Chapter 13 - Davidson Physics

Chapter 13 - Davidson Physics Chapter 13 - Davidson Physics

phy.davidson.edu
from phy.davidson.edu More from this publisher
12.07.2015 Views

CHAPTER 14. FRACTALS AND KINETIC GROWTH MODELS 546(ii.) Use the relaxation method (see Section 11.5) to compute the values of the potential ϕ ifor (empty) sites within the circle.(iii.) Assign a random number r to each empty site within the boundary circle. The randomnumber r i at site i represents a breakdown coefficient and the random inhomogeneousnature of the insulator.(iv.) The growth sites are the nearest neighbor sites of the discharge pattern (the occupiedsites). Form the product r i ϕ a i for each growth site i, where a is an adjustable parameter.Because the potential for the discharge pattern is zero, ϕ i for growth site i can beinterpreted as the magnitude of the potential gradient at site i.(v.) The perimeter site with the maximum value of the product rϕ a breaks down, that is, setϕ for this site equal to zero.(vi.) Use the relaxation method to recompute the values of the potential at the remainingunoccupied sites and repeat steps (iv) and (v).Choose a = 1/4 and analyze the structure of the discharge pattern. Does the pattern appearqualitatively similar to lightning? Does the pattern appear to have a fractal geometry? Estimatethe fractal dimension by counting M(b), the average number of sites belonging to the dischargepattern that are within a b × b box. Consider other values of a, for example, a = 1/6 anda = 1/3, and show that the patterns have a fractal structure with a tunable fractal dimensionthat depends on the parameter a. Published results (Family et al.) are for patterns with 800occupied sites.c. Another version of the dielectric breakdown model associates a growth probability p i = ϕ a i / ∑ j ϕa jwith each growth site i, where the sum is over all the growth sites. One of the growth sites isoccupied with probability p i . That is, choose a growth site at random and generate a randomnumber r between 0 and 1. If r ≤ p i , the growth site i is occupied. As before, the exponent ais a free parameter. Convince yourself that a = 1 corresponds to diffusion-limited aggregation.(The boundary condition used in the latter corresponds to a zero potential at the growth sites.)To what type of cluster does a = 0 correspond? Consider a = 1/2, 1, and 2 and explore thedependence of the visual appearance of the clusters on a. Estimate the fractal dimension of theclusters.d. Consider a deterministic growth model for which all growth sites are tested for occupancy ateach growth step. Adopt the same geometry and boundary conditions as in part (b) and usethe relaxation method to solve Laplace’s equation for ϕ i . Then find the perimeter site with thelargest value of ϕ and set ϕ max equal to this value. Only those perimeter sites for which theratio ϕ i /ϕ max is larger than a parameter p become part of the cluster; ϕ i is set equal to unityfor these sites. After each growth step, the new growth sites are determined and the relaxationmethod is used to recompute the values of ϕ i at each unoccupied site. Choose p = 0.35 anddetermine the nature of the regular fractal pattern. What is the fractal dimension? Considerother values of p and determine the corresponding fractal dimension. These patterns have beentermed Laplace fractal carpets (see Family et al.).Surface growth models. The fractal objects we have discussed so far are self-similar, thatis, if we look at a small piece of the object and magnify it isotropically to the size of the original,the original and the magnified object look similar (on the average). In the following, we introduce

CHAPTER 14. FRACTALS AND KINETIC GROWTH MODELS 547some simple models that generate a class of fractals that are self-similar only for scale changes incertain directions.Suppose that we have a flat surface at time t = 0. How does the surface grow as a result ofvapor deposition and sedimentation? For example, consider a surface that is initially a line of Loccupied sites. Growth is in the vertical direction only (see Figure 14.12).As before, we simply choose a growth site at random and occupy it (the Eden model again).The average height of the surface is given byh = 1 ∑N sh i , (14.11)N swhere h i is the distance of the ith surface site from the substrate, and the sum is over all surfacesites N s . (The precise definition of a surface site is discussed in Problem 14.11.)Each time a particle is deposited, the time t is increased by unity. Our main interest is howthe width of the surface changes with t. We define the width of the surface byi=1w 2 = 1 ∑N s(h i − h) 2 . (14.12)N si=1In general, the width w, which is a measure of the surface roughness, depends on L and t. Forshort times we expect thatw(L, t) ∼ t β . (14.13)The exponent β describes the growth of the correlations with time along the vertical direction.Figure 14.12 illustrates the evolution of the surface generated according to the Eden model.After a characteristic time, the length over which the fluctuations are correlated becomes comparableto L, and the width reaches a steady state value that depends only on L. We writewhere α is known as the roughness exponent.w(L, t ≫ 1) ∼ L α , (14.14)From (14.14) we see that in the steady state, the width of the surface in the direction perpendicularto the substrate grows as L α . This scaling behavior of the width is characteristic ofa self-affine fractal. Such a fractal is invariant (on the average) under anisotropic scale changes,that is, different scaling relations exist along different directions. For example, if we rescale thesurface by a factor b in the horizontal direction, then the surface must be rescaled by a factor ofb α in the direction perpendicular to the surface to preserve the similarity along the original andrescaled surfaces.Note that on short length scales, that is, lengths shorter than the width of the interface, thesurface is rough and its roughness can be characterized by the exponent α. (Imagine an ant walkingon the surface.) For length scales much larger than the width of the surface, the surface appears tobe flat and, in our example, it is a one-dimensional object. The properties of the surface generatedby several growth models are explored in Problem 14.11.Problem 14.11. Growing surfaces

CHAPTER 14. FRACTALS AND KINETIC GROWTH MODELS 546(ii.) Use the relaxation method (see Section 11.5) to compute the values of the potential ϕ ifor (empty) sites within the circle.(iii.) Assign a random number r to each empty site within the boundary circle. The randomnumber r i at site i represents a breakdown coefficient and the random inhomogeneousnature of the insulator.(iv.) The growth sites are the nearest neighbor sites of the discharge pattern (the occupiedsites). Form the product r i ϕ a i for each growth site i, where a is an adjustable parameter.Because the potential for the discharge pattern is zero, ϕ i for growth site i can beinterpreted as the magnitude of the potential gradient at site i.(v.) The perimeter site with the maximum value of the product rϕ a breaks down, that is, setϕ for this site equal to zero.(vi.) Use the relaxation method to recompute the values of the potential at the remainingunoccupied sites and repeat steps (iv) and (v).Choose a = 1/4 and analyze the structure of the discharge pattern. Does the pattern appearqualitatively similar to lightning? Does the pattern appear to have a fractal geometry? Estimatethe fractal dimension by counting M(b), the average number of sites belonging to the dischargepattern that are within a b × b box. Consider other values of a, for example, a = 1/6 anda = 1/3, and show that the patterns have a fractal structure with a tunable fractal dimensionthat depends on the parameter a. Published results (Family et al.) are for patterns with 800occupied sites.c. Another version of the dielectric breakdown model associates a growth probability p i = ϕ a i / ∑ j ϕa jwith each growth site i, where the sum is over all the growth sites. One of the growth sites isoccupied with probability p i . That is, choose a growth site at random and generate a randomnumber r between 0 and 1. If r ≤ p i , the growth site i is occupied. As before, the exponent ais a free parameter. Convince yourself that a = 1 corresponds to diffusion-limited aggregation.(The boundary condition used in the latter corresponds to a zero potential at the growth sites.)To what type of cluster does a = 0 correspond? Consider a = 1/2, 1, and 2 and explore thedependence of the visual appearance of the clusters on a. Estimate the fractal dimension of theclusters.d. Consider a deterministic growth model for which all growth sites are tested for occupancy ateach growth step. Adopt the same geometry and boundary conditions as in part (b) and usethe relaxation method to solve Laplace’s equation for ϕ i . Then find the perimeter site with thelargest value of ϕ and set ϕ max equal to this value. Only those perimeter sites for which theratio ϕ i /ϕ max is larger than a parameter p become part of the cluster; ϕ i is set equal to unityfor these sites. After each growth step, the new growth sites are determined and the relaxationmethod is used to recompute the values of ϕ i at each unoccupied site. Choose p = 0.35 anddetermine the nature of the regular fractal pattern. What is the fractal dimension? Considerother values of p and determine the corresponding fractal dimension. These patterns have beentermed Laplace fractal carpets (see Family et al.).Surface growth models. The fractal objects we have discussed so far are self-similar, thatis, if we look at a small piece of the object and magnify it isotropically to the size of the original,the original and the magnified object look similar (on the average). In the following, we introduce

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

Saved successfully!

Ooh no, something went wrong!