ch. 1). Indeed, this result was confirmed numerically for different systems, including the 3D Anderson model (Rodriguez et al., 2008; Vasquez et al., 2008) and experiments (Faez et al., 2009)