Fractal REU Spring 2007 I. Carl Tams: Introduction II. Masaki Iino: Fractal Theory III. Eric Heisler: Continuous Dependence on Parameters IV. Bill Clark: IFS Random Trees V. Jordan Judkins: Biological Applications VI. Greg Danner: CS Applications VII. Eric Heisler: Music Graduate Assistant: Liz Copene Supervised By Elena Cherkaev Fractals Fractals • What are they? Pictures Definition • A fractal is a geometric figure that can be subdivided in parts, – – – – self similar small parts crazy Euclidean behavior generated by Iterated Function Systems (IFS) Fractals in life • Nature – clouds, mountain ranges, lightning bolts, coastlines, trees, leaves, rivers, mountains, cauliflower, broccoli, blood vessels, pulmonary vessels, and snow flakes • Man made fractals – roads – circuits Analyzing fractals • Fractal Dimension • Hausdorff Dimension • Self Similar Dimension • Box Dimension • Iterated Function Systems History • 1872-Karl Weierstrass • 1904-Helge von Koch – Koch snowflake Sierpinski • 1915-Waclaw Sierpinski-constructed his triangle and, one year later, his carpet Julia Set • Studied fractals without computers – – – – Henri Poincaré Felix Klein Pierre Fatou Gaston Julia Mandelbrot • 1975-Mandelbrot-illustrated fractals with a computer • -These images lead to the popular meaning of the term "fractal". Basic Fractal Theory Masaki Iino Department of Mathematics University of Utah REU Symposium --- April, 2007 Metric Space Def: A metric space (X, d) is a space X together with a real-valued fcn d:X x X -> R which measures the distance bet. pairs of pts x and y∈X. A metric space is considered as complete if every Cauchy seq. has a limit Xn->x for some x in X. Fractal Space Given a metric space (X,d), we define another metric space (H(X),h(d)). H(X) is the set of all nonempty compact subsets of X, and h(d) is the Hausdorff distance between two subsets of H(X). Fractals reside on this space. Def: Hausdorff distance btw. subsets A and B in H(X): H (A, B) = max { h (A, B), h (B, A) } where Example of Hausdorff Distance Take two point sets A {a1,a2,a3} and B{b1,b2,b3} as seen in the picture below. What’s h(A,B)? The Methods of Generating Fractals Escape-time fractals — These are defined by a recurrence relation at each point in a space. Examples of this type are the Mandelbrot set, Julia set. Iterated function systems — These have a fixed geometric replacement rule. Cantor set, Sierpinski carpet, Sierpinski gasket, Peano curve, Koch snowflake, Menger sponge, are some examples of such fractals. Random fractals — Generated by stochastic rather than deterministic processes, for example, trajectories of the Brownian motion, fractal landscapes and the Brownian tree. Iterated Function Systems (IFSs) Def:: An (hyperbolic) iterated function system is a metric space (X,d) together with a finite set of contraction mappings on that space. The notation used for such an IFS is: where each Wn is a contraction map and its contractivity factor is A fractal is made of the union of copies of itself. Each copy is transformed (i.e. scaled, rotated, sheared, reflected) by a fcn. And the fractal is the fixed attractor of an IFS. A transformation can be nonlinear (i.e. polynomial transformation) 2D IFSs and Levy’s Dragon 2D IFSs are expressed as: w(x)=Ax+t where A is a 2x2 real matrix and t is a column vector. As an example of 2D IFS, we introduce the Levy’s Dragon. The right pic shows the 1st to 4th iteration. Continuation of Levy's Dragon Visually, The Pic of the Levy’s Dragon Fractals as Sets of Attractors of IFSs Since the fractal is the fixed attractor of an IFS, we need a theorem for that: Fixed Point Theorem (or Contraction Mapping Theorem). This guarantees the existence and uniqueness of fixed points of some self-maps in metric space, and provides a way to find the fixed pts. Def. (Contraction Map): φ: X->X on a metric space is called a contraction map if there exists a positive constant s < 1 such that d(φ(x),φ(y)) ≤ s*d(x, y) for all x, y ∈X Fixed Point Theorem Let (X,d) be a complete metric space. If φ : X →X is a contraction, then φ has a unique fixed point. Remark: This is required to define a deterministic fractal which is a fixed point of a contractive map on a complete metric space. 3D IFSs and 3D Fern Instead of a 2x2 real matrix A and a column vector t (*,*), we have a 3x3 real matrix A and a column vector t (*,*,*) for a 3D IFS. Again, it can be expressed as w(x)= Ax+t. As an example of 3D, we introduce a 3D Fern, which is the attractor of an IFS of affine maps in 3D. 3D Fern This is an IFS for the 3D Fern [ ] [ ][ ] [ ] ][ ] [ ] [ ][ [ ] [ ][ ] [ ] [ ] [ ][ ] [ ] x1 0 0 0 w1 x 2 = 0 0.18 0 0 0 0 x3 x1 x2 x3 0 0 0 x1 0.85 0 0 w 2 x2 = 0 0.85 0.1 0 − 0.1 0.85 x3 x1 0.2 0.2 0 w 3 x 2 = 0.2 0.2 0 0 0 0.3 x3 x1 − 0.2 0.2 0 w 4 x 2 = 0.2 0.2 0 0 0 0.3 x3 x1 x2 x3 x1 x2 x3 x1 x2 x3 0 1.6 0 0 0.8 0 0 0.8 0 Recommended Books 1. Michael Bransley: “Fractals Everywhere” (Introductory Level) 2. Masaya Yamaguchi et al: “Mathematics of Fractals” (Advanced Level) Properties of IFS fractals Eric Heisler Fractal Dimension The easiest way to characterize a fractal box dimension: log( N ) D lim e 0 log 1 e e Some 3D IFS fractals • Menger sponge log( 20) D 2.73 log 3 0 0 1 / 3 0 w1 ( x) 0 1 / 3 0 x 0 0 0 1 / 3 2 / 3 Another one • Sierpinski pyramid log( 5) D 2.32 log 2 Self similar fractals Looks the same no matter how close you get. A smaller copy of the fern A copy within the copy Parameters • The picture changes continuously as you change the parameters. 0 2 / 3 1 / 3 0 w1 ( x) 0 1 / 3 0 x 0 0 0 1 / 3 0 • An applet that shows this Random Tree Fractals By Bill Clark Transformations cos w1 (v ) sin sin x1 v cos y1 (x1,y1) Each Iteration takes a line segment and creates two branches to replace it. Random Verses Non-Random Trees %Random Tree! %----------------------------MAIN----FUNCTION----------------------------function[]=random_tree(r,theta,iterations,exponent) %here is the original branch in the form of 2 points (x1,y1,x2,y2) l=[0,0,0,1]; x=l; h(1) = plot(l(1:2),l(3:4),'Color',[0.58 0.39 0.39],'Linewidth',1); hold on count = 2; for i=1:iterations n=ceil((rand(1)^exponent)*size(x,1)); l=x(n,:); [L1,L2]=expand(l,r,theta); %this will send the segment for expansion %replace the plotted line with two new lines x(n,:)=[]; x(end+1,:)=L1; x(end+1,:)=L2; %plot here h(count)=plot([L1(1),L1(3)],[L1(2),L1(4)],'Color',[0.58 0.39 0.39],'Linewidth',1); count = count+1; h(count)=plot([L2(1),L2(3)],[L2(2),L2(4)],'Color',[0.58 0.39 0.39],'Linewidth',1); count = count+1; drawnow end; hold off %----------------------------Calculation----FUNCTION----------------------------function [L1,L2]=expand(l,r,theta) t=[l(3);l(4)]; v=[l(1),l(3);l(2),l(4)]; %here are the two transformations A1=r*[cos(theta) , -sin(theta) ; sin(theta) , cos(theta)]; A2=r*[cos(theta) , sin(theta) ; -sin(theta) , cos(theta)]; % Calculate the two new lines L1=A1*(v(:,2)-v(:,1))+t; L2=A2*(v(:,2)-v(:,1))+t; L1=[t(1),t(2),L1(1),L1(2)]; L2=[t(1),t(2),L2(1),L2(2)]; Large Angles verses Smaller Angles Various 3D Objects Fractals In the human body Natural Fractals in the Body SYSTEMS Respiratory Lymphatic Nervous Circulatory (Mandelbrot 1982) Benefits of fractals Increase the Surface Area for absorption and transfer Self-similar (easily constructed) “Packing efficiency” (lungs , small intestines etc) Decrease speed of fluids or air Minimize material used to form structures Blood Vessels and Bile Ducts Branches enhance surface area aiding in distribution and collection. Lungs Surface area130 meters^2=5-6 liters Branching airway of the Lungs Esophagus Trachea Primary Bronchi Smaller Bronchi Bronchioles Alveoli (Porra 2006) Splitting of Air As air is split by the process mentioned above into the Alveoli. The splitting of air occurs on order for rapid diffusion of the gases. (Porra 2006) Different proposed lung models Length=L1= 2n× D 2n =# of branches Material used=SA × ar × L 2 × P × r02 × a × L r= radius of airway ar= width of wall of airway Model 2 L2=D/2+(2n-1) × D) 2n=# of end branches Material used=SA x ar x L 2 × P × a(r02 × D/2+ r12 × D/2 × 2n) r= radius of airway r1= 2-n r0 ar= width of wall of airway D 2 Model 3 L3= D/2+(2)D/4+(2n × D/4) 2n =# of end branches Material used=SA x ar x L 2 × P × a(r02 × D/2+ 2(r12 × D/4) + r22 × D/4 × 2n) r= radius of airway r2= 2-n r0 ar= width of wall of airway Model n Ln= D/2+(2)D/4+((4)D/8)…(2n-1 × D)/ 2n 2n =# of end branches S 2n-1/(2n) i=1…n Material used=SA x ar x L 2 × P × a(r02 × D/2+ 2(r12 × D/4) + 4(r22 × D/8)+… 2n (rn2 × D/(2n-1)) r1,r2,r3..rn= changing radius of airway ar= width of wall of airway Results for models Model 1 L1= 320 M1Z 8042a Model 2 (D=10, r0=2, n=5) L2= 165 M2Z129.591a Model 3 L3= 90 M3Z125.788a Model n Ln= 25.1563 MnZ125.786a Speed of the entering air Mauroy 2006 V=(d12v1+d22v2)/ (d13+d23)2/3 Diffusion of Air (Porra 2006) Air speed must equal the speed of diffusion for air intake to be optimized. n model proves to be the best Minimize material used Fractal dimension is higher (space filling) Security (if failure occurs minimizes effect) Slowing down the air when breathing Artificial organs: The reason to study References 1. Dissecting Fractals in Nature. http://fractal.org/Fractal-Research- and-Products/Dissecting-fractals.htm 2. Frame, Michael, Benoit Mandelbrot, and Nial Neger. Fractal Geometry.Yale University. March 6, 2007. http://classes.yale.edu/Fractals/ 3. Fung, Y.C., G.S. Kassab, and Z.L. Jiang. Diameter-defined Strahler system and connectivity matrix of the pulmonary arterial tree. University of California-San Diego. 1994. 4. Goldberger, Ary. Nonlinear dynamic, fractals,and chaos. March 5, 2007. http://www.anatomiafractal.com/article_fractal_anatomy.htm 5. Losa, Gabriela, Danilo Merlini, Theo F. Nonnenmacher, and Edwald Weibul. Fractals in Biology and Medicine. Springer. Verlag. vol 4. 2005. pp 17-29. References 6. Mandelbrot, B.B, (1982) The Fractal Geometry of Nature. W.H Freeman and Company. ISBN 0-7167-1186-9 7. Mandelbrot, B.B, (1982) The Fractal Geometry of Nature. W.H Freeman and Company. ISBN 0-7167-1186-9 8. Mauroy, Benjamin and Nicolas Meunier. A mechanical model of the human lung. Paris, France. November 2006. 9. Porra, Liisa. Lung structure and function studied by Synchotron Radiation. Report series of Physics. University of Helsinki. Helsinki. 2006. Accessed March 31, 2007. 10. E.R.WEIBEL. Morphometry of the human lung, Springer Verlag, 1963. 11.Xia, Quiglan. Ramified structures in optimal transportation. University of Texas at Austin. Jan 03, 2005. L-Systems Gregory Danner REU Symposium 2007 Formal Grammars Mathematically speaking, a formal grammar G is a 4-tuple of the form G=(T, N, P, S), where T is a set of terminal symbols, N is a set of non-terminal symbols, P is a set of production rules, and S is an element of N. Production rules are of the form: (T U N)*->(T U N)* L-Systems Were developed by a Hungarian botanist, Aristid Lindenmayer, to model plant growth. On the most basic level, an L-system is a formal grammar, often with parameters to allow for environmental influence (like the effects of a plant growing next to a wall or near to light) Since context-sensitive grammars are Turing complete, they have a very great amount of expressive power. What I Did I wrote a simple parser to render context-free grammars in 2D. Quite a few 2D L-systems have important applications – the Sierpinski Curve is used find solutions to Theto USS Makeapproximate Shit Up Lyrics the Traveling Salesman Problem, and the Hilbert Curve is used to preserve locality when mapping a 2D space on to a 1D line. Here are a few examples: A Quick Example Let’s build the Sierpinski Triangle, L-System style. The L-System for the Triangle is Axiom: A Rules: A -> B-A-B B -> A+B+A Where A and B both represent drawing one unit forward, + represents rotating 60⁰ to the right and – represents rotating 60⁰ to the left A Quick Example The first step is easy – it’s just the axiom, A, which is drawing one unit forward A Quick Example Now we apply a rule. We replace A with B-A-B. A Quick Example In the next step, there are two B’s and one A. Each B is replaced by A+B+A, and the A is replaced by BA-B, yielding: A+B+A-B-A-BA+B+A A Quick Example If you continue, the structure begins to look more and more like the famous triangle A few (quicker) examples… Sierpinski Curve Koch Snowflake Tetrahedron Curve ConstantBranch Tree Fractals in Music Eric Heisler Application to art • We know how to find the fractal dimension of various things. – Paintings – Organs – Clouds • Why not music? To visualize music • Make a frequency vs. time plot – Cut the music into short intervals. – Find the frequency spectrum using a Fourier transform. – Set a threshold for black and white. – Put them all together to build a picture! time freq. freq. A picture of music Different types of music Then measure their fractal dimensions. The dimensions • Simple music: d = 1.21 • Complex music: d = 1.36 Compare fractal dimension Why? • Could be a measure of complexity. • Perhaps different types of music have characteristic dimensions. • It’s just interesting!