Resistance metric, and spectral asymptotics, on the graph of the Weierstrass function
aa r X i v : . [ m a t h . F A ] A p r Resistance metric, and spectral asymptotics, on the graphof the Weierstrass functionClaire David
September 12, 2018
Sorbonne Universités, UPMC Univ Paris 06CNRS, UMR 7598, Laboratoire Jacques-Louis Lions, 4, place Jussieu 75005, Paris, France
Following our work on the graph of the Weierstrass function [5], in the spirit of those of J. Kigami[1], [2], and [3], [29], which enabled us to build a Laplacian on the aforementioned graph, it was naturalto go further and give the related explicit resistance metric. In doing so, we made calculations thatdirectly enable one to obtain the box dimension of the graph, in a simpler way than [6]or [7].The aim of this work is twofold. We had a special interest in the study of the spectral propertiesof the Laplacian. In [5], we have given the explicit the spectrum on the graph of the Weierstrassfunction. In the case of Laplacians on post-critically finite fractals, previous works, by J. Kigami andM. Lapidus [10], and R. S. Strichartz [29], make the link between resistance metric, and asymptoticproperties of the spectrum of the Laplacian, by means of an analoguous of Weyl’s formula.So we asked ourselves wether those results were still valid, for the graph of the Weierstrass function.
In this section, we recall results that are developed in [5].
Notation.
In the following, λ and N b are two real numbers such that: < λ < , N b ∈ N and λ N b > We will consider the ( − periodic) Weierstrass function W , defined, for any real number x , by: W ( x ) = + ∞ X n =0 λ n cos (2 π N nb x ) We place ourselves, in the sequel, in the Euclidean plane of dimension 2, referred to a direct or-thonormal frame. The usual Cartesian coordinates are ( x, y ) .1he restriction Γ W to [0 , × R , of the graph of the Weierstrass function, is approximated by means ofa sequence of graphs, built through an iterative process. To this purpose, we introduce the iteratedfunction system of the family of C ∞ contractions from R to R : { T , ..., T N b − } where, for any integer i belonging to { , ..., N b − } , and any ( x, y ) of R : T i ( x, y ) = (cid:18) x + iN b , λ y + cos (cid:18) π (cid:18) x + iN b (cid:19)(cid:19)(cid:19) Property 2.1. Γ W = N b − [ i =0 T i (Γ W ) Definition 2.1.
For any integer i belonging to { , ..., N b − } , let us denote by: P i = ( x i , y i ) = (cid:18) iN b − , − λ cos (cid:18) π iN b − (cid:19)(cid:19) the fixed point of the contraction T i .We will denote by V the ordered set (according to increasing abscissa), of the points: { P , ..., P N b − } The set of points V , where, for any i of { , ..., N b − } , the point P i is linked to the point P i +1 , con-stitutes an oriented graph (according to increasing abscissa)), that we will denote by Γ W . V is calledthe set of vertices of the graph Γ W .For any natural integer m , we set: V m = N b − [ i =0 T i ( V m − ) The set of points V m , where two consecutive points are linked, is an oriented graph (according toincreasing abscissa), which we will denote by Γ W m . V m is called the set of vertices of the graph Γ W m .We will denote, in the sequel, by N S m = 2 N mb + N b − the number of vertices of the graph Γ W m , and we will write: V m = (cid:8) S m , S m , . . . , S m N m − (cid:9) P , , P , , P , , in the case where λ = 12 , and N b = 3 .Figure 2: The graphs Γ W (in green), Γ W (in red), Γ W (in orange), Γ W (in cyan), in the casewhere λ = 12 , and N b = 3 . Definition 2.2. Consecutive vertices on the graph Γ W Two points X et Y de Γ W will be called consecutive vertices of the graph Γ W if there exists anatural integer m , and an integer j of { , ..., N b − } , such that: X = ( T i ◦ . . . ◦ T i m ) ( P j ) et Y = ( T i ◦ . . . ◦ T i m ) ( P j +1 ) { i , . . . , i m } ∈ { , ..., N b − } m X = ( T i ◦ T i ◦ . . . ◦ T i m ) ( P N b − ) et Y = ( T i +1 ◦ T i . . . ◦ T i m ) ( P ) Definition 2.3.
For any natural integer m , the N S m consecutive vertices of the graph Γ W m are, also,the vertices of N mb simple polygons P m,j , j N mb − , with N b sides. For any integer j suchthat j N mb − , one obtains each polygon by linking the point number j to the point num-ber j + 1 if j = i mod N b , i N b − , and the point number j to the point number j − N b + 1 if j = − mod N b . These polygons generate a Borel set of R . Definition 2.4. Polygonal domain delimited by the graph Γ W m , m ∈ N For any natural integer m , well call polygonal domain delimited by the graph Γ W m , and denoteby D (Γ W m ) , the reunion of the N mb polygons P m,j , j N mb − , with N b sides. Definition 2.5. Polygonal domain delimited by the graph Γ W We will call polygonal domain delimited by the graph Γ W , and denote by D (Γ W ) , the limit: D (Γ W ) = lim n → + ∞ D (Γ W m ) Definition 2.6. Word, on the graph Γ W Let m be a strictly positive integer. We will call number-letter any integer M i of { , . . . , N b − } ,and word of length |M| = m , on the graph Γ W , any set of number-letters of the form: M = ( M , . . . , M m ) We will write: T M = T M ◦ . . . ◦ T M m Definition 2.7. Edge relation, on the graph Γ W Given a natural integer m , two points X and Y of Γ W m will be called adjacent if and only if X and Y are two consecutive vertices of Γ W m . We will write: X ∼ m Y This edge relation ensures the existence of a word M = ( M , . . . , M m ) of length m , such that X and Y both belong to the iterate: T M V = ( T M ◦ . . . ◦ T M m ) V Given two points X and Y of the graph Γ W , we will say that X and Y are adjacent if and only ifthere exists a natural integer m such that: X ∼ m Y roposition 2.2. Adresses, on the graph of the Weierstrass function
Given a strictly positive integer m , and a word M = ( M , . . . , M m ) of length m ∈ N ⋆ , on thegraph Γ W m , for any integer j of { , ..., N b − } , any X = T M ( P j ) de V m \ V , i.e. distinct from one ofthe N b fixed point P i , i N b − , has exactly two adjacent vertices, given by: T M ( P j +1 ) et T M ( P j − ) where: T M = T M ◦ . . . ◦ T M m By convention, the adjacent vertices of T M ( P ) are T M ( P ) and T M ( P N b − ) , those of T M ( P N b − ) , T M ( P N b − ) and T M ( P ) . Definition 2.8. m th − order subcell, m ∈ N ⋆ , related to a pair of points of the graph Γ W Given a strictly positive integer m , and two points X and Y of V m such that X ∼ m Y , we will call m th − order subcell, related to the pair of points ( X, Y ) , the polygon, the vertices of whichare X , Y , and the intersection points of the edge between the vertices at the extremities of the polygon,i.e. the respective intersection points of polygons of the type P m,j − and P m,j , j N mb − , onthe one hand, and of the type P m,j and P m,j +1 , j N mb − , on the other hand.Figure 3: A m th − order subcell, in the case where λ = 12 , and N b = 7 . Notation.
For any integer j belonging to { , ..., N b − } , any natural integer m , and any word M oflength m , we set: T M ( P j ) = ( x ( T M ( P j )) , y ( T M ( P j ))) , T M ( P j +1 ) = ( x ( T M ( P j +1 )) , y ( T M ( P j +1 ))) m = x ( T M ( P j +1 )) − x ( T M ( P j )) = 1( N b − N mb Proposition 2.3.
An upper bound and lower bound, for the box-dimension of the graph Γ W For any integer j belonging to { , , . . . , N b − } , each natural integer m , and each word M of length m ,let us consider the rectangle, the width of which is: L m = x ( T M ( P j +1 )) − x ( T M ( P j )) = 1( N b − N mb and height | h j,m | , such that the points T M ( P j +1 ) and T M ( P j +1 ) are two vertices of this rectangle.Then: L − D W m ( N b − − D W (cid:12)(cid:12)(cid:12)(cid:12)(cid:26) − λ min j N b − sin (cid:18) π (2 j + 1) N b − (cid:19) − πN b ( N b −
1) ( λ N b − (cid:27)(cid:12)(cid:12)(cid:12)(cid:12) | h j,m | and: | h j,m | η − D W L − D W m where the real constant η − D W is given by : η − D W = 2 π ( N b − − D W (cid:26) (2 N b − λ ( N b − N b − (1 − λ ) ( λ N b −
1) + 2 N b ( λ N b −
1) ( λ N b − (cid:27) There exists thus a positive constant C = max (cid:26) ( N b − − D W (cid:12)(cid:12)(cid:12)(cid:12)(cid:26) − λ min j N b − sin (cid:18) π (2 j + 1) N b − (cid:19) − πN b ( N b −
1) ( λ N b − (cid:27)(cid:12)(cid:12)(cid:12)(cid:12) , η − D W (cid:27) such that the graph Γ W on L m can be covered by at least and at most: N m ( C (cid:18) L m N m (cid:19) − D W + 1 ) = C L − D W m N D W m + N m squares, the side length of which is L m N m .Proof. For any pair of integers ( i m , j ) of { , ..., N b − } : T i m ( P j ) = (cid:18) x j + i m N b , λ y j + cos (cid:18) π (cid:18) x j + i m N b (cid:19)(cid:19)(cid:19) For any pair of integers ( i m , i m − , j ) of { , ..., N b − } : T i m − ( T i m ( P j )) = x j + i m N b + i m − N b , λ y j + λ cos (cid:18) π (cid:18) x j + i m N b (cid:19)(cid:19) + cos π x j + i m N b + i m − N b !!! = (cid:18) x j + i m N b + i m − N b , λ y j + λ cos (cid:18) π (cid:18) x j + i m N b (cid:19)(cid:19) + cos (cid:18) π (cid:18) x j + i m N b + i m − N b (cid:19)(cid:19)(cid:19) ( i m , i m − , i m − , j ) of { , ..., N b − } : T i m − (cid:0) T i m − ( T i m ( P j )) (cid:1) = (cid:18) x j + i m N b + i m − N b + i m − N b ,λ y j + λ cos (cid:16) π (cid:16) x j + i m N b (cid:17)(cid:17) + λ cos (cid:18) π (cid:18) x j + i m N b + i m − N b (cid:19)(cid:19) + cos (cid:18) π (cid:18) x j + i m N b + i m − N b + i m − N b (cid:19)(cid:19) (cid:19) Given a strictly positive integer m , and two points X and Y of V m such that: X ∼ m Y there exists a word M of length |M| = m , on the graph Γ W , and an integer j of { , ..., N b − } , suchthat: X = T M ( P j ) , Y = T M ( P j +1 ) Let us write T M under the form: T M = T i m ◦ T i m − ◦ . . . ◦ T i where ( i , . . . , i m ) ∈ { , ..., N b − } m .One has then: x ( T M ( P j )) = x j N mb + m X k =1 i k N kb , x ( T M ( P j +1 )) = x j +1 N mb + m X k =1 i k N kb and: y ( T M ( P j )) = λ m y j + m X k =1 λ m − k cos π x j N kb + k X ℓ =0 i m − ℓ N k − ℓb !! y ( T M ( P j +1 )) = λ m y j +1 + m X k =1 λ m − k cos π x j +1 N kb + k X ℓ =0 i m − ℓ N k − ℓb !! This leads to: h j,m − λ m ( y j +1 − y j ) = m X k =1 λ m − k ( cos π x j +1 N kb + k X ℓ =0 i m − ℓ N k − ℓb !! − cos π x j N kb − k X ℓ =0 i m − ℓ N k − ℓb !!) = − m X k =1 λ m − k sin (cid:18) π (cid:18) x j +1 − x j N kb (cid:19)(cid:19) sin π x j +1 + x j N kb + k X ℓ =0 i m − ℓ N k − ℓb !! Taking into account: λ m ( y j +1 − y j ) = λ m − λ (cid:18) cos (cid:18) π ( j + 1) N b − (cid:19) − cos (cid:18) π jN b − (cid:19)(cid:19) = − λ m − λ sin (cid:18) πN b − (cid:19) sin (cid:18) π (2 j + 1) N b − (cid:19) h j,m + 2 λ m − λ sin (cid:18) πN b − (cid:19) sin (cid:18) π (2 j + 1) N b − (cid:19) = − m X k =1 λ m − k sin πN k +1 b ( N b − ! sin π (2 j + 1) N k +1 b ( N b −
1) + 2 π k X ℓ =0 i m − ℓ N k − ℓb ! Thus: (cid:12)(cid:12)(cid:12)(cid:12) y ( T M ( P j )) − y ( T M ( P j +1 )) − λ m − λ sin (cid:18) πN b − (cid:19) sin (cid:18) π (2 j + 1) N b − (cid:19)(cid:12)(cid:12)(cid:12)(cid:12) m X k =1 π λ m − k N k +1 b ( N b − π λ m (cid:16) − λ m N mb (cid:17) λ N b N b ( N b − (cid:16) − λ N b (cid:17) π λ m N b ( N b −
1) ( λ N b − which leads to: y ( T M ( P j )) − y ( T M ( P j +1 )) > λ m − λ sin (cid:18) πN b − (cid:19) sin (cid:18) π (2 j + 1) N b − (cid:19) − π λ m N b ( N b −
1) ( λ N b − or: y ( T M ( P j +1 )) − y ( T M ( P j )) > λ m − λ sin (cid:18) πN b − (cid:19) sin (cid:18) π (2 j + 1) N b − (cid:19) − π λ m N b ( N b −
1) ( λ N b − Due to the symmetric roles played by T M ( P j ) and T M ( P j +1 ) , one may only consider the case when: y ( T M ( P j )) − y ( T M ( P j +1 )) > λ m − λ sin (cid:18) πN b − (cid:19) sin (cid:18) π (2 j + 1) N b − (cid:19) − π λ m N b ( N b −
1) ( λ N b − > > λ m (cid:26) − λ min j N b − sin (cid:18) π (2 j + 1) N b − (cid:19) − πN b ( N b −
1) ( λ N b − (cid:27) The predominant term is thus: λ m = e m ( D W −
2) ln N b = N m ( D W − b = L − D W m ( N b − − D W One also has: 8 h j,m | λ m − λ π (2 j + 1)( N b − + 2 m X k =1 λ m − k π ( j + 1( N b − N kb + 2 k X ℓ =0 i m − ℓ N k − ℓb ) π ( N b − N kb = 2 λ m − λ π (2 j + 1)( N b − + 2 π λ m N b − m X k =1 ( (2 j + 1) λ − k ( N b − N kb + 2 k X ℓ =0 i m − ℓ λ − k N k − ℓb ) = 2 λ m − λ π (2 j + 1)( N b − + 2 π λ m N b − ( λ − N − b (2 j + 1)( N b −
1) (1 − λ − m N − mb )1 − λ − N − b + 2 m X k =1 ( N b − λ − k N kb − N − k − b − N − b ) λ m − λ π (2 N b − N b − + 2 π λ m N b − N b − N b −
1) (1 − λ − m N − mb ) λ N b −
1+ 2 π λ m N b − λ − N − b ( N b −
1) (1 − λ − m N − mb )(1 − N − b ) (1 − λ − N − b ) − π λ m N b − λ − N − b ( N b −
1) (1 − λ − m N − mb )(1 − N − b ) (1 − λ − N − b ) λ m − λ π (2 N b − N b − + 2 π λ m N b − N b − N b −
1) 1 λ N b −
1+ 4 π N b λ m N b − (cid:26) λ N b − − λ N b − (cid:27) = 2 π λ m (cid:26) (2 N b − λ ( N b − N b − (1 − λ ) ( λ N b −
1) + 2 N b ( λ N b −
1) ( λ N b − (cid:27) Since: x ( T M ( P j +1 )) − x ( T M ( P j )) = 1( N b − N mb and: D W = 2 + ln λ ln N b , λ = e ( D W −
2) ln N b = N ( D W − b one has thus: | h j,m | π L − D W m ( N b − − D W (cid:26) (2 N b − λ ( N b − N b − (1 − λ ) ( λ N b −
1) + 2 N b ( λ N b −
1) ( λ N b − (cid:27) roperty 2.4. Exact computation of the measure of the surface of a simple polygon P m,j , j N mb − , with N b sides Let us note that, given a natural integer m , there exists a word M of length m such that the orderedset, according to increasing abscissa, of the N b vertices of a simple polygon P m,j , j N mb − , canbe written as: T M ( P ) , T M ( P ) , . . . , T M ( P N b − ) This enables one to exactly compute the measure, with respect to the standard Lebesgue measure on R ,of any of the aforementioned polygons, as:i. In the case where N b = 3 : A ( P m,j ) = 12 (cid:13)(cid:13)(cid:13) −−−−−−−−−−−−→ T M ( P ) T M ( P ) ∧ −−−−−−−−−−−−→ T M ( P ) T M ( P ) (cid:13)(cid:13)(cid:13) ii. In the case where N b > : A ( P m,j ) = 12 N b − X j =0 (cid:13)(cid:13)(cid:13) −−−−−−−−−−−−−−→ T M ( P j ) T M ( P j +1 ) ∧ −−−−−−−−−−−−−−−→ T M ( P j ) T M ( P N b − ) (cid:13)(cid:13)(cid:13) Remark . One obtains, for j N mb − : (cid:13)(cid:13)(cid:13) −−−−−−−−−−−−−−→ T M ( P j ) T M ( P j +1 ) ∧ −−−−−−−−−−−−−−−→ T M ( P k ) T M ( P N b − ) (cid:13)(cid:13)(cid:13) == (cid:12)(cid:12)(cid:12) x (cid:16) −−−−−−−−−−−−−−→ T M ( P j ) T M ( P j +1 ) (cid:17) y (cid:16) −−−−−−−−−−−−−−−→ T M ( P j ) T M ( P N b − ) (cid:17) − y (cid:16) −−−−−−−−−−−−−−→ T M ( P j ) T M ( P j +1 ) (cid:17) x (cid:16) −−−−−−−−−−−−−−−→ T M ( P j ) T M ( P N b − ) (cid:17)(cid:12)(cid:12)(cid:12) where: x (cid:16) −−−−−−−−−−−−−−−→ T M ( P j ) T M ( P N b − ) (cid:17) = x ( T M ( P j +1 )) − x ( T M ( P N b − ))= x j N mb + m X k =1 i k N kb − x N b − N mb − m X k =1 i k N kb = N b − − j ( N b − N mb and: 10 (cid:16) −−−−−−−−−−−−−−→ T M ( P j ) T M ( P j +1 ) (cid:17) = y ( T M ( P j +1 )) − y ( T M ( P j ))= λ m y j +1 + m − X k =1 λ m − k cos π x j +1 N kb + k X ℓ =0 i m − ℓ N k − ℓb !! + cos π x j +1 N mb + m X k =1 i k N kb !! − λ m y j + m − X k =1 λ m − k cos π x j N kb + k X ℓ =0 i m − ℓ N k − ℓb !! + cos π x j N mb + m X k =1 i k N kb !! = λ m ( y j +1 − y j )+ m − X k =1 λ m − k ( cos π x j +1 N kb + k X ℓ =0 i m − ℓ N k − ℓb !! − cos π x j N kb + k X ℓ =0 i m − ℓ N k − ℓb !!) + cos π x j +1 N mb + m X k =1 i k N kb !! − cos π x j N mb + m X k =1 i k N kb !! = λ m − λ (cid:26) cos (cid:18) π ( j + 1) N b − (cid:19) − cos (cid:18) π jN b − (cid:19)(cid:27) + m − X k =1 λ m − k ( cos π x j +1 N kb + k X ℓ =0 i m − ℓ N k − ℓb !! − cos π x j N kb + k X ℓ =0 i m − ℓ N k − ℓb !!) + cos π x j +1 N mb + m X k =1 i k N kb !! − cos π x j N mb + m X k =1 i k N kb !! y (cid:16) −−−−−−−−−−−−−−−→ T M ( P j ) T M ( P N b − ) (cid:17) = y ( T M ( P N b − )) − y ( T M ( P j ))= λ m ( y N b − − y j )+ m − X k =1 λ m − k ( cos π x N b − N kb + k X ℓ =0 i m − ℓ N k − ℓb !! − cos π x j N kb + k X ℓ =0 i m − ℓ N k − ℓb !!) + cos π x N b − N mb + m X k =1 i k N kb !! − cos π x j N mb + m X k =1 i k N kb !! Thus: 11 N b − N mb (cid:13)(cid:13)(cid:13) −−−−−−−−−−−−−−→ T M ( P j ) T M ( P j +1 ) ∧ −−−−−−−−−−−−−−−→ T M ( P k ) T M ( P N b − ) (cid:13)(cid:13)(cid:13) = (cid:12)(cid:12)(cid:12)(cid:12) λ m − λ (cid:26) − cos (cid:18) π jN b − (cid:19)(cid:27) + m − X k =1 λ m − k ( cos π N kb + k X ℓ =0 i m − ℓ N k − ℓb !! − cos π jN b − N kb + k X ℓ =0 i m − ℓ N k − ℓb !!) + cos π N mb + m X k =1 i k N kb !! − cos π jN b − N mb + m X k =1 i k N kb !! + ( N b − − j ) λ m − λ (cid:26) cos (cid:18) π ( j + 1) N b − (cid:19) − cos (cid:18) π jN b − (cid:19)(cid:27) +( N b − − j ) ( m − X k =1 λ m − k ( cos π ( j + 1) N b − N kb + k X ℓ =0 i m − ℓ N k − ℓb !! − cos π jN b − N kb + k X ℓ =0 i m − ℓ N k − ℓb !!)) +( N b − − j ) ( cos π ( j + 1) N b − N mb + m X k =1 i k N kb !! − cos π j ( N b − N mb + m X k =1 i k N kb !!) (cid:12)(cid:12)(cid:12)(cid:12) Definition 2.9. Measure, on the domain delimited by the graph Γ W We will call domain delimited by the graph Γ W , and denote by D (Γ W ) , the limit: D (Γ W ) = lim n → + ∞ D (Γ W m ) which has to be understood in the following way: given a continuous function u on the graph Γ W , anda measure with full support µ on R , then: Z D (Γ W ) u dµ = lim m → + ∞ N mb − X j =0 X X vertex of P m,j u ( X ) µ ( P m,j ) We will say that µ is a measure, on the domain delimited by the graph Γ W . Proposition 2.5.
Harmonic extension of a function, on the graph of the Weierstrass func-tion
For any strictly positive integer m , if u is a real-valued function defined on V m − , its harmonicextension , denoted by ˜ u , is obtained as the extension of u to V m which minimizes the energy: E Γ W m (˜ u, ˜ u ) = η − − D W N − D W ) mb X X ∼ m Y (˜ u ( X ) − ˜ u ( Y )) The link between E Γ W m and E Γ W m − is obtained through the introduction of two strictly positive con-stants r m and r m +1 such that: η − − D W N − D W ) mb r m X X ∼ m Y (˜ u ( X ) − ˜ u ( Y )) = η − − D W N − D W ) ( m − b r m − X X ∼ m − Y ( u ( X ) − u ( Y )) n particular: η − − D W N − D W ) b r X X ∼ Y (˜ u ( X ) − ˜ u ( Y )) = η − − D W r X X ∼ Y ( u ( X ) − u ( Y )) For the sake of simplicity, we will fix the value of the initial constant: r = 1 . One has then: E Γ W m (˜ u, ˜ u ) = 1 r E Γ W (˜ u, ˜ u ) Let us set: r = 1 r and: E m ( u ) = η − − D W r m X X ∼ m Y (˜ u ( X ) − ˜ u ( Y )) Since the determination of the harmonic extension of a function appears to be a local problem, on thegraph Γ W m − , which is linked to the graph Γ W m by a similar process as the one that links Γ W to Γ W ,one deduces, for any strictly positive integer m : E Γ W m (˜ u, ˜ u ) = 1 r E Γ W m − (˜ u, ˜ u ) By induction, one gets: r m = r m r = r − m If v is a real-valued function, defined on V m − , of harmonic extension ˜ v , we will write: E m ( u, v ) = η − − D W r − m X X ∼ m Y (˜ u ( X ) − ˜ u ( Y )) (˜ v ( X ) − ˜ v ( Y )) For further precision on the construction and existence of harmonic extensions, we refer to [ ? ] . Property 2.6.
Self-similar measure, for the domain delimited by the graph of the Weier-strass function
Let us denote by µ L the Lebesgue measure on R . We set, for any i of { , . . . , N b − } : µ i = µ L ( T i ( P )) µ L ( P ) The measure µ , such that: µ = N b − X i =0 µ i µ ◦ T − i is self-similar, for the domain delimited by the graph of the Weierstrass function. We refer to [5] forfurther details. efinition 2.10. Laplacian of order m ∈ N ⋆ For any strictly positive integer m , and any real-valued function u , defined on the set V m of the verticesof the graph Γ W m , we introduce the Laplacian of order m , ∆ m ( u ) , by: ∆ m u ( X ) = X Y ∈ V m , Y ∼ m X ( u ( Y ) − u ( X )) ∀ X ∈ V m \ V Definition 2.11. Existence domain of the Laplacian, for a continuous function on thegraph Γ W (see [ ? ])We will denote by dom ∆ the existence domain of the Laplacian, on the graph Γ W , as the set offunctions u of dom E such that there exists a continuous function on Γ W , denoted ∆ u , that we will call Laplacian of u , such that : E ( u, v ) = − Z D (Γ W ) v ∆ u dµ for any v ∈ dom E Notation.
In the following, we will denote by H ⊂ dom ∆ the space of harmonic functions, i.e. thespace of functions u ∈ dom ∆ such that: ∆ u = 0 Given a natural integer m , we will denote by S ( H , V m ) the space, of dimension N mb , of spline functions" of level m ", u , defined on Γ W , continuous, such that, for any word M of length m , u ◦ T M is harmonic,i.e.: ∆ m ( u ◦ T M ) = 0 Property 2.7.
Let m be a strictly positive integer, X / ∈ V a vertex of the graph Γ W , and ψ mX ∈ S ( H , V m ) a spline function such that: ψ mX ( Y ) = (cid:26) δ XY ∀ Y ∈ V m ∀ Y / ∈ V m , where δ XY = (cid:26) if X = Y elseFor any function u of dom E , such that its Laplacian exists: ∆ u ( X ) = lim m → + ∞ η − − D W r − m Z D (Γ W ) ψ mX dµ ! − ∆ m u ( X ) Notation.
We will denote by dom E the subspace of continuous functions defined on Γ W , such that: E ( u ) < + ∞ roperty 2.8. Spectrum of the Laplacian (We refer to our work [5])Let us consider the eigenvalues ( − Λ m ) m ∈ N of the sequence of graph Laplacians (∆ m ) m ∈ N , built on thediscrete sequence of graphs (Γ W m ) m ∈ N .The spectral decimation method leads to the following recurrence relations between the eigenvalues oforder m and m + 1 : Λ m = − m − − ε (cid:16) { Λ m − − } − (cid:17) Nb + 1 − m − − ε (cid:16) { Λ m − − } − (cid:17) Nb where ε ∈ {− , } . Property 3.1.
The space dom ∆ , modulo constant functions, is a Hilbert space, included in the spaceof continuous functions on the graph Γ W , modulo constant functions. Definition 3.1. Effective resistance metric, on the graph Γ W Given a pair of points ( X, Y ) of the graph Γ W , we define, as in [28], the effective resistance metricbetween the points X and Y , by: R Γ W ( X, Y ) = (cid:26) min { u | u ( X )=0 ,u ( Y )=1 } E ( u ) (cid:27) − In an equivalent way, R Γ W ( X, Y ) may be defined as the minimum value of the real numbers R suchthat, for any function u of dom ∆ : | u ( X ) − u ( Y ) | R E ( u ) Definition 3.2. Metric, on the graph Γ W Let us define, on the graph Γ W , the distance d Γ W defined, for any pair of points ( X, Y ) of Γ W , by: d Γ W ( X, Y ) = (cid:26) min { u | u ( X )=0 ,u ( Y )=1 } E ( u, u ) (cid:27) − Remark . As it is explained in [29], one may note that the minimum min { u | u ( X )=0 ,u ( Y )=1 } E ( u )
15s reached when the function u is harmonic on the complement set, in Γ W , of the set { X } ∪ { Y } (we re-call that, by definition, a harmonic function u on Γ W minimizes the sequence of energies (cid:0) E Γ W m ( u, u ) (cid:1) m ∈ N .In order to fully apprehend and understand the intrinsic meaning of these functions, one might reasonby analogy with the unit interval [0 , . In this case, one will note that, given two points X and Y of [0 , such that X < Y , the function u is affine by pieces, taking the value zero on [0 , X ] , and thevalue 1 on [ Y, (see the illustration on the following figure): ∀ t ∈ [0 ,
1] : u ( t ) = t − XY − X Figure 4: The graph of the function u where the value min { u | u ( X )=0 ,u ( Y )=1 } E ( u ) is reached.Let us denote by m the natural integer such that: X ∼ m Y One may introduce, the, for any integer p , the sequence of points ( X j ) j p such that: X = X , X p = Y and, for any integer j such that < j < p − : X j ∈ V p +1 , X j ∼ p +1 X j +1 In the case of the unit interval, the normalization constant is: r − = 12 One has then: 16 ( u, u ) = lim p → + ∞ E p ( u, u )= lim m → + ∞ r − p E Γ W p ( u )= lim p → + ∞ X ( X,Y ) ∈ V p , X ∼ p Y r − p (cid:0) u | V p ( X ) − u | V p ( Y ) (cid:1) = lim p → + ∞ X ( X,Y ) ∈ V p , X ∼ p Y p (cid:0) u | V p ( X ) − u | V p ( Y ) (cid:1) = lim p → + ∞ k − X j =0 p (cid:18) u (cid:18) X + j p (cid:19) − u (cid:18) X + j + 12 p (cid:19)(cid:19) = Z YX dt ( Y − X ) = 1 Y − X If d R denotes the usual Euclidean distance on R : ∀ ( X, Y ) ∈ R : d R ( X, Y ) = | Y − X | one has then: min { u | u ( X )=0 ,u ( Y )=1 } E ( u ) = 1 d R ( X, Y ) Let us now consider, more generally, a fractal domain F , in an Euclidean space of dimension d ∈ N ⋆ ,equipped with the distance d R d . If, one has, in advance, defined an energy on F , it is worth searchingwether there exists a real number β such that: ∀ ( X, Y ) ∈ F : (cid:18) min { u | u ( X )=0 ,u ( Y )=1 } E ( u ) (cid:19) − ∼ ( d R d ( X, Y )) β In the case of the Sierpi`nski gasket SG (we refer to [ ? ]), Robert S. Strichartz lays the emphasis uponthe fact that, given X ∼ m Y , one has: min { u | u ( X )=0 ,u ( Y )=1 } E ( u ) . r m SG = (cid:18) (cid:19) m This also corresponds thus to the order of the diameter of the m th − order cells.Since the Sierpi`nski gasket SG is obtained from the initial triangle of diameter 1 by means of threecontractions, the respective ratios of which are equal to , one has simply to look the real number β SG such that: (cid:18) (cid:19) m β SG = (cid:18) (cid:19) m This leads to: β SG = ln ln 2 efinition 3.3. Dimension of the graph Γ W , in the effective resistance metric The dimension of the graph Γ W , in the effective resistance metric, is the strictly positive num-ber d Gamma W such that, given a strictly positive real number r , and a point X ∈ Γ W , for the X − centeredball of radius r , denoted by B r ( X ) : µ ( B r ( X )) = r d Γ W Proposition 3.2.
The dimension of the graph Γ W , in the effective resistance metric, is given by: i . First case: λ > N b . d Γ W = ln N b λ (5 − D W ) ln N b ii . Second case: λ < N b . d Γ W = 25 − D W Proof.Remark . Once again, it is worth having a look at the case of the Sierpiński gasket. Robert S. Strichartzstars from the fact that the measure of m th − order cells is m . Two consecutive points x and y aresuch that, for the effective resistance metric d ( x, y ) ∼ (cid:18) (cid:19) m For the self-similar measure µ SG , which affects the value m to each m th − order cell, one has simplyto look for the real number d SG such that: (cid:18) (cid:19) m d SG = 13 m which leads to: d SG = ln 3ln One may then deduce from the above an estimate, for the effective resistance metric, of the measureof a X − centered ball of radius r , denoted by B r ( X ) : µ SG ( B r ( x )) = r d SG Let us now go back to the graph Γ W . 18iven a natural integer m , and two points X and Y such that X ∼ m Y : min { u | u ( X )=0 ,u ( Y )=1 } E ( u ) . η − − D W r − m = η − − D W N (5 − D W ) mb For the detailed calculations which enable one to obtain the normalization constants, we refer to [27].For the self-similar measure ˜ µ introduced in the above, each m th − order cell, i.e. each simple poly-gon P m,j , j N mb − , with N b sides and N b vertices, has a measure of the order of: ( N b − η m N mb The points X and Y such that X ∼ m Y belong to a m th − order subcell, which is the intersection ofa simple polygon P m,j , j N mb − , with the rectangle of which X and Y are two vertices, ofwidth η m N mb , and height η m . This subcell a has a measure, the order of which is thus: η m L m = η m ( N b − N mb i . First case: λ > N b .One has simply to look for the real number d Γ W such that: (cid:16) N − D W b (cid:17) m d Γ W and λ m N mb are of the same order, which yields: d Γ W = ln N b λ (5 − D W ) ln N b ii . Second case: λ < N b .One has simply to look for the real number d Γ W such that: (cid:16) N − D W b (cid:17) m d Γ W and N mb are of the same order, which yields: d Γ W = 25 − D W As exposed by R. S. Strichartz in [29], one may bear in mind that the eigenvalues can be grouped intotwo categories: i . initial eigenvalues, which a priori belong to the set of forbidden values (as for instance Λ = 2 ) ; ii . continued eigenvalues, obtained by means of spectral decimation.We present, in the sequel, a detailed study of the spectrum of ∆ , in the case where N b = 3 , which canbe easily extended to higher values of the integer N b .19 .1 Eigenvalues and eigenvectors of ∆ Let us recall that the vertices of the graph Γ W are: P , T ( P ) , T ( P ) , T ( P ) P , T ( P ) , T ( P ) , T ( P ) , P One may note that: Card ( V \ V ) = 4 Let us denote by u an eigenfunction, for the eigenvalue − Λ . For the sake of simplicity, we set: u ( T ( P )) = a ∈ R , u ( T ( P )) = b ∈ R , u ( T ( P )) = c ∈ R , u ( T ( P )) = d ∈ R Figure 5: Successive values of an eigenfunction on V , in the case where N b = 3 .One has then: u ( P ) − a + b − a = − Λ aa − b + u ( P ) − b = − Λ bu ( P ) − c + d − c = − Λ cc − d + u ( P ) − d = − Λ d One may note that the only "Dirichlet eigenvalues", i.e. the ones related to the Dirichlet problem: u | V = 0 i.e. u ( P ) = u ( P ) = u ( P ) = 0 are obtained for: b = − (Λ − aa = − (Λ − bd = − (Λ − cc = − (Λ − d i.e.: 20 b = (Λ − ba = (Λ − ad = (Λ − dc = (Λ − c The forbidden eigenvalue
Λ = 2 cannot thus be a Dirichlet one.Let us consider the case where: (Λ − = 1 i.e. Λ = 1 or Λ = 3
The value
Λ = 1 leads to: a = b , c = d which yields a two-dimensional eigenspace. The multiplicity of the eigenvalue Λ = 1 is 2.For the eigenvalue
Λ = 3 : a = − b , c = − d The eigenspace, for the eigenvalue , has dimension 2. The multiplicity of the eigenvalue Λ = 3 is 2.Since the cardinal of V \ V is: N S − N b = 2 N b − one may note that we have the complete spectrum. ∆ Let us now look at the spectrum of ∆ . For the sake of simplicity, we will denote by a , b , c , d , e , f , g , h ,the successive values of an eigenfunction at the N b − points between P and P , and by a ′ , b ′ , c ′ , d ′ , e ′ , f ′ , g ′ , h ′ ,the successive values of an eigenfunction at the N b − points between P and P , as it appears on thefollowing figure.One has then: (2 − Λ) a = − u ( P ) − b (2 − Λ) b = − c − a (2 − Λ) c = − b − d (2 − Λ) d = − c − e (2 − Λ) e = − d − f (2 − Λ) f = − e − g (2 − Λ) g = − f − h (2 − Λ) h = − g − u ( P ) and: 21igure 6: Successive values of an eigenfunction on V , in the case where N b = 3 . (2 − Λ) u ( P ) = − h − a ′ and: (2 − Λ) a ′ = − u ( P ) − b ′ (2 − Λ) b ′ = − c ′ − a ′ (2 − Λ) c ′ = − b ′ − d ′ (2 − Λ) d ′ = − c ′ − e ′ (2 − Λ) e ′ = − d ′ − f ′ (2 − Λ) f ′ = − e ′ − g ′ (2 − Λ) g ′ = − f ′ − h ′ (2 − Λ) h ′ = − g ′ − u ( P ) One may note that the only Dirichlet eigenvalues, in the case where: u | V = 0 i.e. u ( P ) = u ( P ) = u ( P ) = 0 are obtained for: (2 − Λ) a = − b (2 − Λ) b = − a (2 − Λ) c = − b − d (2 − Λ) d = − e (2 − Λ) e = − d (2 − Λ) f = − e − g (2 − Λ) g = − h (2 − Λ) h = − g and − h − a ′ and (2 − Λ) a ′ = − b ′ (2 − Λ) b ′ = − a ′ (2 − Λ) c ′ = − b ′ − d ′ (2 − Λ) d ′ = − e ′ (2 − Λ) e ′ = − d ′ (2 − Λ) f ′ = − e ′ − g ′ (2 − Λ) g ′ = − h ′ (2 − Λ) h ′ = − g ′ i.e.: 22 (2 − Λ) a = − b (2 − Λ) b = b (cid:8) − (2 − Λ) (cid:9) a = 0(2 − Λ) c = − b − d (2 − Λ) d = − e (2 − Λ) e = e (2 − Λ) f = − e − g (2 − Λ) h = h (2 − Λ) h = − g and − h − a ′ and (2 − Λ) a ′ = − b ′ (2 − Λ) b ′ = b ′ (cid:8) − (2 − Λ) (cid:9) a ′ = 0(2 − Λ) c ′ = − b ′ − d ′ (2 − Λ) d ′ = − e ′ (2 − Λ) e ′ = e ′ (2 − Λ) f ′ = − e ′ − g ′ (2 − Λ) h ′ = h ′ (2 − Λ) h ′ = − g ′ The forbidden eigenvalue
Λ = 2 is not therefore a Dirichlet one.Let us consider the case where: (Λ − = 1 i.e. Λ = 3 or Λ = 1
For
Λ = 1 , one has: a = − bc = − b − dd = − ef = − e − gh = − g and − h − a ′ and a ′ = − b ′ c ′ = − b ′ − d ′ d ′ = − e ′ f ′ = − e ′ − g ′ h ′ = − g ′ The eigenspace, for
Λ = 1 , has thus dimension 5. The multiplicity of the eigenvalue
Λ = 1 is 5.For
Λ = 3 : (cid:8) − (2 − Λ) (cid:9) a = 0 c = b + dd = ef = e + gh = g and − h − a ′ and a ′ = b ′ c ′ = b ′ + d ′ d ′ = e ′ f ′ = e ′ + g ′ h ′ = g ′ The eigenspace, for
Λ = 3 , has thus dimension 5. The multiplicity of the eigenvalue
Λ = 3 is 5.Let us now look at the continued eigenvalues, i.e. the ones obtained from the eigenvalues Λ = 1 and Λ = 3 by means of spectral decimation: Λ = φ − (cid:16) ( φ (Λ )) Nb (cid:17) = n ( φ (Λ )) Nb + 1 o ( φ (Λ )) Nb = − − ε q { Λ − } − Nb + 1 − − ε q { Λ − } − Nb ε ∈ {− , } , for the values: Λ ∈ { , } As in [29], let us get rid, temporarily, of the Dirichlet conditions. We have thus: u ( P ) + b = − (Λ − aa + c = − (Λ − bb + d = − (Λ − ce + f = − Λ de + g = − (Λ − ff + h = − (Λ − gg + u ( P ) = − (Λ − h and (2 − Λ) u ( P ) = − h − a ′ and u ( P ) + b ′ = − (Λ − a ′ a ′ + c ′ = − (Λ − b ′ b ′ + d ′ = − (Λ − c ′ e ′ + f ′ = − Λ d ′ e ′ + g ′ = − (Λ − f ′ f ′ + h ′ = − (Λ − g ′ g ′ + u ( P ) = − (Λ − h ′ For the initial eigenvalue Λ = 1 , it is worth noticing that the restriction of the associated eigenvalue Λ to V \ V must satisfy the eigensystem associated to the eigenvalue Λ = 1 , i.e.: (cid:26) u ( P ) + f = − (Λ − cu ( P ) + c = − (Λ − f and − (Λ − u ( P ) = f + c ′ and (cid:26) u ( P ) + f ′ = − (Λ − c ′ u ( P ) + c ′ = − (Λ − f ′ or: (cid:26) u ( P ) + f = cu ( P ) + c = f and − (Λ − u ( P ) = f + c ′ and (cid:26) u ( P ) + f ′ = c ′ u ( P ) + c ′ = f ′ i.e.: u ( P ) + u ( P ) = 0 and u ( P ) + u ( P ) = 0 For u ( P ) = u ( P ) = u ( P ) = 0 , it works, and the Dirichlet conditions appear to be satisfied. One hasthen: b = − (Λ − ac = (cid:8) − − (cid:9) ad = (Λ − (cid:8) − (cid:8) − (Λ − (cid:9)(cid:9) ae + f = − Λ (Λ − (cid:8) − (cid:8) − (Λ − (cid:9)(cid:9) ae = (Λ − (cid:8) − (cid:8) − − (cid:9)(cid:9) hf = (cid:8) − − (cid:9) hg = − (Λ − h , c ′ = − f , b ′ = − (Λ − a ′ c ′ = (cid:8) − − (cid:9) a ′ d ′ = (Λ − (cid:8) − (cid:8) − (Λ − (cid:9)(cid:9) a ′ e ′ + f ′ = − Λ (Λ − (cid:8) − (cid:8) − (Λ − (cid:9)(cid:9) a ′ e ′ = (Λ − (cid:8) − (cid:8) − − (cid:9)(cid:9) h ′ f ′ = (cid:8) − − (cid:9) h ′ g ′ = − (Λ − h ′ We obtain thus an eigenspace, the dimension of which is .For the eigenvalue Λ = 1 , the spectral decimation spectral leads to: − Λ + ε ρ ( ω ) e i θω ! N b = 1 + ε √ e i π which leads to the quadruple eigenvalue: Λ = 2 + cos π √ π For the eigenvalue Λ = 3 , the spectral decimation leads to the quadruple eigenvalue: Λ = 2 n π o Since the cardinal of V \ V is: N S − N S = 16 one may note that we have the complete spectrum.24 .3 Eigenvalues of ∆ As previously, one can easily check that the forbidden eigenvalue
Λ = 2 is not therefore a Dirichletone.One can also check that Λ = 1 and Λ = 3 are eigenvalues of ∆ , both with multiplicity 8.From: Λ ֒ → , = 2 n π o the spectral decimation leads then to the quadruple eigenvalue: Λ ֒ → = 4 cos π From: Λ ֒ → = 2 + cos π √ π the spectral decimation leads then to the quadruple eigenvalue: Λ ֒ → = 4 cos π ∆ m , m ∈ N , m > As previously, one can easily check that the forbidden eigenvalue
Λ = 2 is not therefore a Dirichletone.One can also check that Λ m = 1 and Λ m = 3 are eigenvalues of ∆ m , both with multiplicity 2.By induction, one may note that, due to the spectral decimation, the initial eigenvalue Λ = 1 givesbirth, at this m th step, to an eigenvalue Λ m , of multiplicity m − . In the same way, the initial eigen-value Λ = 3 gives birth, at this m th step, to an eigenvalue Λ m , of multiplicity m − .Results are summarized in the following array:Initial eigenvalue Λ continued eigenvalue Λ continued eigenvalue Λ ֒ → continued eigenvalue Λ ֒ → π √ π π
27 4 cos π
813 2 n π o π
54 2 n π o Property 4.1.
Let us introduce:
Λ = lim m → + ∞ η − − D W N (5 − D W ) mb One may note that, due to the definition of the Laplacian ∆ , the limit exists. .5 Eigenvalue counting function Definition 4.1. Eigenvalue counting function
Let us introduce the eigenvalue counting function, related to Γ W \ V , such that, for any positivenumber x : N Γ W \ V ( x ) = Card { Λ Dirichlet eigenvalue of − ∆ : Λ x } Property 4.2.
Given a strictly positive integer, the cardinal of V m \ V is: N S m − N S = 2 N mb − Let us denote by Λ sm the largest eigenvalue, which is such that: η − D W N mb × Λ sm η − D W N mb × This leads to: N Γ W ( C N mb ) = 2 N mb − , C If one looks for an asymptotic growth rate of the form N Γ W ( x ) ∼ x α one obtains: α = 1 By following [29], one may note that the ratio N Γ W ( x ) x is bounded above and away from zero, and admits a limit along any sequence of the form C N mb , C > , m ∈ N ⋆ .This enables one to deduce the existence of a periodic function g , the period of which is equal to ln N b ,discontinuous at the value , such that: lim x → + ∞ (cid:26) N Γ W ( x ) x − g (ln x ) (cid:27) = 0 Level Cardinal of the Dirichlet spectrum, in the case where N b = 3 m N mb −
21 42 163 524 160 m ln N Γ W ( C N mb ) m in the case where λ = 12 , and N b = 3 . Remark . Existing results of J. Kigami and M. Lapidus [10], and also of R. S. Strichartz [29], yield: N Γ W ( x ) = G ( x ) x α Γ W + O (1) with: α Γ W = d Γ W d Γ W + 1 = ln N b λ ln N b λ + (5 − D W ) ln N b where: d Γ W = ln N b λ (5 − D W ) ln N b is the dimension of the graph Γ W for the resistance metric. Thanks
The author would like to thank R. Str., who suggested for our previous work, the introduction ofspecific energies to fully take into account the very specific geometry of the problem.
References [1] J. Kigami, A harmonic calculus on the Sierpiński spaces, Japan J. Appl. Math., (1989), pages259-290. 272] J. Kigami, Harmonic calculus on p.c.f. self-similar sets, Trans. Amer. Math. Soc., (1993), pages721-755.[3] R. S. Strichartz, Analysis on fractals, Notices of the AMS, (8), 1999, pages 1199-1208.[4] J. Kigami, R. S. Strichartz, K. C. Walker, Constructing a Laplacian on the Diamond Fractal, A.K. Peters, Ltd, Experimental Mathematics, (3), pages 437-448.[5] Cl. David, Laplacian, on the graph of the Weierstrass function, arXiv:1703.03371v1.[6] J. Kaplan, J. Mallet-Paret and J. Yorke, The Lyapunov dimension of a nowhere differentiableattracting torus, Ergodic Theory Dynam. Systems , 1984, pages 261âĂŞ281.[7] T.-Y. Hu, K.-S. Lau, Fractal Dimensions and Singularities of the Weierstrass Type Functions,Transactions of the American Mathematical Society, 1993, (2), pages 649-665.[8] Weixiao Shen, Hausdorff dimension of the graphs of the classical Weierstrass functions,arXiv:1505.03986.[9] G. Keller, A simpler proof for the dimension of the graph of the classical Weierstrass function,Ann. Inst. Poincaré, 53(1), 2017, pages 169-181.[10] J. Kigami and M. Lapidus, Weyl’s problem for the spectral distribution of Laplacians on P.C.F.self-similar fractalss, Communications in Mathematical Physics, , 2003, pages 399-444.[11] K. Weierstrass, Über continuirliche Funktionen eines reellen Arguments, die für keinen Werth desletzteren einen bestimmten Differentialquotienten besitzen, 1967, in Karl Weiertrass Mathematis-che Werke, Abhandlungen II, Johnson, Gelesen in der KÃűnigl. Akademie der Wissenchaften am18 Juli 1872, , pages 71-74.[12] E. C. Titschmarsh, The theory of functions, Second edition, Oxford University Press, 1939, pages351-353.[13] G. H. Hardy, Theorems Connected with Maclaurin’s Test for the Convergence of Series, Proc.London Math. Soc., 1911, s2-9 (1), pages 126-144.[14] A. S. Besicovitch, H. D. Ursell, Sets of Fractional Dimensions, Journal of the London MathematicalSociety, (1), 1937, pages 18-25.[15] B. B. Mandelbrot, Fractals: form, chance, and dimension, San Francisco: Freeman, 1977.[16] K. Falconer, The Geometry of Fractal Sets, 1985, Cambridge University Press, pages 114-149.[17] B. Hunt, The Hausdorff dimension of graphs of Weierstrass functions, Proc. Amer. Math. Soc., (3), 1998, pages 791-800.[18] K. Barańsky, B. Bárány, J. Romanowska, On the dimension of the graph of the classical Weierstrassfunction, Advances in Math., , 2014, pages 32-59.[19] G. Keller, A simpler proof for the dimension of the graph of the classical Weierstrass function,Ann. Inst. Poincaré, (1), 2017, pages 169-181.[20] M. F. Barnsley, S. Demko, Iterated Function Systems and the Global Construction of Fractals,The Proceedings of the Royal Society of London, A (399), 1985, pages 243-275.[21] M.V. Berry, and Z. V. Lewis, On the Weierstrass-Mandelbrot function, Proc. R. Soc. Lond., A (370), 1980, pages 459-484.[22] C. Sabot, Existence and uniqueness of diffusions on finitely ramified self-similar fractals, Annalesscientifiques de l’É.N.S. 4 e série, (4), 1997, pages 605-673.2823] A. Beurling, J. Deny, Espaces de Dirichlet. I. Le cas élémentaire, Acta Mathematica, (1), 1985,pages 203-224.[24] M. Fukushima, Y. Oshima, and M. Takeda, Dirichlet forms and symmetric Markov processes,1994, Walter de Gruyter & Co.[25] J. Kigami, Harmonic Analysis for Resistance Forms, Journal of Functional Analysis, , 2003,pages 399-444.[26] N. Riane, 2016, Autour du Laplacien sur des domaines présentant un caractère fractal, Mémoirede recherche, M2 Mathématiques de la modélisation, Université Pierre et Marie Curie-Paris 6.[27] Cl. David et N. Riane, Formes de Dirichlet et fonctions harmoniques sur le graphe de la fonctionde Weierstrass, preprint, HAL.[28] R. S. Strichartz, Function spaces on fractals, Journal of Functional Analysis, 198(1), 2003, pages43-83.[29] R. S. Strichartz, Differential Equations on Fractals, A tutorial, Princeton University Press, 2006.[30] J. E. Hutchinson, Fractals and self similarity, Indiana University Mathematics Journal 30, 1981,pages 713-747.[31] R. S. Strichartz, A. Taylor and T. Zhang, Densities of Self-Similar Measures on the Line, Experi-mental Mathematics, (2), 1995, pages 101-128.[32] M. Fukushima and T. Shima, On a spectral analysis for the Sierpinski gasket, Potential Anal.,1